fix: remove unused $exception_message and $exception_type (#383)

* fix: remove unused

* fix: remove exception type

* fix: wip

.env.example

@@ -0,0 +1,11 @@
+# PostHog API Configuration
+# Copy this file to .env and update with your actual values
+
+# Your project API key (found on the /setup page in PostHog)
+POSTHOG_PROJECT_API_KEY=phc_your_project_api_key_here
+
+# Your personal API key (for local evaluation and other advanced features)
+POSTHOG_PERSONAL_API_KEY=phx_your_personal_api_key_here
+
+# PostHog host URL (remove this line if using posthog.com)
+POSTHOG_HOST=http://localhost:8000

.github/dependabot.yml

@@ -0,0 +1,36 @@
+version: 2
+updates:
+  - package-ecosystem: "pip"
+    directory: "/"
+    schedule:
+      interval: "daily"
+      time: "10:00"
+      timezone: "UTC"
+    groups:
+      ai-providers:
+        patterns:
+          - "openai"
+          - "anthropic"
+          - "google-genai"
+          - "langchain-core"
+          - "langchain-community"
+          - "langchain-openai"
+          - "langchain-anthropic"
+          - "langgraph"
+    allow:
+      - dependency-name: "openai"
+      - dependency-name: "anthropic"
+      - dependency-name: "google-genai"
+      - dependency-name: "langchain-core"
+      - dependency-name: "langchain-community"
+      - dependency-name: "langchain-openai"
+      - dependency-name: "langchain-anthropic"
+      - dependency-name: "langgraph"
+    open-pull-requests-limit: 1
+    reviewers:
+      - "PostHog/team-llm-analytics"
+    # Uncomment below to enable auto-merge for minor updates when CI passes
+    # pull-request-branch-name:
+    #   separator: "/"
+    # assignees:
+    #   - "PostHog/ai-team"

.github/workflows/ci.yml

@@ -3,6 +3,9 @@ name: CI
 on:
     - pull_request
 
+permissions:
+    contents: read
+
 jobs:
     code-quality:
         name: Code quality checks
@@ -33,6 +36,10 @@ jobs:
               run: |
                   ruff format --check .
 
+            - name: Lint with ruff
+              run: |
+                  ruff check .
+
             - name: Check types with mypy
               run: |
                   mypy --no-site-packages --config-file mypy.ini  . | mypy-baseline filter
@@ -42,7 +49,7 @@ jobs:
         runs-on: ubuntu-latest
         strategy:
             matrix:
-                python-version: ['3.9', '3.10', '3.11', '3.12', '3.13']
+                python-version: ['3.10', '3.11', '3.12', '3.13', '3.14']
 
         steps:
             - uses: actions/checkout@85e6279cec87321a52edac9c87bce653a07cf6c2
@@ -68,3 +75,34 @@ jobs:
             - name: Run posthog tests
               run: |
                   pytest --verbose --timeout=30
+
+    django5-integration:
+        name: Django 5 integration tests
+        runs-on: ubuntu-latest
+
+        steps:
+            - uses: actions/checkout@85e6279cec87321a52edac9c87bce653a07cf6c2
+              with:
+                  fetch-depth: 1
+
+            - name: Set up Python 3.12
+              uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55
+              with:
+                  python-version: 3.12
+
+            - name: Install uv
+              uses: astral-sh/setup-uv@0c5e2b8115b80b4c7c5ddf6ffdd634974642d182 # v5.4.1
+              with:
+                  enable-cache: true
+                  pyproject-file: 'integration_tests/django5/pyproject.toml'
+
+            - name: Install Django 5 test project dependencies
+              shell: bash
+              working-directory: integration_tests/django5
+              run: |
+                  UV_PROJECT_ENVIRONMENT=$pythonLocation uv sync
+
+            - name: Run Django 5 middleware integration tests
+              working-directory: integration_tests/django5
+              run: |
+                  uv run pytest test_middleware.py test_exception_capture.py --verbose

.github/workflows/generate-references.yml

@@ -0,0 +1,48 @@
+name: "Generate References"
+
+on:
+  workflow_dispatch:
+
+jobs:
+  docs-generation:
+    name: Generate references
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout the repository
+        uses: actions/checkout@85e6279cec87321a52edac9c87bce653a07cf6c2
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.POSTHOG_BOT_PAT }}
+
+      - name: Set up Python
+        uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55
+        with:
+          python-version: 3.11.11
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@0c5e2b8115b80b4c7c5ddf6ffdd634974642d182 # v5.4.1
+        with:
+            enable-cache: true
+            pyproject-file: 'pyproject.toml'
+
+      - name: Generate references
+        run: |
+          uv run bin/docs generate-references
+
+      - name: Check for changes in references
+        id: changes
+        run: |
+          if [ -n "$(git status --porcelain references/)" ]; then
+            echo "changed=true" >> $GITHUB_OUTPUT
+            echo "New references generated in references directory:"
+            git status --porcelain references/
+          else
+            echo "changed=false" >> $GITHUB_OUTPUT
+            echo "No new references generated in references directory"
+          fi
+
+      - uses: stefanzweifel/git-auto-commit-action@778341af668090896ca464160c2def5d1d1a3eb0
+        if: steps.changes.outputs.changed == 'true'
+        with:
+          commit_message: "Update generated references"
+          file_pattern: references/

.github/workflows/release.yml

@@ -20,7 +20,7 @@ jobs:
         uses: actions/checkout@85e6279cec87321a52edac9c87bce653a07cf6c2
         with:
           fetch-depth: 0
-          token: ${{ secrets.POSTHOG_BOT_GITHUB_TOKEN }}
+          token: ${{ secrets.POSTHOG_BOT_PAT }}
 
       - name: Set up Python
         uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55
@@ -45,7 +45,13 @@ jobs:
       - name: Create GitHub release
         uses: actions/create-release@0cb9c9b65d5d1901c1f53e5e66eaf4afd303e70e # v1
         env:
-          GITHUB_TOKEN: ${{ secrets.POSTHOG_BOT_GITHUB_TOKEN }}
+          GITHUB_TOKEN: ${{ secrets.POSTHOG_BOT_PAT }}
         with:
           tag_name: v${{ env.REPO_VERSION }}
           release_name: ${{ env.REPO_VERSION }}
+      
+      - name: Dispatch generate-references for posthog-python
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          gh workflow run generate-references.yml --ref master

.gitignore

@@ -19,3 +19,4 @@ pyrightconfig.json
 .env
 .DS_Store
 posthog-python-references.json
+.claude/settings.local.json

CHANGELOG.md

@@ -1,6 +1,149 @@
+# 7.3.1 - 2025-12-06
+
+fix: remove unused $exception_message and $exception_type
+
+# 7.3.0 - 2025-12-05
+
+feat: improve code variables capture masking
+
+# 7.2.0 - 2025-12-01
+
+feat: add $feature_flag_evaluated_at properties to $feature_flag_called events
+
+# 7.1.0 - 2025-11-26
+
+Add support for the async version of Gemini.
+
+# 7.0.2 - 2025-11-18
+
+Add support for Python 3.14.
+Projects upgrading to Python 3.14 should ensure any Pydantic models passed into the SDK use Pydantic v2, as Pydantic v1 is not compatible with Python 3.14.
+
+# 7.0.1 - 2025-11-15
+
+Try to use repr() when formatting code variables
+
+# 7.0.0 - 2025-11-11
+
+NB Python 3.9 is no longer supported
+
+- chore(llma): update LLM provider SDKs to latest major versions
+  - openai: 1.102.0 → 2.7.1
+  - anthropic: 0.64.0 → 0.72.0
+  - google-genai: 1.32.0 → 1.49.0
+  - langchain-core: 0.3.75 → 1.0.3
+  - langchain-openai: 0.3.32 → 1.0.2
+  - langchain-anthropic: 0.3.19 → 1.0.1
+  - langchain-community: 0.3.29 → 0.4.1
+  - langgraph: 0.6.6 → 1.0.2
+
+# 6.9.3 - 2025-11-10
+
+- feat(ph-ai): PostHog properties dict in GenerationMetadata
+
+# 6.9.2 - 2025-11-10
+
+- fix(llma): fix cache token double subtraction in Langchain for non-Anthropic providers causing negative costs
+
+# 6.9.1 - 2025-11-07
+
+- fix(error-tracking): pass code variables config from init to client
+
+# 6.9.0 - 2025-11-06
+
+- feat(error-tracking): add local variables capture
+
+# 6.8.0 - 2025-11-03
+
+- feat(llma): send web search calls to be used for LLM cost calculations
+
+# 6.7.14 - 2025-11-03
+
+- fix(django): Handle request.user access in async middleware context to prevent SynchronousOnlyOperation errors in Django 5+ (fixes #355)
+- test(django): Add Django 5 integration test suite with real ASGI application testing async middleware behavior
+
+# 6.7.13 - 2025-11-02
+
+- fix(llma): cache cost calculation in the LangChain callback
+
+# 6.7.12 - 2025-11-02
+
+- fix(django): Restore process_exception method to capture view and downstream middleware exceptions (fixes #329)
+- fix(ai/langchain): Add LangChain 1.0+ compatibility for CallbackHandler imports (fixes #362)
+
+# 6.7.11 - 2025-10-28
+
+- feat(ai): Add `$ai_framework` property for framework integrations (e.g. LangChain)
+
+# 6.7.10 - 2025-10-24
+
+- fix(django): Make middleware truly hybrid - compatible with both sync (WSGI) and async (ASGI) Django stacks without breaking sync-only deployments
+
+# 6.7.9 - 2025-10-22
+
+- fix(flags): multi-condition flags with static cohorts returning wrong variants
+
+# 6.7.8 - 2025-10-16
+
+- fix(llma): missing async for OpenAI's streaming implementation
+
+# 6.7.7 - 2025-10-14
+
+- fix: remove deprecated attribute $exception_personURL from exception events
+
+# 6.7.6 - 2025-09-16
+
+- fix: don't sort condition sets with variant overrides to the top
+- fix: Prevent core Client methods from raising exceptions
+
+# 6.7.5 - 2025-09-16
+
+- feat: Django middleware now supports async request handling.
+
+# 6.7.4 - 2025-09-05
+
+- fix: Missing system prompts for some providers
+
+# 6.7.3 - 2025-09-04
+
+- fix: missing usage tokens in Gemini
+
+# 6.7.2 - 2025-09-03
+
+- fix: tool call results in streaming providers
+
+# 6.7.1 - 2025-09-01
+
+- fix: Add base64 inline image sanitization
+
+# 6.7.0 - 2025-08-26
+
+- feat: Add support for feature flag dependencies
+
+# 6.6.1 - 2025-08-21
+
+- fix: Prevent `NoneType` error when `group_properties` is `None`
+
+# 6.6.0 - 2025-08-15
+
+- feat: Add `flag_keys_to_evaluate` parameter to optimize feature flag evaluation performance by only evaluating specified flags
+- feat: Add `flag_keys_filter` option to `send_feature_flags` for selective flag evaluation in capture events
+
+# 6.5.0 - 2025-08-08
+
+- feat: Add `$context_tags` to an event to know which properties were included as tags
+
+# 6.4.1 - 2025-08-06
+
+- fix: Always pass project API key in `remote_config` requests for deterministic project routing
+
+# 6.4.0 - 2025-08-05
+
+- feat: support Vertex AI for Gemini
+
 # 6.3.4 - 2025-08-04
 
-- fix: Set `$ai_tools` for all providers and `$ai_output_choices` for all non-streaming provider flows properly
+- fix: set `$ai_tools` for all providers and `$ai_output_choices` for all non-streaming provider flows properly
 
 # 6.3.3 - 2025-08-01
 

README.md

@@ -30,8 +30,8 @@ We recommend using [uv](https://docs.astral.sh/uv/). It's super fast.
 ## PostHog recommends `uv` so...
 
 ```bash
-uv python install 3.9.19
-uv python pin 3.9.19
+uv python install 3.12
+uv python pin 3.12
 uv venv
 source env/bin/activate
 uv sync --extra dev --extra test

bin/docs_scripts/doc_constant.py

@@ -3,50 +3,11 @@ Constants for PostHog Python SDK documentation generation.
 """
 
 from typing import Dict, Union
-
-# Types that are built-in to Python and don't need to be documented
-NO_DOCS_TYPES = [
-    "Client",
-    "any",
-    "int",
-    "float",
-    "bool",
-    "dict",
-    "list",
-    "str",
-    "tuple",
-    "set",
-    "frozenset",
-    "bytes",
-    "bytearray",
-    "memoryview",
-    "range",
-    "slice",
-    "complex",
-    "Union",
-    "Optional",
-    "Any",
-    "Callable",
-    "Type",
-    "TypeVar",
-    "Generic",
-    "Literal",
-    "ClassVar",
-    "Final",
-    "Annotated",
-    "NotRequired",
-    "Required",
-    "None",
-    "NoneType",
-    "object",
-    "Unpack",
-    "BaseException",
-    "Exception",
-]
+from posthog.version import VERSION
 
 # Documentation generation metadata
 DOCUMENTATION_METADATA = {
-    "hogRef": "0.1",
+    "hogRef": "0.3",
     "slugPrefix": "posthog-python",
     "specUrl": "https://github.com/PostHog/posthog-python",
 }
@@ -67,8 +28,9 @@ DOCSTRING_PATTERNS = {
 
 # Output file configuration
 OUTPUT_CONFIG: Dict[str, Union[str, int]] = {
-    "output_dir": ".",
-    "filename": "posthog-python-references.json",
+    "output_dir": "./references",
+    "filename": f"posthog-python-references-{VERSION}.json",
+    "filename_latest": "posthog-python-references-latest.json",
     "indent": 2,
 }
 

bin/docs_scripts/generate_json_schemas.py

@@ -11,7 +11,6 @@ from dataclasses import is_dataclass, fields
 from typing import get_origin, get_args, Union
 from textwrap import dedent
 from doc_constant import (
-    NO_DOCS_TYPES,
     DOCUMENTATION_METADATA,
     DOCSTRING_PATTERNS,
     OUTPUT_CONFIG,
@@ -187,7 +186,7 @@ def analyze_parameter(param: inspect.Parameter, docstring: str = "") -> dict:
         param_type = get_type_name(type(param.default))
 
     # Extract parameter description from Args section
-    param_description = f"Parameter: {param.name}"
+    param_description = ""
     if docstring:
         # Look for Args section and extract description for this parameter
         args_section_match = re.search(
@@ -378,6 +377,14 @@ def generate_sdk_documentation():
             except Exception as e:
                 print(f"Error analyzing type {name}: {e}")
 
+    # Clean types of empty types
+
+    # Remove types that have no properties and no examples
+    # Remove types that have no properties and no examples
+    types_list = [
+        t for t in types_list if len(t["properties"]) > 0 or t["example"] != ""
+    ]
+
     # Collect classes
     classes_list = []
 
@@ -420,14 +427,28 @@ def generate_sdk_documentation():
             }
         )
 
+    # Collect categories from functions
+    categories = ["Initialization", "Identification", "Capture"]
+    seen_categories = set(categories)
+    for class_info in classes_list:
+        if "functions" in class_info:
+            for func in class_info["functions"]:
+                if (
+                    "category" in func
+                    and func["category"] not in seen_categories
+                    and func["category"]
+                ):
+                    categories.append(func["category"])
+                    seen_categories.add(func["category"])
+
     # Create the final structure
     result = {
         "id": "posthog-python",
         "hogRef": DOCUMENTATION_METADATA["hogRef"],
         "info": sdk_info,
-        "noDocsTypes": NO_DOCS_TYPES,
         "types": types_list,
         "classes": classes_list,
+        "categories": categories,
     }
 
     return result
@@ -439,12 +460,23 @@ if __name__ == "__main__":
     try:
         documentation = generate_sdk_documentation()
 
-        # Write to file
+        # Ensure output directory exists
+        output_dir = str(OUTPUT_CONFIG["output_dir"])
+        os.makedirs(output_dir, exist_ok=True)
+
         output_file = os.path.join(
             str(OUTPUT_CONFIG["output_dir"]), str(OUTPUT_CONFIG["filename"])
         )
+        output_file_latest = os.path.join(
+            str(OUTPUT_CONFIG["output_dir"]), str(OUTPUT_CONFIG["filename_latest"])
+        )
+
+        # Write to current version
         with open(output_file, "w") as f:
             json.dump(documentation, f, indent=int(OUTPUT_CONFIG["indent"]))
+        # Write to latest
+        with open(output_file_latest, "w") as f:
+            json.dump(documentation, f, indent=int(OUTPUT_CONFIG["indent"]))
 
         print(f"✓ Generated {output_file}")
 
@@ -459,12 +491,6 @@ if __name__ == "__main__":
         print(f"   • {classes_count} classes documented")
         print(f"   • {total_functions} functions documented")
 
-        no_docs = documentation["noDocsTypes"]
-        if no_docs:
-            print(
-                f"   • {len(no_docs)} types without documentation: {', '.join(no_docs[:5])}{'...' if len(no_docs) > 5 else ''}"
-            )
-
     except Exception as e:
         print(f"❌ Error generating documentation: {e}")
         import traceback

bin/fmt

@@ -6,9 +6,7 @@ set_source_and_root_dir
 ensure_virtual_env
 
 if [[ "$1" == "--check" ]]; then
-  black --check .
-  isort --check-only .
+  ruff format --check .
 else
-  black .
-  isort .
+  ruff format .
 fi

example.py

@@ -1,186 +1,502 @@
 # PostHog Python library example
-import argparse
+#
+# This script demonstrates various PostHog Python SDK capabilities including:
+# - Basic event capture and user identification
+# - Feature flag local evaluation
+# - Feature flag payloads
+# - Context management and tagging
+#
+# Setup:
+# 1. Copy .env.example to .env and fill in your PostHog credentials
+# 2. Run this script and choose from the interactive menu
+
+import os
 
 import posthog
 
-# Add argument parsing
-parser = argparse.ArgumentParser(description="PostHog Python library example")
-parser.add_argument(
-    "--flag",
-    default="person-on-events-enabled",
-    help="Feature flag key to check (default: person-on-events-enabled)",
-)
-args = parser.parse_args()
 
-posthog.debug = True
+def load_env_file():
+    """Load environment variables from .env file if it exists."""
+    env_path = os.path.join(os.path.dirname(__file__), ".env")
+    if os.path.exists(env_path):
+        with open(env_path, "r") as f:
+            for line in f:
+                line = line.strip()
+                if line and not line.startswith("#") and "=" in line:
+                    key, value = line.split("=", 1)
+                    os.environ.setdefault(key.strip(), value.strip())
 
-# You can find this key on the /setup page in PostHog
-posthog.project_api_key = "phc_gtWmTq3Pgl06u4sZY3TRcoQfp42yfuXHKoe8ZVSR6Kh"
-posthog.personal_api_key = "phx_fiRCOQkTA3o2ePSdLrFDAILLHjMu2Mv52vUi8MNruIm"
 
-# Where you host PostHog, with no trailing /.
-# You can remove this line if you're using posthog.com
-posthog.host = "http://localhost:8000"
-posthog.poll_interval = 10
+# Load .env file if it exists
+load_env_file()
 
-print(
-    posthog.feature_enabled(
-        args.flag,  # Use the flag from command line arguments
-        "12345",
-        groups={"organization": str("0182ee91-8ef7-0000-4cb9-fedc5f00926a")},
-        group_properties={
-            "organization": {
-                "id": "0182ee91-8ef7-0000-4cb9-fedc5f00926a",
-                "created_at": "2022-06-30 11:44:52.984121+00:00",
-            }
-        },
+# Get configuration
+project_key = os.getenv("POSTHOG_PROJECT_API_KEY", "")
+personal_api_key = os.getenv("POSTHOG_PERSONAL_API_KEY", "")
+host = os.getenv("POSTHOG_HOST", "http://localhost:8000")
+
+# Check if credentials are provided
+if not project_key or not personal_api_key:
+    print("❌ Missing PostHog credentials!")
+    print(
+        "   Please set POSTHOG_PROJECT_API_KEY and POSTHOG_PERSONAL_API_KEY environment variables"
+    )
+    print("   or copy .env.example to .env and fill in your values")
+    exit(1)
+
+# Test authentication before proceeding
+print("🔑 Testing PostHog authentication...")
+
+try:
+    # Configure PostHog with credentials
+    posthog.debug = False  # Keep quiet during auth test
+    posthog.api_key = project_key
+    posthog.project_api_key = project_key
+    posthog.personal_api_key = personal_api_key
+    posthog.host = host
+    posthog.poll_interval = 10
+
+    # Test by attempting to get feature flags (this validates both keys)
+    # This will fail if credentials are invalid
+    test_flags = posthog.get_all_flags("test_user", only_evaluate_locally=True)
+
+    # If we get here without exception, credentials work
+    print("✅ Authentication successful!")
+    print(f"   Project API Key: {project_key[:9]}...")
+    print("   Personal API Key: [REDACTED]")
+    print(f"   Host: {host}\n\n")
+
+except Exception as e:
+    print("❌ Authentication failed!")
+    print(f"   Error: {e}")
+    print("\n   Please check your credentials:")
+    print("   - POSTHOG_PROJECT_API_KEY: Project API key from PostHog settings")
+    print(
+        "   - POSTHOG_PERSONAL_API_KEY: Personal API key (required for local evaluation)"
+    )
+    print("   - POSTHOG_HOST: Your PostHog instance URL")
+    exit(1)
+
+# Display menu and get user choice
+print("🚀 PostHog Python SDK Demo - Choose an example to run:\n")
+print("1. Identify and capture examples")
+print("2. Feature flag local evaluation examples")
+print("3. Feature flag payload examples")
+print("4. Flag dependencies examples")
+print("5. Context management and tagging examples")
+print("6. Run all examples")
+print("7. Exit")
+choice = input("\nEnter your choice (1-7): ").strip()
+
+if choice == "1":
+    print("\n" + "=" * 60)
+    print("IDENTIFY AND CAPTURE EXAMPLES")
+    print("=" * 60)
+
+    posthog.debug = True
+
+    # Capture an event
+    print("📊 Capturing events...")
+    posthog.capture(
+        "event",
+        distinct_id="distinct_id",
+        properties={"property1": "value", "property2": "value"},
+        send_feature_flags=True,
+    )
+
+    # Alias a previous distinct id with a new one
+    print("🔗 Creating alias...")
+    posthog.alias("distinct_id", "new_distinct_id")
+
+    posthog.capture(
+        "event2",
+        distinct_id="new_distinct_id",
+        properties={"property1": "value", "property2": "value"},
+    )
+    posthog.capture(
+        "event-with-groups",
+        distinct_id="new_distinct_id",
+        properties={"property1": "value", "property2": "value"},
+        groups={"company": "id:5"},
+    )
+
+    # Add properties to the person
+    print("👤 Identifying user...")
+    posthog.set(
+        distinct_id="new_distinct_id", properties={"email": "something@something.com"}
+    )
+
+    # Add properties to a group
+    print("🏢 Identifying group...")
+    posthog.group_identify("company", "id:5", {"employees": 11})
+
+    # Properties set only once to the person
+    print("🔒 Setting properties once...")
+    posthog.set_once(
+        distinct_id="new_distinct_id", properties={"self_serve_signup": True}
+    )
+
+    # This will not change the property (because it was already set)
+    posthog.set_once(
+        distinct_id="new_distinct_id", properties={"self_serve_signup": False}
+    )
+
+    print("🔄 Updating properties...")
+    posthog.set(distinct_id="new_distinct_id", properties={"current_browser": "Chrome"})
+    posthog.set(
+        distinct_id="new_distinct_id", properties={"current_browser": "Firefox"}
+    )
+
+elif choice == "2":
+    print("\n" + "=" * 60)
+    print("FEATURE FLAG LOCAL EVALUATION EXAMPLES")
+    print("=" * 60)
+
+    posthog.debug = True
+
+    print("🏁 Testing basic feature flags...")
+    print(
+        f"beta-feature for 'distinct_id': {posthog.feature_enabled('beta-feature', 'distinct_id')}"
+    )
+    print(
+        f"beta-feature for 'new_distinct_id': {posthog.feature_enabled('beta-feature', 'new_distinct_id')}"
+    )
+    print(
+        f"beta-feature with groups: {posthog.feature_enabled('beta-feature-groups', 'distinct_id', groups={'company': 'id:5'})}"
+    )
+
+    print("\n🌍 Testing location-based flags...")
+    # Assume test-flag has `City Name = Sydney` as a person property set
+    print(
+        f"Sydney user: {posthog.feature_enabled('test-flag', 'random_id_12345', person_properties={'$geoip_city_name': 'Sydney'})}"
+    )
+
+    print(
+        f"Sydney user (local only): {posthog.feature_enabled('test-flag', 'distinct_id_random_22', person_properties={'$geoip_city_name': 'Sydney'}, only_evaluate_locally=True)}"
+    )
+
+    print("\n📋 Getting all flags...")
+    print(f"All flags: {posthog.get_all_flags('distinct_id_random_22')}")
+    print(
+        f"All flags (local): {posthog.get_all_flags('distinct_id_random_22', only_evaluate_locally=True)}"
+    )
+    print(
+        f"All flags with properties: {posthog.get_all_flags('distinct_id_random_22', person_properties={'$geoip_city_name': 'Sydney'}, only_evaluate_locally=True)}"
+    )
+
+elif choice == "3":
+    print("\n" + "=" * 60)
+    print("FEATURE FLAG PAYLOAD EXAMPLES")
+    print("=" * 60)
+
+    posthog.debug = True
+
+    print("📦 Testing feature flag payloads...")
+    print(
+        f"beta-feature payload: {posthog.get_feature_flag_payload('beta-feature', 'distinct_id')}"
+    )
+    print(
+        f"All flags and payloads: {posthog.get_all_flags_and_payloads('distinct_id')}"
+    )
+    print(
+        f"Remote config payload: {posthog.get_remote_config_payload('encrypted_payload_flag_key')}"
+    )
+
+    # Get feature flag result with all details (enabled, variant, payload, key, reason)
+    print("\n🔍 Getting detailed flag result...")
+    result = posthog.get_feature_flag_result("beta-feature", "distinct_id")
+    if result:
+        print(f"Flag key: {result.key}")
+        print(f"Flag enabled: {result.enabled}")
+        print(f"Variant: {result.variant}")
+        print(f"Payload: {result.payload}")
+        print(f"Reason: {result.reason}")
+        # get_value() returns the variant if it exists, otherwise the enabled value
+        print(f"Value (variant or enabled): {result.get_value()}")
+
+elif choice == "4":
+    print("\n" + "=" * 60)
+    print("FLAG DEPENDENCIES EXAMPLES")
+    print("=" * 60)
+    print("🔗 Testing flag dependencies with local evaluation...")
+    print(
+        "   Flag structure: 'test-flag-dependency' depends on 'beta-feature' being enabled"
+    )
+    print("")
+    print("📋 Required setup (if 'test-flag-dependency' doesn't exist):")
+    print("   1. Create feature flag 'beta-feature':")
+    print("      - Condition: email contains '@example.com'")
+    print("      - Rollout: 100%")
+    print("   2. Create feature flag 'test-flag-dependency':")
+    print("      - Condition: flag 'beta-feature' is enabled")
+    print("      - Rollout: 100%")
+    print("")
+
+    posthog.debug = True
+
+    # Test @example.com user (should satisfy dependency if flags exist)
+    result1 = posthog.feature_enabled(
+        "test-flag-dependency",
+        "example_user",
+        person_properties={"email": "user@example.com"},
         only_evaluate_locally=True,
     )
-)
+    print(f"✅ @example.com user (test-flag-dependency): {result1}")
 
-
-# Capture an event
-posthog.capture(
-    "event",
-    distinct_id="distinct_id",
-    properties={"property1": "value", "property2": "value"},
-    send_feature_flags=True,
-)
-
-print(posthog.feature_enabled("beta-feature", "distinct_id"))
-print(
-    posthog.feature_enabled(
-        "beta-feature-groups", "distinct_id", groups={"company": "id:5"}
-    )
-)
-
-print(posthog.feature_enabled("beta-feature", "distinct_id"))
-
-# get payload
-print(posthog.get_feature_flag_payload("beta-feature", "distinct_id"))
-print(posthog.get_all_flags_and_payloads("distinct_id"))
-
-# get feature flag result with all details (enabled, variant, payload, key, reason)
-result = posthog.get_feature_flag_result("beta-feature", "distinct_id")
-if result:
-    print(f"Flag key: {result.key}")
-    print(f"Flag enabled: {result.enabled}")
-    print(f"Variant: {result.variant}")
-    print(f"Payload: {result.payload}")
-    print(f"Reason: {result.reason}")
-    # get_value() returns the variant if it exists, otherwise the enabled value
-    print(f"Value (variant or enabled): {result.get_value()}")
-
-# Alias a previous distinct id with a new one
-
-posthog.alias("distinct_id", "new_distinct_id")
-
-posthog.capture(
-    "event2",
-    distinct_id="new_distinct_id",
-    properties={"property1": "value", "property2": "value"},
-)
-posthog.capture(
-    "event-with-groups",
-    distinct_id="new_distinct_id",
-    properties={"property1": "value", "property2": "value"},
-    groups={"company": "id:5"},
-)
-
-# # Add properties to the person
-posthog.set(
-    distinct_id="new_distinct_id", properties={"email": "something@something.com"}
-)
-
-# Add properties to a group
-posthog.group_identify("company", "id:5", {"employees": 11})
-
-# properties set only once to the person
-posthog.set_once(distinct_id="new_distinct_id", properties={"self_serve_signup": True})
-
-
-posthog.set_once(
-    distinct_id="new_distinct_id", properties={"self_serve_signup": False}
-)  # this will not change the property (because it was already set)
-
-posthog.set(distinct_id="new_distinct_id", properties={"current_browser": "Chrome"})
-posthog.set(distinct_id="new_distinct_id", properties={"current_browser": "Firefox"})
-
-
-# #############################################################################
-# Make sure you have a personal API key for the examples below
-
-# Local Evaluation
-
-# If flag has City=Sydney, this call doesn't go to `/decide`
-print(
-    posthog.feature_enabled(
-        "test-flag",
-        "distinct_id_random_22",
-        person_properties={"$geoip_city_name": "Sydney"},
-    )
-)
-
-print(
-    posthog.feature_enabled(
-        "test-flag",
-        "distinct_id_random_22",
-        person_properties={"$geoip_city_name": "Sydney"},
+    # Test non-example.com user (dependency should not be satisfied)
+    result2 = posthog.feature_enabled(
+        "test-flag-dependency",
+        "regular_user",
+        person_properties={"email": "user@other.com"},
         only_evaluate_locally=True,
     )
-)
+    print(f"❌ Regular user (test-flag-dependency): {result2}")
 
-
-print(posthog.get_all_flags("distinct_id_random_22"))
-print(posthog.get_all_flags("distinct_id_random_22", only_evaluate_locally=True))
-print(
-    posthog.get_all_flags(
-        "distinct_id_random_22",
-        person_properties={"$geoip_city_name": "Sydney"},
+    # Test beta-feature directly for comparison
+    beta1 = posthog.feature_enabled(
+        "beta-feature",
+        "example_user",
+        person_properties={"email": "user@example.com"},
         only_evaluate_locally=True,
     )
-)
-print(posthog.get_remote_config_payload("encrypted_payload_flag_key"))
+    beta2 = posthog.feature_enabled(
+        "beta-feature",
+        "regular_user",
+        person_properties={"email": "user@other.com"},
+        only_evaluate_locally=True,
+    )
+    print(f"📊 Beta feature comparison - @example.com: {beta1}, regular: {beta2}")
 
+    print("\n🎯 Results Summary:")
+    print(
+        f"   - Flag dependencies evaluated locally: {'✅ YES' if result1 != result2 else '❌ NO'}"
+    )
+    print("   - Zero API calls needed: ✅ YES (all evaluated locally)")
+    print("   - Python SDK supports flag dependencies: ✅ YES")
 
-# You can add tags to a context, and these are automatically added to any events (including exceptions) captured
-# within that context.
+    print("\n" + "-" * 60)
+    print("PRODUCTION-STYLE MULTIVARIATE DEPENDENCY CHAIN")
+    print("-" * 60)
+    print("🔗 Testing complex multivariate flag dependencies...")
+    print(
+        "   Structure: multivariate-root-flag -> multivariate-intermediate-flag -> multivariate-leaf-flag"
+    )
+    print("")
+    print("📋 Required setup (if flags don't exist):")
+    print(
+        "   1. Create 'multivariate-leaf-flag' with fruit variants (pineapple, mango, papaya, kiwi)"
+    )
+    print("      - pineapple: email = 'pineapple@example.com'")
+    print("      - mango: email = 'mango@example.com'")
+    print(
+        "   2. Create 'multivariate-intermediate-flag' with color variants (blue, red)"
+    )
+    print("      - blue: depends on multivariate-leaf-flag = 'pineapple'")
+    print("      - red: depends on multivariate-leaf-flag = 'mango'")
+    print(
+        "   3. Create 'multivariate-root-flag' with show variants (breaking-bad, the-wire)"
+    )
+    print("      - breaking-bad: depends on multivariate-intermediate-flag = 'blue'")
+    print("      - the-wire: depends on multivariate-intermediate-flag = 'red'")
+    print("")
 
-# You can enter a new context using a with statement. Any exceptions thrown in the context will be captured,
-# and tagged with the context tags. Other events captured will also be tagged with the context tags. By default,
-# the new context inherits tags from the parent context.
-with posthog.new_context():
-    posthog.tag("transaction_id", "abc123")
-    posthog.tag("some_arbitrary_value", {"tags": "can be dicts"})
+    # Test pineapple -> blue -> breaking-bad chain
+    dependent_result3 = posthog.get_feature_flag(
+        "multivariate-root-flag",
+        "regular_user",
+        person_properties={"email": "pineapple@example.com"},
+        only_evaluate_locally=True,
+    )
+    if str(dependent_result3) != "breaking-bad":
+        print(
+            f"     ❌ Something went wrong evaluating 'multivariate-root-flag' with pineapple@example.com. Expected 'breaking-bad', got '{dependent_result3}'"
+        )
+    else:
+        print("✅ 'multivariate-root-flag' with email pineapple@example.com succeeded")
 
-    # This event will be captured with the tags set above
-    posthog.capture("order_processed")
-    # This exception will be captured with the tags set above
-    raise Exception("Order processing failed")
+    # Test mango -> red -> the-wire chain
+    dependent_result4 = posthog.get_feature_flag(
+        "multivariate-root-flag",
+        "regular_user",
+        person_properties={"email": "mango@example.com"},
+        only_evaluate_locally=True,
+    )
+    if str(dependent_result4) != "the-wire":
+        print(
+            f"     ❌ Something went wrong evaluating multivariate-root-flag with mango@example.com. Expected 'the-wire', got '{dependent_result4}'"
+        )
+    else:
+        print("✅ 'multivariate-root-flag' with email mango@example.com succeeded")
 
+    # Show the complete chain evaluation
+    print("\n🔍 Complete dependency chain evaluation:")
+    for email, expected_chain in [
+        ("pineapple@example.com", ["pineapple", "blue", "breaking-bad"]),
+        ("mango@example.com", ["mango", "red", "the-wire"]),
+    ]:
+        leaf = posthog.get_feature_flag(
+            "multivariate-leaf-flag",
+            "regular_user",
+            person_properties={"email": email},
+            only_evaluate_locally=True,
+        )
+        intermediate = posthog.get_feature_flag(
+            "multivariate-intermediate-flag",
+            "regular_user",
+            person_properties={"email": email},
+            only_evaluate_locally=True,
+        )
+        root = posthog.get_feature_flag(
+            "multivariate-root-flag",
+            "regular_user",
+            person_properties={"email": email},
+            only_evaluate_locally=True,
+        )
 
-# Use fresh=True to start with a clean context (no inherited tags)
-with posthog.new_context(fresh=True):
-    posthog.tag("session_id", "xyz789")
-    # Only session_id tag will be present, no inherited tags
-    raise Exception("Session handling failed")
+        actual_chain = [str(leaf), str(intermediate), str(root)]
+        chain_success = actual_chain == expected_chain
 
+        print(f"   📧 {email}:")
+        print(f"      Expected: {' -> '.join(map(str, expected_chain))}")
+        print(f"      Actual:   {' -> '.join(map(str, actual_chain))}")
+        print(f"      Status:   {'✅ SUCCESS' if chain_success else '❌ FAILED'}")
 
-# You can also use the `@posthog.scoped()` decorator to enter a new context.
-# By default, it inherits tags from the parent context
-@posthog.scoped()
-def process_order(order_id):
-    posthog.tag("order_id", order_id)
-    # Exception will be captured and tagged automatically
-    raise Exception("Order processing failed")
+    print("\n🎯 Multivariate Chain Summary:")
+    print("   - Complex dependency chains: ✅ SUPPORTED")
+    print("   - Multivariate flag dependencies: ✅ SUPPORTED")
+    print("   - Local evaluation of chains: ✅ WORKING")
 
+elif choice == "5":
+    print("\n" + "=" * 60)
+    print("CONTEXT MANAGEMENT AND TAGGING EXAMPLES")
+    print("=" * 60)
 
-# Use fresh=True to start with a clean context (no inherited tags)
-@posthog.scoped(fresh=True)
-def process_payment(payment_id):
-    posthog.tag("payment_id", payment_id)
-    # Only payment_id tag will be present, no inherited tags
-    raise Exception("Payment processing failed")
+    posthog.debug = True
 
+    print("🏷️ Testing context management...")
+    print(
+        "You can add tags to a context, and these are automatically added to any events captured within that context."
+    )
+
+    # You can enter a new context using a with statement. Any exceptions thrown in the context will be captured,
+    # and tagged with the context tags. Other events captured will also be tagged with the context tags. By default,
+    # the new context inherits tags from the parent context.
+    try:
+        with posthog.new_context():
+            posthog.tag("transaction_id", "abc123")
+            posthog.tag("some_arbitrary_value", {"tags": "can be dicts"})
+
+            # This event will be captured with the tags set above
+            posthog.capture("order_processed")
+            print("✅ Event captured with inherited context tags")
+            # This exception will be captured with the tags set above
+            # raise Exception("Order processing failed")
+    except Exception as e:
+        print(f"Exception captured: {e}")
+
+    # Use fresh=True to start with a clean context (no inherited tags)
+    try:
+        with posthog.new_context(fresh=True):
+            posthog.tag("session_id", "xyz789")
+            # Only session_id tag will be present, no inherited tags
+            posthog.capture("session_event")
+            print("✅ Event captured with fresh context tags")
+            # raise Exception("Session handling failed")
+    except Exception as e:
+        print(f"Exception captured: {e}")
+
+    # You can also use the `@posthog.scoped()` decorator to enter a new context.
+    # By default, it inherits tags from the parent context
+    @posthog.scoped()
+    def process_order(order_id):
+        posthog.tag("order_id", order_id)
+        posthog.capture("order_step_completed")
+        print(f"✅ Order {order_id} processed with scoped context")
+        # Exception will be captured and tagged automatically
+        # raise Exception("Order processing failed")
+
+    # Use fresh=True to start with a clean context (no inherited tags)
+    @posthog.scoped(fresh=True)
+    def process_payment(payment_id):
+        posthog.tag("payment_id", payment_id)
+        posthog.capture("payment_processed")
+        print(f"✅ Payment {payment_id} processed with fresh scoped context")
+        # Only payment_id tag will be present, no inherited tags
+        # raise Exception("Payment processing failed")
+
+    process_order("12345")
+    process_payment("67890")
+
+elif choice == "6":
+    print("\n🔄 Running all examples...")
+
+    # Run example 1
+    print(f"\n{'🔸' * 20} IDENTIFY AND CAPTURE {'🔸' * 20}")
+    posthog.debug = True
+    print("📊 Capturing events...")
+    posthog.capture(
+        "event",
+        distinct_id="distinct_id",
+        properties={"property1": "value", "property2": "value"},
+        send_feature_flags=True,
+    )
+    print("🔗 Creating alias...")
+    posthog.alias("distinct_id", "new_distinct_id")
+    print("👤 Identifying user...")
+    posthog.set(
+        distinct_id="new_distinct_id", properties={"email": "something@something.com"}
+    )
+
+    # Run example 2
+    print(f"\n{'🔸' * 20} FEATURE FLAGS {'🔸' * 20}")
+    print("🏁 Testing basic feature flags...")
+    print(f"beta-feature: {posthog.feature_enabled('beta-feature', 'distinct_id')}")
+    print(
+        f"Sydney user: {posthog.feature_enabled('test-flag', 'random_id_12345', person_properties={'$geoip_city_name': 'Sydney'})}"
+    )
+
+    # Run example 3
+    print(f"\n{'🔸' * 20} PAYLOADS {'🔸' * 20}")
+    print("📦 Testing payloads...")
+    print(f"Payload: {posthog.get_feature_flag_payload('beta-feature', 'distinct_id')}")
+
+    # Run example 4
+    print(f"\n{'🔸' * 20} FLAG DEPENDENCIES {'🔸' * 20}")
+    print("🔗 Testing flag dependencies...")
+    result1 = posthog.feature_enabled(
+        "test-flag-dependency",
+        "demo_user",
+        person_properties={"email": "user@example.com"},
+        only_evaluate_locally=True,
+    )
+    result2 = posthog.feature_enabled(
+        "test-flag-dependency",
+        "demo_user2",
+        person_properties={"email": "user@other.com"},
+        only_evaluate_locally=True,
+    )
+    print(f"✅ @example.com user: {result1}, regular user: {result2}")
+
+    # Run example 5
+    print(f"\n{'🔸' * 20} CONTEXT MANAGEMENT {'🔸' * 20}")
+    print("🏷️ Testing context management...")
+    with posthog.new_context():
+        posthog.tag("demo_run", "all_examples")
+        posthog.capture("demo_completed")
+        print("✅ Demo completed with context tags")
+
+elif choice == "7":
+    print("👋 Goodbye!")
+    posthog.shutdown()
+    exit()
+
+else:
+    print("❌ Invalid choice. Please run again and select 1-7.")
+    posthog.shutdown()
+    exit()
+
+print("\n" + "=" * 60)
+print("✅ Example completed!")
+print("=" * 60)
 
 posthog.shutdown()

integration_tests/django5/.gitignore

@@ -0,0 +1,4 @@
+db.sqlite3
+*.pyc
+__pycache__/
+.pytest_cache/

integration_tests/django5/manage.py

@@ -0,0 +1,23 @@
+#!/usr/bin/env python
+"""Django's command-line utility for administrative tasks."""
+
+import os
+import sys
+
+
+def main():
+    """Run administrative tasks."""
+    os.environ.setdefault("DJANGO_SETTINGS_MODULE", "testdjango.settings")
+    try:
+        from django.core.management import execute_from_command_line
+    except ImportError as exc:
+        raise ImportError(
+            "Couldn't import Django. Are you sure it's installed and "
+            "available on your PYTHONPATH environment variable? Did you "
+            "forget to activate a virtual environment?"
+        ) from exc
+    execute_from_command_line(sys.argv)
+
+
+if __name__ == "__main__":
+    main()

integration_tests/django5/pyproject.toml

@@ -0,0 +1,19 @@
+[project]
+name = "test-django5"
+version = "0.1.0"
+requires-python = ">=3.12"
+dependencies = [
+    "django~=5.2.7",
+    "uvicorn[standard]~=0.38.0",
+    "posthog",
+    "pytest~=8.4.2",
+    "pytest-asyncio~=1.2.0",
+    "pytest-django~=4.11.1",
+    "httpx~=0.28.1",
+]
+
+[tool.uv]
+required-version = ">=0.5"
+
+[tool.uv.sources]
+posthog = { path = "../..", editable = true }

integration_tests/django5/test_exception_capture.py

@@ -0,0 +1,111 @@
+"""
+Test that verifies exception capture functionality.
+
+These tests verify that exceptions are actually captured to PostHog, not just that
+500 responses are returned.
+
+Without process_exception(), view exceptions are NOT captured to PostHog (v6.7.11 and earlier).
+With process_exception(), Django calls this method to capture exceptions before
+converting them to 500 responses.
+"""
+
+import os
+import django
+
+# Setup Django before importing anything else
+os.environ.setdefault("DJANGO_SETTINGS_MODULE", "testdjango.settings")
+django.setup()
+
+import pytest  # noqa: E402
+from httpx import AsyncClient, ASGITransport  # noqa: E402
+from django.core.asgi import get_asgi_application  # noqa: E402
+
+
+@pytest.fixture(scope="session")
+def asgi_app():
+    """Shared ASGI application for all tests."""
+    return get_asgi_application()
+
+
+@pytest.mark.asyncio
+async def test_async_exception_is_captured(asgi_app):
+    """
+    Test that async view exceptions are captured to PostHog.
+
+    The middleware's process_exception() method ensures exceptions are captured.
+    Without it (v6.7.11 and earlier), exceptions are NOT captured even though 500 is returned.
+    """
+    from unittest.mock import patch
+
+    # Track captured exceptions
+    captured = []
+
+    def mock_capture(exception, **kwargs):
+        """Mock capture_exception to record calls."""
+        captured.append(
+            {
+                "exception": exception,
+                "type": type(exception).__name__,
+                "message": str(exception),
+            }
+        )
+
+    # Patch at the posthog module level where middleware imports from
+    with patch("posthog.capture_exception", side_effect=mock_capture):
+        async with AsyncClient(
+            transport=ASGITransport(app=asgi_app), base_url="http://testserver"
+        ) as ac:
+            response = await ac.get("/test/async-exception")
+
+        # Django returns 500
+        assert response.status_code == 500
+
+        # CRITICAL: Verify PostHog captured the exception
+        assert len(captured) > 0, "Exception was NOT captured to PostHog!"
+
+        # Verify it's the right exception
+        exception_data = captured[0]
+        assert exception_data["type"] == "ValueError"
+        assert "Test exception from Django 5 async view" in exception_data["message"]
+
+
+@pytest.mark.asyncio
+async def test_sync_exception_is_captured(asgi_app):
+    """
+    Test that sync view exceptions are captured to PostHog.
+
+    The middleware's process_exception() method ensures exceptions are captured.
+    Without it (v6.7.11 and earlier), exceptions are NOT captured even though 500 is returned.
+    """
+    from unittest.mock import patch
+
+    # Track captured exceptions
+    captured = []
+
+    def mock_capture(exception, **kwargs):
+        """Mock capture_exception to record calls."""
+        captured.append(
+            {
+                "exception": exception,
+                "type": type(exception).__name__,
+                "message": str(exception),
+            }
+        )
+
+    # Patch at the posthog module level where middleware imports from
+    with patch("posthog.capture_exception", side_effect=mock_capture):
+        async with AsyncClient(
+            transport=ASGITransport(app=asgi_app), base_url="http://testserver"
+        ) as ac:
+            response = await ac.get("/test/sync-exception")
+
+        # Django returns 500
+        assert response.status_code == 500
+
+        # CRITICAL: Verify PostHog captured the exception
+        assert len(captured) > 0, "Exception was NOT captured to PostHog!"
+
+        # Verify it's the right exception
+        exception_data = captured[0]
+        assert exception_data["type"] == "ValueError"
+        assert "Test exception from Django 5 sync view" in exception_data["message"]

integration_tests/django5/test_middleware.py

@@ -0,0 +1,170 @@
+"""
+Tests for PostHog Django middleware in async context.
+
+These tests verify that the middleware correctly handles:
+1. Async user access (request.auser() in Django 5)
+2. Exception capture in both sync and async views
+3. No SynchronousOnlyOperation errors in async context
+
+Tests run directly against the ASGI application without needing a server.
+"""
+
+import os
+import django
+
+# Setup Django before importing anything else
+os.environ.setdefault("DJANGO_SETTINGS_MODULE", "testdjango.settings")
+django.setup()
+
+import pytest  # noqa: E402
+from httpx import AsyncClient, ASGITransport  # noqa: E402
+from django.core.asgi import get_asgi_application  # noqa: E402
+
+
+@pytest.fixture(scope="session")
+def asgi_app():
+    """Shared ASGI application for all tests."""
+    return get_asgi_application()
+
+
+@pytest.mark.asyncio
+async def test_async_user_access(asgi_app):
+    """
+    Test that middleware can access request.user in async context.
+
+    In Django 5, this requires using await request.auser() instead of request.user
+    to avoid SynchronousOnlyOperation error.
+
+    Without authentication, request.user is AnonymousUser which doesn't
+    trigger the lazy loading bug. This test verifies the middleware works
+    in the common case.
+    """
+    async with AsyncClient(
+        transport=ASGITransport(app=asgi_app), base_url="http://testserver"
+    ) as ac:
+        response = await ac.get("/test/async-user")
+
+    assert response.status_code == 200
+    data = response.json()
+    assert data["status"] == "success"
+    assert "django_version" in data
+
+
+@pytest.mark.django_db(transaction=True)
+@pytest.mark.asyncio
+async def test_async_authenticated_user_access(asgi_app):
+    """
+    Test that middleware can access an authenticated user in async context.
+
+    This is the critical test that triggers the SynchronousOnlyOperation bug
+    in v6.7.11. When AuthenticationMiddleware sets request.user to a
+    SimpleLazyObject wrapping a database query, accessing user.pk or user.email
+    in async context causes the error.
+
+    In v6.7.11, extract_request_user() does getattr(user, "is_authenticated", False)
+    which triggers the lazy object evaluation synchronously.
+
+    The fix uses await request.auser() instead to avoid this.
+    """
+    from django.contrib.auth import get_user_model
+    from django.test import Client
+    from asgiref.sync import sync_to_async
+    from django.test import override_settings
+
+    # Create a test user (must use sync_to_async since we're in async test)
+    User = get_user_model()
+
+    @sync_to_async
+    def create_or_get_user():
+        user, created = User.objects.get_or_create(
+            username="testuser",
+            defaults={
+                "email": "test@example.com",
+            },
+        )
+        if created:
+            user.set_password("testpass123")
+            user.save()
+        return user
+
+    user = await create_or_get_user()
+
+    # Create a session with authenticated user (sync operation)
+    @sync_to_async
+    def create_session():
+        client = Client()
+        client.force_login(user)
+        return client.cookies.get("sessionid")
+
+    session_cookie = await create_session()
+
+    if not session_cookie:
+        pytest.skip("Could not create authenticated session")
+
+    # Make request with session cookie - this should trigger the bug in v6.7.11
+    # Disable exception capture to see the SynchronousOnlyOperation clearly
+    with override_settings(POSTHOG_MW_CAPTURE_EXCEPTIONS=False):
+        async with AsyncClient(
+            transport=ASGITransport(app=asgi_app),
+            base_url="http://testserver",
+            cookies={"sessionid": session_cookie.value},
+        ) as ac:
+            response = await ac.get("/test/async-user")
+
+        assert response.status_code == 200
+        data = response.json()
+        assert data["status"] == "success"
+        assert data["user_authenticated"]
+
+
+@pytest.mark.asyncio
+async def test_sync_user_access(asgi_app):
+    """
+    Test that middleware works with sync views.
+
+    This should always work regardless of middleware version.
+    """
+    async with AsyncClient(
+        transport=ASGITransport(app=asgi_app), base_url="http://testserver"
+    ) as ac:
+        response = await ac.get("/test/sync-user")
+
+    assert response.status_code == 200
+    data = response.json()
+    assert data["status"] == "success"
+
+
+@pytest.mark.asyncio
+async def test_async_exception_capture(asgi_app):
+    """
+    Test that middleware handles exceptions from async views.
+
+    The middleware's process_exception() method captures view exceptions to PostHog
+    before Django converts them to 500 responses. This test verifies the exception
+    causes a 500 response. See test_exception_capture.py for tests that verify
+    actual exception capture to PostHog.
+    """
+    async with AsyncClient(
+        transport=ASGITransport(app=asgi_app), base_url="http://testserver"
+    ) as ac:
+        response = await ac.get("/test/async-exception")
+
+    # Django returns 500 for unhandled exceptions
+    assert response.status_code == 500
+
+
+@pytest.mark.asyncio
+async def test_sync_exception_capture(asgi_app):
+    """
+    Test that middleware handles exceptions from sync views.
+
+    The middleware's process_exception() method captures view exceptions to PostHog.
+    This test verifies the exception causes a 500 response.
+    """
+    async with AsyncClient(
+        transport=ASGITransport(app=asgi_app), base_url="http://testserver"
+    ) as ac:
+        response = await ac.get("/test/sync-exception")
+
+    # Django returns 500 for unhandled exceptions
+    assert response.status_code == 500

integration_tests/django5/testdjango/asgi.py

@@ -0,0 +1,16 @@
+"""
+ASGI config for testdjango project.
+
+It exposes the ASGI callable as a module-level variable named ``application``.
+
+For more information on this file, see
+https://docs.djangoproject.com/en/5.2/howto/deployment/asgi/
+"""
+
+import os
+
+from django.core.asgi import get_asgi_application
+
+os.environ.setdefault("DJANGO_SETTINGS_MODULE", "testdjango.settings")
+
+application = get_asgi_application()

integration_tests/django5/testdjango/settings.py

@@ -0,0 +1,129 @@
+"""
+Django settings for testdjango project.
+
+Generated by 'django-admin startproject' using Django 5.2.7.
+
+For more information on this file, see
+https://docs.djangoproject.com/en/5.2/topics/settings/
+
+For the full list of settings and their values, see
+https://docs.djangoproject.com/en/5.2/ref/settings/
+"""
+
+from pathlib import Path
+
+# Build paths inside the project like this: BASE_DIR / 'subdir'.
+BASE_DIR = Path(__file__).resolve().parent.parent
+
+
+# Quick-start development settings - unsuitable for production
+# See https://docs.djangoproject.com/en/5.2/howto/deployment/checklist/
+
+# SECURITY WARNING: keep the secret key used in production secret!
+SECRET_KEY = "django-insecure-q5(&wfw@_lb)noyowbfl$2ls8c82hl__0f9s5(mohlh2)aas#3"
+
+# SECURITY WARNING: don't run with debug turned on in production!
+DEBUG = True
+
+ALLOWED_HOSTS = ["*"]
+
+
+# Application definition
+
+INSTALLED_APPS = [
+    "django.contrib.admin",
+    "django.contrib.auth",
+    "django.contrib.contenttypes",
+    "django.contrib.sessions",
+    "django.contrib.messages",
+    "django.contrib.staticfiles",
+]
+
+MIDDLEWARE = [
+    "django.middleware.security.SecurityMiddleware",
+    "django.contrib.sessions.middleware.SessionMiddleware",
+    "django.middleware.common.CommonMiddleware",
+    "django.middleware.csrf.CsrfViewMiddleware",
+    "django.contrib.auth.middleware.AuthenticationMiddleware",
+    "django.contrib.messages.middleware.MessageMiddleware",
+    "django.middleware.clickjacking.XFrameOptionsMiddleware",
+    "posthog.integrations.django.PosthogContextMiddleware",  # Test PostHog middleware
+]
+
+ROOT_URLCONF = "testdjango.urls"
+
+TEMPLATES = [
+    {
+        "BACKEND": "django.template.backends.django.DjangoTemplates",
+        "DIRS": [],
+        "APP_DIRS": True,
+        "OPTIONS": {
+            "context_processors": [
+                "django.template.context_processors.request",
+                "django.contrib.auth.context_processors.auth",
+                "django.contrib.messages.context_processors.messages",
+            ],
+        },
+    },
+]
+
+WSGI_APPLICATION = "testdjango.wsgi.application"
+
+
+# Database
+# https://docs.djangoproject.com/en/5.2/ref/settings/#databases
+
+DATABASES = {
+    "default": {
+        "ENGINE": "django.db.backends.sqlite3",
+        "NAME": BASE_DIR / "db.sqlite3",
+    }
+}
+
+
+# Password validation
+# https://docs.djangoproject.com/en/5.2/ref/settings/#auth-password-validators
+
+AUTH_PASSWORD_VALIDATORS = [
+    {
+        "NAME": "django.contrib.auth.password_validation.UserAttributeSimilarityValidator",
+    },
+    {
+        "NAME": "django.contrib.auth.password_validation.MinimumLengthValidator",
+    },
+    {
+        "NAME": "django.contrib.auth.password_validation.CommonPasswordValidator",
+    },
+    {
+        "NAME": "django.contrib.auth.password_validation.NumericPasswordValidator",
+    },
+]
+
+
+# Internationalization
+# https://docs.djangoproject.com/en/5.2/topics/i18n/
+
+LANGUAGE_CODE = "en-us"
+
+TIME_ZONE = "UTC"
+
+USE_I18N = True
+
+USE_TZ = True
+
+
+# Static files (CSS, JavaScript, Images)
+# https://docs.djangoproject.com/en/5.2/howto/static-files/
+
+STATIC_URL = "static/"
+
+# Default primary key field type
+# https://docs.djangoproject.com/en/5.2/ref/settings/#default-auto-field
+
+DEFAULT_AUTO_FIELD = "django.db.models.BigAutoField"
+
+
+# PostHog settings for testing
+POSTHOG_API_KEY = "test-key"
+POSTHOG_HOST = "https://app.posthog.com"
+POSTHOG_MW_CAPTURE_EXCEPTIONS = True

integration_tests/django5/testdjango/urls.py

@@ -0,0 +1,28 @@
+"""
+URL configuration for testdjango project.
+
+The `urlpatterns` list routes URLs to views. For more information please see:
+    https://docs.djangoproject.com/en/5.2/topics/http/urls/
+Examples:
+Function views
+    1. Add an import:  from my_app import views
+    2. Add a URL to urlpatterns:  path('', views.home, name='home')
+Class-based views
+    1. Add an import:  from other_app.views import Home
+    2. Add a URL to urlpatterns:  path('', Home.as_view(), name='home')
+Including another URLconf
+    1. Import the include() function: from django.urls import include, path
+    2. Add a URL to urlpatterns:  path('blog/', include('blog.urls'))
+"""
+
+from django.contrib import admin
+from django.urls import path
+from testdjango import views
+
+urlpatterns = [
+    path("admin/", admin.site.urls),
+    path("test/async-user", views.test_async_user),
+    path("test/sync-user", views.test_sync_user),
+    path("test/async-exception", views.test_async_exception),
+    path("test/sync-exception", views.test_sync_exception),
+]

integration_tests/django5/testdjango/views.py

@@ -0,0 +1,50 @@
+"""
+Test views for validating PostHog middleware with Django 5 ASGI.
+"""
+
+from django.http import JsonResponse
+
+
+async def test_async_user(request):
+    """
+    Async view that tests middleware with request.user access.
+
+    The middleware will access request.user (SimpleLazyObject) via auser()
+    in async context. Without the fix, this causes SynchronousOnlyOperation.
+    """
+    # The middleware has already accessed request.user via auser()
+    # If we got here, the fix works!
+    user = await request.auser()
+
+    return JsonResponse(
+        {
+            "status": "success",
+            "message": "Django 5 async middleware test passed!",
+            "django_version": "5.x",
+            "user_authenticated": user.is_authenticated if user else False,
+            "note": "Middleware used await request.auser() successfully",
+        }
+    )
+
+
+def test_sync_user(request):
+    """Sync view for comparison."""
+    return JsonResponse(
+        {
+            "status": "success",
+            "message": "Sync view works",
+            "user_authenticated": request.user.is_authenticated
+            if hasattr(request, "user")
+            else False,
+        }
+    )
+
+
+async def test_async_exception(request):
+    """Async view that raises an exception for testing exception capture."""
+    raise ValueError("Test exception from Django 5 async view")
+
+
+def test_sync_exception(request):
+    """Sync view that raises an exception for testing exception capture."""
+    raise ValueError("Test exception from Django 5 sync view")

integration_tests/django5/testdjango/wsgi.py

@@ -0,0 +1,16 @@
+"""
+WSGI config for testdjango project.
+
+It exposes the WSGI callable as a module-level variable named ``application``.
+
+For more information on this file, see
+https://docs.djangoproject.com/en/5.2/howto/deployment/wsgi/
+"""
+
+import os
+
+from django.core.wsgi import get_wsgi_application
+
+os.environ.setdefault("DJANGO_SETTINGS_MODULE", "testdjango.settings")
+
+application = get_wsgi_application()

integration_tests/django5/uv.lock

@@ -0,0 +1,674 @@
+version = 1
+revision = 3
+requires-python = ">=3.12"
+
+[[package]]
+name = "anyio"
+version = "4.11.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "idna" },
+    { name = "sniffio" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/c6/78/7d432127c41b50bccba979505f272c16cbcadcc33645d5fa3a738110ae75/anyio-4.11.0.tar.gz", hash = "sha256:82a8d0b81e318cc5ce71a5f1f8b5c4e63619620b63141ef8c995fa0db95a57c4", size = 219094, upload-time = "2025-09-23T09:19:12.58Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/15/b3/9b1a8074496371342ec1e796a96f99c82c945a339cd81a8e73de28b4cf9e/anyio-4.11.0-py3-none-any.whl", hash = "sha256:0287e96f4d26d4149305414d4e3bc32f0dcd0862365a4bddea19d7a1ec38c4fc", size = 109097, upload-time = "2025-09-23T09:19:10.601Z" },
+]
+
+[[package]]
+name = "asgiref"
+version = "3.10.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/46/08/4dfec9b90758a59acc6be32ac82e98d1fbfc321cb5cfa410436dbacf821c/asgiref-3.10.0.tar.gz", hash = "sha256:d89f2d8cd8b56dada7d52fa7dc8075baa08fb836560710d38c292a7a3f78c04e", size = 37483, upload-time = "2025-10-05T09:15:06.557Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/17/9c/fc2331f538fbf7eedba64b2052e99ccf9ba9d6888e2f41441ee28847004b/asgiref-3.10.0-py3-none-any.whl", hash = "sha256:aef8a81283a34d0ab31630c9b7dfe70c812c95eba78171367ca8745e88124734", size = 24050, upload-time = "2025-10-05T09:15:05.11Z" },
+]
+
+[[package]]
+name = "backoff"
+version = "2.2.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/47/d7/5bbeb12c44d7c4f2fb5b56abce497eb5ed9f34d85701de869acedd602619/backoff-2.2.1.tar.gz", hash = "sha256:03f829f5bb1923180821643f8753b0502c3b682293992485b0eef2807afa5cba", size = 17001, upload-time = "2022-10-05T19:19:32.061Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/df/73/b6e24bd22e6720ca8ee9a85a0c4a2971af8497d8f3193fa05390cbd46e09/backoff-2.2.1-py3-none-any.whl", hash = "sha256:63579f9a0628e06278f7e47b7d7d5b6ce20dc65c5e96a6f3ca99a6adca0396e8", size = 15148, upload-time = "2022-10-05T19:19:30.546Z" },
+]
+
+[[package]]
+name = "certifi"
+version = "2025.10.5"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/4c/5b/b6ce21586237c77ce67d01dc5507039d444b630dd76611bbca2d8e5dcd91/certifi-2025.10.5.tar.gz", hash = "sha256:47c09d31ccf2acf0be3f701ea53595ee7e0b8fa08801c6624be771df09ae7b43", size = 164519, upload-time = "2025-10-05T04:12:15.808Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e4/37/af0d2ef3967ac0d6113837b44a4f0bfe1328c2b9763bd5b1744520e5cfed/certifi-2025.10.5-py3-none-any.whl", hash = "sha256:0f212c2744a9bb6de0c56639a6f68afe01ecd92d91f14ae897c4fe7bbeeef0de", size = 163286, upload-time = "2025-10-05T04:12:14.03Z" },
+]
+
+[[package]]
+name = "charset-normalizer"
+version = "3.4.4"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/13/69/33ddede1939fdd074bce5434295f38fae7136463422fe4fd3e0e89b98062/charset_normalizer-3.4.4.tar.gz", hash = "sha256:94537985111c35f28720e43603b8e7b43a6ecfb2ce1d3058bbe955b73404e21a", size = 129418, upload-time = "2025-10-14T04:42:32.879Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/f3/85/1637cd4af66fa687396e757dec650f28025f2a2f5a5531a3208dc0ec43f2/charset_normalizer-3.4.4-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:0a98e6759f854bd25a58a73fa88833fba3b7c491169f86ce1180c948ab3fd394", size = 208425, upload-time = "2025-10-14T04:40:53.353Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/6a/04130023fef2a0d9c62d0bae2649b69f7b7d8d24ea5536feef50551029df/charset_normalizer-3.4.4-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:b5b290ccc2a263e8d185130284f8501e3e36c5e02750fc6b6bdeb2e9e96f1e25", size = 148162, upload-time = "2025-10-14T04:40:54.558Z" },
+    { url = "https://files.pythonhosted.org/packages/78/29/62328d79aa60da22c9e0b9a66539feae06ca0f5a4171ac4f7dc285b83688/charset_normalizer-3.4.4-cp312-cp312-manylinux2014_armv7l.manylinux_2_17_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:74bb723680f9f7a6234dcf67aea57e708ec1fbdf5699fb91dfd6f511b0a320ef", size = 144558, upload-time = "2025-10-14T04:40:55.677Z" },
+    { url = "https://files.pythonhosted.org/packages/86/bb/b32194a4bf15b88403537c2e120b817c61cd4ecffa9b6876e941c3ee38fe/charset_normalizer-3.4.4-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:f1e34719c6ed0b92f418c7c780480b26b5d9c50349e9a9af7d76bf757530350d", size = 161497, upload-time = "2025-10-14T04:40:57.217Z" },
+    { url = "https://files.pythonhosted.org/packages/19/89/a54c82b253d5b9b111dc74aca196ba5ccfcca8242d0fb64146d4d3183ff1/charset_normalizer-3.4.4-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:2437418e20515acec67d86e12bf70056a33abdacb5cb1655042f6538d6b085a8", size = 159240, upload-time = "2025-10-14T04:40:58.358Z" },
+    { url = "https://files.pythonhosted.org/packages/c0/10/d20b513afe03acc89ec33948320a5544d31f21b05368436d580dec4e234d/charset_normalizer-3.4.4-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:11d694519d7f29d6cd09f6ac70028dba10f92f6cdd059096db198c283794ac86", size = 153471, upload-time = "2025-10-14T04:40:59.468Z" },
+    { url = "https://files.pythonhosted.org/packages/61/fa/fbf177b55bdd727010f9c0a3c49eefa1d10f960e5f09d1d887bf93c2e698/charset_normalizer-3.4.4-cp312-cp312-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:ac1c4a689edcc530fc9d9aa11f5774b9e2f33f9a0c6a57864e90908f5208d30a", size = 150864, upload-time = "2025-10-14T04:41:00.623Z" },
+    { url = "https://files.pythonhosted.org/packages/05/12/9fbc6a4d39c0198adeebbde20b619790e9236557ca59fc40e0e3cebe6f40/charset_normalizer-3.4.4-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:21d142cc6c0ec30d2efee5068ca36c128a30b0f2c53c1c07bd78cb6bc1d3be5f", size = 150647, upload-time = "2025-10-14T04:41:01.754Z" },
+    { url = "https://files.pythonhosted.org/packages/ad/1f/6a9a593d52e3e8c5d2b167daf8c6b968808efb57ef4c210acb907c365bc4/charset_normalizer-3.4.4-cp312-cp312-musllinux_1_2_armv7l.whl", hash = "sha256:5dbe56a36425d26d6cfb40ce79c314a2e4dd6211d51d6d2191c00bed34f354cc", size = 145110, upload-time = "2025-10-14T04:41:03.231Z" },
+    { url = "https://files.pythonhosted.org/packages/30/42/9a52c609e72471b0fc54386dc63c3781a387bb4fe61c20231a4ebcd58bdd/charset_normalizer-3.4.4-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:5bfbb1b9acf3334612667b61bd3002196fe2a1eb4dd74d247e0f2a4d50ec9bbf", size = 162839, upload-time = "2025-10-14T04:41:04.715Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/5b/c0682bbf9f11597073052628ddd38344a3d673fda35a36773f7d19344b23/charset_normalizer-3.4.4-cp312-cp312-musllinux_1_2_riscv64.whl", hash = "sha256:d055ec1e26e441f6187acf818b73564e6e6282709e9bcb5b63f5b23068356a15", size = 150667, upload-time = "2025-10-14T04:41:05.827Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/24/a41afeab6f990cf2daf6cb8c67419b63b48cf518e4f56022230840c9bfb2/charset_normalizer-3.4.4-cp312-cp312-musllinux_1_2_s390x.whl", hash = "sha256:af2d8c67d8e573d6de5bc30cdb27e9b95e49115cd9baad5ddbd1a6207aaa82a9", size = 160535, upload-time = "2025-10-14T04:41:06.938Z" },
+    { url = "https://files.pythonhosted.org/packages/2a/e5/6a4ce77ed243c4a50a1fecca6aaaab419628c818a49434be428fe24c9957/charset_normalizer-3.4.4-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:780236ac706e66881f3b7f2f32dfe90507a09e67d1d454c762cf642e6e1586e0", size = 154816, upload-time = "2025-10-14T04:41:08.101Z" },
+    { url = "https://files.pythonhosted.org/packages/a8/ef/89297262b8092b312d29cdb2517cb1237e51db8ecef2e9af5edbe7b683b1/charset_normalizer-3.4.4-cp312-cp312-win32.whl", hash = "sha256:5833d2c39d8896e4e19b689ffc198f08ea58116bee26dea51e362ecc7cd3ed26", size = 99694, upload-time = "2025-10-14T04:41:09.23Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/2d/1e5ed9dd3b3803994c155cd9aacb60c82c331bad84daf75bcb9c91b3295e/charset_normalizer-3.4.4-cp312-cp312-win_amd64.whl", hash = "sha256:a79cfe37875f822425b89a82333404539ae63dbdddf97f84dcbc3d339aae9525", size = 107131, upload-time = "2025-10-14T04:41:10.467Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/d9/0ed4c7098a861482a7b6a95603edce4c0d9db2311af23da1fb2b75ec26fc/charset_normalizer-3.4.4-cp312-cp312-win_arm64.whl", hash = "sha256:376bec83a63b8021bb5c8ea75e21c4ccb86e7e45ca4eb81146091b56599b80c3", size = 100390, upload-time = "2025-10-14T04:41:11.915Z" },
+    { url = "https://files.pythonhosted.org/packages/97/45/4b3a1239bbacd321068ea6e7ac28875b03ab8bc0aa0966452db17cd36714/charset_normalizer-3.4.4-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:e1f185f86a6f3403aa2420e815904c67b2f9ebc443f045edd0de921108345794", size = 208091, upload-time = "2025-10-14T04:41:13.346Z" },
+    { url = "https://files.pythonhosted.org/packages/7d/62/73a6d7450829655a35bb88a88fca7d736f9882a27eacdca2c6d505b57e2e/charset_normalizer-3.4.4-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:6b39f987ae8ccdf0d2642338faf2abb1862340facc796048b604ef14919e55ed", size = 147936, upload-time = "2025-10-14T04:41:14.461Z" },
+    { url = "https://files.pythonhosted.org/packages/89/c5/adb8c8b3d6625bef6d88b251bbb0d95f8205831b987631ab0c8bb5d937c2/charset_normalizer-3.4.4-cp313-cp313-manylinux2014_armv7l.manylinux_2_17_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:3162d5d8ce1bb98dd51af660f2121c55d0fa541b46dff7bb9b9f86ea1d87de72", size = 144180, upload-time = "2025-10-14T04:41:15.588Z" },
+    { url = "https://files.pythonhosted.org/packages/91/ed/9706e4070682d1cc219050b6048bfd293ccf67b3d4f5a4f39207453d4b99/charset_normalizer-3.4.4-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:81d5eb2a312700f4ecaa977a8235b634ce853200e828fbadf3a9c50bab278328", size = 161346, upload-time = "2025-10-14T04:41:16.738Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/0d/031f0d95e4972901a2f6f09ef055751805ff541511dc1252ba3ca1f80cf5/charset_normalizer-3.4.4-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:5bd2293095d766545ec1a8f612559f6b40abc0eb18bb2f5d1171872d34036ede", size = 158874, upload-time = "2025-10-14T04:41:17.923Z" },
+    { url = "https://files.pythonhosted.org/packages/f5/83/6ab5883f57c9c801ce5e5677242328aa45592be8a00644310a008d04f922/charset_normalizer-3.4.4-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:a8a8b89589086a25749f471e6a900d3f662d1d3b6e2e59dcecf787b1cc3a1894", size = 153076, upload-time = "2025-10-14T04:41:19.106Z" },
+    { url = "https://files.pythonhosted.org/packages/75/1e/5ff781ddf5260e387d6419959ee89ef13878229732732ee73cdae01800f2/charset_normalizer-3.4.4-cp313-cp313-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:bc7637e2f80d8530ee4a78e878bce464f70087ce73cf7c1caf142416923b98f1", size = 150601, upload-time = "2025-10-14T04:41:20.245Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/57/71be810965493d3510a6ca79b90c19e48696fb1ff964da319334b12677f0/charset_normalizer-3.4.4-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:f8bf04158c6b607d747e93949aa60618b61312fe647a6369f88ce2ff16043490", size = 150376, upload-time = "2025-10-14T04:41:21.398Z" },
+    { url = "https://files.pythonhosted.org/packages/e5/d5/c3d057a78c181d007014feb7e9f2e65905a6c4ef182c0ddf0de2924edd65/charset_normalizer-3.4.4-cp313-cp313-musllinux_1_2_armv7l.whl", hash = "sha256:554af85e960429cf30784dd47447d5125aaa3b99a6f0683589dbd27e2f45da44", size = 144825, upload-time = "2025-10-14T04:41:22.583Z" },
+    { url = "https://files.pythonhosted.org/packages/e6/8c/d0406294828d4976f275ffbe66f00266c4b3136b7506941d87c00cab5272/charset_normalizer-3.4.4-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:74018750915ee7ad843a774364e13a3db91682f26142baddf775342c3f5b1133", size = 162583, upload-time = "2025-10-14T04:41:23.754Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/24/e2aa1f18c8f15c4c0e932d9287b8609dd30ad56dbe41d926bd846e22fb8d/charset_normalizer-3.4.4-cp313-cp313-musllinux_1_2_riscv64.whl", hash = "sha256:c0463276121fdee9c49b98908b3a89c39be45d86d1dbaa22957e38f6321d4ce3", size = 150366, upload-time = "2025-10-14T04:41:25.27Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/5b/1e6160c7739aad1e2df054300cc618b06bf784a7a164b0f238360721ab86/charset_normalizer-3.4.4-cp313-cp313-musllinux_1_2_s390x.whl", hash = "sha256:362d61fd13843997c1c446760ef36f240cf81d3ebf74ac62652aebaf7838561e", size = 160300, upload-time = "2025-10-14T04:41:26.725Z" },
+    { url = "https://files.pythonhosted.org/packages/7a/10/f882167cd207fbdd743e55534d5d9620e095089d176d55cb22d5322f2afd/charset_normalizer-3.4.4-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:9a26f18905b8dd5d685d6d07b0cdf98a79f3c7a918906af7cc143ea2e164c8bc", size = 154465, upload-time = "2025-10-14T04:41:28.322Z" },
+    { url = "https://files.pythonhosted.org/packages/89/66/c7a9e1b7429be72123441bfdbaf2bc13faab3f90b933f664db506dea5915/charset_normalizer-3.4.4-cp313-cp313-win32.whl", hash = "sha256:9b35f4c90079ff2e2edc5b26c0c77925e5d2d255c42c74fdb70fb49b172726ac", size = 99404, upload-time = "2025-10-14T04:41:29.95Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/26/b9924fa27db384bdcd97ab83b4f0a8058d96ad9626ead570674d5e737d90/charset_normalizer-3.4.4-cp313-cp313-win_amd64.whl", hash = "sha256:b435cba5f4f750aa6c0a0d92c541fb79f69a387c91e61f1795227e4ed9cece14", size = 107092, upload-time = "2025-10-14T04:41:31.188Z" },
+    { url = "https://files.pythonhosted.org/packages/af/8f/3ed4bfa0c0c72a7ca17f0380cd9e4dd842b09f664e780c13cff1dcf2ef1b/charset_normalizer-3.4.4-cp313-cp313-win_arm64.whl", hash = "sha256:542d2cee80be6f80247095cc36c418f7bddd14f4a6de45af91dfad36d817bba2", size = 100408, upload-time = "2025-10-14T04:41:32.624Z" },
+    { url = "https://files.pythonhosted.org/packages/2a/35/7051599bd493e62411d6ede36fd5af83a38f37c4767b92884df7301db25d/charset_normalizer-3.4.4-cp314-cp314-macosx_10_13_universal2.whl", hash = "sha256:da3326d9e65ef63a817ecbcc0df6e94463713b754fe293eaa03da99befb9a5bd", size = 207746, upload-time = "2025-10-14T04:41:33.773Z" },
+    { url = "https://files.pythonhosted.org/packages/10/9a/97c8d48ef10d6cd4fcead2415523221624bf58bcf68a802721a6bc807c8f/charset_normalizer-3.4.4-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:8af65f14dc14a79b924524b1e7fffe304517b2bff5a58bf64f30b98bbc5079eb", size = 147889, upload-time = "2025-10-14T04:41:34.897Z" },
+    { url = "https://files.pythonhosted.org/packages/10/bf/979224a919a1b606c82bd2c5fa49b5c6d5727aa47b4312bb27b1734f53cd/charset_normalizer-3.4.4-cp314-cp314-manylinux2014_armv7l.manylinux_2_17_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:74664978bb272435107de04e36db5a9735e78232b85b77d45cfb38f758efd33e", size = 143641, upload-time = "2025-10-14T04:41:36.116Z" },
+    { url = "https://files.pythonhosted.org/packages/ba/33/0ad65587441fc730dc7bd90e9716b30b4702dc7b617e6ba4997dc8651495/charset_normalizer-3.4.4-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:752944c7ffbfdd10c074dc58ec2d5a8a4cd9493b314d367c14d24c17684ddd14", size = 160779, upload-time = "2025-10-14T04:41:37.229Z" },
+    { url = "https://files.pythonhosted.org/packages/67/ed/331d6b249259ee71ddea93f6f2f0a56cfebd46938bde6fcc6f7b9a3d0e09/charset_normalizer-3.4.4-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:d1f13550535ad8cff21b8d757a3257963e951d96e20ec82ab44bc64aeb62a191", size = 159035, upload-time = "2025-10-14T04:41:38.368Z" },
+    { url = "https://files.pythonhosted.org/packages/67/ff/f6b948ca32e4f2a4576aa129d8bed61f2e0543bf9f5f2b7fc3758ed005c9/charset_normalizer-3.4.4-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:ecaae4149d99b1c9e7b88bb03e3221956f68fd6d50be2ef061b2381b61d20838", size = 152542, upload-time = "2025-10-14T04:41:39.862Z" },
+    { url = "https://files.pythonhosted.org/packages/16/85/276033dcbcc369eb176594de22728541a925b2632f9716428c851b149e83/charset_normalizer-3.4.4-cp314-cp314-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:cb6254dc36b47a990e59e1068afacdcd02958bdcce30bb50cc1700a8b9d624a6", size = 149524, upload-time = "2025-10-14T04:41:41.319Z" },
+    { url = "https://files.pythonhosted.org/packages/9e/f2/6a2a1f722b6aba37050e626530a46a68f74e63683947a8acff92569f979a/charset_normalizer-3.4.4-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:c8ae8a0f02f57a6e61203a31428fa1d677cbe50c93622b4149d5c0f319c1d19e", size = 150395, upload-time = "2025-10-14T04:41:42.539Z" },
+    { url = "https://files.pythonhosted.org/packages/60/bb/2186cb2f2bbaea6338cad15ce23a67f9b0672929744381e28b0592676824/charset_normalizer-3.4.4-cp314-cp314-musllinux_1_2_armv7l.whl", hash = "sha256:47cc91b2f4dd2833fddaedd2893006b0106129d4b94fdb6af1f4ce5a9965577c", size = 143680, upload-time = "2025-10-14T04:41:43.661Z" },
+    { url = "https://files.pythonhosted.org/packages/7d/a5/bf6f13b772fbb2a90360eb620d52ed8f796f3c5caee8398c3b2eb7b1c60d/charset_normalizer-3.4.4-cp314-cp314-musllinux_1_2_ppc64le.whl", hash = "sha256:82004af6c302b5d3ab2cfc4cc5f29db16123b1a8417f2e25f9066f91d4411090", size = 162045, upload-time = "2025-10-14T04:41:44.821Z" },
+    { url = "https://files.pythonhosted.org/packages/df/c5/d1be898bf0dc3ef9030c3825e5d3b83f2c528d207d246cbabe245966808d/charset_normalizer-3.4.4-cp314-cp314-musllinux_1_2_riscv64.whl", hash = "sha256:2b7d8f6c26245217bd2ad053761201e9f9680f8ce52f0fcd8d0755aeae5b2152", size = 149687, upload-time = "2025-10-14T04:41:46.442Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/42/90c1f7b9341eef50c8a1cb3f098ac43b0508413f33affd762855f67a410e/charset_normalizer-3.4.4-cp314-cp314-musllinux_1_2_s390x.whl", hash = "sha256:799a7a5e4fb2d5898c60b640fd4981d6a25f1c11790935a44ce38c54e985f828", size = 160014, upload-time = "2025-10-14T04:41:47.631Z" },
+    { url = "https://files.pythonhosted.org/packages/76/be/4d3ee471e8145d12795ab655ece37baed0929462a86e72372fd25859047c/charset_normalizer-3.4.4-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:99ae2cffebb06e6c22bdc25801d7b30f503cc87dbd283479e7b606f70aff57ec", size = 154044, upload-time = "2025-10-14T04:41:48.81Z" },
+    { url = "https://files.pythonhosted.org/packages/b0/6f/8f7af07237c34a1defe7defc565a9bc1807762f672c0fde711a4b22bf9c0/charset_normalizer-3.4.4-cp314-cp314-win32.whl", hash = "sha256:f9d332f8c2a2fcbffe1378594431458ddbef721c1769d78e2cbc06280d8155f9", size = 99940, upload-time = "2025-10-14T04:41:49.946Z" },
+    { url = "https://files.pythonhosted.org/packages/4b/51/8ade005e5ca5b0d80fb4aff72a3775b325bdc3d27408c8113811a7cbe640/charset_normalizer-3.4.4-cp314-cp314-win_amd64.whl", hash = "sha256:8a6562c3700cce886c5be75ade4a5db4214fda19fede41d9792d100288d8f94c", size = 107104, upload-time = "2025-10-14T04:41:51.051Z" },
+    { url = "https://files.pythonhosted.org/packages/da/5f/6b8f83a55bb8278772c5ae54a577f3099025f9ade59d0136ac24a0df4bde/charset_normalizer-3.4.4-cp314-cp314-win_arm64.whl", hash = "sha256:de00632ca48df9daf77a2c65a484531649261ec9f25489917f09e455cb09ddb2", size = 100743, upload-time = "2025-10-14T04:41:52.122Z" },
+    { url = "https://files.pythonhosted.org/packages/0a/4c/925909008ed5a988ccbb72dcc897407e5d6d3bd72410d69e051fc0c14647/charset_normalizer-3.4.4-py3-none-any.whl", hash = "sha256:7a32c560861a02ff789ad905a2fe94e3f840803362c84fecf1851cb4cf3dc37f", size = 53402, upload-time = "2025-10-14T04:42:31.76Z" },
+]
+
+[[package]]
+name = "click"
+version = "8.3.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/46/61/de6cd827efad202d7057d93e0fed9294b96952e188f7384832791c7b2254/click-8.3.0.tar.gz", hash = "sha256:e7b8232224eba16f4ebe410c25ced9f7875cb5f3263ffc93cc3e8da705e229c4", size = 276943, upload-time = "2025-09-18T17:32:23.696Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/db/d3/9dcc0f5797f070ec8edf30fbadfb200e71d9db6b84d211e3b2085a7589a0/click-8.3.0-py3-none-any.whl", hash = "sha256:9b9f285302c6e3064f4330c05f05b81945b2a39544279343e6e7c5f27a9baddc", size = 107295, upload-time = "2025-09-18T17:32:22.42Z" },
+]
+
+[[package]]
+name = "colorama"
+version = "0.4.6"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/d8/53/6f443c9a4a8358a93a6792e2acffb9d9d5cb0a5cfd8802644b7b1c9a02e4/colorama-0.4.6.tar.gz", hash = "sha256:08695f5cb7ed6e0531a20572697297273c47b8cae5a63ffc6d6ed5c201be6e44", size = 27697, upload-time = "2022-10-25T02:36:22.414Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d1/d6/3965ed04c63042e047cb6a3e6ed1a63a35087b6a609aa3a15ed8ac56c221/colorama-0.4.6-py2.py3-none-any.whl", hash = "sha256:4f1d9991f5acc0ca119f9d443620b77f9d6b33703e51011c16baf57afb285fc6", size = 25335, upload-time = "2022-10-25T02:36:20.889Z" },
+]
+
+[[package]]
+name = "distro"
+version = "1.9.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/fc/f8/98eea607f65de6527f8a2e8885fc8015d3e6f5775df186e443e0964a11c3/distro-1.9.0.tar.gz", hash = "sha256:2fa77c6fd8940f116ee1d6b94a2f90b13b5ea8d019b98bc8bafdcabcdd9bdbed", size = 60722, upload-time = "2023-12-24T09:54:32.31Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/12/b3/231ffd4ab1fc9d679809f356cebee130ac7daa00d6d6f3206dd4fd137e9e/distro-1.9.0-py3-none-any.whl", hash = "sha256:7bffd925d65168f85027d8da9af6bddab658135b840670a223589bc0c8ef02b2", size = 20277, upload-time = "2023-12-24T09:54:30.421Z" },
+]
+
+[[package]]
+name = "django"
+version = "5.2.7"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "asgiref" },
+    { name = "sqlparse" },
+    { name = "tzdata", marker = "sys_platform == 'win32'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b1/96/bd84e2bb997994de8bcda47ae4560991084e86536541d7214393880f01a8/django-5.2.7.tar.gz", hash = "sha256:e0f6f12e2551b1716a95a63a1366ca91bbcd7be059862c1b18f989b1da356cdd", size = 10865812, upload-time = "2025-10-01T14:22:12.081Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/8f/ef/81f3372b5dd35d8d354321155d1a38894b2b766f576d0abffac4d8ae78d9/django-5.2.7-py3-none-any.whl", hash = "sha256:59a13a6515f787dec9d97a0438cd2efac78c8aca1c80025244b0fe507fe0754b", size = 8307145, upload-time = "2025-10-01T14:22:49.476Z" },
+]
+
+[[package]]
+name = "h11"
+version = "0.16.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/01/ee/02a2c011bdab74c6fb3c75474d40b3052059d95df7e73351460c8588d963/h11-0.16.0.tar.gz", hash = "sha256:4e35b956cf45792e4caa5885e69fba00bdbc6ffafbfa020300e549b208ee5ff1", size = 101250, upload-time = "2025-04-24T03:35:25.427Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/04/4b/29cac41a4d98d144bf5f6d33995617b185d14b22401f75ca86f384e87ff1/h11-0.16.0-py3-none-any.whl", hash = "sha256:63cf8bbe7522de3bf65932fda1d9c2772064ffb3dae62d55932da54b31cb6c86", size = 37515, upload-time = "2025-04-24T03:35:24.344Z" },
+]
+
+[[package]]
+name = "httpcore"
+version = "1.0.9"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "certifi" },
+    { name = "h11" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/06/94/82699a10bca87a5556c9c59b5963f2d039dbd239f25bc2a63907a05a14cb/httpcore-1.0.9.tar.gz", hash = "sha256:6e34463af53fd2ab5d807f399a9b45ea31c3dfa2276f15a2c3f00afff6e176e8", size = 85484, upload-time = "2025-04-24T22:06:22.219Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7e/f5/f66802a942d491edb555dd61e3a9961140fd64c90bce1eafd741609d334d/httpcore-1.0.9-py3-none-any.whl", hash = "sha256:2d400746a40668fc9dec9810239072b40b4484b640a8c38fd654a024c7a1bf55", size = 78784, upload-time = "2025-04-24T22:06:20.566Z" },
+]
+
+[[package]]
+name = "httptools"
+version = "0.7.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/b5/46/120a669232c7bdedb9d52d4aeae7e6c7dfe151e99dc70802e2fc7a5e1993/httptools-0.7.1.tar.gz", hash = "sha256:abd72556974f8e7c74a259655924a717a2365b236c882c3f6f8a45fe94703ac9", size = 258961, upload-time = "2025-10-10T03:55:08.559Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/53/7f/403e5d787dc4942316e515e949b0c8a013d84078a915910e9f391ba9b3ed/httptools-0.7.1-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:38e0c83a2ea9746ebbd643bdfb521b9aa4a91703e2cd705c20443405d2fd16a5", size = 206280, upload-time = "2025-10-10T03:54:39.274Z" },
+    { url = "https://files.pythonhosted.org/packages/2a/0d/7f3fd28e2ce311ccc998c388dd1c53b18120fda3b70ebb022b135dc9839b/httptools-0.7.1-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:f25bbaf1235e27704f1a7b86cd3304eabc04f569c828101d94a0e605ef7205a5", size = 110004, upload-time = "2025-10-10T03:54:40.403Z" },
+    { url = "https://files.pythonhosted.org/packages/84/a6/b3965e1e146ef5762870bbe76117876ceba51a201e18cc31f5703e454596/httptools-0.7.1-cp312-cp312-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl", hash = "sha256:2c15f37ef679ab9ecc06bfc4e6e8628c32a8e4b305459de7cf6785acd57e4d03", size = 517655, upload-time = "2025-10-10T03:54:41.347Z" },
+    { url = "https://files.pythonhosted.org/packages/11/7d/71fee6f1844e6fa378f2eddde6c3e41ce3a1fb4b2d81118dd544e3441ec0/httptools-0.7.1-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:7fe6e96090df46b36ccfaf746f03034e5ab723162bc51b0a4cf58305324036f2", size = 511440, upload-time = "2025-10-10T03:54:42.452Z" },
+    { url = "https://files.pythonhosted.org/packages/22/a5/079d216712a4f3ffa24af4a0381b108aa9c45b7a5cc6eb141f81726b1823/httptools-0.7.1-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:f72fdbae2dbc6e68b8239defb48e6a5937b12218e6ffc2c7846cc37befa84362", size = 495186, upload-time = "2025-10-10T03:54:43.937Z" },
+    { url = "https://files.pythonhosted.org/packages/e9/9e/025ad7b65278745dee3bd0ebf9314934c4592560878308a6121f7f812084/httptools-0.7.1-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:e99c7b90a29fd82fea9ef57943d501a16f3404d7b9ee81799d41639bdaae412c", size = 499192, upload-time = "2025-10-10T03:54:45.003Z" },
+    { url = "https://files.pythonhosted.org/packages/6d/de/40a8f202b987d43afc4d54689600ff03ce65680ede2f31df348d7f368b8f/httptools-0.7.1-cp312-cp312-win_amd64.whl", hash = "sha256:3e14f530fefa7499334a79b0cf7e7cd2992870eb893526fb097d51b4f2d0f321", size = 86694, upload-time = "2025-10-10T03:54:45.923Z" },
+    { url = "https://files.pythonhosted.org/packages/09/8f/c77b1fcbfd262d422f12da02feb0d218fa228d52485b77b953832105bb90/httptools-0.7.1-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:6babce6cfa2a99545c60bfef8bee0cc0545413cb0018f617c8059a30ad985de3", size = 202889, upload-time = "2025-10-10T03:54:47.089Z" },
+    { url = "https://files.pythonhosted.org/packages/0a/1a/22887f53602feaa066354867bc49a68fc295c2293433177ee90870a7d517/httptools-0.7.1-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:601b7628de7504077dd3dcb3791c6b8694bbd967148a6d1f01806509254fb1ca", size = 108180, upload-time = "2025-10-10T03:54:48.052Z" },
+    { url = "https://files.pythonhosted.org/packages/32/6a/6aaa91937f0010d288d3d124ca2946d48d60c3a5ee7ca62afe870e3ea011/httptools-0.7.1-cp313-cp313-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl", hash = "sha256:04c6c0e6c5fb0739c5b8a9eb046d298650a0ff38cf42537fc372b28dc7e4472c", size = 478596, upload-time = "2025-10-10T03:54:48.919Z" },
+    { url = "https://files.pythonhosted.org/packages/6d/70/023d7ce117993107be88d2cbca566a7c1323ccbaf0af7eabf2064fe356f6/httptools-0.7.1-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:69d4f9705c405ae3ee83d6a12283dc9feba8cc6aaec671b412917e644ab4fa66", size = 473268, upload-time = "2025-10-10T03:54:49.993Z" },
+    { url = "https://files.pythonhosted.org/packages/32/4d/9dd616c38da088e3f436e9a616e1d0cc66544b8cdac405cc4e81c8679fc7/httptools-0.7.1-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:44c8f4347d4b31269c8a9205d8a5ee2df5322b09bbbd30f8f862185bb6b05346", size = 455517, upload-time = "2025-10-10T03:54:51.066Z" },
+    { url = "https://files.pythonhosted.org/packages/1d/3a/a6c595c310b7df958e739aae88724e24f9246a514d909547778d776799be/httptools-0.7.1-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:465275d76db4d554918aba40bf1cbebe324670f3dfc979eaffaa5d108e2ed650", size = 458337, upload-time = "2025-10-10T03:54:52.196Z" },
+    { url = "https://files.pythonhosted.org/packages/fd/82/88e8d6d2c51edc1cc391b6e044c6c435b6aebe97b1abc33db1b0b24cd582/httptools-0.7.1-cp313-cp313-win_amd64.whl", hash = "sha256:322d00c2068d125bd570f7bf78b2d367dad02b919d8581d7476d8b75b294e3e6", size = 85743, upload-time = "2025-10-10T03:54:53.448Z" },
+    { url = "https://files.pythonhosted.org/packages/34/50/9d095fcbb6de2d523e027a2f304d4551855c2f46e0b82befd718b8b20056/httptools-0.7.1-cp314-cp314-macosx_10_13_universal2.whl", hash = "sha256:c08fe65728b8d70b6923ce31e3956f859d5e1e8548e6f22ec520a962c6757270", size = 203619, upload-time = "2025-10-10T03:54:54.321Z" },
+    { url = "https://files.pythonhosted.org/packages/07/f0/89720dc5139ae54b03f861b5e2c55a37dba9a5da7d51e1e824a1f343627f/httptools-0.7.1-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:7aea2e3c3953521c3c51106ee11487a910d45586e351202474d45472db7d72d3", size = 108714, upload-time = "2025-10-10T03:54:55.163Z" },
+    { url = "https://files.pythonhosted.org/packages/b3/cb/eea88506f191fb552c11787c23f9a405f4c7b0c5799bf73f2249cd4f5228/httptools-0.7.1-cp314-cp314-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl", hash = "sha256:0e68b8582f4ea9166be62926077a3334064d422cf08ab87d8b74664f8e9058e1", size = 472909, upload-time = "2025-10-10T03:54:56.056Z" },
+    { url = "https://files.pythonhosted.org/packages/e0/4a/a548bdfae6369c0d078bab5769f7b66f17f1bfaa6fa28f81d6be6959066b/httptools-0.7.1-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:df091cf961a3be783d6aebae963cc9b71e00d57fa6f149025075217bc6a55a7b", size = 470831, upload-time = "2025-10-10T03:54:57.219Z" },
+    { url = "https://files.pythonhosted.org/packages/4d/31/14df99e1c43bd132eec921c2e7e11cda7852f65619bc0fc5bdc2d0cb126c/httptools-0.7.1-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:f084813239e1eb403ddacd06a30de3d3e09a9b76e7894dcda2b22f8a726e9c60", size = 452631, upload-time = "2025-10-10T03:54:58.219Z" },
+    { url = "https://files.pythonhosted.org/packages/22/d2/b7e131f7be8d854d48cb6d048113c30f9a46dca0c9a8b08fcb3fcd588cdc/httptools-0.7.1-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:7347714368fb2b335e9063bc2b96f2f87a9ceffcd9758ac295f8bbcd3ffbc0ca", size = 452910, upload-time = "2025-10-10T03:54:59.366Z" },
+    { url = "https://files.pythonhosted.org/packages/53/cf/878f3b91e4e6e011eff6d1fa9ca39f7eb17d19c9d7971b04873734112f30/httptools-0.7.1-cp314-cp314-win_amd64.whl", hash = "sha256:cfabda2a5bb85aa2a904ce06d974a3f30fb36cc63d7feaddec05d2050acede96", size = 88205, upload-time = "2025-10-10T03:55:00.389Z" },
+]
+
+[[package]]
+name = "httpx"
+version = "0.28.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio" },
+    { name = "certifi" },
+    { name = "httpcore" },
+    { name = "idna" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b1/df/48c586a5fe32a0f01324ee087459e112ebb7224f646c0b5023f5e79e9956/httpx-0.28.1.tar.gz", hash = "sha256:75e98c5f16b0f35b567856f597f06ff2270a374470a5c2392242528e3e3e42fc", size = 141406, upload-time = "2024-12-06T15:37:23.222Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/2a/39/e50c7c3a983047577ee07d2a9e53faf5a69493943ec3f6a384bdc792deb2/httpx-0.28.1-py3-none-any.whl", hash = "sha256:d909fcccc110f8c7faf814ca82a9a4d816bc5a6dbfea25d6591d6985b8ba59ad", size = 73517, upload-time = "2024-12-06T15:37:21.509Z" },
+]
+
+[[package]]
+name = "idna"
+version = "3.11"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/6f/6d/0703ccc57f3a7233505399edb88de3cbd678da106337b9fcde432b65ed60/idna-3.11.tar.gz", hash = "sha256:795dafcc9c04ed0c1fb032c2aa73654d8e8c5023a7df64a53f39190ada629902", size = 194582, upload-time = "2025-10-12T14:55:20.501Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/0e/61/66938bbb5fc52dbdf84594873d5b51fb1f7c7794e9c0f5bd885f30bc507b/idna-3.11-py3-none-any.whl", hash = "sha256:771a87f49d9defaf64091e6e6fe9c18d4833f140bd19464795bc32d966ca37ea", size = 71008, upload-time = "2025-10-12T14:55:18.883Z" },
+]
+
+[[package]]
+name = "iniconfig"
+version = "2.3.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/72/34/14ca021ce8e5dfedc35312d08ba8bf51fdd999c576889fc2c24cb97f4f10/iniconfig-2.3.0.tar.gz", hash = "sha256:c76315c77db068650d49c5b56314774a7804df16fee4402c1f19d6d15d8c4730", size = 20503, upload-time = "2025-10-18T21:55:43.219Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/cb/b1/3846dd7f199d53cb17f49cba7e651e9ce294d8497c8c150530ed11865bb8/iniconfig-2.3.0-py3-none-any.whl", hash = "sha256:f631c04d2c48c52b84d0d0549c99ff3859c98df65b3101406327ecc7d53fbf12", size = 7484, upload-time = "2025-10-18T21:55:41.639Z" },
+]
+
+[[package]]
+name = "packaging"
+version = "25.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/a1/d4/1fc4078c65507b51b96ca8f8c3ba19e6a61c8253c72794544580a7b6c24d/packaging-25.0.tar.gz", hash = "sha256:d443872c98d677bf60f6a1f2f8c1cb748e8fe762d2bf9d3148b5599295b0fc4f", size = 165727, upload-time = "2025-04-19T11:48:59.673Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/20/12/38679034af332785aac8774540895e234f4d07f7545804097de4b666afd8/packaging-25.0-py3-none-any.whl", hash = "sha256:29572ef2b1f17581046b3a2227d5c611fb25ec70ca1ba8554b24b0e69331a484", size = 66469, upload-time = "2025-04-19T11:48:57.875Z" },
+]
+
+[[package]]
+name = "pluggy"
+version = "1.6.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/f9/e2/3e91f31a7d2b083fe6ef3fa267035b518369d9511ffab804f839851d2779/pluggy-1.6.0.tar.gz", hash = "sha256:7dcc130b76258d33b90f61b658791dede3486c3e6bfb003ee5c9bfb396dd22f3", size = 69412, upload-time = "2025-05-15T12:30:07.975Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/54/20/4d324d65cc6d9205fabedc306948156824eb9f0ee1633355a8f7ec5c66bf/pluggy-1.6.0-py3-none-any.whl", hash = "sha256:e920276dd6813095e9377c0bc5566d94c932c33b27a3e3945d8389c374dd4746", size = 20538, upload-time = "2025-05-15T12:30:06.134Z" },
+]
+
+[[package]]
+name = "posthog"
+source = { editable = "../" }
+dependencies = [
+    { name = "backoff" },
+    { name = "distro" },
+    { name = "python-dateutil" },
+    { name = "requests" },
+    { name = "six" },
+    { name = "typing-extensions" },
+]
+
+[package.metadata]
+requires-dist = [
+    { name = "anthropic", marker = "extra == 'test'" },
+    { name = "backoff", specifier = ">=1.10.0" },
+    { name = "coverage", marker = "extra == 'test'" },
+    { name = "distro", specifier = ">=1.5.0" },
+    { name = "django", marker = "extra == 'test'" },
+    { name = "django-stubs", marker = "extra == 'dev'" },
+    { name = "freezegun", marker = "extra == 'test'", specifier = "==1.5.1" },
+    { name = "google-genai", marker = "extra == 'test'" },
+    { name = "langchain", marker = "extra == 'langchain'", specifier = ">=0.2.0" },
+    { name = "langchain-anthropic", marker = "extra == 'test'", specifier = ">=0.3.15" },
+    { name = "langchain-community", marker = "extra == 'test'", specifier = ">=0.3.25" },
+    { name = "langchain-core", marker = "extra == 'test'", specifier = ">=0.3.65" },
+    { name = "langchain-openai", marker = "extra == 'test'", specifier = ">=0.3.22" },
+    { name = "langgraph", marker = "extra == 'test'", specifier = ">=0.4.8" },
+    { name = "lxml", marker = "extra == 'dev'" },
+    { name = "mock", marker = "extra == 'test'", specifier = ">=2.0.0" },
+    { name = "mypy", marker = "extra == 'dev'" },
+    { name = "mypy-baseline", marker = "extra == 'dev'" },
+    { name = "openai", marker = "extra == 'test'" },
+    { name = "packaging", marker = "extra == 'dev'" },
+    { name = "parameterized", marker = "extra == 'test'", specifier = ">=0.8.1" },
+    { name = "pre-commit", marker = "extra == 'dev'" },
+    { name = "pydantic", marker = "extra == 'dev'" },
+    { name = "pydantic", marker = "extra == 'test'" },
+    { name = "pytest", marker = "extra == 'test'" },
+    { name = "pytest-asyncio", marker = "extra == 'test'" },
+    { name = "pytest-timeout", marker = "extra == 'test'" },
+    { name = "python-dateutil", specifier = ">=2.2" },
+    { name = "requests", specifier = ">=2.7,<3.0" },
+    { name = "ruff", marker = "extra == 'dev'" },
+    { name = "setuptools", marker = "extra == 'dev'" },
+    { name = "six", specifier = ">=1.5" },
+    { name = "tomli", marker = "extra == 'dev'" },
+    { name = "tomli-w", marker = "extra == 'dev'" },
+    { name = "twine", marker = "extra == 'dev'" },
+    { name = "types-mock", marker = "extra == 'dev'" },
+    { name = "types-python-dateutil", marker = "extra == 'dev'" },
+    { name = "types-requests", marker = "extra == 'dev'" },
+    { name = "types-setuptools", marker = "extra == 'dev'" },
+    { name = "types-six", marker = "extra == 'dev'" },
+    { name = "typing-extensions", specifier = ">=4.2.0" },
+    { name = "wheel", marker = "extra == 'dev'" },
+]
+provides-extras = ["langchain", "dev", "test"]
+
+[[package]]
+name = "pygments"
+version = "2.19.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/b0/77/a5b8c569bf593b0140bde72ea885a803b82086995367bf2037de0159d924/pygments-2.19.2.tar.gz", hash = "sha256:636cb2477cec7f8952536970bc533bc43743542f70392ae026374600add5b887", size = 4968631, upload-time = "2025-06-21T13:39:12.283Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c7/21/705964c7812476f378728bdf590ca4b771ec72385c533964653c68e86bdc/pygments-2.19.2-py3-none-any.whl", hash = "sha256:86540386c03d588bb81d44bc3928634ff26449851e99741617ecb9037ee5ec0b", size = 1225217, upload-time = "2025-06-21T13:39:07.939Z" },
+]
+
+[[package]]
+name = "pytest"
+version = "8.4.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+    { name = "iniconfig" },
+    { name = "packaging" },
+    { name = "pluggy" },
+    { name = "pygments" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/a3/5c/00a0e072241553e1a7496d638deababa67c5058571567b92a7eaa258397c/pytest-8.4.2.tar.gz", hash = "sha256:86c0d0b93306b961d58d62a4db4879f27fe25513d4b969df351abdddb3c30e01", size = 1519618, upload-time = "2025-09-04T14:34:22.711Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a8/a4/20da314d277121d6534b3a980b29035dcd51e6744bd79075a6ce8fa4eb8d/pytest-8.4.2-py3-none-any.whl", hash = "sha256:872f880de3fc3a5bdc88a11b39c9710c3497a547cfa9320bc3c5e62fbf272e79", size = 365750, upload-time = "2025-09-04T14:34:20.226Z" },
+]
+
+[[package]]
+name = "pytest-asyncio"
+version = "1.2.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "pytest" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/42/86/9e3c5f48f7b7b638b216e4b9e645f54d199d7abbbab7a64a13b4e12ba10f/pytest_asyncio-1.2.0.tar.gz", hash = "sha256:c609a64a2a8768462d0c99811ddb8bd2583c33fd33cf7f21af1c142e824ffb57", size = 50119, upload-time = "2025-09-12T07:33:53.816Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/04/93/2fa34714b7a4ae72f2f8dad66ba17dd9a2c793220719e736dda28b7aec27/pytest_asyncio-1.2.0-py3-none-any.whl", hash = "sha256:8e17ae5e46d8e7efe51ab6494dd2010f4ca8dae51652aa3c8d55acf50bfb2e99", size = 15095, upload-time = "2025-09-12T07:33:52.639Z" },
+]
+
+[[package]]
+name = "pytest-django"
+version = "4.11.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "pytest" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b1/fb/55d580352db26eb3d59ad50c64321ddfe228d3d8ac107db05387a2fadf3a/pytest_django-4.11.1.tar.gz", hash = "sha256:a949141a1ee103cb0e7a20f1451d355f83f5e4a5d07bdd4dcfdd1fd0ff227991", size = 86202, upload-time = "2025-04-03T18:56:09.338Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/be/ac/bd0608d229ec808e51a21044f3f2f27b9a37e7a0ebaca7247882e67876af/pytest_django-4.11.1-py3-none-any.whl", hash = "sha256:1b63773f648aa3d8541000c26929c1ea63934be1cfa674c76436966d73fe6a10", size = 25281, upload-time = "2025-04-03T18:56:07.678Z" },
+]
+
+[[package]]
+name = "python-dateutil"
+version = "2.9.0.post0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "six" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/66/c0/0c8b6ad9f17a802ee498c46e004a0eb49bc148f2fd230864601a86dcf6db/python-dateutil-2.9.0.post0.tar.gz", hash = "sha256:37dd54208da7e1cd875388217d5e00ebd4179249f90fb72437e91a35459a0ad3", size = 342432, upload-time = "2024-03-01T18:36:20.211Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/ec/57/56b9bcc3c9c6a792fcbaf139543cee77261f3651ca9da0c93f5c1221264b/python_dateutil-2.9.0.post0-py2.py3-none-any.whl", hash = "sha256:a8b2bc7bffae282281c8140a97d3aa9c14da0b136dfe83f850eea9a5f7470427", size = 229892, upload-time = "2024-03-01T18:36:18.57Z" },
+]
+
+[[package]]
+name = "python-dotenv"
+version = "1.2.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/f0/26/19cadc79a718c5edbec86fd4919a6b6d3f681039a2f6d66d14be94e75fb9/python_dotenv-1.2.1.tar.gz", hash = "sha256:42667e897e16ab0d66954af0e60a9caa94f0fd4ecf3aaf6d2d260eec1aa36ad6", size = 44221, upload-time = "2025-10-26T15:12:10.434Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/14/1b/a298b06749107c305e1fe0f814c6c74aea7b2f1e10989cb30f544a1b3253/python_dotenv-1.2.1-py3-none-any.whl", hash = "sha256:b81ee9561e9ca4004139c6cbba3a238c32b03e4894671e181b671e8cb8425d61", size = 21230, upload-time = "2025-10-26T15:12:09.109Z" },
+]
+
+[[package]]
+name = "pyyaml"
+version = "6.0.3"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/05/8e/961c0007c59b8dd7729d542c61a4d537767a59645b82a0b521206e1e25c2/pyyaml-6.0.3.tar.gz", hash = "sha256:d76623373421df22fb4cf8817020cbb7ef15c725b9d5e45f17e189bfc384190f", size = 130960, upload-time = "2025-09-25T21:33:16.546Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d1/33/422b98d2195232ca1826284a76852ad5a86fe23e31b009c9886b2d0fb8b2/pyyaml-6.0.3-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:7f047e29dcae44602496db43be01ad42fc6f1cc0d8cd6c83d342306c32270196", size = 182063, upload-time = "2025-09-25T21:32:11.445Z" },
+    { url = "https://files.pythonhosted.org/packages/89/a0/6cf41a19a1f2f3feab0e9c0b74134aa2ce6849093d5517a0c550fe37a648/pyyaml-6.0.3-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:fc09d0aa354569bc501d4e787133afc08552722d3ab34836a80547331bb5d4a0", size = 173973, upload-time = "2025-09-25T21:32:12.492Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/23/7a778b6bd0b9a8039df8b1b1d80e2e2ad78aa04171592c8a5c43a56a6af4/pyyaml-6.0.3-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:9149cad251584d5fb4981be1ecde53a1ca46c891a79788c0df828d2f166bda28", size = 775116, upload-time = "2025-09-25T21:32:13.652Z" },
+    { url = "https://files.pythonhosted.org/packages/65/30/d7353c338e12baef4ecc1b09e877c1970bd3382789c159b4f89d6a70dc09/pyyaml-6.0.3-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:5fdec68f91a0c6739b380c83b951e2c72ac0197ace422360e6d5a959d8d97b2c", size = 844011, upload-time = "2025-09-25T21:32:15.21Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/9d/b3589d3877982d4f2329302ef98a8026e7f4443c765c46cfecc8858c6b4b/pyyaml-6.0.3-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:ba1cc08a7ccde2d2ec775841541641e4548226580ab850948cbfda66a1befcdc", size = 807870, upload-time = "2025-09-25T21:32:16.431Z" },
+    { url = "https://files.pythonhosted.org/packages/05/c0/b3be26a015601b822b97d9149ff8cb5ead58c66f981e04fedf4e762f4bd4/pyyaml-6.0.3-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:8dc52c23056b9ddd46818a57b78404882310fb473d63f17b07d5c40421e47f8e", size = 761089, upload-time = "2025-09-25T21:32:17.56Z" },
+    { url = "https://files.pythonhosted.org/packages/be/8e/98435a21d1d4b46590d5459a22d88128103f8da4c2d4cb8f14f2a96504e1/pyyaml-6.0.3-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:41715c910c881bc081f1e8872880d3c650acf13dfa8214bad49ed4cede7c34ea", size = 790181, upload-time = "2025-09-25T21:32:18.834Z" },
+    { url = "https://files.pythonhosted.org/packages/74/93/7baea19427dcfbe1e5a372d81473250b379f04b1bd3c4c5ff825e2327202/pyyaml-6.0.3-cp312-cp312-win32.whl", hash = "sha256:96b533f0e99f6579b3d4d4995707cf36df9100d67e0c8303a0c55b27b5f99bc5", size = 137658, upload-time = "2025-09-25T21:32:20.209Z" },
+    { url = "https://files.pythonhosted.org/packages/86/bf/899e81e4cce32febab4fb42bb97dcdf66bc135272882d1987881a4b519e9/pyyaml-6.0.3-cp312-cp312-win_amd64.whl", hash = "sha256:5fcd34e47f6e0b794d17de1b4ff496c00986e1c83f7ab2fb8fcfe9616ff7477b", size = 154003, upload-time = "2025-09-25T21:32:21.167Z" },
+    { url = "https://files.pythonhosted.org/packages/1a/08/67bd04656199bbb51dbed1439b7f27601dfb576fb864099c7ef0c3e55531/pyyaml-6.0.3-cp312-cp312-win_arm64.whl", hash = "sha256:64386e5e707d03a7e172c0701abfb7e10f0fb753ee1d773128192742712a98fd", size = 140344, upload-time = "2025-09-25T21:32:22.617Z" },
+    { url = "https://files.pythonhosted.org/packages/d1/11/0fd08f8192109f7169db964b5707a2f1e8b745d4e239b784a5a1dd80d1db/pyyaml-6.0.3-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:8da9669d359f02c0b91ccc01cac4a67f16afec0dac22c2ad09f46bee0697eba8", size = 181669, upload-time = "2025-09-25T21:32:23.673Z" },
+    { url = "https://files.pythonhosted.org/packages/b1/16/95309993f1d3748cd644e02e38b75d50cbc0d9561d21f390a76242ce073f/pyyaml-6.0.3-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:2283a07e2c21a2aa78d9c4442724ec1eb15f5e42a723b99cb3d822d48f5f7ad1", size = 173252, upload-time = "2025-09-25T21:32:25.149Z" },
+    { url = "https://files.pythonhosted.org/packages/50/31/b20f376d3f810b9b2371e72ef5adb33879b25edb7a6d072cb7ca0c486398/pyyaml-6.0.3-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:ee2922902c45ae8ccada2c5b501ab86c36525b883eff4255313a253a3160861c", size = 767081, upload-time = "2025-09-25T21:32:26.575Z" },
+    { url = "https://files.pythonhosted.org/packages/49/1e/a55ca81e949270d5d4432fbbd19dfea5321eda7c41a849d443dc92fd1ff7/pyyaml-6.0.3-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:a33284e20b78bd4a18c8c2282d549d10bc8408a2a7ff57653c0cf0b9be0afce5", size = 841159, upload-time = "2025-09-25T21:32:27.727Z" },
+    { url = "https://files.pythonhosted.org/packages/74/27/e5b8f34d02d9995b80abcef563ea1f8b56d20134d8f4e5e81733b1feceb2/pyyaml-6.0.3-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:0f29edc409a6392443abf94b9cf89ce99889a1dd5376d94316ae5145dfedd5d6", size = 801626, upload-time = "2025-09-25T21:32:28.878Z" },
+    { url = "https://files.pythonhosted.org/packages/f9/11/ba845c23988798f40e52ba45f34849aa8a1f2d4af4b798588010792ebad6/pyyaml-6.0.3-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:f7057c9a337546edc7973c0d3ba84ddcdf0daa14533c2065749c9075001090e6", size = 753613, upload-time = "2025-09-25T21:32:30.178Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/e0/7966e1a7bfc0a45bf0a7fb6b98ea03fc9b8d84fa7f2229e9659680b69ee3/pyyaml-6.0.3-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:eda16858a3cab07b80edaf74336ece1f986ba330fdb8ee0d6c0d68fe82bc96be", size = 794115, upload-time = "2025-09-25T21:32:31.353Z" },
+    { url = "https://files.pythonhosted.org/packages/de/94/980b50a6531b3019e45ddeada0626d45fa85cbe22300844a7983285bed3b/pyyaml-6.0.3-cp313-cp313-win32.whl", hash = "sha256:d0eae10f8159e8fdad514efdc92d74fd8d682c933a6dd088030f3834bc8e6b26", size = 137427, upload-time = "2025-09-25T21:32:32.58Z" },
+    { url = "https://files.pythonhosted.org/packages/97/c9/39d5b874e8b28845e4ec2202b5da735d0199dbe5b8fb85f91398814a9a46/pyyaml-6.0.3-cp313-cp313-win_amd64.whl", hash = "sha256:79005a0d97d5ddabfeeea4cf676af11e647e41d81c9a7722a193022accdb6b7c", size = 154090, upload-time = "2025-09-25T21:32:33.659Z" },
+    { url = "https://files.pythonhosted.org/packages/73/e8/2bdf3ca2090f68bb3d75b44da7bbc71843b19c9f2b9cb9b0f4ab7a5a4329/pyyaml-6.0.3-cp313-cp313-win_arm64.whl", hash = "sha256:5498cd1645aa724a7c71c8f378eb29ebe23da2fc0d7a08071d89469bf1d2defb", size = 140246, upload-time = "2025-09-25T21:32:34.663Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/8c/f4bd7f6465179953d3ac9bc44ac1a8a3e6122cf8ada906b4f96c60172d43/pyyaml-6.0.3-cp314-cp314-macosx_10_13_x86_64.whl", hash = "sha256:8d1fab6bb153a416f9aeb4b8763bc0f22a5586065f86f7664fc23339fc1c1fac", size = 181814, upload-time = "2025-09-25T21:32:35.712Z" },
+    { url = "https://files.pythonhosted.org/packages/bd/9c/4d95bb87eb2063d20db7b60faa3840c1b18025517ae857371c4dd55a6b3a/pyyaml-6.0.3-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:34d5fcd24b8445fadc33f9cf348c1047101756fd760b4dacb5c3e99755703310", size = 173809, upload-time = "2025-09-25T21:32:36.789Z" },
+    { url = "https://files.pythonhosted.org/packages/92/b5/47e807c2623074914e29dabd16cbbdd4bf5e9b2db9f8090fa64411fc5382/pyyaml-6.0.3-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:501a031947e3a9025ed4405a168e6ef5ae3126c59f90ce0cd6f2bfc477be31b7", size = 766454, upload-time = "2025-09-25T21:32:37.966Z" },
+    { url = "https://files.pythonhosted.org/packages/02/9e/e5e9b168be58564121efb3de6859c452fccde0ab093d8438905899a3a483/pyyaml-6.0.3-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:b3bc83488de33889877a0f2543ade9f70c67d66d9ebb4ac959502e12de895788", size = 836355, upload-time = "2025-09-25T21:32:39.178Z" },
+    { url = "https://files.pythonhosted.org/packages/88/f9/16491d7ed2a919954993e48aa941b200f38040928474c9e85ea9e64222c3/pyyaml-6.0.3-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:c458b6d084f9b935061bc36216e8a69a7e293a2f1e68bf956dcd9e6cbcd143f5", size = 794175, upload-time = "2025-09-25T21:32:40.865Z" },
+    { url = "https://files.pythonhosted.org/packages/dd/3f/5989debef34dc6397317802b527dbbafb2b4760878a53d4166579111411e/pyyaml-6.0.3-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:7c6610def4f163542a622a73fb39f534f8c101d690126992300bf3207eab9764", size = 755228, upload-time = "2025-09-25T21:32:42.084Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/ce/af88a49043cd2e265be63d083fc75b27b6ed062f5f9fd6cdc223ad62f03e/pyyaml-6.0.3-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:5190d403f121660ce8d1d2c1bb2ef1bd05b5f68533fc5c2ea899bd15f4399b35", size = 789194, upload-time = "2025-09-25T21:32:43.362Z" },
+    { url = "https://files.pythonhosted.org/packages/23/20/bb6982b26a40bb43951265ba29d4c246ef0ff59c9fdcdf0ed04e0687de4d/pyyaml-6.0.3-cp314-cp314-win_amd64.whl", hash = "sha256:4a2e8cebe2ff6ab7d1050ecd59c25d4c8bd7e6f400f5f82b96557ac0abafd0ac", size = 156429, upload-time = "2025-09-25T21:32:57.844Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/f4/a4541072bb9422c8a883ab55255f918fa378ecf083f5b85e87fc2b4eda1b/pyyaml-6.0.3-cp314-cp314-win_arm64.whl", hash = "sha256:93dda82c9c22deb0a405ea4dc5f2d0cda384168e466364dec6255b293923b2f3", size = 143912, upload-time = "2025-09-25T21:32:59.247Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/f9/07dd09ae774e4616edf6cda684ee78f97777bdd15847253637a6f052a62f/pyyaml-6.0.3-cp314-cp314t-macosx_10_13_x86_64.whl", hash = "sha256:02893d100e99e03eda1c8fd5c441d8c60103fd175728e23e431db1b589cf5ab3", size = 189108, upload-time = "2025-09-25T21:32:44.377Z" },
+    { url = "https://files.pythonhosted.org/packages/4e/78/8d08c9fb7ce09ad8c38ad533c1191cf27f7ae1effe5bb9400a46d9437fcf/pyyaml-6.0.3-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:c1ff362665ae507275af2853520967820d9124984e0f7466736aea23d8611fba", size = 183641, upload-time = "2025-09-25T21:32:45.407Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/5b/3babb19104a46945cf816d047db2788bcaf8c94527a805610b0289a01c6b/pyyaml-6.0.3-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:6adc77889b628398debc7b65c073bcb99c4a0237b248cacaf3fe8a557563ef6c", size = 831901, upload-time = "2025-09-25T21:32:48.83Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/cc/dff0684d8dc44da4d22a13f35f073d558c268780ce3c6ba1b87055bb0b87/pyyaml-6.0.3-cp314-cp314t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:a80cb027f6b349846a3bf6d73b5e95e782175e52f22108cfa17876aaeff93702", size = 861132, upload-time = "2025-09-25T21:32:50.149Z" },
+    { url = "https://files.pythonhosted.org/packages/b1/5e/f77dc6b9036943e285ba76b49e118d9ea929885becb0a29ba8a7c75e29fe/pyyaml-6.0.3-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:00c4bdeba853cc34e7dd471f16b4114f4162dc03e6b7afcc2128711f0eca823c", size = 839261, upload-time = "2025-09-25T21:32:51.808Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/88/a9db1376aa2a228197c58b37302f284b5617f56a5d959fd1763fb1675ce6/pyyaml-6.0.3-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:66e1674c3ef6f541c35191caae2d429b967b99e02040f5ba928632d9a7f0f065", size = 805272, upload-time = "2025-09-25T21:32:52.941Z" },
+    { url = "https://files.pythonhosted.org/packages/da/92/1446574745d74df0c92e6aa4a7b0b3130706a4142b2d1a5869f2eaa423c6/pyyaml-6.0.3-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:16249ee61e95f858e83976573de0f5b2893b3677ba71c9dd36b9cf8be9ac6d65", size = 829923, upload-time = "2025-09-25T21:32:54.537Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/7a/1c7270340330e575b92f397352af856a8c06f230aa3e76f86b39d01b416a/pyyaml-6.0.3-cp314-cp314t-win_amd64.whl", hash = "sha256:4ad1906908f2f5ae4e5a8ddfce73c320c2a1429ec52eafd27138b7f1cbe341c9", size = 174062, upload-time = "2025-09-25T21:32:55.767Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/12/de94a39c2ef588c7e6455cfbe7343d3b2dc9d6b6b2f40c4c6565744c873d/pyyaml-6.0.3-cp314-cp314t-win_arm64.whl", hash = "sha256:ebc55a14a21cb14062aa4162f906cd962b28e2e9ea38f9b4391244cd8de4ae0b", size = 149341, upload-time = "2025-09-25T21:32:56.828Z" },
+]
+
+[[package]]
+name = "requests"
+version = "2.32.5"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "certifi" },
+    { name = "charset-normalizer" },
+    { name = "idna" },
+    { name = "urllib3" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/c9/74/b3ff8e6c8446842c3f5c837e9c3dfcfe2018ea6ecef224c710c85ef728f4/requests-2.32.5.tar.gz", hash = "sha256:dbba0bac56e100853db0ea71b82b4dfd5fe2bf6d3754a8893c3af500cec7d7cf", size = 134517, upload-time = "2025-08-18T20:46:02.573Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/1e/db/4254e3eabe8020b458f1a747140d32277ec7a271daf1d235b70dc0b4e6e3/requests-2.32.5-py3-none-any.whl", hash = "sha256:2462f94637a34fd532264295e186976db0f5d453d1cdd31473c85a6a161affb6", size = 64738, upload-time = "2025-08-18T20:46:00.542Z" },
+]
+
+[[package]]
+name = "six"
+version = "1.17.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/94/e7/b2c673351809dca68a0e064b6af791aa332cf192da575fd474ed7d6f16a2/six-1.17.0.tar.gz", hash = "sha256:ff70335d468e7eb6ec65b95b99d3a2836546063f63acc5171de367e834932a81", size = 34031, upload-time = "2024-12-04T17:35:28.174Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/b7/ce/149a00dd41f10bc29e5921b496af8b574d8413afcd5e30dfa0ed46c2cc5e/six-1.17.0-py2.py3-none-any.whl", hash = "sha256:4721f391ed90541fddacab5acf947aa0d3dc7d27b2e1e8eda2be8970586c3274", size = 11050, upload-time = "2024-12-04T17:35:26.475Z" },
+]
+
+[[package]]
+name = "sniffio"
+version = "1.3.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/a2/87/a6771e1546d97e7e041b6ae58d80074f81b7d5121207425c964ddf5cfdbd/sniffio-1.3.1.tar.gz", hash = "sha256:f4324edc670a0f49750a81b895f35c3adb843cca46f0530f79fc1babb23789dc", size = 20372, upload-time = "2024-02-25T23:20:04.057Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e9/44/75a9c9421471a6c4805dbf2356f7c181a29c1879239abab1ea2cc8f38b40/sniffio-1.3.1-py3-none-any.whl", hash = "sha256:2f6da418d1f1e0fddd844478f41680e794e6051915791a034ff65e5f100525a2", size = 10235, upload-time = "2024-02-25T23:20:01.196Z" },
+]
+
+[[package]]
+name = "sqlparse"
+version = "0.5.3"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/e5/40/edede8dd6977b0d3da179a342c198ed100dd2aba4be081861ee5911e4da4/sqlparse-0.5.3.tar.gz", hash = "sha256:09f67787f56a0b16ecdbde1bfc7f5d9c3371ca683cfeaa8e6ff60b4807ec9272", size = 84999, upload-time = "2024-12-10T12:05:30.728Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a9/5c/bfd6bd0bf979426d405cc6e71eceb8701b148b16c21d2dc3c261efc61c7b/sqlparse-0.5.3-py3-none-any.whl", hash = "sha256:cf2196ed3418f3ba5de6af7e82c694a9fbdbfecccdfc72e281548517081f16ca", size = 44415, upload-time = "2024-12-10T12:05:27.824Z" },
+]
+
+[[package]]
+name = "test-django5"
+version = "0.1.0"
+source = { virtual = "." }
+dependencies = [
+    { name = "django" },
+    { name = "httpx" },
+    { name = "posthog" },
+    { name = "pytest" },
+    { name = "pytest-asyncio" },
+    { name = "pytest-django" },
+    { name = "uvicorn", extra = ["standard"] },
+]
+
+[package.metadata]
+requires-dist = [
+    { name = "django", specifier = "~=5.2.7" },
+    { name = "httpx", specifier = "~=0.28.1" },
+    { name = "posthog", editable = "../" },
+    { name = "pytest", specifier = "~=8.4.2" },
+    { name = "pytest-asyncio", specifier = "~=1.2.0" },
+    { name = "pytest-django", specifier = "~=4.11.1" },
+    { name = "uvicorn", extras = ["standard"], specifier = "~=0.38.0" },
+]
+
+[[package]]
+name = "typing-extensions"
+version = "4.15.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/72/94/1a15dd82efb362ac84269196e94cf00f187f7ed21c242792a923cdb1c61f/typing_extensions-4.15.0.tar.gz", hash = "sha256:0cea48d173cc12fa28ecabc3b837ea3cf6f38c6d1136f85cbaaf598984861466", size = 109391, upload-time = "2025-08-25T13:49:26.313Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/18/67/36e9267722cc04a6b9f15c7f3441c2363321a3ea07da7ae0c0707beb2a9c/typing_extensions-4.15.0-py3-none-any.whl", hash = "sha256:f0fa19c6845758ab08074a0cfa8b7aecb71c999ca73d62883bc25cc018c4e548", size = 44614, upload-time = "2025-08-25T13:49:24.86Z" },
+]
+
+[[package]]
+name = "tzdata"
+version = "2025.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/95/32/1a225d6164441be760d75c2c42e2780dc0873fe382da3e98a2e1e48361e5/tzdata-2025.2.tar.gz", hash = "sha256:b60a638fcc0daffadf82fe0f57e53d06bdec2f36c4df66280ae79bce6bd6f2b9", size = 196380, upload-time = "2025-03-23T13:54:43.652Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/5c/23/c7abc0ca0a1526a0774eca151daeb8de62ec457e77262b66b359c3c7679e/tzdata-2025.2-py2.py3-none-any.whl", hash = "sha256:1a403fada01ff9221ca8044d701868fa132215d84beb92242d9acd2147f667a8", size = 347839, upload-time = "2025-03-23T13:54:41.845Z" },
+]
+
+[[package]]
+name = "urllib3"
+version = "2.5.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/15/22/9ee70a2574a4f4599c47dd506532914ce044817c7752a79b6a51286319bc/urllib3-2.5.0.tar.gz", hash = "sha256:3fc47733c7e419d4bc3f6b3dc2b4f890bb743906a30d56ba4a5bfa4bbff92760", size = 393185, upload-time = "2025-06-18T14:07:41.644Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a7/c2/fe1e52489ae3122415c51f387e221dd0773709bad6c6cdaa599e8a2c5185/urllib3-2.5.0-py3-none-any.whl", hash = "sha256:e6b01673c0fa6a13e374b50871808eb3bf7046c4b125b216f6bf1cc604cff0dc", size = 129795, upload-time = "2025-06-18T14:07:40.39Z" },
+]
+
+[[package]]
+name = "uvicorn"
+version = "0.38.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "click" },
+    { name = "h11" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/cb/ce/f06b84e2697fef4688ca63bdb2fdf113ca0a3be33f94488f2cadb690b0cf/uvicorn-0.38.0.tar.gz", hash = "sha256:fd97093bdd120a2609fc0d3afe931d4d4ad688b6e75f0f929fde1bc36fe0e91d", size = 80605, upload-time = "2025-10-18T13:46:44.63Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/ee/d9/d88e73ca598f4f6ff671fb5fde8a32925c2e08a637303a1d12883c7305fa/uvicorn-0.38.0-py3-none-any.whl", hash = "sha256:48c0afd214ceb59340075b4a052ea1ee91c16fbc2a9b1469cca0e54566977b02", size = 68109, upload-time = "2025-10-18T13:46:42.958Z" },
+]
+
+[package.optional-dependencies]
+standard = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+    { name = "httptools" },
+    { name = "python-dotenv" },
+    { name = "pyyaml" },
+    { name = "uvloop", marker = "platform_python_implementation != 'PyPy' and sys_platform != 'cygwin' and sys_platform != 'win32'" },
+    { name = "watchfiles" },
+    { name = "websockets" },
+]
+
+[[package]]
+name = "uvloop"
+version = "0.22.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/06/f0/18d39dbd1971d6d62c4629cc7fa67f74821b0dc1f5a77af43719de7936a7/uvloop-0.22.1.tar.gz", hash = "sha256:6c84bae345b9147082b17371e3dd5d42775bddce91f885499017f4607fdaf39f", size = 2443250, upload-time = "2025-10-16T22:17:19.342Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/3d/ff/7f72e8170be527b4977b033239a83a68d5c881cc4775fca255c677f7ac5d/uvloop-0.22.1-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:fe94b4564e865d968414598eea1a6de60adba0c040ba4ed05ac1300de402cd42", size = 1359936, upload-time = "2025-10-16T22:16:29.436Z" },
+    { url = "https://files.pythonhosted.org/packages/c3/c6/e5d433f88fd54d81ef4be58b2b7b0cea13c442454a1db703a1eea0db1a59/uvloop-0.22.1-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:51eb9bd88391483410daad430813d982010f9c9c89512321f5b60e2cddbdddd6", size = 752769, upload-time = "2025-10-16T22:16:30.493Z" },
+    { url = "https://files.pythonhosted.org/packages/24/68/a6ac446820273e71aa762fa21cdcc09861edd3536ff47c5cd3b7afb10eeb/uvloop-0.22.1-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:700e674a166ca5778255e0e1dc4e9d79ab2acc57b9171b79e65feba7184b3370", size = 4317413, upload-time = "2025-10-16T22:16:31.644Z" },
+    { url = "https://files.pythonhosted.org/packages/5f/6f/e62b4dfc7ad6518e7eff2516f680d02a0f6eb62c0c212e152ca708a0085e/uvloop-0.22.1-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:7b5b1ac819a3f946d3b2ee07f09149578ae76066d70b44df3fa990add49a82e4", size = 4426307, upload-time = "2025-10-16T22:16:32.917Z" },
+    { url = "https://files.pythonhosted.org/packages/90/60/97362554ac21e20e81bcef1150cb2a7e4ffdaf8ea1e5b2e8bf7a053caa18/uvloop-0.22.1-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:e047cc068570bac9866237739607d1313b9253c3051ad84738cbb095be0537b2", size = 4131970, upload-time = "2025-10-16T22:16:34.015Z" },
+    { url = "https://files.pythonhosted.org/packages/99/39/6b3f7d234ba3964c428a6e40006340f53ba37993f46ed6e111c6e9141d18/uvloop-0.22.1-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:512fec6815e2dd45161054592441ef76c830eddaad55c8aa30952e6fe1ed07c0", size = 4296343, upload-time = "2025-10-16T22:16:35.149Z" },
+    { url = "https://files.pythonhosted.org/packages/89/8c/182a2a593195bfd39842ea68ebc084e20c850806117213f5a299dfc513d9/uvloop-0.22.1-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:561577354eb94200d75aca23fbde86ee11be36b00e52a4eaf8f50fb0c86b7705", size = 1358611, upload-time = "2025-10-16T22:16:36.833Z" },
+    { url = "https://files.pythonhosted.org/packages/d2/14/e301ee96a6dc95224b6f1162cd3312f6d1217be3907b79173b06785f2fe7/uvloop-0.22.1-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:1cdf5192ab3e674ca26da2eada35b288d2fa49fdd0f357a19f0e7c4e7d5077c8", size = 751811, upload-time = "2025-10-16T22:16:38.275Z" },
+    { url = "https://files.pythonhosted.org/packages/b7/02/654426ce265ac19e2980bfd9ea6590ca96a56f10c76e63801a2df01c0486/uvloop-0.22.1-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:6e2ea3d6190a2968f4a14a23019d3b16870dd2190cd69c8180f7c632d21de68d", size = 4288562, upload-time = "2025-10-16T22:16:39.375Z" },
+    { url = "https://files.pythonhosted.org/packages/15/c0/0be24758891ef825f2065cd5db8741aaddabe3e248ee6acc5e8a80f04005/uvloop-0.22.1-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:0530a5fbad9c9e4ee3f2b33b148c6a64d47bbad8000ea63704fa8260f4cf728e", size = 4366890, upload-time = "2025-10-16T22:16:40.547Z" },
+    { url = "https://files.pythonhosted.org/packages/d2/53/8369e5219a5855869bcee5f4d317f6da0e2c669aecf0ef7d371e3d084449/uvloop-0.22.1-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:bc5ef13bbc10b5335792360623cc378d52d7e62c2de64660616478c32cd0598e", size = 4119472, upload-time = "2025-10-16T22:16:41.694Z" },
+    { url = "https://files.pythonhosted.org/packages/f8/ba/d69adbe699b768f6b29a5eec7b47dd610bd17a69de51b251126a801369ea/uvloop-0.22.1-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:1f38ec5e3f18c8a10ded09742f7fb8de0108796eb673f30ce7762ce1b8550cad", size = 4239051, upload-time = "2025-10-16T22:16:43.224Z" },
+    { url = "https://files.pythonhosted.org/packages/90/cd/b62bdeaa429758aee8de8b00ac0dd26593a9de93d302bff3d21439e9791d/uvloop-0.22.1-cp314-cp314-macosx_10_13_universal2.whl", hash = "sha256:3879b88423ec7e97cd4eba2a443aa26ed4e59b45e6b76aabf13fe2f27023a142", size = 1362067, upload-time = "2025-10-16T22:16:44.503Z" },
+    { url = "https://files.pythonhosted.org/packages/0d/f8/a132124dfda0777e489ca86732e85e69afcd1ff7686647000050ba670689/uvloop-0.22.1-cp314-cp314-macosx_10_13_x86_64.whl", hash = "sha256:4baa86acedf1d62115c1dc6ad1e17134476688f08c6efd8a2ab076e815665c74", size = 752423, upload-time = "2025-10-16T22:16:45.968Z" },
+    { url = "https://files.pythonhosted.org/packages/a3/94/94af78c156f88da4b3a733773ad5ba0b164393e357cc4bd0ab2e2677a7d6/uvloop-0.22.1-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:297c27d8003520596236bdb2335e6b3f649480bd09e00d1e3a99144b691d2a35", size = 4272437, upload-time = "2025-10-16T22:16:47.451Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/35/60249e9fd07b32c665192cec7af29e06c7cd96fa1d08b84f012a56a0b38e/uvloop-0.22.1-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:c1955d5a1dd43198244d47664a5858082a3239766a839b2102a269aaff7a4e25", size = 4292101, upload-time = "2025-10-16T22:16:49.318Z" },
+    { url = "https://files.pythonhosted.org/packages/02/62/67d382dfcb25d0a98ce73c11ed1a6fba5037a1a1d533dcbb7cab033a2636/uvloop-0.22.1-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:b31dc2fccbd42adc73bc4e7cdbae4fc5086cf378979e53ca5d0301838c5682c6", size = 4114158, upload-time = "2025-10-16T22:16:50.517Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/7a/f1171b4a882a5d13c8b7576f348acfe6074d72eaf52cccef752f748d4a9f/uvloop-0.22.1-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:93f617675b2d03af4e72a5333ef89450dfaa5321303ede6e67ba9c9d26878079", size = 4177360, upload-time = "2025-10-16T22:16:52.646Z" },
+    { url = "https://files.pythonhosted.org/packages/79/7b/b01414f31546caf0919da80ad57cbfe24c56b151d12af68cee1b04922ca8/uvloop-0.22.1-cp314-cp314t-macosx_10_13_universal2.whl", hash = "sha256:37554f70528f60cad66945b885eb01f1bb514f132d92b6eeed1c90fd54ed6289", size = 1454790, upload-time = "2025-10-16T22:16:54.355Z" },
+    { url = "https://files.pythonhosted.org/packages/d4/31/0bb232318dd838cad3fa8fb0c68c8b40e1145b32025581975e18b11fab40/uvloop-0.22.1-cp314-cp314t-macosx_10_13_x86_64.whl", hash = "sha256:b76324e2dc033a0b2f435f33eb88ff9913c156ef78e153fb210e03c13da746b3", size = 796783, upload-time = "2025-10-16T22:16:55.906Z" },
+    { url = "https://files.pythonhosted.org/packages/42/38/c9b09f3271a7a723a5de69f8e237ab8e7803183131bc57c890db0b6bb872/uvloop-0.22.1-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:badb4d8e58ee08dad957002027830d5c3b06aea446a6a3744483c2b3b745345c", size = 4647548, upload-time = "2025-10-16T22:16:57.008Z" },
+    { url = "https://files.pythonhosted.org/packages/c1/37/945b4ca0ac27e3dc4952642d4c900edd030b3da6c9634875af6e13ae80e5/uvloop-0.22.1-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:b91328c72635f6f9e0282e4a57da7470c7350ab1c9f48546c0f2866205349d21", size = 4467065, upload-time = "2025-10-16T22:16:58.206Z" },
+    { url = "https://files.pythonhosted.org/packages/97/cc/48d232f33d60e2e2e0b42f4e73455b146b76ebe216487e862700457fbf3c/uvloop-0.22.1-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:daf620c2995d193449393d6c62131b3fbd40a63bf7b307a1527856ace637fe88", size = 4328384, upload-time = "2025-10-16T22:16:59.36Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/16/c1fd27e9549f3c4baf1dc9c20c456cd2f822dbf8de9f463824b0c0357e06/uvloop-0.22.1-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:6cde23eeda1a25c75b2e07d39970f3374105d5eafbaab2a4482be82f272d5a5e", size = 4296730, upload-time = "2025-10-16T22:17:00.744Z" },
+]
+
+[[package]]
+name = "watchfiles"
+version = "1.1.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/c2/c9/8869df9b2a2d6c59d79220a4db37679e74f807c559ffe5265e08b227a210/watchfiles-1.1.1.tar.gz", hash = "sha256:a173cb5c16c4f40ab19cecf48a534c409f7ea983ab8fed0741304a1c0a31b3f2", size = 94440, upload-time = "2025-10-14T15:06:21.08Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/74/d5/f039e7e3c639d9b1d09b07ea412a6806d38123f0508e5f9b48a87b0a76cc/watchfiles-1.1.1-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:8c89f9f2f740a6b7dcc753140dd5e1ab9215966f7a3530d0c0705c83b401bd7d", size = 404745, upload-time = "2025-10-14T15:04:46.731Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/96/a881a13aa1349827490dab2d363c8039527060cfcc2c92cc6d13d1b1049e/watchfiles-1.1.1-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:bd404be08018c37350f0d6e34676bd1e2889990117a2b90070b3007f172d0610", size = 391769, upload-time = "2025-10-14T15:04:48.003Z" },
+    { url = "https://files.pythonhosted.org/packages/4b/5b/d3b460364aeb8da471c1989238ea0e56bec24b6042a68046adf3d9ddb01c/watchfiles-1.1.1-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:8526e8f916bb5b9a0a777c8317c23ce65de259422bba5b31325a6fa6029d33af", size = 449374, upload-time = "2025-10-14T15:04:49.179Z" },
+    { url = "https://files.pythonhosted.org/packages/b9/44/5769cb62d4ed055cb17417c0a109a92f007114a4e07f30812a73a4efdb11/watchfiles-1.1.1-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:2edc3553362b1c38d9f06242416a5d8e9fe235c204a4072e988ce2e5bb1f69f6", size = 459485, upload-time = "2025-10-14T15:04:50.155Z" },
+    { url = "https://files.pythonhosted.org/packages/19/0c/286b6301ded2eccd4ffd0041a1b726afda999926cf720aab63adb68a1e36/watchfiles-1.1.1-cp312-cp312-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:30f7da3fb3f2844259cba4720c3fc7138eb0f7b659c38f3bfa65084c7fc7abce", size = 488813, upload-time = "2025-10-14T15:04:51.059Z" },
+    { url = "https://files.pythonhosted.org/packages/c7/2b/8530ed41112dd4a22f4dcfdb5ccf6a1baad1ff6eed8dc5a5f09e7e8c41c7/watchfiles-1.1.1-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f8979280bdafff686ba5e4d8f97840f929a87ed9cdf133cbbd42f7766774d2aa", size = 594816, upload-time = "2025-10-14T15:04:52.031Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/d2/f5f9fb49489f184f18470d4f99f4e862a4b3e9ac2865688eb2099e3d837a/watchfiles-1.1.1-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:dcc5c24523771db3a294c77d94771abcfcb82a0e0ee8efd910c37c59ec1b31bb", size = 475186, upload-time = "2025-10-14T15:04:53.064Z" },
+    { url = "https://files.pythonhosted.org/packages/cf/68/5707da262a119fb06fbe214d82dd1fe4a6f4af32d2d14de368d0349eb52a/watchfiles-1.1.1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:1db5d7ae38ff20153d542460752ff397fcf5c96090c1230803713cf3147a6803", size = 456812, upload-time = "2025-10-14T15:04:55.174Z" },
+    { url = "https://files.pythonhosted.org/packages/66/ab/3cbb8756323e8f9b6f9acb9ef4ec26d42b2109bce830cc1f3468df20511d/watchfiles-1.1.1-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:28475ddbde92df1874b6c5c8aaeb24ad5be47a11f87cde5a28ef3835932e3e94", size = 630196, upload-time = "2025-10-14T15:04:56.22Z" },
+    { url = "https://files.pythonhosted.org/packages/78/46/7152ec29b8335f80167928944a94955015a345440f524d2dfe63fc2f437b/watchfiles-1.1.1-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:36193ed342f5b9842edd3532729a2ad55c4160ffcfa3700e0d54be496b70dd43", size = 622657, upload-time = "2025-10-14T15:04:57.521Z" },
+    { url = "https://files.pythonhosted.org/packages/0a/bf/95895e78dd75efe9a7f31733607f384b42eb5feb54bd2eb6ed57cc2e94f4/watchfiles-1.1.1-cp312-cp312-win32.whl", hash = "sha256:859e43a1951717cc8de7f4c77674a6d389b106361585951d9e69572823f311d9", size = 272042, upload-time = "2025-10-14T15:04:59.046Z" },
+    { url = "https://files.pythonhosted.org/packages/87/0a/90eb755f568de2688cb220171c4191df932232c20946966c27a59c400850/watchfiles-1.1.1-cp312-cp312-win_amd64.whl", hash = "sha256:91d4c9a823a8c987cce8fa2690923b069966dabb196dd8d137ea2cede885fde9", size = 288410, upload-time = "2025-10-14T15:05:00.081Z" },
+    { url = "https://files.pythonhosted.org/packages/36/76/f322701530586922fbd6723c4f91ace21364924822a8772c549483abed13/watchfiles-1.1.1-cp312-cp312-win_arm64.whl", hash = "sha256:a625815d4a2bdca61953dbba5a39d60164451ef34c88d751f6c368c3ea73d404", size = 278209, upload-time = "2025-10-14T15:05:01.168Z" },
+    { url = "https://files.pythonhosted.org/packages/bb/f4/f750b29225fe77139f7ae5de89d4949f5a99f934c65a1f1c0b248f26f747/watchfiles-1.1.1-cp313-cp313-macosx_10_12_x86_64.whl", hash = "sha256:130e4876309e8686a5e37dba7d5e9bc77e6ed908266996ca26572437a5271e18", size = 404321, upload-time = "2025-10-14T15:05:02.063Z" },
+    { url = "https://files.pythonhosted.org/packages/2b/f9/f07a295cde762644aa4c4bb0f88921d2d141af45e735b965fb2e87858328/watchfiles-1.1.1-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:5f3bde70f157f84ece3765b42b4a52c6ac1a50334903c6eaf765362f6ccca88a", size = 391783, upload-time = "2025-10-14T15:05:03.052Z" },
+    { url = "https://files.pythonhosted.org/packages/bc/11/fc2502457e0bea39a5c958d86d2cb69e407a4d00b85735ca724bfa6e0d1a/watchfiles-1.1.1-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:14e0b1fe858430fc0251737ef3824c54027bedb8c37c38114488b8e131cf8219", size = 449279, upload-time = "2025-10-14T15:05:04.004Z" },
+    { url = "https://files.pythonhosted.org/packages/e3/1f/d66bc15ea0b728df3ed96a539c777acfcad0eb78555ad9efcaa1274688f0/watchfiles-1.1.1-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:f27db948078f3823a6bb3b465180db8ebecf26dd5dae6f6180bd87383b6b4428", size = 459405, upload-time = "2025-10-14T15:05:04.942Z" },
+    { url = "https://files.pythonhosted.org/packages/be/90/9f4a65c0aec3ccf032703e6db02d89a157462fbb2cf20dd415128251cac0/watchfiles-1.1.1-cp313-cp313-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:059098c3a429f62fc98e8ec62b982230ef2c8df68c79e826e37b895bc359a9c0", size = 488976, upload-time = "2025-10-14T15:05:05.905Z" },
+    { url = "https://files.pythonhosted.org/packages/37/57/ee347af605d867f712be7029bb94c8c071732a4b44792e3176fa3c612d39/watchfiles-1.1.1-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:bfb5862016acc9b869bb57284e6cb35fdf8e22fe59f7548858e2f971d045f150", size = 595506, upload-time = "2025-10-14T15:05:06.906Z" },
+    { url = "https://files.pythonhosted.org/packages/a8/78/cc5ab0b86c122047f75e8fc471c67a04dee395daf847d3e59381996c8707/watchfiles-1.1.1-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:319b27255aacd9923b8a276bb14d21a5f7ff82564c744235fc5eae58d95422ae", size = 474936, upload-time = "2025-10-14T15:05:07.906Z" },
+    { url = "https://files.pythonhosted.org/packages/62/da/def65b170a3815af7bd40a3e7010bf6ab53089ef1b75d05dd5385b87cf08/watchfiles-1.1.1-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:c755367e51db90e75b19454b680903631d41f9e3607fbd941d296a020c2d752d", size = 456147, upload-time = "2025-10-14T15:05:09.138Z" },
+    { url = "https://files.pythonhosted.org/packages/57/99/da6573ba71166e82d288d4df0839128004c67d2778d3b566c138695f5c0b/watchfiles-1.1.1-cp313-cp313-musllinux_1_1_aarch64.whl", hash = "sha256:c22c776292a23bfc7237a98f791b9ad3144b02116ff10d820829ce62dff46d0b", size = 630007, upload-time = "2025-10-14T15:05:10.117Z" },
+    { url = "https://files.pythonhosted.org/packages/a8/51/7439c4dd39511368849eb1e53279cd3454b4a4dbace80bab88feeb83c6b5/watchfiles-1.1.1-cp313-cp313-musllinux_1_1_x86_64.whl", hash = "sha256:3a476189be23c3686bc2f4321dd501cb329c0a0469e77b7b534ee10129ae6374", size = 622280, upload-time = "2025-10-14T15:05:11.146Z" },
+    { url = "https://files.pythonhosted.org/packages/95/9c/8ed97d4bba5db6fdcdb2b298d3898f2dd5c20f6b73aee04eabe56c59677e/watchfiles-1.1.1-cp313-cp313-win32.whl", hash = "sha256:bf0a91bfb5574a2f7fc223cf95eeea79abfefa404bf1ea5e339c0c1560ae99a0", size = 272056, upload-time = "2025-10-14T15:05:12.156Z" },
+    { url = "https://files.pythonhosted.org/packages/1f/f3/c14e28429f744a260d8ceae18bf58c1d5fa56b50d006a7a9f80e1882cb0d/watchfiles-1.1.1-cp313-cp313-win_amd64.whl", hash = "sha256:52e06553899e11e8074503c8e716d574adeeb7e68913115c4b3653c53f9bae42", size = 288162, upload-time = "2025-10-14T15:05:13.208Z" },
+    { url = "https://files.pythonhosted.org/packages/dc/61/fe0e56c40d5cd29523e398d31153218718c5786b5e636d9ae8ae79453d27/watchfiles-1.1.1-cp313-cp313-win_arm64.whl", hash = "sha256:ac3cc5759570cd02662b15fbcd9d917f7ecd47efe0d6b40474eafd246f91ea18", size = 277909, upload-time = "2025-10-14T15:05:14.49Z" },
+    { url = "https://files.pythonhosted.org/packages/79/42/e0a7d749626f1e28c7108a99fb9bf524b501bbbeb9b261ceecde644d5a07/watchfiles-1.1.1-cp313-cp313t-macosx_10_12_x86_64.whl", hash = "sha256:563b116874a9a7ce6f96f87cd0b94f7faf92d08d0021e837796f0a14318ef8da", size = 403389, upload-time = "2025-10-14T15:05:15.777Z" },
+    { url = "https://files.pythonhosted.org/packages/15/49/08732f90ce0fbbc13913f9f215c689cfc9ced345fb1bcd8829a50007cc8d/watchfiles-1.1.1-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:3ad9fe1dae4ab4212d8c91e80b832425e24f421703b5a42ef2e4a1e215aff051", size = 389964, upload-time = "2025-10-14T15:05:16.85Z" },
+    { url = "https://files.pythonhosted.org/packages/27/0d/7c315d4bd5f2538910491a0393c56bf70d333d51bc5b34bee8e68e8cea19/watchfiles-1.1.1-cp313-cp313t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:ce70f96a46b894b36eba678f153f052967a0d06d5b5a19b336ab0dbbd029f73e", size = 448114, upload-time = "2025-10-14T15:05:17.876Z" },
+    { url = "https://files.pythonhosted.org/packages/c3/24/9e096de47a4d11bc4df41e9d1e61776393eac4cb6eb11b3e23315b78b2cc/watchfiles-1.1.1-cp313-cp313t-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:cb467c999c2eff23a6417e58d75e5828716f42ed8289fe6b77a7e5a91036ca70", size = 460264, upload-time = "2025-10-14T15:05:18.962Z" },
+    { url = "https://files.pythonhosted.org/packages/cc/0f/e8dea6375f1d3ba5fcb0b3583e2b493e77379834c74fd5a22d66d85d6540/watchfiles-1.1.1-cp313-cp313t-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:836398932192dae4146c8f6f737d74baeac8b70ce14831a239bdb1ca882fc261", size = 487877, upload-time = "2025-10-14T15:05:20.094Z" },
+    { url = "https://files.pythonhosted.org/packages/ac/5b/df24cfc6424a12deb41503b64d42fbea6b8cb357ec62ca84a5a3476f654a/watchfiles-1.1.1-cp313-cp313t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:743185e7372b7bc7c389e1badcc606931a827112fbbd37f14c537320fca08620", size = 595176, upload-time = "2025-10-14T15:05:21.134Z" },
+    { url = "https://files.pythonhosted.org/packages/8f/b5/853b6757f7347de4e9b37e8cc3289283fb983cba1ab4d2d7144694871d9c/watchfiles-1.1.1-cp313-cp313t-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:afaeff7696e0ad9f02cbb8f56365ff4686ab205fcf9c4c5b6fdfaaa16549dd04", size = 473577, upload-time = "2025-10-14T15:05:22.306Z" },
+    { url = "https://files.pythonhosted.org/packages/e1/f7/0a4467be0a56e80447c8529c9fce5b38eab4f513cb3d9bf82e7392a5696b/watchfiles-1.1.1-cp313-cp313t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:3f7eb7da0eb23aa2ba036d4f616d46906013a68caf61b7fdbe42fc8b25132e77", size = 455425, upload-time = "2025-10-14T15:05:23.348Z" },
+    { url = "https://files.pythonhosted.org/packages/8e/e0/82583485ea00137ddf69bc84a2db88bd92ab4a6e3c405e5fb878ead8d0e7/watchfiles-1.1.1-cp313-cp313t-musllinux_1_1_aarch64.whl", hash = "sha256:831a62658609f0e5c64178211c942ace999517f5770fe9436be4c2faeba0c0ef", size = 628826, upload-time = "2025-10-14T15:05:24.398Z" },
+    { url = "https://files.pythonhosted.org/packages/28/9a/a785356fccf9fae84c0cc90570f11702ae9571036fb25932f1242c82191c/watchfiles-1.1.1-cp313-cp313t-musllinux_1_1_x86_64.whl", hash = "sha256:f9a2ae5c91cecc9edd47e041a930490c31c3afb1f5e6d71de3dc671bfaca02bf", size = 622208, upload-time = "2025-10-14T15:05:25.45Z" },
+    { url = "https://files.pythonhosted.org/packages/c3/f4/0872229324ef69b2c3edec35e84bd57a1289e7d3fe74588048ed8947a323/watchfiles-1.1.1-cp314-cp314-macosx_10_12_x86_64.whl", hash = "sha256:d1715143123baeeaeadec0528bb7441103979a1d5f6fd0e1f915383fea7ea6d5", size = 404315, upload-time = "2025-10-14T15:05:26.501Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/22/16d5331eaed1cb107b873f6ae1b69e9ced582fcf0c59a50cd84f403b1c32/watchfiles-1.1.1-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:39574d6370c4579d7f5d0ad940ce5b20db0e4117444e39b6d8f99db5676c52fd", size = 390869, upload-time = "2025-10-14T15:05:27.649Z" },
+    { url = "https://files.pythonhosted.org/packages/b2/7e/5643bfff5acb6539b18483128fdc0ef2cccc94a5b8fbda130c823e8ed636/watchfiles-1.1.1-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:7365b92c2e69ee952902e8f70f3ba6360d0d596d9299d55d7d386df84b6941fb", size = 449919, upload-time = "2025-10-14T15:05:28.701Z" },
+    { url = "https://files.pythonhosted.org/packages/51/2e/c410993ba5025a9f9357c376f48976ef0e1b1aefb73b97a5ae01a5972755/watchfiles-1.1.1-cp314-cp314-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:bfff9740c69c0e4ed32416f013f3c45e2ae42ccedd1167ef2d805c000b6c71a5", size = 460845, upload-time = "2025-10-14T15:05:30.064Z" },
+    { url = "https://files.pythonhosted.org/packages/8e/a4/2df3b404469122e8680f0fcd06079317e48db58a2da2950fb45020947734/watchfiles-1.1.1-cp314-cp314-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:b27cf2eb1dda37b2089e3907d8ea92922b673c0c427886d4edc6b94d8dfe5db3", size = 489027, upload-time = "2025-10-14T15:05:31.064Z" },
+    { url = "https://files.pythonhosted.org/packages/ea/84/4587ba5b1f267167ee715b7f66e6382cca6938e0a4b870adad93e44747e6/watchfiles-1.1.1-cp314-cp314-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:526e86aced14a65a5b0ec50827c745597c782ff46b571dbfe46192ab9e0b3c33", size = 595615, upload-time = "2025-10-14T15:05:32.074Z" },
+    { url = "https://files.pythonhosted.org/packages/6a/0f/c6988c91d06e93cd0bb3d4a808bcf32375ca1904609835c3031799e3ecae/watchfiles-1.1.1-cp314-cp314-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:04e78dd0b6352db95507fd8cb46f39d185cf8c74e4cf1e4fbad1d3df96faf510", size = 474836, upload-time = "2025-10-14T15:05:33.209Z" },
+    { url = "https://files.pythonhosted.org/packages/b4/36/ded8aebea91919485b7bbabbd14f5f359326cb5ec218cd67074d1e426d74/watchfiles-1.1.1-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:5c85794a4cfa094714fb9c08d4a218375b2b95b8ed1666e8677c349906246c05", size = 455099, upload-time = "2025-10-14T15:05:34.189Z" },
+    { url = "https://files.pythonhosted.org/packages/98/e0/8c9bdba88af756a2fce230dd365fab2baf927ba42cd47521ee7498fd5211/watchfiles-1.1.1-cp314-cp314-musllinux_1_1_aarch64.whl", hash = "sha256:74d5012b7630714b66be7b7b7a78855ef7ad58e8650c73afc4c076a1f480a8d6", size = 630626, upload-time = "2025-10-14T15:05:35.216Z" },
+    { url = "https://files.pythonhosted.org/packages/2a/84/a95db05354bf2d19e438520d92a8ca475e578c647f78f53197f5a2f17aaf/watchfiles-1.1.1-cp314-cp314-musllinux_1_1_x86_64.whl", hash = "sha256:8fbe85cb3201c7d380d3d0b90e63d520f15d6afe217165d7f98c9c649654db81", size = 622519, upload-time = "2025-10-14T15:05:36.259Z" },
+    { url = "https://files.pythonhosted.org/packages/1d/ce/d8acdc8de545de995c339be67711e474c77d643555a9bb74a9334252bd55/watchfiles-1.1.1-cp314-cp314-win32.whl", hash = "sha256:3fa0b59c92278b5a7800d3ee7733da9d096d4aabcfabb9a928918bd276ef9b9b", size = 272078, upload-time = "2025-10-14T15:05:37.63Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/c9/a74487f72d0451524be827e8edec251da0cc1fcf111646a511ae752e1a3d/watchfiles-1.1.1-cp314-cp314-win_amd64.whl", hash = "sha256:c2047d0b6cea13b3316bdbafbfa0c4228ae593d995030fda39089d36e64fc03a", size = 287664, upload-time = "2025-10-14T15:05:38.95Z" },
+    { url = "https://files.pythonhosted.org/packages/df/b8/8ac000702cdd496cdce998c6f4ee0ca1f15977bba51bdf07d872ebdfc34c/watchfiles-1.1.1-cp314-cp314-win_arm64.whl", hash = "sha256:842178b126593addc05acf6fce960d28bc5fae7afbaa2c6c1b3a7b9460e5be02", size = 277154, upload-time = "2025-10-14T15:05:39.954Z" },
+    { url = "https://files.pythonhosted.org/packages/47/a8/e3af2184707c29f0f14b1963c0aace6529f9d1b8582d5b99f31bbf42f59e/watchfiles-1.1.1-cp314-cp314t-macosx_10_12_x86_64.whl", hash = "sha256:88863fbbc1a7312972f1c511f202eb30866370ebb8493aef2812b9ff28156a21", size = 403820, upload-time = "2025-10-14T15:05:40.932Z" },
+    { url = "https://files.pythonhosted.org/packages/c0/ec/e47e307c2f4bd75f9f9e8afbe3876679b18e1bcec449beca132a1c5ffb2d/watchfiles-1.1.1-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:55c7475190662e202c08c6c0f4d9e345a29367438cf8e8037f3155e10a88d5a5", size = 390510, upload-time = "2025-10-14T15:05:41.945Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/a0/ad235642118090f66e7b2f18fd5c42082418404a79205cdfca50b6309c13/watchfiles-1.1.1-cp314-cp314t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:3f53fa183d53a1d7a8852277c92b967ae99c2d4dcee2bfacff8868e6e30b15f7", size = 448408, upload-time = "2025-10-14T15:05:43.385Z" },
+    { url = "https://files.pythonhosted.org/packages/df/85/97fa10fd5ff3332ae17e7e40e20784e419e28521549780869f1413742e9d/watchfiles-1.1.1-cp314-cp314t-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:6aae418a8b323732fa89721d86f39ec8f092fc2af67f4217a2b07fd3e93c6101", size = 458968, upload-time = "2025-10-14T15:05:44.404Z" },
+    { url = "https://files.pythonhosted.org/packages/47/c2/9059c2e8966ea5ce678166617a7f75ecba6164375f3b288e50a40dc6d489/watchfiles-1.1.1-cp314-cp314t-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:f096076119da54a6080e8920cbdaac3dbee667eb91dcc5e5b78840b87415bd44", size = 488096, upload-time = "2025-10-14T15:05:45.398Z" },
+    { url = "https://files.pythonhosted.org/packages/94/44/d90a9ec8ac309bc26db808a13e7bfc0e4e78b6fc051078a554e132e80160/watchfiles-1.1.1-cp314-cp314t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:00485f441d183717038ed2e887a7c868154f216877653121068107b227a2f64c", size = 596040, upload-time = "2025-10-14T15:05:46.502Z" },
+    { url = "https://files.pythonhosted.org/packages/95/68/4e3479b20ca305cfc561db3ed207a8a1c745ee32bf24f2026a129d0ddb6e/watchfiles-1.1.1-cp314-cp314t-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:a55f3e9e493158d7bfdb60a1165035f1cf7d320914e7b7ea83fe22c6023b58fc", size = 473847, upload-time = "2025-10-14T15:05:47.484Z" },
+    { url = "https://files.pythonhosted.org/packages/4f/55/2af26693fd15165c4ff7857e38330e1b61ab8c37d15dc79118cdba115b7a/watchfiles-1.1.1-cp314-cp314t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8c91ed27800188c2ae96d16e3149f199d62f86c7af5f5f4d2c61a3ed8cd3666c", size = 455072, upload-time = "2025-10-14T15:05:48.928Z" },
+    { url = "https://files.pythonhosted.org/packages/66/1d/d0d200b10c9311ec25d2273f8aad8c3ef7cc7ea11808022501811208a750/watchfiles-1.1.1-cp314-cp314t-musllinux_1_1_aarch64.whl", hash = "sha256:311ff15a0bae3714ffb603e6ba6dbfba4065ab60865d15a6ec544133bdb21099", size = 629104, upload-time = "2025-10-14T15:05:49.908Z" },
+    { url = "https://files.pythonhosted.org/packages/e3/bd/fa9bb053192491b3867ba07d2343d9f2252e00811567d30ae8d0f78136fe/watchfiles-1.1.1-cp314-cp314t-musllinux_1_1_x86_64.whl", hash = "sha256:a916a2932da8f8ab582f242c065f5c81bed3462849ca79ee357dd9551b0e9b01", size = 622112, upload-time = "2025-10-14T15:05:50.941Z" },
+]
+
+[[package]]
+name = "websockets"
+version = "15.0.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/21/e6/26d09fab466b7ca9c7737474c52be4f76a40301b08362eb2dbc19dcc16c1/websockets-15.0.1.tar.gz", hash = "sha256:82544de02076bafba038ce055ee6412d68da13ab47f0c60cab827346de828dee", size = 177016, upload-time = "2025-03-05T20:03:41.606Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/51/6b/4545a0d843594f5d0771e86463606a3988b5a09ca5123136f8a76580dd63/websockets-15.0.1-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:3e90baa811a5d73f3ca0bcbf32064d663ed81318ab225ee4f427ad4e26e5aff3", size = 175437, upload-time = "2025-03-05T20:02:16.706Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/71/809a0f5f6a06522af902e0f2ea2757f71ead94610010cf570ab5c98e99ed/websockets-15.0.1-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:592f1a9fe869c778694f0aa806ba0374e97648ab57936f092fd9d87f8bc03665", size = 173096, upload-time = "2025-03-05T20:02:18.832Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/69/1a681dd6f02180916f116894181eab8b2e25b31e484c5d0eae637ec01f7c/websockets-15.0.1-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:0701bc3cfcb9164d04a14b149fd74be7347a530ad3bbf15ab2c678a2cd3dd9a2", size = 173332, upload-time = "2025-03-05T20:02:20.187Z" },
+    { url = "https://files.pythonhosted.org/packages/a6/02/0073b3952f5bce97eafbb35757f8d0d54812b6174ed8dd952aa08429bcc3/websockets-15.0.1-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:e8b56bdcdb4505c8078cb6c7157d9811a85790f2f2b3632c7d1462ab5783d215", size = 183152, upload-time = "2025-03-05T20:02:22.286Z" },
+    { url = "https://files.pythonhosted.org/packages/74/45/c205c8480eafd114b428284840da0b1be9ffd0e4f87338dc95dc6ff961a1/websockets-15.0.1-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:0af68c55afbd5f07986df82831c7bff04846928ea8d1fd7f30052638788bc9b5", size = 182096, upload-time = "2025-03-05T20:02:24.368Z" },
+    { url = "https://files.pythonhosted.org/packages/14/8f/aa61f528fba38578ec553c145857a181384c72b98156f858ca5c8e82d9d3/websockets-15.0.1-cp312-cp312-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:64dee438fed052b52e4f98f76c5790513235efaa1ef7f3f2192c392cd7c91b65", size = 182523, upload-time = "2025-03-05T20:02:25.669Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/6d/0267396610add5bc0d0d3e77f546d4cd287200804fe02323797de77dbce9/websockets-15.0.1-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:d5f6b181bb38171a8ad1d6aa58a67a6aa9d4b38d0f8c5f496b9e42561dfc62fe", size = 182790, upload-time = "2025-03-05T20:02:26.99Z" },
+    { url = "https://files.pythonhosted.org/packages/02/05/c68c5adbf679cf610ae2f74a9b871ae84564462955d991178f95a1ddb7dd/websockets-15.0.1-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:5d54b09eba2bada6011aea5375542a157637b91029687eb4fdb2dab11059c1b4", size = 182165, upload-time = "2025-03-05T20:02:30.291Z" },
+    { url = "https://files.pythonhosted.org/packages/29/93/bb672df7b2f5faac89761cb5fa34f5cec45a4026c383a4b5761c6cea5c16/websockets-15.0.1-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:3be571a8b5afed347da347bfcf27ba12b069d9d7f42cb8c7028b5e98bbb12597", size = 182160, upload-time = "2025-03-05T20:02:31.634Z" },
+    { url = "https://files.pythonhosted.org/packages/ff/83/de1f7709376dc3ca9b7eeb4b9a07b4526b14876b6d372a4dc62312bebee0/websockets-15.0.1-cp312-cp312-win32.whl", hash = "sha256:c338ffa0520bdb12fbc527265235639fb76e7bc7faafbb93f6ba80d9c06578a9", size = 176395, upload-time = "2025-03-05T20:02:33.017Z" },
+    { url = "https://files.pythonhosted.org/packages/7d/71/abf2ebc3bbfa40f391ce1428c7168fb20582d0ff57019b69ea20fa698043/websockets-15.0.1-cp312-cp312-win_amd64.whl", hash = "sha256:fcd5cf9e305d7b8338754470cf69cf81f420459dbae8a3b40cee57417f4614a7", size = 176841, upload-time = "2025-03-05T20:02:34.498Z" },
+    { url = "https://files.pythonhosted.org/packages/cb/9f/51f0cf64471a9d2b4d0fc6c534f323b664e7095640c34562f5182e5a7195/websockets-15.0.1-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:ee443ef070bb3b6ed74514f5efaa37a252af57c90eb33b956d35c8e9c10a1931", size = 175440, upload-time = "2025-03-05T20:02:36.695Z" },
+    { url = "https://files.pythonhosted.org/packages/8a/05/aa116ec9943c718905997412c5989f7ed671bc0188ee2ba89520e8765d7b/websockets-15.0.1-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:5a939de6b7b4e18ca683218320fc67ea886038265fd1ed30173f5ce3f8e85675", size = 173098, upload-time = "2025-03-05T20:02:37.985Z" },
+    { url = "https://files.pythonhosted.org/packages/ff/0b/33cef55ff24f2d92924923c99926dcce78e7bd922d649467f0eda8368923/websockets-15.0.1-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:746ee8dba912cd6fc889a8147168991d50ed70447bf18bcda7039f7d2e3d9151", size = 173329, upload-time = "2025-03-05T20:02:39.298Z" },
+    { url = "https://files.pythonhosted.org/packages/31/1d/063b25dcc01faa8fada1469bdf769de3768b7044eac9d41f734fd7b6ad6d/websockets-15.0.1-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:595b6c3969023ecf9041b2936ac3827e4623bfa3ccf007575f04c5a6aa318c22", size = 183111, upload-time = "2025-03-05T20:02:40.595Z" },
+    { url = "https://files.pythonhosted.org/packages/93/53/9a87ee494a51bf63e4ec9241c1ccc4f7c2f45fff85d5bde2ff74fcb68b9e/websockets-15.0.1-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:3c714d2fc58b5ca3e285461a4cc0c9a66bd0e24c5da9911e30158286c9b5be7f", size = 182054, upload-time = "2025-03-05T20:02:41.926Z" },
+    { url = "https://files.pythonhosted.org/packages/ff/b2/83a6ddf56cdcbad4e3d841fcc55d6ba7d19aeb89c50f24dd7e859ec0805f/websockets-15.0.1-cp313-cp313-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:0f3c1e2ab208db911594ae5b4f79addeb3501604a165019dd221c0bdcabe4db8", size = 182496, upload-time = "2025-03-05T20:02:43.304Z" },
+    { url = "https://files.pythonhosted.org/packages/98/41/e7038944ed0abf34c45aa4635ba28136f06052e08fc2168520bb8b25149f/websockets-15.0.1-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:229cf1d3ca6c1804400b0a9790dc66528e08a6a1feec0d5040e8b9eb14422375", size = 182829, upload-time = "2025-03-05T20:02:48.812Z" },
+    { url = "https://files.pythonhosted.org/packages/e0/17/de15b6158680c7623c6ef0db361da965ab25d813ae54fcfeae2e5b9ef910/websockets-15.0.1-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:756c56e867a90fb00177d530dca4b097dd753cde348448a1012ed6c5131f8b7d", size = 182217, upload-time = "2025-03-05T20:02:50.14Z" },
+    { url = "https://files.pythonhosted.org/packages/33/2b/1f168cb6041853eef0362fb9554c3824367c5560cbdaad89ac40f8c2edfc/websockets-15.0.1-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:558d023b3df0bffe50a04e710bc87742de35060580a293c2a984299ed83bc4e4", size = 182195, upload-time = "2025-03-05T20:02:51.561Z" },
+    { url = "https://files.pythonhosted.org/packages/86/eb/20b6cdf273913d0ad05a6a14aed4b9a85591c18a987a3d47f20fa13dcc47/websockets-15.0.1-cp313-cp313-win32.whl", hash = "sha256:ba9e56e8ceeeedb2e080147ba85ffcd5cd0711b89576b83784d8605a7df455fa", size = 176393, upload-time = "2025-03-05T20:02:53.814Z" },
+    { url = "https://files.pythonhosted.org/packages/1b/6c/c65773d6cab416a64d191d6ee8a8b1c68a09970ea6909d16965d26bfed1e/websockets-15.0.1-cp313-cp313-win_amd64.whl", hash = "sha256:e09473f095a819042ecb2ab9465aee615bd9c2028e4ef7d933600a8401c79561", size = 176837, upload-time = "2025-03-05T20:02:55.237Z" },
+    { url = "https://files.pythonhosted.org/packages/fa/a8/5b41e0da817d64113292ab1f8247140aac61cbf6cfd085d6a0fa77f4984f/websockets-15.0.1-py3-none-any.whl", hash = "sha256:f7a866fbc1e97b5c617ee4116daaa09b722101d4a3c170c787450ba409f9736f", size = 169743, upload-time = "2025-03-05T20:03:39.41Z" },
+]

mypy-baseline.txt

@@ -36,11 +36,5 @@ posthog/client.py:0: error: "None" has no attribute "start"  [attr-defined]
 posthog/client.py:0: error: "None" has no attribute "get"  [attr-defined]
 posthog/client.py:0: error: Statement is unreachable  [unreachable]
 posthog/client.py:0: error: Statement is unreachable  [unreachable]
-example.py:0: error: Statement is unreachable  [unreachable]
-posthog/ai/utils.py:0: error: Need type annotation for "output" (hint: "output: list[<type>] = ...")  [var-annotated]
-posthog/ai/utils.py:0: error: Function "builtins.any" is not valid as a type  [valid-type]
-posthog/ai/utils.py:0: note: Perhaps you meant "typing.Any" instead of "any"?
-posthog/ai/utils.py:0: error: Function "builtins.any" is not valid as a type  [valid-type]
-posthog/ai/utils.py:0: note: Perhaps you meant "typing.Any" instead of "any"?
 posthog/client.py:0: error: Name "urlparse" already defined (possibly by an import)  [no-redef]
 posthog/client.py:0: error: Name "parse_qs" already defined (possibly by an import)  [no-redef]

posthog/__init__.py

@@ -10,8 +10,29 @@ from posthog.contexts import (
     tag as inner_tag,
     set_context_session as inner_set_context_session,
     identify_context as inner_identify_context,
+    set_capture_exception_code_variables_context as inner_set_capture_exception_code_variables_context,
+    set_code_variables_mask_patterns_context as inner_set_code_variables_mask_patterns_context,
+    set_code_variables_ignore_patterns_context as inner_set_code_variables_ignore_patterns_context,
+)
+from posthog.exception_utils import (
+    DEFAULT_CODE_VARIABLES_IGNORE_PATTERNS,
+    DEFAULT_CODE_VARIABLES_MASK_PATTERNS,
+)
+from posthog.feature_flags import (
+    InconclusiveMatchError as InconclusiveMatchError,
+    RequiresServerEvaluation as RequiresServerEvaluation,
+)
+from posthog.request import (
+    disable_connection_reuse as disable_connection_reuse,
+    enable_keep_alive as enable_keep_alive,
+    set_socket_options as set_socket_options,
+    SocketOptions as SocketOptions,
+)
+from posthog.types import (
+    FeatureFlag,
+    FlagsAndPayloads,
+    FeatureFlagResult as FeatureFlagResult,
 )
-from posthog.types import FeatureFlag, FlagsAndPayloads, FeatureFlagResult
 from posthog.version import VERSION
 
 __version__ = VERSION
@@ -19,13 +40,14 @@ __version__ = VERSION
 """Context management."""
 
 
-def new_context(fresh=False, capture_exceptions=True):
+def new_context(fresh=False, capture_exceptions=True, client=None):
     """
     Create a new context scope that will be active for the duration of the with block.
 
     Args:
         fresh: Whether to start with a fresh context (default: False)
         capture_exceptions: Whether to capture exceptions raised within the context (default: True)
+        client: Optional Posthog client instance to use for this context (default: None)
 
     Examples:
         ```python
@@ -38,7 +60,9 @@ def new_context(fresh=False, capture_exceptions=True):
     Category:
         Contexts
     """
-    return inner_new_context(fresh=fresh, capture_exceptions=capture_exceptions)
+    return inner_new_context(
+        fresh=fresh, capture_exceptions=capture_exceptions, client=client
+    )
 
 
 def scoped(fresh=False, capture_exceptions=True):
@@ -102,6 +126,27 @@ def identify_context(distinct_id: str):
     return inner_identify_context(distinct_id)
 
 
+def set_capture_exception_code_variables_context(enabled: bool):
+    """
+    Set whether code variables are captured for the current context.
+    """
+    return inner_set_capture_exception_code_variables_context(enabled)
+
+
+def set_code_variables_mask_patterns_context(mask_patterns: list):
+    """
+    Variable names matching these patterns will be masked with *** when capturing code variables.
+    """
+    return inner_set_code_variables_mask_patterns_context(mask_patterns)
+
+
+def set_code_variables_ignore_patterns_context(ignore_patterns: list):
+    """
+    Variable names matching these patterns will be ignored completely when capturing code variables.
+    """
+    return inner_set_code_variables_ignore_patterns_context(ignore_patterns)
+
+
 def tag(name: str, value: Any):
     """
     Add a tag to the current context.
@@ -149,6 +194,10 @@ enable_local_evaluation = True  # type: bool
 
 default_client = None  # type: Optional[Client]
 
+capture_exception_code_variables = False
+code_variables_mask_patterns = DEFAULT_CODE_VARIABLES_MASK_PATTERNS
+code_variables_ignore_patterns = DEFAULT_CODE_VARIABLES_IGNORE_PATTERNS
+
 
 # NOTE - this and following functions take unpacked kwargs because we needed to make
 # it impossible to write `posthog.capture(distinct-id, event-name)` - basically, to enforce
@@ -743,6 +792,9 @@ def setup() -> Client:
             enable_exception_autocapture=enable_exception_autocapture,
             log_captured_exceptions=log_captured_exceptions,
             enable_local_evaluation=enable_local_evaluation,
+            capture_exception_code_variables=capture_exception_code_variables,
+            code_variables_mask_patterns=code_variables_mask_patterns,
+            code_variables_ignore_patterns=code_variables_ignore_patterns,
         )
 
     # always set incase user changes it

posthog/ai/anthropic/__init__.py

@@ -6,6 +6,12 @@ from .anthropic_providers import (
     AsyncAnthropicBedrock,
     AsyncAnthropicVertex,
 )
+from .anthropic_converter import (
+    format_anthropic_response,
+    format_anthropic_input,
+    extract_anthropic_tools,
+    format_anthropic_streaming_content,
+)
 
 __all__ = [
     "Anthropic",
@@ -14,4 +20,8 @@ __all__ = [
     "AsyncAnthropicBedrock",
     "AnthropicVertex",
     "AsyncAnthropicVertex",
+    "format_anthropic_response",
+    "format_anthropic_input",
+    "extract_anthropic_tools",
+    "format_anthropic_streaming_content",
 ]

posthog/ai/anthropic/anthropic.py

@@ -8,14 +8,21 @@ except ImportError:
 
 import time
 import uuid
-from typing import Any, Dict, Optional, cast
+from typing import Any, Dict, List, Optional
 
+from posthog.ai.types import StreamingContentBlock, TokenUsage, ToolInProgress
 from posthog.ai.utils import (
     call_llm_and_track_usage,
-    get_model_params,
-    merge_system_prompt,
-    with_privacy_mode,
+    merge_usage_stats,
 )
+from posthog.ai.anthropic.anthropic_converter import (
+    extract_anthropic_usage_from_event,
+    handle_anthropic_content_block_start,
+    handle_anthropic_text_delta,
+    handle_anthropic_tool_delta,
+    finalize_anthropic_tool_input,
+)
+from posthog.ai.sanitization import sanitize_anthropic
 from posthog.client import Client as PostHogClient
 from posthog import setup
 
@@ -61,6 +68,7 @@ class WrappedMessages(Messages):
             posthog_groups: Optional group analytics properties
             **kwargs: Arguments passed to Anthropic's messages.create
         """
+
         if posthog_trace_id is None:
             posthog_trace_id = str(uuid.uuid4())
 
@@ -118,35 +126,66 @@ class WrappedMessages(Messages):
         **kwargs: Any,
     ):
         start_time = time.time()
-        usage_stats: Dict[str, int] = {"input_tokens": 0, "output_tokens": 0}
-        accumulated_content = []
+        usage_stats: TokenUsage = TokenUsage(input_tokens=0, output_tokens=0)
+        accumulated_content = ""
+        content_blocks: List[StreamingContentBlock] = []
+        tools_in_progress: Dict[str, ToolInProgress] = {}
+        current_text_block: Optional[StreamingContentBlock] = None
         response = super().create(**kwargs)
 
         def generator():
             nonlocal usage_stats
-            nonlocal accumulated_content  # noqa: F824
+            nonlocal accumulated_content
+            nonlocal content_blocks
+            nonlocal tools_in_progress
+            nonlocal current_text_block
+
             try:
                 for event in response:
-                    if hasattr(event, "usage") and event.usage:
-                        usage_stats = {
-                            k: getattr(event.usage, k, 0)
-                            for k in [
-                                "input_tokens",
-                                "output_tokens",
-                                "cache_read_input_tokens",
-                                "cache_creation_input_tokens",
-                            ]
-                        }
+                    # Extract usage stats from event
+                    event_usage = extract_anthropic_usage_from_event(event)
+                    merge_usage_stats(usage_stats, event_usage)
 
-                    if hasattr(event, "content") and event.content:
-                        accumulated_content.append(event.content)
+                    # Handle content block start events
+                    if hasattr(event, "type") and event.type == "content_block_start":
+                        block, tool = handle_anthropic_content_block_start(event)
+
+                        if block:
+                            content_blocks.append(block)
+
+                            if block.get("type") == "text":
+                                current_text_block = block
+                            else:
+                                current_text_block = None
+
+                        if tool:
+                            tool_id = tool["block"].get("id")
+                            if tool_id:
+                                tools_in_progress[tool_id] = tool
+
+                    # Handle text delta events
+                    delta_text = handle_anthropic_text_delta(event, current_text_block)
+
+                    if delta_text:
+                        accumulated_content += delta_text
+
+                    # Handle tool input delta events
+                    handle_anthropic_tool_delta(
+                        event, content_blocks, tools_in_progress
+                    )
+
+                    # Handle content block stop events
+                    if hasattr(event, "type") and event.type == "content_block_stop":
+                        current_text_block = None
+                        finalize_anthropic_tool_input(
+                            event, content_blocks, tools_in_progress
+                        )
 
                     yield event
 
             finally:
                 end_time = time.time()
                 latency = end_time - start_time
-                output = "".join(accumulated_content)
 
                 self._capture_streaming_event(
                     posthog_distinct_id,
@@ -157,7 +196,8 @@ class WrappedMessages(Messages):
                     kwargs,
                     usage_stats,
                     latency,
-                    output,
+                    content_blocks,
+                    accumulated_content,
                 )
 
         return generator()
@@ -170,49 +210,39 @@ class WrappedMessages(Messages):
         posthog_privacy_mode: bool,
         posthog_groups: Optional[Dict[str, Any]],
         kwargs: Dict[str, Any],
-        usage_stats: Dict[str, int],
+        usage_stats: TokenUsage,
         latency: float,
-        output: str,
+        content_blocks: List[StreamingContentBlock],
+        accumulated_content: str,
     ):
-        if posthog_trace_id is None:
-            posthog_trace_id = str(uuid.uuid4())
+        from posthog.ai.types import StreamingEventData
+        from posthog.ai.anthropic.anthropic_converter import (
+            format_anthropic_streaming_input,
+            format_anthropic_streaming_output_complete,
+        )
+        from posthog.ai.utils import capture_streaming_event
 
-        event_properties = {
-            "$ai_provider": "anthropic",
-            "$ai_model": kwargs.get("model"),
-            "$ai_model_parameters": get_model_params(kwargs),
-            "$ai_input": with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                merge_system_prompt(kwargs, "anthropic"),
-            ),
-            "$ai_output_choices": with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                [{"content": output, "role": "assistant"}],
-            ),
-            "$ai_http_status": 200,
-            "$ai_input_tokens": usage_stats.get("input_tokens", 0),
-            "$ai_output_tokens": usage_stats.get("output_tokens", 0),
-            "$ai_cache_read_input_tokens": usage_stats.get(
-                "cache_read_input_tokens", 0
-            ),
-            "$ai_cache_creation_input_tokens": usage_stats.get(
-                "cache_creation_input_tokens", 0
-            ),
-            "$ai_latency": latency,
-            "$ai_trace_id": posthog_trace_id,
-            "$ai_base_url": str(self._client.base_url),
-            **(posthog_properties or {}),
-        }
+        # Prepare standardized event data
+        formatted_input = format_anthropic_streaming_input(kwargs)
+        sanitized_input = sanitize_anthropic(formatted_input)
 
-        if posthog_distinct_id is None:
-            event_properties["$process_person_profile"] = False
+        event_data = StreamingEventData(
+            provider="anthropic",
+            model=kwargs.get("model", "unknown"),
+            base_url=str(self._client.base_url),
+            kwargs=kwargs,
+            formatted_input=sanitized_input,
+            formatted_output=format_anthropic_streaming_output_complete(
+                content_blocks, accumulated_content
+            ),
+            usage_stats=usage_stats,
+            latency=latency,
+            distinct_id=posthog_distinct_id,
+            trace_id=posthog_trace_id,
+            properties=posthog_properties,
+            privacy_mode=posthog_privacy_mode,
+            groups=posthog_groups,
+        )
 
-        if hasattr(self._client._ph_client, "capture"):
-            self._client._ph_client.capture(
-                distinct_id=posthog_distinct_id or posthog_trace_id,
-                event="$ai_generation",
-                properties=event_properties,
-                groups=posthog_groups,
-            )
+        # Use the common capture function
+        capture_streaming_event(self._client._ph_client, event_data)

posthog/ai/anthropic/anthropic_async.py

@@ -8,15 +8,22 @@ except ImportError:
 
 import time
 import uuid
-from typing import Any, Dict, Optional
+from typing import Any, Dict, List, Optional
 
 from posthog import setup
+from posthog.ai.types import StreamingContentBlock, TokenUsage, ToolInProgress
 from posthog.ai.utils import (
     call_llm_and_track_usage_async,
-    get_model_params,
-    merge_system_prompt,
-    with_privacy_mode,
+    merge_usage_stats,
 )
+from posthog.ai.anthropic.anthropic_converter import (
+    extract_anthropic_usage_from_event,
+    handle_anthropic_content_block_start,
+    handle_anthropic_text_delta,
+    handle_anthropic_tool_delta,
+    finalize_anthropic_tool_input,
+)
+from posthog.ai.sanitization import sanitize_anthropic
 from posthog.client import Client as PostHogClient
 
 
@@ -61,6 +68,7 @@ class AsyncWrappedMessages(AsyncMessages):
             posthog_groups: Optional group analytics properties
             **kwargs: Arguments passed to Anthropic's messages.create
         """
+
         if posthog_trace_id is None:
             posthog_trace_id = str(uuid.uuid4())
 
@@ -118,35 +126,66 @@ class AsyncWrappedMessages(AsyncMessages):
         **kwargs: Any,
     ):
         start_time = time.time()
-        usage_stats: Dict[str, int] = {"input_tokens": 0, "output_tokens": 0}
-        accumulated_content = []
+        usage_stats: TokenUsage = TokenUsage(input_tokens=0, output_tokens=0)
+        accumulated_content = ""
+        content_blocks: List[StreamingContentBlock] = []
+        tools_in_progress: Dict[str, ToolInProgress] = {}
+        current_text_block: Optional[StreamingContentBlock] = None
         response = await super().create(**kwargs)
 
         async def generator():
             nonlocal usage_stats
-            nonlocal accumulated_content  # noqa: F824
+            nonlocal accumulated_content
+            nonlocal content_blocks
+            nonlocal tools_in_progress
+            nonlocal current_text_block
+
             try:
                 async for event in response:
-                    if hasattr(event, "usage") and event.usage:
-                        usage_stats = {
-                            k: getattr(event.usage, k, 0)
-                            for k in [
-                                "input_tokens",
-                                "output_tokens",
-                                "cache_read_input_tokens",
-                                "cache_creation_input_tokens",
-                            ]
-                        }
+                    # Extract usage stats from event
+                    event_usage = extract_anthropic_usage_from_event(event)
+                    merge_usage_stats(usage_stats, event_usage)
 
-                    if hasattr(event, "content") and event.content:
-                        accumulated_content.append(event.content)
+                    # Handle content block start events
+                    if hasattr(event, "type") and event.type == "content_block_start":
+                        block, tool = handle_anthropic_content_block_start(event)
+
+                        if block:
+                            content_blocks.append(block)
+
+                            if block.get("type") == "text":
+                                current_text_block = block
+                            else:
+                                current_text_block = None
+
+                        if tool:
+                            tool_id = tool["block"].get("id")
+                            if tool_id:
+                                tools_in_progress[tool_id] = tool
+
+                    # Handle text delta events
+                    delta_text = handle_anthropic_text_delta(event, current_text_block)
+
+                    if delta_text:
+                        accumulated_content += delta_text
+
+                    # Handle tool input delta events
+                    handle_anthropic_tool_delta(
+                        event, content_blocks, tools_in_progress
+                    )
+
+                    # Handle content block stop events
+                    if hasattr(event, "type") and event.type == "content_block_stop":
+                        current_text_block = None
+                        finalize_anthropic_tool_input(
+                            event, content_blocks, tools_in_progress
+                        )
 
                     yield event
 
             finally:
                 end_time = time.time()
                 latency = end_time - start_time
-                output = "".join(accumulated_content)
 
                 await self._capture_streaming_event(
                     posthog_distinct_id,
@@ -157,7 +196,8 @@ class AsyncWrappedMessages(AsyncMessages):
                     kwargs,
                     usage_stats,
                     latency,
-                    output,
+                    content_blocks,
+                    accumulated_content,
                 )
 
         return generator()
@@ -170,49 +210,39 @@ class AsyncWrappedMessages(AsyncMessages):
         posthog_privacy_mode: bool,
         posthog_groups: Optional[Dict[str, Any]],
         kwargs: Dict[str, Any],
-        usage_stats: Dict[str, int],
+        usage_stats: TokenUsage,
         latency: float,
-        output: str,
+        content_blocks: List[StreamingContentBlock],
+        accumulated_content: str,
     ):
-        if posthog_trace_id is None:
-            posthog_trace_id = str(uuid.uuid4())
+        from posthog.ai.types import StreamingEventData
+        from posthog.ai.anthropic.anthropic_converter import (
+            format_anthropic_streaming_input,
+            format_anthropic_streaming_output_complete,
+        )
+        from posthog.ai.utils import capture_streaming_event
 
-        event_properties = {
-            "$ai_provider": "anthropic",
-            "$ai_model": kwargs.get("model"),
-            "$ai_model_parameters": get_model_params(kwargs),
-            "$ai_input": with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                merge_system_prompt(kwargs, "anthropic"),
-            ),
-            "$ai_output_choices": with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                [{"content": output, "role": "assistant"}],
-            ),
-            "$ai_http_status": 200,
-            "$ai_input_tokens": usage_stats.get("input_tokens", 0),
-            "$ai_output_tokens": usage_stats.get("output_tokens", 0),
-            "$ai_cache_read_input_tokens": usage_stats.get(
-                "cache_read_input_tokens", 0
-            ),
-            "$ai_cache_creation_input_tokens": usage_stats.get(
-                "cache_creation_input_tokens", 0
-            ),
-            "$ai_latency": latency,
-            "$ai_trace_id": posthog_trace_id,
-            "$ai_base_url": str(self._client.base_url),
-            **(posthog_properties or {}),
-        }
+        # Prepare standardized event data
+        formatted_input = format_anthropic_streaming_input(kwargs)
+        sanitized_input = sanitize_anthropic(formatted_input)
 
-        if posthog_distinct_id is None:
-            event_properties["$process_person_profile"] = False
+        event_data = StreamingEventData(
+            provider="anthropic",
+            model=kwargs.get("model", "unknown"),
+            base_url=str(self._client.base_url),
+            kwargs=kwargs,
+            formatted_input=sanitized_input,
+            formatted_output=format_anthropic_streaming_output_complete(
+                content_blocks, accumulated_content
+            ),
+            usage_stats=usage_stats,
+            latency=latency,
+            distinct_id=posthog_distinct_id,
+            trace_id=posthog_trace_id,
+            properties=posthog_properties,
+            privacy_mode=posthog_privacy_mode,
+            groups=posthog_groups,
+        )
 
-        if hasattr(self._client._ph_client, "capture"):
-            self._client._ph_client.capture(
-                distinct_id=posthog_distinct_id or posthog_trace_id,
-                event="$ai_generation",
-                properties=event_properties,
-                groups=posthog_groups,
-            )
+        # Use the common capture function
+        capture_streaming_event(self._client._ph_client, event_data)

posthog/ai/anthropic/anthropic_converter.py

@@ -0,0 +1,443 @@
+"""
+Anthropic-specific conversion utilities.
+
+This module handles the conversion of Anthropic API responses and inputs
+into standardized formats for PostHog tracking.
+"""
+
+import json
+from typing import Any, Dict, List, Optional, Tuple
+
+from posthog.ai.types import (
+    FormattedContentItem,
+    FormattedFunctionCall,
+    FormattedMessage,
+    FormattedTextContent,
+    StreamingContentBlock,
+    TokenUsage,
+    ToolInProgress,
+)
+
+
+def format_anthropic_response(response: Any) -> List[FormattedMessage]:
+    """
+    Format an Anthropic response into standardized message format.
+
+    Args:
+        response: The response object from Anthropic API
+
+    Returns:
+        List of formatted messages with role and content
+    """
+
+    output: List[FormattedMessage] = []
+
+    if response is None:
+        return output
+
+    content: List[FormattedContentItem] = []
+
+    # Process content blocks from the response
+    if hasattr(response, "content"):
+        for choice in response.content:
+            if (
+                hasattr(choice, "type")
+                and choice.type == "text"
+                and hasattr(choice, "text")
+                and choice.text
+            ):
+                text_content: FormattedTextContent = {
+                    "type": "text",
+                    "text": choice.text,
+                }
+                content.append(text_content)
+
+            elif (
+                hasattr(choice, "type")
+                and choice.type == "tool_use"
+                and hasattr(choice, "name")
+                and hasattr(choice, "id")
+            ):
+                function_call: FormattedFunctionCall = {
+                    "type": "function",
+                    "id": choice.id,
+                    "function": {
+                        "name": choice.name,
+                        "arguments": getattr(choice, "input", {}),
+                    },
+                }
+                content.append(function_call)
+
+    if content:
+        message: FormattedMessage = {
+            "role": "assistant",
+            "content": content,
+        }
+        output.append(message)
+
+    return output
+
+
+def format_anthropic_input(
+    messages: List[Dict[str, Any]], system: Optional[str] = None
+) -> List[FormattedMessage]:
+    """
+    Format Anthropic input messages with optional system prompt.
+
+    Args:
+        messages: List of message dictionaries
+        system: Optional system prompt to prepend
+
+    Returns:
+        List of formatted messages
+    """
+
+    formatted_messages: List[FormattedMessage] = []
+
+    # Add system message if provided
+    if system is not None:
+        formatted_messages.append({"role": "system", "content": system})
+
+    # Add user messages
+    if messages:
+        for msg in messages:
+            # Messages are already in the correct format, just ensure type safety
+            formatted_msg: FormattedMessage = {
+                "role": msg.get("role", "user"),
+                "content": msg.get("content", ""),
+            }
+            formatted_messages.append(formatted_msg)
+
+    return formatted_messages
+
+
+def extract_anthropic_tools(kwargs: Dict[str, Any]) -> Optional[Any]:
+    """
+    Extract tool definitions from Anthropic API kwargs.
+
+    Args:
+        kwargs: Keyword arguments passed to Anthropic API
+
+    Returns:
+        Tool definitions if present, None otherwise
+    """
+
+    return kwargs.get("tools", None)
+
+
+def format_anthropic_streaming_content(
+    content_blocks: List[StreamingContentBlock],
+) -> List[FormattedContentItem]:
+    """
+    Format content blocks from Anthropic streaming response.
+
+    Used by streaming handlers to format accumulated content blocks.
+
+    Args:
+        content_blocks: List of content block dictionaries from streaming
+
+    Returns:
+        List of formatted content items
+    """
+
+    formatted: List[FormattedContentItem] = []
+
+    for block in content_blocks:
+        if block.get("type") == "text":
+            formatted.append(
+                {
+                    "type": "text",
+                    "text": block.get("text") or "",
+                }
+            )
+
+        elif block.get("type") == "function":
+            formatted.append(
+                {
+                    "type": "function",
+                    "id": block.get("id"),
+                    "function": block.get("function") or {},
+                }
+            )
+
+    return formatted
+
+
+def extract_anthropic_web_search_count(response: Any) -> int:
+    """
+    Extract web search count from Anthropic response.
+
+    Anthropic provides exact web search counts via usage.server_tool_use.web_search_requests.
+
+    Args:
+        response: The response from Anthropic API
+
+    Returns:
+        Number of web search requests (0 if none)
+    """
+    if not hasattr(response, "usage"):
+        return 0
+
+    if not hasattr(response.usage, "server_tool_use"):
+        return 0
+
+    server_tool_use = response.usage.server_tool_use
+
+    if hasattr(server_tool_use, "web_search_requests"):
+        return max(0, int(getattr(server_tool_use, "web_search_requests", 0)))
+
+    return 0
+
+
+def extract_anthropic_usage_from_response(response: Any) -> TokenUsage:
+    """
+    Extract usage from a full Anthropic response (non-streaming).
+
+    Args:
+        response: The complete response from Anthropic API
+
+    Returns:
+        TokenUsage with standardized usage
+    """
+    if not hasattr(response, "usage"):
+        return TokenUsage(input_tokens=0, output_tokens=0)
+
+    result = TokenUsage(
+        input_tokens=getattr(response.usage, "input_tokens", 0),
+        output_tokens=getattr(response.usage, "output_tokens", 0),
+    )
+
+    if hasattr(response.usage, "cache_read_input_tokens"):
+        cache_read = response.usage.cache_read_input_tokens
+        if cache_read and cache_read > 0:
+            result["cache_read_input_tokens"] = cache_read
+
+    if hasattr(response.usage, "cache_creation_input_tokens"):
+        cache_creation = response.usage.cache_creation_input_tokens
+        if cache_creation and cache_creation > 0:
+            result["cache_creation_input_tokens"] = cache_creation
+
+    web_search_count = extract_anthropic_web_search_count(response)
+    if web_search_count > 0:
+        result["web_search_count"] = web_search_count
+
+    return result
+
+
+def extract_anthropic_usage_from_event(event: Any) -> TokenUsage:
+    """
+    Extract usage statistics from an Anthropic streaming event.
+
+    Args:
+        event: Streaming event from Anthropic API
+
+    Returns:
+        Dictionary of usage statistics
+    """
+
+    usage: TokenUsage = TokenUsage()
+
+    # Handle usage stats from message_start event
+    if hasattr(event, "type") and event.type == "message_start":
+        if hasattr(event, "message") and hasattr(event.message, "usage"):
+            usage["input_tokens"] = getattr(event.message.usage, "input_tokens", 0)
+            usage["cache_creation_input_tokens"] = getattr(
+                event.message.usage, "cache_creation_input_tokens", 0
+            )
+            usage["cache_read_input_tokens"] = getattr(
+                event.message.usage, "cache_read_input_tokens", 0
+            )
+
+    # Handle usage stats from message_delta event
+    if hasattr(event, "usage") and event.usage:
+        usage["output_tokens"] = getattr(event.usage, "output_tokens", 0)
+
+        # Extract web search count from usage
+        if hasattr(event.usage, "server_tool_use"):
+            server_tool_use = event.usage.server_tool_use
+            if hasattr(server_tool_use, "web_search_requests"):
+                web_search_count = int(
+                    getattr(server_tool_use, "web_search_requests", 0)
+                )
+                if web_search_count > 0:
+                    usage["web_search_count"] = web_search_count
+
+    return usage
+
+
+def handle_anthropic_content_block_start(
+    event: Any,
+) -> Tuple[Optional[StreamingContentBlock], Optional[ToolInProgress]]:
+    """
+    Handle content block start event from Anthropic streaming.
+
+    Args:
+        event: Content block start event
+
+    Returns:
+        Tuple of (content_block, tool_in_progress)
+    """
+
+    if not (hasattr(event, "type") and event.type == "content_block_start"):
+        return None, None
+
+    if not hasattr(event, "content_block"):
+        return None, None
+
+    block = event.content_block
+
+    if not hasattr(block, "type"):
+        return None, None
+
+    if block.type == "text":
+        content_block: StreamingContentBlock = {"type": "text", "text": ""}
+        return content_block, None
+
+    elif block.type == "tool_use":
+        tool_block: StreamingContentBlock = {
+            "type": "function",
+            "id": getattr(block, "id", ""),
+            "function": {"name": getattr(block, "name", ""), "arguments": {}},
+        }
+        tool_in_progress: ToolInProgress = {"block": tool_block, "input_string": ""}
+        return tool_block, tool_in_progress
+
+    return None, None
+
+
+def handle_anthropic_text_delta(
+    event: Any, current_block: Optional[StreamingContentBlock]
+) -> Optional[str]:
+    """
+    Handle text delta event from Anthropic streaming.
+
+    Args:
+        event: Delta event
+        current_block: Current text block being accumulated
+
+    Returns:
+        Text delta if present
+    """
+
+    if hasattr(event, "delta") and hasattr(event.delta, "text"):
+        delta_text = event.delta.text or ""
+
+        if current_block is not None and current_block.get("type") == "text":
+            text_val = current_block.get("text")
+            if text_val is not None:
+                current_block["text"] = text_val + delta_text
+            else:
+                current_block["text"] = delta_text
+
+        return delta_text
+
+    return None
+
+
+def handle_anthropic_tool_delta(
+    event: Any,
+    content_blocks: List[StreamingContentBlock],
+    tools_in_progress: Dict[str, ToolInProgress],
+) -> None:
+    """
+    Handle tool input delta event from Anthropic streaming.
+
+    Args:
+        event: Tool delta event
+        content_blocks: List of content blocks
+        tools_in_progress: Dictionary tracking tools being accumulated
+    """
+
+    if not (hasattr(event, "type") and event.type == "content_block_delta"):
+        return
+
+    if not (
+        hasattr(event, "delta")
+        and hasattr(event.delta, "type")
+        and event.delta.type == "input_json_delta"
+    ):
+        return
+
+    if hasattr(event, "index") and event.index < len(content_blocks):
+        block = content_blocks[event.index]
+
+        if block.get("type") == "function" and block.get("id") in tools_in_progress:
+            tool = tools_in_progress[block["id"]]
+            partial_json = getattr(event.delta, "partial_json", "")
+            tool["input_string"] += partial_json
+
+
+def finalize_anthropic_tool_input(
+    event: Any,
+    content_blocks: List[StreamingContentBlock],
+    tools_in_progress: Dict[str, ToolInProgress],
+) -> None:
+    """
+    Finalize tool input when content block stops.
+
+    Args:
+        event: Content block stop event
+        content_blocks: List of content blocks
+        tools_in_progress: Dictionary tracking tools being accumulated
+    """
+
+    if not (hasattr(event, "type") and event.type == "content_block_stop"):
+        return
+
+    if hasattr(event, "index") and event.index < len(content_blocks):
+        block = content_blocks[event.index]
+
+        if block.get("type") == "function" and block.get("id") in tools_in_progress:
+            tool = tools_in_progress[block["id"]]
+
+            try:
+                block["function"]["arguments"] = json.loads(tool["input_string"])
+            except (json.JSONDecodeError, Exception):
+                # Keep empty dict if parsing fails
+                pass
+
+            del tools_in_progress[block["id"]]
+
+
+def format_anthropic_streaming_input(kwargs: Dict[str, Any]) -> Any:
+    """
+    Format Anthropic streaming input using system prompt merging.
+
+    Args:
+        kwargs: Keyword arguments passed to Anthropic API
+
+    Returns:
+        Formatted input ready for PostHog tracking
+    """
+    from posthog.ai.utils import merge_system_prompt
+
+    return merge_system_prompt(kwargs, "anthropic")
+
+
+def format_anthropic_streaming_output_complete(
+    content_blocks: List[StreamingContentBlock], accumulated_content: str
+) -> List[FormattedMessage]:
+    """
+    Format complete Anthropic streaming output.
+
+    Combines existing logic for formatting content blocks with fallback to accumulated content.
+
+    Args:
+        content_blocks: List of content blocks accumulated during streaming
+        accumulated_content: Raw accumulated text content as fallback
+
+    Returns:
+        Formatted messages ready for PostHog tracking
+    """
+    formatted_content = format_anthropic_streaming_content(content_blocks)
+
+    if formatted_content:
+        return [{"role": "assistant", "content": formatted_content}]
+    else:
+        # Fallback to accumulated content if no blocks
+        return [
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": accumulated_content}],
+            }
+        ]

posthog/ai/gemini/__init__.py

@@ -1,11 +1,25 @@
 from .gemini import Client
+from .gemini_async import AsyncClient
+from .gemini_converter import (
+    format_gemini_input,
+    format_gemini_response,
+    extract_gemini_tools,
+)
 
 
 # Create a genai-like module for perfect drop-in replacement
 class _GenAI:
     Client = Client
+    AsyncClient = AsyncClient
 
 
 genai = _GenAI()
 
-__all__ = ["Client", "genai"]
+__all__ = [
+    "Client",
+    "AsyncClient",
+    "genai",
+    "format_gemini_input",
+    "format_gemini_response",
+    "extract_gemini_tools",
+]

posthog/ai/gemini/gemini.py

@@ -3,6 +3,9 @@ import time
 import uuid
 from typing import Any, Dict, Optional
 
+from posthog.ai.types import TokenUsage, StreamingEventData
+from posthog.ai.utils import merge_system_prompt
+
 try:
     from google import genai
 except ImportError:
@@ -13,9 +16,15 @@ except ImportError:
 from posthog import setup
 from posthog.ai.utils import (
     call_llm_and_track_usage,
-    get_model_params,
-    with_privacy_mode,
+    capture_streaming_event,
+    merge_usage_stats,
 )
+from posthog.ai.gemini.gemini_converter import (
+    extract_gemini_usage_from_chunk,
+    extract_gemini_content_from_chunk,
+    format_gemini_streaming_output,
+)
+from posthog.ai.sanitization import sanitize_gemini
 from posthog.client import Client as PostHogClient
 
 
@@ -42,6 +51,12 @@ class Client:
     def __init__(
         self,
         api_key: Optional[str] = None,
+        vertexai: Optional[bool] = None,
+        credentials: Optional[Any] = None,
+        project: Optional[str] = None,
+        location: Optional[str] = None,
+        debug_config: Optional[Any] = None,
+        http_options: Optional[Any] = None,
         posthog_client: Optional[PostHogClient] = None,
         posthog_distinct_id: Optional[str] = None,
         posthog_properties: Optional[Dict[str, Any]] = None,
@@ -51,7 +66,13 @@ class Client:
     ):
         """
         Args:
-            api_key: Google AI API key. If not provided, will use GOOGLE_API_KEY or API_KEY environment variable
+            api_key: Google AI API key. If not provided, will use GOOGLE_API_KEY or API_KEY environment variable (not required for Vertex AI)
+            vertexai: Whether to use Vertex AI authentication
+            credentials: Vertex AI credentials object
+            project: GCP project ID for Vertex AI
+            location: GCP location for Vertex AI
+            debug_config: Debug configuration for the client
+            http_options: HTTP options for the client
             posthog_client: PostHog client for tracking usage
             posthog_distinct_id: Default distinct ID for all calls (can be overridden per call)
             posthog_properties: Default properties for all calls (can be overridden per call)
@@ -59,6 +80,7 @@ class Client:
             posthog_groups: Default groups for all calls (can be overridden per call)
             **kwargs: Additional arguments (for future compatibility)
         """
+
         self._ph_client = posthog_client or setup()
 
         if self._ph_client is None:
@@ -66,6 +88,12 @@ class Client:
 
         self.models = Models(
             api_key=api_key,
+            vertexai=vertexai,
+            credentials=credentials,
+            project=project,
+            location=location,
+            debug_config=debug_config,
+            http_options=http_options,
             posthog_client=self._ph_client,
             posthog_distinct_id=posthog_distinct_id,
             posthog_properties=posthog_properties,
@@ -85,6 +113,12 @@ class Models:
     def __init__(
         self,
         api_key: Optional[str] = None,
+        vertexai: Optional[bool] = None,
+        credentials: Optional[Any] = None,
+        project: Optional[str] = None,
+        location: Optional[str] = None,
+        debug_config: Optional[Any] = None,
+        http_options: Optional[Any] = None,
         posthog_client: Optional[PostHogClient] = None,
         posthog_distinct_id: Optional[str] = None,
         posthog_properties: Optional[Dict[str, Any]] = None,
@@ -94,7 +128,13 @@ class Models:
     ):
         """
         Args:
-            api_key: Google AI API key. If not provided, will use GOOGLE_API_KEY or API_KEY environment variable
+            api_key: Google AI API key. If not provided, will use GOOGLE_API_KEY or API_KEY environment variable (not required for Vertex AI)
+            vertexai: Whether to use Vertex AI authentication
+            credentials: Vertex AI credentials object
+            project: GCP project ID for Vertex AI
+            location: GCP location for Vertex AI
+            debug_config: Debug configuration for the client
+            http_options: HTTP options for the client
             posthog_client: PostHog client for tracking usage
             posthog_distinct_id: Default distinct ID for all calls
             posthog_properties: Default properties for all calls
@@ -102,6 +142,7 @@ class Models:
             posthog_groups: Default groups for all calls
             **kwargs: Additional arguments (for future compatibility)
         """
+
         self._ph_client = posthog_client or setup()
 
         if self._ph_client is None:
@@ -113,16 +154,46 @@ class Models:
         self._default_privacy_mode = posthog_privacy_mode
         self._default_groups = posthog_groups
 
-        # Handle API key - try parameter first, then environment variables
-        if api_key is None:
-            api_key = os.environ.get("GOOGLE_API_KEY") or os.environ.get("API_KEY")
+        # Build genai.Client arguments
+        client_args: Dict[str, Any] = {}
 
-        if api_key is None:
-            raise ValueError(
-                "API key must be provided either as parameter or via GOOGLE_API_KEY/API_KEY environment variable"
-            )
+        # Add Vertex AI parameters if provided
+        if vertexai is not None:
+            client_args["vertexai"] = vertexai
 
-        self._client = genai.Client(api_key=api_key)
+        if credentials is not None:
+            client_args["credentials"] = credentials
+
+        if project is not None:
+            client_args["project"] = project
+
+        if location is not None:
+            client_args["location"] = location
+
+        if debug_config is not None:
+            client_args["debug_config"] = debug_config
+
+        if http_options is not None:
+            client_args["http_options"] = http_options
+
+        # Handle API key authentication
+        if vertexai:
+            # For Vertex AI, api_key is optional
+            if api_key is not None:
+                client_args["api_key"] = api_key
+        else:
+            # For non-Vertex AI mode, api_key is required (backwards compatibility)
+            if api_key is None:
+                api_key = os.environ.get("GOOGLE_API_KEY") or os.environ.get("API_KEY")
+
+            if api_key is None:
+                raise ValueError(
+                    "API key must be provided either as parameter or via GOOGLE_API_KEY/API_KEY environment variable"
+                )
+
+            client_args["api_key"] = api_key
+
+        self._client = genai.Client(**client_args)
         self._base_url = "https://generativelanguage.googleapis.com"
 
     def _merge_posthog_params(
@@ -134,6 +205,7 @@ class Models:
         call_groups: Optional[Dict[str, Any]],
     ):
         """Merge call-level PostHog parameters with client defaults."""
+
         # Use call-level values if provided, otherwise fall back to defaults
         distinct_id = (
             call_distinct_id
@@ -149,6 +221,7 @@ class Models:
 
         # Merge properties: default properties + call properties (call properties override)
         properties = dict(self._default_properties)
+
         if call_properties:
             properties.update(call_properties)
 
@@ -184,6 +257,7 @@ class Models:
             posthog_groups: Group analytics properties (overrides client default)
             **kwargs: Arguments passed to Gemini's generate_content
         """
+
         # Merge PostHog parameters
         distinct_id, trace_id, properties, privacy_mode, groups = (
             self._merge_posthog_params(
@@ -222,7 +296,7 @@ class Models:
         **kwargs: Any,
     ):
         start_time = time.time()
-        usage_stats: Dict[str, int] = {"input_tokens": 0, "output_tokens": 0}
+        usage_stats: TokenUsage = TokenUsage(input_tokens=0, output_tokens=0)
         accumulated_content = []
 
         kwargs_without_stream = {"model": model, "contents": contents, **kwargs}
@@ -230,28 +304,27 @@ class Models:
 
         def generator():
             nonlocal usage_stats
-            nonlocal accumulated_content  # noqa: F824
+            nonlocal accumulated_content
             try:
                 for chunk in response:
-                    if hasattr(chunk, "usage_metadata") and chunk.usage_metadata:
-                        usage_stats = {
-                            "input_tokens": getattr(
-                                chunk.usage_metadata, "prompt_token_count", 0
-                            ),
-                            "output_tokens": getattr(
-                                chunk.usage_metadata, "candidates_token_count", 0
-                            ),
-                        }
+                    # Extract usage stats from chunk
+                    chunk_usage = extract_gemini_usage_from_chunk(chunk)
 
-                    if hasattr(chunk, "text") and chunk.text:
-                        accumulated_content.append(chunk.text)
+                    if chunk_usage:
+                        # Gemini reports cumulative totals, not incremental values
+                        merge_usage_stats(usage_stats, chunk_usage, mode="cumulative")
+
+                    # Extract content from chunk (now returns content blocks)
+                    content_block = extract_gemini_content_from_chunk(chunk)
+
+                    if content_block is not None:
+                        accumulated_content.append(content_block)
 
                     yield chunk
 
             finally:
                 end_time = time.time()
                 latency = end_time - start_time
-                output = "".join(accumulated_content)
 
                 self._capture_streaming_event(
                     model,
@@ -264,7 +337,7 @@ class Models:
                     kwargs,
                     usage_stats,
                     latency,
-                    output,
+                    accumulated_content,
                 )
 
         return generator()
@@ -279,63 +352,39 @@ class Models:
         privacy_mode: bool,
         groups: Optional[Dict[str, Any]],
         kwargs: Dict[str, Any],
-        usage_stats: Dict[str, int],
+        usage_stats: TokenUsage,
         latency: float,
-        output: str,
+        output: Any,
     ):
-        if trace_id is None:
-            trace_id = str(uuid.uuid4())
+        # Prepare standardized event data
+        formatted_input = self._format_input(contents, **kwargs)
+        sanitized_input = sanitize_gemini(formatted_input)
 
-        event_properties = {
-            "$ai_provider": "gemini",
-            "$ai_model": model,
-            "$ai_model_parameters": get_model_params(kwargs),
-            "$ai_input": with_privacy_mode(
-                self._ph_client,
-                privacy_mode,
-                self._format_input(contents),
-            ),
-            "$ai_output_choices": with_privacy_mode(
-                self._ph_client,
-                privacy_mode,
-                [{"content": output, "role": "assistant"}],
-            ),
-            "$ai_http_status": 200,
-            "$ai_input_tokens": usage_stats.get("input_tokens", 0),
-            "$ai_output_tokens": usage_stats.get("output_tokens", 0),
-            "$ai_latency": latency,
-            "$ai_trace_id": trace_id,
-            "$ai_base_url": self._base_url,
-            **(properties or {}),
-        }
+        event_data = StreamingEventData(
+            provider="gemini",
+            model=model,
+            base_url=self._base_url,
+            kwargs=kwargs,
+            formatted_input=sanitized_input,
+            formatted_output=format_gemini_streaming_output(output),
+            usage_stats=usage_stats,
+            latency=latency,
+            distinct_id=distinct_id,
+            trace_id=trace_id,
+            properties=properties,
+            privacy_mode=privacy_mode,
+            groups=groups,
+        )
 
-        if distinct_id is None:
-            event_properties["$process_person_profile"] = False
+        # Use the common capture function
+        capture_streaming_event(self._ph_client, event_data)
 
-        if hasattr(self._ph_client, "capture"):
-            self._ph_client.capture(
-                distinct_id=distinct_id,
-                event="$ai_generation",
-                properties=event_properties,
-                groups=groups,
-            )
-
-    def _format_input(self, contents):
+    def _format_input(self, contents, **kwargs):
         """Format input contents for PostHog tracking"""
-        if isinstance(contents, str):
-            return [{"role": "user", "content": contents}]
-        elif isinstance(contents, list):
-            formatted = []
-            for item in contents:
-                if isinstance(item, str):
-                    formatted.append({"role": "user", "content": item})
-                elif hasattr(item, "text"):
-                    formatted.append({"role": "user", "content": item.text})
-                else:
-                    formatted.append({"role": "user", "content": str(item)})
-            return formatted
-        else:
-            return [{"role": "user", "content": str(contents)}]
+
+        # Create kwargs dict with contents for merge_system_prompt
+        input_kwargs = {"contents": contents, **kwargs}
+        return merge_system_prompt(input_kwargs, "gemini")
 
     def generate_content_stream(
         self,

posthog/ai/gemini/gemini_async.py

@@ -0,0 +1,423 @@
+import os
+import time
+import uuid
+from typing import Any, Dict, Optional
+
+from posthog.ai.types import TokenUsage, StreamingEventData
+from posthog.ai.utils import merge_system_prompt
+
+try:
+    from google import genai
+except ImportError:
+    raise ModuleNotFoundError(
+        "Please install the Google Gemini SDK to use this feature: 'pip install google-genai'"
+    )
+
+from posthog import setup
+from posthog.ai.utils import (
+    call_llm_and_track_usage_async,
+    capture_streaming_event,
+    merge_usage_stats,
+)
+from posthog.ai.gemini.gemini_converter import (
+    extract_gemini_usage_from_chunk,
+    extract_gemini_content_from_chunk,
+    format_gemini_streaming_output,
+)
+from posthog.ai.sanitization import sanitize_gemini
+from posthog.client import Client as PostHogClient
+
+
+class AsyncClient:
+    """
+    An async drop-in replacement for genai.Client that automatically sends LLM usage events to PostHog.
+
+    Usage:
+        client = AsyncClient(
+            api_key="your_api_key",
+            posthog_client=posthog_client,
+            posthog_distinct_id="default_user",  # Optional defaults
+            posthog_properties={"team": "ai"}    # Optional defaults
+        )
+        response = await client.models.generate_content(
+            model="gemini-2.0-flash",
+            contents=["Hello world"],
+            posthog_distinct_id="specific_user"  # Override default
+        )
+    """
+
+    _ph_client: PostHogClient
+
+    def __init__(
+        self,
+        api_key: Optional[str] = None,
+        vertexai: Optional[bool] = None,
+        credentials: Optional[Any] = None,
+        project: Optional[str] = None,
+        location: Optional[str] = None,
+        debug_config: Optional[Any] = None,
+        http_options: Optional[Any] = None,
+        posthog_client: Optional[PostHogClient] = None,
+        posthog_distinct_id: Optional[str] = None,
+        posthog_properties: Optional[Dict[str, Any]] = None,
+        posthog_privacy_mode: bool = False,
+        posthog_groups: Optional[Dict[str, Any]] = None,
+        **kwargs,
+    ):
+        """
+        Args:
+            api_key: Google AI API key. If not provided, will use GOOGLE_API_KEY or API_KEY environment variable (not required for Vertex AI)
+            vertexai: Whether to use Vertex AI authentication
+            credentials: Vertex AI credentials object
+            project: GCP project ID for Vertex AI
+            location: GCP location for Vertex AI
+            debug_config: Debug configuration for the client
+            http_options: HTTP options for the client
+            posthog_client: PostHog client for tracking usage
+            posthog_distinct_id: Default distinct ID for all calls (can be overridden per call)
+            posthog_properties: Default properties for all calls (can be overridden per call)
+            posthog_privacy_mode: Default privacy mode for all calls (can be overridden per call)
+            posthog_groups: Default groups for all calls (can be overridden per call)
+            **kwargs: Additional arguments (for future compatibility)
+        """
+
+        self._ph_client = posthog_client or setup()
+
+        if self._ph_client is None:
+            raise ValueError("posthog_client is required for PostHog tracking")
+
+        self.models = AsyncModels(
+            api_key=api_key,
+            vertexai=vertexai,
+            credentials=credentials,
+            project=project,
+            location=location,
+            debug_config=debug_config,
+            http_options=http_options,
+            posthog_client=self._ph_client,
+            posthog_distinct_id=posthog_distinct_id,
+            posthog_properties=posthog_properties,
+            posthog_privacy_mode=posthog_privacy_mode,
+            posthog_groups=posthog_groups,
+            **kwargs,
+        )
+
+
+class AsyncModels:
+    """
+    Async Models interface that mimics genai.Client().aio.models with PostHog tracking.
+    """
+
+    _ph_client: PostHogClient  # Not None after __init__ validation
+
+    def __init__(
+        self,
+        api_key: Optional[str] = None,
+        vertexai: Optional[bool] = None,
+        credentials: Optional[Any] = None,
+        project: Optional[str] = None,
+        location: Optional[str] = None,
+        debug_config: Optional[Any] = None,
+        http_options: Optional[Any] = None,
+        posthog_client: Optional[PostHogClient] = None,
+        posthog_distinct_id: Optional[str] = None,
+        posthog_properties: Optional[Dict[str, Any]] = None,
+        posthog_privacy_mode: bool = False,
+        posthog_groups: Optional[Dict[str, Any]] = None,
+        **kwargs,
+    ):
+        """
+        Args:
+            api_key: Google AI API key. If not provided, will use GOOGLE_API_KEY or API_KEY environment variable (not required for Vertex AI)
+            vertexai: Whether to use Vertex AI authentication
+            credentials: Vertex AI credentials object
+            project: GCP project ID for Vertex AI
+            location: GCP location for Vertex AI
+            debug_config: Debug configuration for the client
+            http_options: HTTP options for the client
+            posthog_client: PostHog client for tracking usage
+            posthog_distinct_id: Default distinct ID for all calls
+            posthog_properties: Default properties for all calls
+            posthog_privacy_mode: Default privacy mode for all calls
+            posthog_groups: Default groups for all calls
+            **kwargs: Additional arguments (for future compatibility)
+        """
+
+        self._ph_client = posthog_client or setup()
+
+        if self._ph_client is None:
+            raise ValueError("posthog_client is required for PostHog tracking")
+
+        # Store default PostHog settings
+        self._default_distinct_id = posthog_distinct_id
+        self._default_properties = posthog_properties or {}
+        self._default_privacy_mode = posthog_privacy_mode
+        self._default_groups = posthog_groups
+
+        # Build genai.Client arguments
+        client_args: Dict[str, Any] = {}
+
+        # Add Vertex AI parameters if provided
+        if vertexai is not None:
+            client_args["vertexai"] = vertexai
+
+        if credentials is not None:
+            client_args["credentials"] = credentials
+
+        if project is not None:
+            client_args["project"] = project
+
+        if location is not None:
+            client_args["location"] = location
+
+        if debug_config is not None:
+            client_args["debug_config"] = debug_config
+
+        if http_options is not None:
+            client_args["http_options"] = http_options
+
+        # Handle API key authentication
+        if vertexai:
+            # For Vertex AI, api_key is optional
+            if api_key is not None:
+                client_args["api_key"] = api_key
+        else:
+            # For non-Vertex AI mode, api_key is required (backwards compatibility)
+            if api_key is None:
+                api_key = os.environ.get("GOOGLE_API_KEY") or os.environ.get("API_KEY")
+
+            if api_key is None:
+                raise ValueError(
+                    "API key must be provided either as parameter or via GOOGLE_API_KEY/API_KEY environment variable"
+                )
+
+            client_args["api_key"] = api_key
+
+        self._client = genai.Client(**client_args)
+        self._base_url = "https://generativelanguage.googleapis.com"
+
+    def _merge_posthog_params(
+        self,
+        call_distinct_id: Optional[str],
+        call_trace_id: Optional[str],
+        call_properties: Optional[Dict[str, Any]],
+        call_privacy_mode: Optional[bool],
+        call_groups: Optional[Dict[str, Any]],
+    ):
+        """Merge call-level PostHog parameters with client defaults."""
+
+        # Use call-level values if provided, otherwise fall back to defaults
+        distinct_id = (
+            call_distinct_id
+            if call_distinct_id is not None
+            else self._default_distinct_id
+        )
+        privacy_mode = (
+            call_privacy_mode
+            if call_privacy_mode is not None
+            else self._default_privacy_mode
+        )
+        groups = call_groups if call_groups is not None else self._default_groups
+
+        # Merge properties: default properties + call properties (call properties override)
+        properties = dict(self._default_properties)
+
+        if call_properties:
+            properties.update(call_properties)
+
+        if call_trace_id is None:
+            call_trace_id = str(uuid.uuid4())
+
+        return distinct_id, call_trace_id, properties, privacy_mode, groups
+
+    async def generate_content(
+        self,
+        model: str,
+        contents,
+        posthog_distinct_id: Optional[str] = None,
+        posthog_trace_id: Optional[str] = None,
+        posthog_properties: Optional[Dict[str, Any]] = None,
+        posthog_privacy_mode: Optional[bool] = None,
+        posthog_groups: Optional[Dict[str, Any]] = None,
+        **kwargs: Any,
+    ):
+        """
+        Generate content using Gemini's API while tracking usage in PostHog.
+
+        This method signature exactly matches genai.Client().aio.models.generate_content()
+        with additional PostHog tracking parameters.
+
+        Args:
+            model: The model to use (e.g., 'gemini-2.0-flash')
+            contents: The input content for generation
+            posthog_distinct_id: ID to associate with the usage event (overrides client default)
+            posthog_trace_id: Trace UUID for linking events (auto-generated if not provided)
+            posthog_properties: Extra properties to include in the event (merged with client defaults)
+            posthog_privacy_mode: Whether to redact sensitive information (overrides client default)
+            posthog_groups: Group analytics properties (overrides client default)
+            **kwargs: Arguments passed to Gemini's generate_content
+        """
+
+        # Merge PostHog parameters
+        distinct_id, trace_id, properties, privacy_mode, groups = (
+            self._merge_posthog_params(
+                posthog_distinct_id,
+                posthog_trace_id,
+                posthog_properties,
+                posthog_privacy_mode,
+                posthog_groups,
+            )
+        )
+
+        kwargs_with_contents = {"model": model, "contents": contents, **kwargs}
+
+        return await call_llm_and_track_usage_async(
+            distinct_id,
+            self._ph_client,
+            "gemini",
+            trace_id,
+            properties,
+            privacy_mode,
+            groups,
+            self._base_url,
+            self._client.aio.models.generate_content,
+            **kwargs_with_contents,
+        )
+
+    async def _generate_content_streaming(
+        self,
+        model: str,
+        contents,
+        distinct_id: Optional[str],
+        trace_id: Optional[str],
+        properties: Optional[Dict[str, Any]],
+        privacy_mode: bool,
+        groups: Optional[Dict[str, Any]],
+        **kwargs: Any,
+    ):
+        start_time = time.time()
+        usage_stats: TokenUsage = TokenUsage(input_tokens=0, output_tokens=0)
+        accumulated_content = []
+
+        kwargs_without_stream = {"model": model, "contents": contents, **kwargs}
+        response = await self._client.aio.models.generate_content_stream(
+            **kwargs_without_stream
+        )
+
+        async def async_generator():
+            nonlocal usage_stats
+            nonlocal accumulated_content
+
+            try:
+                async for chunk in response:
+                    # Extract usage stats from chunk
+                    chunk_usage = extract_gemini_usage_from_chunk(chunk)
+
+                    if chunk_usage:
+                        # Gemini reports cumulative totals, not incremental values
+                        merge_usage_stats(usage_stats, chunk_usage, mode="cumulative")
+
+                    # Extract content from chunk (now returns content blocks)
+                    content_block = extract_gemini_content_from_chunk(chunk)
+
+                    if content_block is not None:
+                        accumulated_content.append(content_block)
+
+                    yield chunk
+
+            finally:
+                end_time = time.time()
+                latency = end_time - start_time
+
+                self._capture_streaming_event(
+                    model,
+                    contents,
+                    distinct_id,
+                    trace_id,
+                    properties,
+                    privacy_mode,
+                    groups,
+                    kwargs,
+                    usage_stats,
+                    latency,
+                    accumulated_content,
+                )
+
+        return async_generator()
+
+    def _capture_streaming_event(
+        self,
+        model: str,
+        contents,
+        distinct_id: Optional[str],
+        trace_id: Optional[str],
+        properties: Optional[Dict[str, Any]],
+        privacy_mode: bool,
+        groups: Optional[Dict[str, Any]],
+        kwargs: Dict[str, Any],
+        usage_stats: TokenUsage,
+        latency: float,
+        output: Any,
+    ):
+        # Prepare standardized event data
+        formatted_input = self._format_input(contents, **kwargs)
+        sanitized_input = sanitize_gemini(formatted_input)
+
+        event_data = StreamingEventData(
+            provider="gemini",
+            model=model,
+            base_url=self._base_url,
+            kwargs=kwargs,
+            formatted_input=sanitized_input,
+            formatted_output=format_gemini_streaming_output(output),
+            usage_stats=usage_stats,
+            latency=latency,
+            distinct_id=distinct_id,
+            trace_id=trace_id,
+            properties=properties,
+            privacy_mode=privacy_mode,
+            groups=groups,
+        )
+
+        # Use the common capture function
+        capture_streaming_event(self._ph_client, event_data)
+
+    def _format_input(self, contents, **kwargs):
+        """Format input contents for PostHog tracking"""
+
+        # Create kwargs dict with contents for merge_system_prompt
+        input_kwargs = {"contents": contents, **kwargs}
+        return merge_system_prompt(input_kwargs, "gemini")
+
+    async def generate_content_stream(
+        self,
+        model: str,
+        contents,
+        posthog_distinct_id: Optional[str] = None,
+        posthog_trace_id: Optional[str] = None,
+        posthog_properties: Optional[Dict[str, Any]] = None,
+        posthog_privacy_mode: Optional[bool] = None,
+        posthog_groups: Optional[Dict[str, Any]] = None,
+        **kwargs: Any,
+    ):
+        # Merge PostHog parameters
+        distinct_id, trace_id, properties, privacy_mode, groups = (
+            self._merge_posthog_params(
+                posthog_distinct_id,
+                posthog_trace_id,
+                posthog_properties,
+                posthog_privacy_mode,
+                posthog_groups,
+            )
+        )
+
+        return await self._generate_content_streaming(
+            model,
+            contents,
+            distinct_id,
+            trace_id,
+            properties,
+            privacy_mode,
+            groups,
+            **kwargs,
+        )

posthog/ai/gemini/gemini_converter.py

@@ -0,0 +1,586 @@
+"""
+Gemini-specific conversion utilities.
+
+This module handles the conversion of Gemini API responses and inputs
+into standardized formats for PostHog tracking.
+"""
+
+from typing import Any, Dict, List, Optional, TypedDict, Union
+
+from posthog.ai.types import (
+    FormattedContentItem,
+    FormattedMessage,
+    TokenUsage,
+)
+
+
+class GeminiPart(TypedDict, total=False):
+    """Represents a part in a Gemini message."""
+
+    text: str
+
+
+class GeminiMessage(TypedDict, total=False):
+    """Represents a Gemini message with various possible fields."""
+
+    role: str
+    parts: List[Union[GeminiPart, Dict[str, Any]]]
+    content: Union[str, List[Any]]
+    text: str
+
+
+def _extract_text_from_parts(parts: List[Any]) -> str:
+    """
+    Extract and concatenate text from a parts array.
+
+    Args:
+        parts: List of parts that may contain text content
+
+    Returns:
+        Concatenated text from all parts
+    """
+
+    content_parts = []
+
+    for part in parts:
+        if isinstance(part, dict) and "text" in part:
+            content_parts.append(part["text"])
+
+        elif isinstance(part, str):
+            content_parts.append(part)
+
+        elif hasattr(part, "text"):
+            # Get the text attribute value
+            text_value = getattr(part, "text", "")
+            content_parts.append(text_value if text_value else str(part))
+
+        else:
+            content_parts.append(str(part))
+
+    return "".join(content_parts)
+
+
+def _format_dict_message(item: Dict[str, Any]) -> FormattedMessage:
+    """
+    Format a dictionary message into standardized format.
+
+    Args:
+        item: Dictionary containing message data
+
+    Returns:
+        Formatted message with role and content
+    """
+
+    # Handle dict format with parts array (Gemini-specific format)
+    if "parts" in item and isinstance(item["parts"], list):
+        content = _extract_text_from_parts(item["parts"])
+        return {"role": item.get("role", "user"), "content": content}
+
+    # Handle dict with content field
+    if "content" in item:
+        content = item["content"]
+
+        if isinstance(content, list):
+            # If content is a list, extract text from it
+            content = _extract_text_from_parts(content)
+
+        elif not isinstance(content, str):
+            content = str(content)
+
+        return {"role": item.get("role", "user"), "content": content}
+
+    # Handle dict with text field
+    if "text" in item:
+        return {"role": item.get("role", "user"), "content": item["text"]}
+
+    # Fallback to string representation
+    return {"role": "user", "content": str(item)}
+
+
+def _format_object_message(item: Any) -> FormattedMessage:
+    """
+    Format an object (with attributes) into standardized format.
+
+    Args:
+        item: Object that may have text or parts attributes
+
+    Returns:
+        Formatted message with role and content
+    """
+
+    # Handle object with parts attribute
+    if hasattr(item, "parts") and hasattr(item.parts, "__iter__"):
+        content = _extract_text_from_parts(item.parts)
+        role = getattr(item, "role", "user") if hasattr(item, "role") else "user"
+
+        # Ensure role is a string
+        if not isinstance(role, str):
+            role = "user"
+
+        return {"role": role, "content": content}
+
+    # Handle object with text attribute
+    if hasattr(item, "text"):
+        role = getattr(item, "role", "user") if hasattr(item, "role") else "user"
+
+        # Ensure role is a string
+        if not isinstance(role, str):
+            role = "user"
+
+        return {"role": role, "content": item.text}
+
+    # Handle object with content attribute
+    if hasattr(item, "content"):
+        role = getattr(item, "role", "user") if hasattr(item, "role") else "user"
+
+        # Ensure role is a string
+        if not isinstance(role, str):
+            role = "user"
+
+        content = item.content
+
+        if isinstance(content, list):
+            content = _extract_text_from_parts(content)
+
+        elif not isinstance(content, str):
+            content = str(content)
+        return {"role": role, "content": content}
+
+    # Fallback to string representation
+    return {"role": "user", "content": str(item)}
+
+
+def format_gemini_response(response: Any) -> List[FormattedMessage]:
+    """
+    Format a Gemini response into standardized message format.
+
+    Args:
+        response: The response object from Gemini API
+
+    Returns:
+        List of formatted messages with role and content
+    """
+
+    output: List[FormattedMessage] = []
+
+    if response is None:
+        return output
+
+    if hasattr(response, "candidates") and response.candidates:
+        for candidate in response.candidates:
+            if hasattr(candidate, "content") and candidate.content:
+                content: List[FormattedContentItem] = []
+
+                if hasattr(candidate.content, "parts") and candidate.content.parts:
+                    for part in candidate.content.parts:
+                        if hasattr(part, "text") and part.text:
+                            content.append(
+                                {
+                                    "type": "text",
+                                    "text": part.text,
+                                }
+                            )
+
+                        elif hasattr(part, "function_call") and part.function_call:
+                            function_call = part.function_call
+                            content.append(
+                                {
+                                    "type": "function",
+                                    "function": {
+                                        "name": function_call.name,
+                                        "arguments": function_call.args,
+                                    },
+                                }
+                            )
+
+                if content:
+                    output.append(
+                        {
+                            "role": "assistant",
+                            "content": content,
+                        }
+                    )
+
+            elif hasattr(candidate, "text") and candidate.text:
+                output.append(
+                    {
+                        "role": "assistant",
+                        "content": [{"type": "text", "text": candidate.text}],
+                    }
+                )
+
+    elif hasattr(response, "text") and response.text:
+        output.append(
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": response.text}],
+            }
+        )
+
+    return output
+
+
+def extract_gemini_system_instruction(config: Any) -> Optional[str]:
+    """
+    Extract system instruction from Gemini config parameter.
+
+    Args:
+        config: Config object or dict that may contain system instruction
+
+    Returns:
+        System instruction string if present, None otherwise
+    """
+    if config is None:
+        return None
+
+    # Handle different config formats
+    if hasattr(config, "system_instruction"):
+        return config.system_instruction
+    elif isinstance(config, dict) and "system_instruction" in config:
+        return config["system_instruction"]
+    elif isinstance(config, dict) and "systemInstruction" in config:
+        return config["systemInstruction"]
+
+    return None
+
+
+def extract_gemini_tools(kwargs: Dict[str, Any]) -> Optional[Any]:
+    """
+    Extract tool definitions from Gemini API kwargs.
+
+    Args:
+        kwargs: Keyword arguments passed to Gemini API
+
+    Returns:
+        Tool definitions if present, None otherwise
+    """
+
+    if "config" in kwargs and hasattr(kwargs["config"], "tools"):
+        return kwargs["config"].tools
+
+    return None
+
+
+def format_gemini_input_with_system(
+    contents: Any, config: Any = None
+) -> List[FormattedMessage]:
+    """
+    Format Gemini input contents into standardized message format, including system instruction handling.
+
+    Args:
+        contents: Input contents in various possible formats
+        config: Config object or dict that may contain system instruction
+
+    Returns:
+        List of formatted messages with role and content fields, with system message prepended if needed
+    """
+    formatted_messages = format_gemini_input(contents)
+
+    # Check if system instruction is provided in config parameter
+    system_instruction = extract_gemini_system_instruction(config)
+
+    if system_instruction is not None:
+        has_system = any(msg.get("role") == "system" for msg in formatted_messages)
+        if not has_system:
+            from posthog.ai.types import FormattedMessage
+
+            system_message: FormattedMessage = {
+                "role": "system",
+                "content": system_instruction,
+            }
+            formatted_messages = [system_message] + list(formatted_messages)
+
+    return formatted_messages
+
+
+def format_gemini_input(contents: Any) -> List[FormattedMessage]:
+    """
+    Format Gemini input contents into standardized message format for PostHog tracking.
+
+    This function handles various input formats:
+    - String inputs
+    - List of strings, dicts, or objects
+    - Single dict or object
+    - Gemini-specific format with parts array
+
+    Args:
+        contents: Input contents in various possible formats
+
+    Returns:
+        List of formatted messages with role and content fields
+    """
+
+    # Handle string input
+    if isinstance(contents, str):
+        return [{"role": "user", "content": contents}]
+
+    # Handle list input
+    if isinstance(contents, list):
+        formatted: List[FormattedMessage] = []
+
+        for item in contents:
+            if isinstance(item, str):
+                formatted.append({"role": "user", "content": item})
+
+            elif isinstance(item, dict):
+                formatted.append(_format_dict_message(item))
+
+            else:
+                formatted.append(_format_object_message(item))
+
+        return formatted
+
+    # Handle single dict input
+    if isinstance(contents, dict):
+        return [_format_dict_message(contents)]
+
+    # Handle single object input
+    return [_format_object_message(contents)]
+
+
+def extract_gemini_web_search_count(response: Any) -> int:
+    """
+    Extract web search count from Gemini response.
+
+    Gemini bills per request that uses grounding, not per query.
+    Returns 1 if grounding_metadata is present with actual search data, 0 otherwise.
+
+    Args:
+        response: The response from Gemini API
+
+    Returns:
+        1 if web search/grounding was used, 0 otherwise
+    """
+
+    # Check for grounding_metadata in candidates
+    if hasattr(response, "candidates"):
+        for candidate in response.candidates:
+            if (
+                hasattr(candidate, "grounding_metadata")
+                and candidate.grounding_metadata
+            ):
+                grounding_metadata = candidate.grounding_metadata
+
+                # Check if web_search_queries exists and is non-empty
+                if hasattr(grounding_metadata, "web_search_queries"):
+                    queries = grounding_metadata.web_search_queries
+
+                    if queries is not None and len(queries) > 0:
+                        return 1
+
+                # Check if grounding_chunks exists and is non-empty
+                if hasattr(grounding_metadata, "grounding_chunks"):
+                    chunks = grounding_metadata.grounding_chunks
+
+                    if chunks is not None and len(chunks) > 0:
+                        return 1
+
+            # Also check for google_search or grounding in function call names
+            if hasattr(candidate, "content") and candidate.content:
+                if hasattr(candidate.content, "parts") and candidate.content.parts:
+                    for part in candidate.content.parts:
+                        if hasattr(part, "function_call") and part.function_call:
+                            function_name = getattr(
+                                part.function_call, "name", ""
+                            ).lower()
+
+                            if (
+                                "google_search" in function_name
+                                or "grounding" in function_name
+                            ):
+                                return 1
+
+    return 0
+
+
+def _extract_usage_from_metadata(metadata: Any) -> TokenUsage:
+    """
+    Common logic to extract usage from Gemini metadata.
+    Used by both streaming and non-streaming paths.
+
+    Args:
+        metadata: usage_metadata from Gemini response or chunk
+
+    Returns:
+        TokenUsage with standardized usage
+    """
+    usage = TokenUsage(
+        input_tokens=getattr(metadata, "prompt_token_count", 0),
+        output_tokens=getattr(metadata, "candidates_token_count", 0),
+    )
+
+    # Add cache tokens if present (don't add if 0)
+    if hasattr(metadata, "cached_content_token_count"):
+        cache_tokens = metadata.cached_content_token_count
+        if cache_tokens and cache_tokens > 0:
+            usage["cache_read_input_tokens"] = cache_tokens
+
+    # Add reasoning tokens if present (don't add if 0)
+    if hasattr(metadata, "thoughts_token_count"):
+        reasoning_tokens = metadata.thoughts_token_count
+        if reasoning_tokens and reasoning_tokens > 0:
+            usage["reasoning_tokens"] = reasoning_tokens
+
+    return usage
+
+
+def extract_gemini_usage_from_response(response: Any) -> TokenUsage:
+    """
+    Extract usage statistics from a full Gemini response (non-streaming).
+
+    Args:
+        response: The complete response from Gemini API
+
+    Returns:
+        TokenUsage with standardized usage statistics
+    """
+    if not hasattr(response, "usage_metadata") or not response.usage_metadata:
+        return TokenUsage(input_tokens=0, output_tokens=0)
+
+    usage = _extract_usage_from_metadata(response.usage_metadata)
+
+    # Add web search count if present
+    web_search_count = extract_gemini_web_search_count(response)
+    if web_search_count > 0:
+        usage["web_search_count"] = web_search_count
+
+    return usage
+
+
+def extract_gemini_usage_from_chunk(chunk: Any) -> TokenUsage:
+    """
+    Extract usage statistics from a Gemini streaming chunk.
+
+    Args:
+        chunk: Streaming chunk from Gemini API
+
+    Returns:
+        TokenUsage with standardized usage statistics
+    """
+
+    usage: TokenUsage = TokenUsage()
+
+    # Extract web search count from the chunk before checking for usage_metadata
+    # Web search indicators can appear on any chunk, not just those with usage data
+    web_search_count = extract_gemini_web_search_count(chunk)
+    if web_search_count > 0:
+        usage["web_search_count"] = web_search_count
+
+    if not hasattr(chunk, "usage_metadata") or not chunk.usage_metadata:
+        return usage
+
+    usage_from_metadata = _extract_usage_from_metadata(chunk.usage_metadata)
+
+    # Merge the usage from metadata with any web search count we found
+    usage.update(usage_from_metadata)
+
+    return usage
+
+
+def extract_gemini_content_from_chunk(chunk: Any) -> Optional[Dict[str, Any]]:
+    """
+    Extract content (text or function call) from a Gemini streaming chunk.
+
+    Args:
+        chunk: Streaming chunk from Gemini API
+
+    Returns:
+        Content block dictionary if present, None otherwise
+    """
+
+    # Check for text content
+    if hasattr(chunk, "text") and chunk.text:
+        return {"type": "text", "text": chunk.text}
+
+    # Check for function calls in candidates
+    if hasattr(chunk, "candidates") and chunk.candidates:
+        for candidate in chunk.candidates:
+            if hasattr(candidate, "content") and candidate.content:
+                if hasattr(candidate.content, "parts") and candidate.content.parts:
+                    for part in candidate.content.parts:
+                        # Check for function_call part
+                        if hasattr(part, "function_call") and part.function_call:
+                            function_call = part.function_call
+                            return {
+                                "type": "function",
+                                "function": {
+                                    "name": function_call.name,
+                                    "arguments": function_call.args,
+                                },
+                            }
+                        # Also check for text in parts
+                        elif hasattr(part, "text") and part.text:
+                            return {"type": "text", "text": part.text}
+
+    return None
+
+
+def format_gemini_streaming_output(
+    accumulated_content: Union[str, List[Any]],
+) -> List[FormattedMessage]:
+    """
+    Format the final output from Gemini streaming.
+
+    Args:
+        accumulated_content: Accumulated content from streaming (string, list of strings, or list of content blocks)
+
+    Returns:
+        List of formatted messages
+    """
+
+    # Handle legacy string input (backward compatibility)
+    if isinstance(accumulated_content, str):
+        return [
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": accumulated_content}],
+            }
+        ]
+
+    # Handle list input
+    if isinstance(accumulated_content, list):
+        content: List[FormattedContentItem] = []
+        text_parts = []
+
+        for item in accumulated_content:
+            if isinstance(item, str):
+                # Legacy support: accumulate strings
+                text_parts.append(item)
+            elif isinstance(item, dict):
+                # New format: content blocks
+                if item.get("type") == "text":
+                    text_parts.append(item.get("text", ""))
+                elif item.get("type") == "function":
+                    # If we have accumulated text, add it first
+                    if text_parts:
+                        content.append(
+                            {
+                                "type": "text",
+                                "text": "".join(text_parts),
+                            }
+                        )
+                        text_parts = []
+
+                    # Add the function call
+                    content.append(
+                        {
+                            "type": "function",
+                            "function": item.get("function", {}),
+                        }
+                    )
+
+        # Add any remaining text
+        if text_parts:
+            content.append(
+                {
+                    "type": "text",
+                    "text": "".join(text_parts),
+                }
+            )
+
+        # If we have content, return it
+        if content:
+            return [{"role": "assistant", "content": content}]
+
+    # Fallback for empty or unexpected input
+    return [{"role": "assistant", "content": [{"type": "text", "text": ""}]}]

posthog/ai/langchain/callbacks.py

@@ -1,8 +1,8 @@
 try:
-    import langchain  # noqa: F401
+    import langchain_core  # noqa: F401
 except ImportError:
     raise ModuleNotFoundError(
-        "Please install LangChain to use this feature: 'pip install langchain'"
+        "Please install LangChain to use this feature: 'pip install langchain-core'"
     )
 
 import json
@@ -20,8 +20,14 @@ from typing import (
 )
 from uuid import UUID
 
-from langchain.callbacks.base import BaseCallbackHandler
-from langchain.schema.agent import AgentAction, AgentFinish
+try:
+    # LangChain 1.0+ and modern 0.x with langchain-core
+    from langchain_core.callbacks.base import BaseCallbackHandler
+    from langchain_core.agents import AgentAction, AgentFinish
+except (ImportError, ModuleNotFoundError):
+    # Fallback for older LangChain versions
+    from langchain.callbacks.base import BaseCallbackHandler
+    from langchain.schema.agent import AgentAction, AgentFinish
 from langchain_core.documents import Document
 from langchain_core.messages import (
     AIMessage,
@@ -37,6 +43,7 @@ from pydantic import BaseModel
 
 from posthog import setup
 from posthog.ai.utils import get_model_params, with_privacy_mode
+from posthog.ai.sanitization import sanitize_langchain
 from posthog.client import Client
 
 log = logging.getLogger("posthog")
@@ -72,6 +79,8 @@ class GenerationMetadata(SpanMetadata):
     """Base URL of the provider's API used in the run."""
     tools: Optional[List[Dict[str, Any]]] = None
     """Tools provided to the model."""
+    posthog_properties: Optional[Dict[str, Any]] = None
+    """PostHog properties of the run."""
 
 
 RunMetadata = Union[SpanMetadata, GenerationMetadata]
@@ -413,6 +422,8 @@ class CallbackHandler(BaseCallbackHandler):
                 generation.model = model
             if provider := metadata.get("ls_provider"):
                 generation.provider = provider
+
+            generation.posthog_properties = metadata.get("posthog_properties")
         try:
             base_url = serialized["kwargs"]["openai_api_base"]
             if base_url is not None:
@@ -480,11 +491,12 @@ class CallbackHandler(BaseCallbackHandler):
         event_properties = {
             "$ai_trace_id": trace_id,
             "$ai_input_state": with_privacy_mode(
-                self._ph_client, self._privacy_mode, run.input
+                self._ph_client, self._privacy_mode, sanitize_langchain(run.input)
             ),
             "$ai_latency": run.latency,
             "$ai_span_name": run.name,
             "$ai_span_id": run_id,
+            "$ai_framework": "langchain",
         }
         if parent_run_id is not None:
             event_properties["$ai_parent_id"] = parent_run_id
@@ -550,13 +562,17 @@ class CallbackHandler(BaseCallbackHandler):
             "$ai_model": run.model,
             "$ai_model_parameters": run.model_params,
             "$ai_input": with_privacy_mode(
-                self._ph_client, self._privacy_mode, run.input
+                self._ph_client, self._privacy_mode, sanitize_langchain(run.input)
             ),
             "$ai_http_status": 200,
             "$ai_latency": run.latency,
             "$ai_base_url": run.base_url,
+            "$ai_framework": "langchain",
         }
 
+        if isinstance(run.posthog_properties, dict):
+            event_properties.update(run.posthog_properties)
+
         if run.tools:
             event_properties["$ai_tools"] = run.tools
 
@@ -566,7 +582,7 @@ class CallbackHandler(BaseCallbackHandler):
             event_properties["$ai_is_error"] = True
         else:
             # Add usage
-            usage = _parse_usage(output)
+            usage = _parse_usage(output, run.provider, run.model)
             event_properties["$ai_input_tokens"] = usage.input_tokens
             event_properties["$ai_output_tokens"] = usage.output_tokens
             event_properties["$ai_cache_creation_input_tokens"] = (
@@ -687,6 +703,8 @@ class ModelUsage:
 
 def _parse_usage_model(
     usage: Union[BaseModel, dict],
+    provider: Optional[str] = None,
+    model: Optional[str] = None,
 ) -> ModelUsage:
     if isinstance(usage, BaseModel):
         usage = usage.__dict__
@@ -749,15 +767,36 @@ def _parse_usage_model(
         "cache_read": "cache_read_tokens",
         "reasoning": "reasoning_tokens",
     }
-    return ModelUsage(
+    normalized_usage = ModelUsage(
         **{
             dataclass_key: parsed_usage.get(mapped_key) or 0
             for mapped_key, dataclass_key in field_mapping.items()
         },
     )
+    # For Anthropic providers, LangChain reports input_tokens as the sum of input and cache read tokens.
+    # Our cost calculation expects them to be separate for Anthropic, so we subtract cache tokens.
+    # For other providers (OpenAI, etc.), input_tokens already includes cache tokens as expected.
+    # Match logic consistent with plugin-server: exact match on provider OR substring match on model
+    is_anthropic = False
+    if provider and provider.lower() == "anthropic":
+        is_anthropic = True
+    elif model and "anthropic" in model.lower():
+        is_anthropic = True
+
+    if (
+        is_anthropic
+        and normalized_usage.input_tokens
+        and normalized_usage.cache_read_tokens
+    ):
+        normalized_usage.input_tokens = max(
+            normalized_usage.input_tokens - normalized_usage.cache_read_tokens, 0
+        )
+    return normalized_usage
 
 
-def _parse_usage(response: LLMResult) -> ModelUsage:
+def _parse_usage(
+    response: LLMResult, provider: Optional[str] = None, model: Optional[str] = None
+) -> ModelUsage:
     # langchain-anthropic uses the usage field
     llm_usage_keys = ["token_usage", "usage"]
     llm_usage: ModelUsage = ModelUsage(
@@ -771,13 +810,15 @@ def _parse_usage(response: LLMResult) -> ModelUsage:
     if response.llm_output is not None:
         for key in llm_usage_keys:
             if response.llm_output.get(key):
-                llm_usage = _parse_usage_model(response.llm_output[key])
+                llm_usage = _parse_usage_model(
+                    response.llm_output[key], provider, model
+                )
                 break
 
     if hasattr(response, "generations"):
         for generation in response.generations:
             if "usage" in generation:
-                llm_usage = _parse_usage_model(generation["usage"])
+                llm_usage = _parse_usage_model(generation["usage"], provider, model)
                 break
 
             for generation_chunk in generation:
@@ -785,7 +826,9 @@ def _parse_usage(response: LLMResult) -> ModelUsage:
                     "usage_metadata" in generation_chunk.generation_info
                 ):
                     llm_usage = _parse_usage_model(
-                        generation_chunk.generation_info["usage_metadata"]
+                        generation_chunk.generation_info["usage_metadata"],
+                        provider,
+                        model,
                     )
                     break
 
@@ -812,7 +855,7 @@ def _parse_usage(response: LLMResult) -> ModelUsage:
                     bedrock_anthropic_usage or bedrock_titan_usage or ollama_usage
                 )
                 if chunk_usage:
-                    llm_usage = _parse_usage_model(chunk_usage)
+                    llm_usage = _parse_usage_model(chunk_usage, provider, model)
                     break
 
     return llm_usage

posthog/ai/openai/__init__.py

@@ -1,5 +1,20 @@
 from .openai import OpenAI
 from .openai_async import AsyncOpenAI
 from .openai_providers import AsyncAzureOpenAI, AzureOpenAI
+from .openai_converter import (
+    format_openai_response,
+    format_openai_input,
+    extract_openai_tools,
+    format_openai_streaming_content,
+)
 
-__all__ = ["OpenAI", "AsyncOpenAI", "AzureOpenAI", "AsyncAzureOpenAI"]
+__all__ = [
+    "OpenAI",
+    "AsyncOpenAI",
+    "AzureOpenAI",
+    "AsyncAzureOpenAI",
+    "format_openai_response",
+    "format_openai_input",
+    "extract_openai_tools",
+    "format_openai_streaming_content",
+]

posthog/ai/openai/openai.py

@@ -2,6 +2,8 @@ import time
 import uuid
 from typing import Any, Dict, List, Optional
 
+from posthog.ai.types import TokenUsage
+
 try:
     import openai
 except ImportError:
@@ -12,9 +14,16 @@ except ImportError:
 from posthog.ai.utils import (
     call_llm_and_track_usage,
     extract_available_tool_calls,
-    get_model_params,
+    merge_usage_stats,
     with_privacy_mode,
 )
+from posthog.ai.openai.openai_converter import (
+    extract_openai_usage_from_chunk,
+    extract_openai_content_from_chunk,
+    extract_openai_tool_calls_from_chunk,
+    accumulate_openai_tool_calls,
+)
+from posthog.ai.sanitization import sanitize_openai, sanitize_openai_response
 from posthog.client import Client as PostHogClient
 from posthog import setup
 
@@ -33,6 +42,7 @@ class OpenAI(openai.OpenAI):
             posthog_client: If provided, events will be captured via this client instead of the global `posthog`.
             **openai_config: Any additional keyword args to set on openai (e.g. organization="xxx").
         """
+
         super().__init__(**kwargs)
         self._ph_client = posthog_client or setup()
 
@@ -112,7 +122,7 @@ class WrappedResponses:
         **kwargs: Any,
     ):
         start_time = time.time()
-        usage_stats: Dict[str, int] = {}
+        usage_stats: TokenUsage = TokenUsage()
         final_content = []
         response = self._original.create(**kwargs)
 
@@ -122,35 +132,17 @@ class WrappedResponses:
 
             try:
                 for chunk in response:
-                    if hasattr(chunk, "type") and chunk.type == "response.completed":
-                        res = chunk.response
-                        if res.output and len(res.output) > 0:
-                            final_content.append(res.output[0])
+                    # Extract usage stats from chunk
+                    chunk_usage = extract_openai_usage_from_chunk(chunk, "responses")
 
-                    if hasattr(chunk, "usage") and chunk.usage:
-                        usage_stats = {
-                            k: getattr(chunk.usage, k, 0)
-                            for k in [
-                                "input_tokens",
-                                "output_tokens",
-                                "total_tokens",
-                            ]
-                        }
+                    if chunk_usage:
+                        merge_usage_stats(usage_stats, chunk_usage)
 
-                        # Add support for cached tokens
-                        if hasattr(chunk.usage, "output_tokens_details") and hasattr(
-                            chunk.usage.output_tokens_details, "reasoning_tokens"
-                        ):
-                            usage_stats["reasoning_tokens"] = (
-                                chunk.usage.output_tokens_details.reasoning_tokens
-                            )
+                    # Extract content from chunk
+                    content = extract_openai_content_from_chunk(chunk, "responses")
 
-                        if hasattr(chunk.usage, "input_tokens_details") and hasattr(
-                            chunk.usage.input_tokens_details, "cached_tokens"
-                        ):
-                            usage_stats["cache_read_input_tokens"] = (
-                                chunk.usage.input_tokens_details.cached_tokens
-                            )
+                    if content is not None:
+                        final_content.append(content)
 
                     yield chunk
 
@@ -168,7 +160,7 @@ class WrappedResponses:
                     usage_stats,
                     latency,
                     output,
-                    extract_available_tool_calls("openai", kwargs),
+                    None,  # Responses API doesn't have tools
                 )
 
         return generator()
@@ -181,52 +173,40 @@ class WrappedResponses:
         posthog_privacy_mode: bool,
         posthog_groups: Optional[Dict[str, Any]],
         kwargs: Dict[str, Any],
-        usage_stats: Dict[str, int],
+        usage_stats: TokenUsage,
         latency: float,
         output: Any,
         available_tool_calls: Optional[List[Dict[str, Any]]] = None,
     ):
-        if posthog_trace_id is None:
-            posthog_trace_id = str(uuid.uuid4())
+        from posthog.ai.types import StreamingEventData
+        from posthog.ai.openai.openai_converter import (
+            format_openai_streaming_input,
+            format_openai_streaming_output,
+        )
+        from posthog.ai.utils import capture_streaming_event
 
-        event_properties = {
-            "$ai_provider": "openai",
-            "$ai_model": kwargs.get("model"),
-            "$ai_model_parameters": get_model_params(kwargs),
-            "$ai_input": with_privacy_mode(
-                self._client._ph_client, posthog_privacy_mode, kwargs.get("input")
-            ),
-            "$ai_output_choices": with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                output,
-            ),
-            "$ai_http_status": 200,
-            "$ai_input_tokens": usage_stats.get("input_tokens", 0),
-            "$ai_output_tokens": usage_stats.get("output_tokens", 0),
-            "$ai_cache_read_input_tokens": usage_stats.get(
-                "cache_read_input_tokens", 0
-            ),
-            "$ai_reasoning_tokens": usage_stats.get("reasoning_tokens", 0),
-            "$ai_latency": latency,
-            "$ai_trace_id": posthog_trace_id,
-            "$ai_base_url": str(self._client.base_url),
-            **(posthog_properties or {}),
-        }
+        # Prepare standardized event data
+        formatted_input = format_openai_streaming_input(kwargs, "responses")
+        sanitized_input = sanitize_openai_response(formatted_input)
 
-        if available_tool_calls:
-            event_properties["$ai_tools"] = available_tool_calls
+        event_data = StreamingEventData(
+            provider="openai",
+            model=kwargs.get("model", "unknown"),
+            base_url=str(self._client.base_url),
+            kwargs=kwargs,
+            formatted_input=sanitized_input,
+            formatted_output=format_openai_streaming_output(output, "responses"),
+            usage_stats=usage_stats,
+            latency=latency,
+            distinct_id=posthog_distinct_id,
+            trace_id=posthog_trace_id,
+            properties=posthog_properties,
+            privacy_mode=posthog_privacy_mode,
+            groups=posthog_groups,
+        )
 
-        if posthog_distinct_id is None:
-            event_properties["$process_person_profile"] = False
-
-        if hasattr(self._client._ph_client, "capture"):
-            self._client._ph_client.capture(
-                distinct_id=posthog_distinct_id or posthog_trace_id,
-                event="$ai_generation",
-                properties=event_properties,
-                groups=posthog_groups,
-            )
+        # Use the common capture function
+        capture_streaming_event(self._client._ph_client, event_data)
 
     def parse(
         self,
@@ -337,8 +317,9 @@ class WrappedCompletions:
         **kwargs: Any,
     ):
         start_time = time.time()
-        usage_stats: Dict[str, int] = {}
+        usage_stats: TokenUsage = TokenUsage()
         accumulated_content = []
+        accumulated_tool_calls: Dict[int, Dict[str, Any]] = {}
         if "stream_options" not in kwargs:
             kwargs["stream_options"] = {}
         kwargs["stream_options"]["include_usage"] = True
@@ -347,50 +328,42 @@ class WrappedCompletions:
         def generator():
             nonlocal usage_stats
             nonlocal accumulated_content  # noqa: F824
+            nonlocal accumulated_tool_calls
 
             try:
                 for chunk in response:
-                    if hasattr(chunk, "usage") and chunk.usage:
-                        usage_stats = {
-                            k: getattr(chunk.usage, k, 0)
-                            for k in [
-                                "prompt_tokens",
-                                "completion_tokens",
-                                "total_tokens",
-                            ]
-                        }
+                    # Extract usage stats from chunk
+                    chunk_usage = extract_openai_usage_from_chunk(chunk, "chat")
 
-                        # Add support for cached tokens
-                        if hasattr(chunk.usage, "prompt_tokens_details") and hasattr(
-                            chunk.usage.prompt_tokens_details, "cached_tokens"
-                        ):
-                            usage_stats["cache_read_input_tokens"] = (
-                                chunk.usage.prompt_tokens_details.cached_tokens
-                            )
+                    if chunk_usage:
+                        merge_usage_stats(usage_stats, chunk_usage)
 
-                        if hasattr(chunk.usage, "output_tokens_details") and hasattr(
-                            chunk.usage.output_tokens_details, "reasoning_tokens"
-                        ):
-                            usage_stats["reasoning_tokens"] = (
-                                chunk.usage.output_tokens_details.reasoning_tokens
-                            )
+                    # Extract content from chunk
+                    content = extract_openai_content_from_chunk(chunk, "chat")
 
-                    if (
-                        hasattr(chunk, "choices")
-                        and chunk.choices
-                        and len(chunk.choices) > 0
-                    ):
-                        if chunk.choices[0].delta and chunk.choices[0].delta.content:
-                            content = chunk.choices[0].delta.content
-                            if content:
-                                accumulated_content.append(content)
+                    if content is not None:
+                        accumulated_content.append(content)
+
+                    # Extract and accumulate tool calls from chunk
+                    chunk_tool_calls = extract_openai_tool_calls_from_chunk(chunk)
+                    if chunk_tool_calls:
+                        accumulate_openai_tool_calls(
+                            accumulated_tool_calls, chunk_tool_calls
+                        )
 
                     yield chunk
 
             finally:
                 end_time = time.time()
                 latency = end_time - start_time
-                output = "".join(accumulated_content)
+
+                # Convert accumulated tool calls dict to list
+                tool_calls_list = (
+                    list(accumulated_tool_calls.values())
+                    if accumulated_tool_calls
+                    else None
+                )
+
                 self._capture_streaming_event(
                     posthog_distinct_id,
                     posthog_trace_id,
@@ -400,7 +373,8 @@ class WrappedCompletions:
                     kwargs,
                     usage_stats,
                     latency,
-                    output,
+                    accumulated_content,
+                    tool_calls_list,
                     extract_available_tool_calls("openai", kwargs),
                 )
 
@@ -414,52 +388,41 @@ class WrappedCompletions:
         posthog_privacy_mode: bool,
         posthog_groups: Optional[Dict[str, Any]],
         kwargs: Dict[str, Any],
-        usage_stats: Dict[str, int],
+        usage_stats: TokenUsage,
         latency: float,
         output: Any,
+        tool_calls: Optional[List[Dict[str, Any]]] = None,
         available_tool_calls: Optional[List[Dict[str, Any]]] = None,
     ):
-        if posthog_trace_id is None:
-            posthog_trace_id = str(uuid.uuid4())
+        from posthog.ai.types import StreamingEventData
+        from posthog.ai.openai.openai_converter import (
+            format_openai_streaming_input,
+            format_openai_streaming_output,
+        )
+        from posthog.ai.utils import capture_streaming_event
 
-        event_properties = {
-            "$ai_provider": "openai",
-            "$ai_model": kwargs.get("model"),
-            "$ai_model_parameters": get_model_params(kwargs),
-            "$ai_input": with_privacy_mode(
-                self._client._ph_client, posthog_privacy_mode, kwargs.get("messages")
-            ),
-            "$ai_output_choices": with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                [{"content": output, "role": "assistant"}],
-            ),
-            "$ai_http_status": 200,
-            "$ai_input_tokens": usage_stats.get("prompt_tokens", 0),
-            "$ai_output_tokens": usage_stats.get("completion_tokens", 0),
-            "$ai_cache_read_input_tokens": usage_stats.get(
-                "cache_read_input_tokens", 0
-            ),
-            "$ai_reasoning_tokens": usage_stats.get("reasoning_tokens", 0),
-            "$ai_latency": latency,
-            "$ai_trace_id": posthog_trace_id,
-            "$ai_base_url": str(self._client.base_url),
-            **(posthog_properties or {}),
-        }
+        # Prepare standardized event data
+        formatted_input = format_openai_streaming_input(kwargs, "chat")
+        sanitized_input = sanitize_openai(formatted_input)
 
-        if available_tool_calls:
-            event_properties["$ai_tools"] = available_tool_calls
+        event_data = StreamingEventData(
+            provider="openai",
+            model=kwargs.get("model", "unknown"),
+            base_url=str(self._client.base_url),
+            kwargs=kwargs,
+            formatted_input=sanitized_input,
+            formatted_output=format_openai_streaming_output(output, "chat", tool_calls),
+            usage_stats=usage_stats,
+            latency=latency,
+            distinct_id=posthog_distinct_id,
+            trace_id=posthog_trace_id,
+            properties=posthog_properties,
+            privacy_mode=posthog_privacy_mode,
+            groups=posthog_groups,
+        )
 
-        if posthog_distinct_id is None:
-            event_properties["$process_person_profile"] = False
-
-        if hasattr(self._client._ph_client, "capture"):
-            self._client._ph_client.capture(
-                distinct_id=posthog_distinct_id or posthog_trace_id,
-                event="$ai_generation",
-                properties=event_properties,
-                groups=posthog_groups,
-            )
+        # Use the common capture function
+        capture_streaming_event(self._client._ph_client, event_data)
 
 
 class WrappedEmbeddings:
@@ -496,6 +459,7 @@ class WrappedEmbeddings:
         Returns:
             The response from OpenAI's embeddings.create call.
         """
+
         if posthog_trace_id is None:
             posthog_trace_id = str(uuid.uuid4())
 
@@ -518,7 +482,9 @@ class WrappedEmbeddings:
             "$ai_provider": "openai",
             "$ai_model": kwargs.get("model"),
             "$ai_input": with_privacy_mode(
-                self._client._ph_client, posthog_privacy_mode, kwargs.get("input")
+                self._client._ph_client,
+                posthog_privacy_mode,
+                sanitize_openai_response(kwargs.get("input")),
             ),
             "$ai_http_status": 200,
             "$ai_input_tokens": usage_stats.get("prompt_tokens", 0),

posthog/ai/openai/openai_async.py

@@ -1,6 +1,8 @@
 import time
 import uuid
-from typing import Any, Dict, List, Optional, cast
+from typing import Any, Dict, List, Optional
+
+from posthog.ai.types import TokenUsage
 
 try:
     import openai
@@ -14,8 +16,17 @@ from posthog.ai.utils import (
     call_llm_and_track_usage_async,
     extract_available_tool_calls,
     get_model_params,
+    merge_usage_stats,
     with_privacy_mode,
 )
+from posthog.ai.openai.openai_converter import (
+    extract_openai_usage_from_chunk,
+    extract_openai_content_from_chunk,
+    extract_openai_tool_calls_from_chunk,
+    accumulate_openai_tool_calls,
+    format_openai_streaming_output,
+)
+from posthog.ai.sanitization import sanitize_openai, sanitize_openai_response
 from posthog.client import Client as PostHogClient
 
 
@@ -34,6 +45,7 @@ class AsyncOpenAI(openai.AsyncOpenAI):
                             of the global posthog.
             **openai_config: Any additional keyword args to set on openai (e.g. organization="xxx").
         """
+
         super().__init__(**kwargs)
         self._ph_client = posthog_client or setup()
 
@@ -66,6 +78,7 @@ class WrappedResponses:
 
     def __getattr__(self, name):
         """Fallback to original responses object for any methods we don't explicitly handle."""
+
         return getattr(self._original, name)
 
     async def create(
@@ -113,7 +126,7 @@ class WrappedResponses:
         **kwargs: Any,
     ):
         start_time = time.time()
-        usage_stats: Dict[str, int] = {}
+        usage_stats: TokenUsage = TokenUsage()
         final_content = []
         response = await self._original.create(**kwargs)
 
@@ -123,35 +136,17 @@ class WrappedResponses:
 
             try:
                 async for chunk in response:
-                    if hasattr(chunk, "type") and chunk.type == "response.completed":
-                        res = chunk.response
-                        if res.output and len(res.output) > 0:
-                            final_content.append(res.output[0])
+                    # Extract usage stats from chunk
+                    chunk_usage = extract_openai_usage_from_chunk(chunk, "responses")
 
-                    if hasattr(chunk, "usage") and chunk.usage:
-                        usage_stats = {
-                            k: getattr(chunk.usage, k, 0)
-                            for k in [
-                                "input_tokens",
-                                "output_tokens",
-                                "total_tokens",
-                            ]
-                        }
+                    if chunk_usage:
+                        merge_usage_stats(usage_stats, chunk_usage)
 
-                        # Add support for cached tokens
-                        if hasattr(chunk.usage, "output_tokens_details") and hasattr(
-                            chunk.usage.output_tokens_details, "reasoning_tokens"
-                        ):
-                            usage_stats["reasoning_tokens"] = (
-                                chunk.usage.output_tokens_details.reasoning_tokens
-                            )
+                    # Extract content from chunk
+                    content = extract_openai_content_from_chunk(chunk, "responses")
 
-                        if hasattr(chunk.usage, "input_tokens_details") and hasattr(
-                            chunk.usage.input_tokens_details, "cached_tokens"
-                        ):
-                            usage_stats["cache_read_input_tokens"] = (
-                                chunk.usage.input_tokens_details.cached_tokens
-                            )
+                    if content is not None:
+                        final_content.append(content)
 
                     yield chunk
 
@@ -159,6 +154,7 @@ class WrappedResponses:
                 end_time = time.time()
                 latency = end_time - start_time
                 output = final_content
+
                 await self._capture_streaming_event(
                     posthog_distinct_id,
                     posthog_trace_id,
@@ -182,7 +178,7 @@ class WrappedResponses:
         posthog_privacy_mode: bool,
         posthog_groups: Optional[Dict[str, Any]],
         kwargs: Dict[str, Any],
-        usage_stats: Dict[str, int],
+        usage_stats: TokenUsage,
         latency: float,
         output: Any,
         available_tool_calls: Optional[List[Dict[str, Any]]] = None,
@@ -195,12 +191,14 @@ class WrappedResponses:
             "$ai_model": kwargs.get("model"),
             "$ai_model_parameters": get_model_params(kwargs),
             "$ai_input": with_privacy_mode(
-                self._client._ph_client, posthog_privacy_mode, kwargs.get("input")
+                self._client._ph_client,
+                posthog_privacy_mode,
+                sanitize_openai_response(kwargs.get("input")),
             ),
             "$ai_output_choices": with_privacy_mode(
                 self._client._ph_client,
                 posthog_privacy_mode,
-                output,
+                format_openai_streaming_output(output, "responses"),
             ),
             "$ai_http_status": 200,
             "$ai_input_tokens": usage_stats.get("input_tokens", 0),
@@ -215,6 +213,15 @@ class WrappedResponses:
             **(posthog_properties or {}),
         }
 
+        # Add web search count if present
+        web_search_count = usage_stats.get("web_search_count")
+        if (
+            web_search_count is not None
+            and isinstance(web_search_count, int)
+            and web_search_count > 0
+        ):
+            event_properties["$ai_web_search_count"] = web_search_count
+
         if available_tool_calls:
             event_properties["$ai_tools"] = available_tool_calls
 
@@ -340,8 +347,9 @@ class WrappedCompletions:
         **kwargs: Any,
     ):
         start_time = time.time()
-        usage_stats: Dict[str, int] = {}
+        usage_stats: TokenUsage = TokenUsage()
         accumulated_content = []
+        accumulated_tool_calls: Dict[int, Dict[str, Any]] = {}
 
         if "stream_options" not in kwargs:
             kwargs["stream_options"] = {}
@@ -351,50 +359,40 @@ class WrappedCompletions:
         async def async_generator():
             nonlocal usage_stats
             nonlocal accumulated_content  # noqa: F824
+            nonlocal accumulated_tool_calls
 
             try:
                 async for chunk in response:
-                    if hasattr(chunk, "usage") and chunk.usage:
-                        usage_stats = {
-                            k: getattr(chunk.usage, k, 0)
-                            for k in [
-                                "prompt_tokens",
-                                "completion_tokens",
-                                "total_tokens",
-                            ]
-                        }
+                    # Extract usage stats from chunk
+                    chunk_usage = extract_openai_usage_from_chunk(chunk, "chat")
+                    if chunk_usage:
+                        merge_usage_stats(usage_stats, chunk_usage)
 
-                        # Add support for cached tokens
-                        if hasattr(chunk.usage, "prompt_tokens_details") and hasattr(
-                            chunk.usage.prompt_tokens_details, "cached_tokens"
-                        ):
-                            usage_stats["cache_read_input_tokens"] = (
-                                chunk.usage.prompt_tokens_details.cached_tokens
-                            )
+                    # Extract content from chunk
+                    content = extract_openai_content_from_chunk(chunk, "chat")
+                    if content is not None:
+                        accumulated_content.append(content)
 
-                        if hasattr(chunk.usage, "output_tokens_details") and hasattr(
-                            chunk.usage.output_tokens_details, "reasoning_tokens"
-                        ):
-                            usage_stats["reasoning_tokens"] = (
-                                chunk.usage.output_tokens_details.reasoning_tokens
-                            )
-
-                    if (
-                        hasattr(chunk, "choices")
-                        and chunk.choices
-                        and len(chunk.choices) > 0
-                    ):
-                        if chunk.choices[0].delta and chunk.choices[0].delta.content:
-                            content = chunk.choices[0].delta.content
-                            if content:
-                                accumulated_content.append(content)
+                    # Extract and accumulate tool calls from chunk
+                    chunk_tool_calls = extract_openai_tool_calls_from_chunk(chunk)
+                    if chunk_tool_calls:
+                        accumulate_openai_tool_calls(
+                            accumulated_tool_calls, chunk_tool_calls
+                        )
 
                     yield chunk
 
             finally:
                 end_time = time.time()
                 latency = end_time - start_time
-                output = "".join(accumulated_content)
+
+                # Convert accumulated tool calls dict to list
+                tool_calls_list = (
+                    list(accumulated_tool_calls.values())
+                    if accumulated_tool_calls
+                    else None
+                )
+
                 await self._capture_streaming_event(
                     posthog_distinct_id,
                     posthog_trace_id,
@@ -404,7 +402,8 @@ class WrappedCompletions:
                     kwargs,
                     usage_stats,
                     latency,
-                    output,
+                    accumulated_content,
+                    tool_calls_list,
                     extract_available_tool_calls("openai", kwargs),
                 )
 
@@ -418,9 +417,10 @@ class WrappedCompletions:
         posthog_privacy_mode: bool,
         posthog_groups: Optional[Dict[str, Any]],
         kwargs: Dict[str, Any],
-        usage_stats: Dict[str, int],
+        usage_stats: TokenUsage,
         latency: float,
         output: Any,
+        tool_calls: Optional[List[Dict[str, Any]]] = None,
         available_tool_calls: Optional[List[Dict[str, Any]]] = None,
     ):
         if posthog_trace_id is None:
@@ -431,16 +431,18 @@ class WrappedCompletions:
             "$ai_model": kwargs.get("model"),
             "$ai_model_parameters": get_model_params(kwargs),
             "$ai_input": with_privacy_mode(
-                self._client._ph_client, posthog_privacy_mode, kwargs.get("messages")
+                self._client._ph_client,
+                posthog_privacy_mode,
+                sanitize_openai(kwargs.get("messages")),
             ),
             "$ai_output_choices": with_privacy_mode(
                 self._client._ph_client,
                 posthog_privacy_mode,
-                [{"content": output, "role": "assistant"}],
+                format_openai_streaming_output(output, "chat", tool_calls),
             ),
             "$ai_http_status": 200,
-            "$ai_input_tokens": usage_stats.get("prompt_tokens", 0),
-            "$ai_output_tokens": usage_stats.get("completion_tokens", 0),
+            "$ai_input_tokens": usage_stats.get("input_tokens", 0),
+            "$ai_output_tokens": usage_stats.get("output_tokens", 0),
             "$ai_cache_read_input_tokens": usage_stats.get(
                 "cache_read_input_tokens", 0
             ),
@@ -451,6 +453,16 @@ class WrappedCompletions:
             **(posthog_properties or {}),
         }
 
+        # Add web search count if present
+        web_search_count = usage_stats.get("web_search_count")
+
+        if (
+            web_search_count is not None
+            and isinstance(web_search_count, int)
+            and web_search_count > 0
+        ):
+            event_properties["$ai_web_search_count"] = web_search_count
+
         if available_tool_calls:
             event_properties["$ai_tools"] = available_tool_calls
 
@@ -475,6 +487,7 @@ class WrappedEmbeddings:
 
     def __getattr__(self, name):
         """Fallback to original embeddings object for any methods we don't explicitly handle."""
+
         return getattr(self._original, name)
 
     async def create(
@@ -500,6 +513,7 @@ class WrappedEmbeddings:
         Returns:
             The response from OpenAI's embeddings.create call.
         """
+
         if posthog_trace_id is None:
             posthog_trace_id = str(uuid.uuid4())
 
@@ -508,12 +522,13 @@ class WrappedEmbeddings:
         end_time = time.time()
 
         # Extract usage statistics if available
-        usage_stats = {}
+        usage_stats: TokenUsage = TokenUsage()
+
         if hasattr(response, "usage") and response.usage:
-            usage_stats = {
-                "prompt_tokens": getattr(response.usage, "prompt_tokens", 0),
-                "total_tokens": getattr(response.usage, "total_tokens", 0),
-            }
+            usage_stats = TokenUsage(
+                input_tokens=getattr(response.usage, "prompt_tokens", 0),
+                output_tokens=getattr(response.usage, "completion_tokens", 0),
+            )
 
         latency = end_time - start_time
 
@@ -522,10 +537,12 @@ class WrappedEmbeddings:
             "$ai_provider": "openai",
             "$ai_model": kwargs.get("model"),
             "$ai_input": with_privacy_mode(
-                self._client._ph_client, posthog_privacy_mode, kwargs.get("input")
+                self._client._ph_client,
+                posthog_privacy_mode,
+                sanitize_openai_response(kwargs.get("input")),
             ),
             "$ai_http_status": 200,
-            "$ai_input_tokens": usage_stats.get("prompt_tokens", 0),
+            "$ai_input_tokens": usage_stats.get("input_tokens", 0),
             "$ai_latency": latency,
             "$ai_trace_id": posthog_trace_id,
             "$ai_base_url": str(self._client.base_url),
@@ -556,6 +573,7 @@ class WrappedBeta:
 
     def __getattr__(self, name):
         """Fallback to original beta object for any methods we don't explicitly handle."""
+
         return getattr(self._original, name)
 
     @property
@@ -572,6 +590,7 @@ class WrappedBetaChat:
 
     def __getattr__(self, name):
         """Fallback to original beta chat object for any methods we don't explicitly handle."""
+
         return getattr(self._original, name)
 
     @property
@@ -588,6 +607,7 @@ class WrappedBetaCompletions:
 
     def __getattr__(self, name):
         """Fallback to original beta completions object for any methods we don't explicitly handle."""
+
         return getattr(self._original, name)
 
     async def parse(

posthog/ai/openai/openai_converter.py

@@ -0,0 +1,735 @@
+"""
+OpenAI-specific conversion utilities.
+
+This module handles the conversion of OpenAI API responses and inputs
+into standardized formats for PostHog tracking. It supports both
+Chat Completions API and Responses API formats.
+"""
+
+from typing import Any, Dict, List, Optional
+
+from posthog.ai.types import (
+    FormattedContentItem,
+    FormattedFunctionCall,
+    FormattedImageContent,
+    FormattedMessage,
+    FormattedTextContent,
+    TokenUsage,
+)
+
+
+def format_openai_response(response: Any) -> List[FormattedMessage]:
+    """
+    Format an OpenAI response into standardized message format.
+
+    Handles both Chat Completions API and Responses API formats.
+
+    Args:
+        response: The response object from OpenAI API
+
+    Returns:
+        List of formatted messages with role and content
+    """
+
+    output: List[FormattedMessage] = []
+
+    if response is None:
+        return output
+
+    # Handle Chat Completions response format
+    if hasattr(response, "choices"):
+        content: List[FormattedContentItem] = []
+        role = "assistant"
+
+        for choice in response.choices:
+            if hasattr(choice, "message") and choice.message:
+                if choice.message.role:
+                    role = choice.message.role
+
+                if choice.message.content:
+                    content.append(
+                        {
+                            "type": "text",
+                            "text": choice.message.content,
+                        }
+                    )
+
+                if hasattr(choice.message, "tool_calls") and choice.message.tool_calls:
+                    for tool_call in choice.message.tool_calls:
+                        content.append(
+                            {
+                                "type": "function",
+                                "id": tool_call.id,
+                                "function": {
+                                    "name": tool_call.function.name,
+                                    "arguments": tool_call.function.arguments,
+                                },
+                            }
+                        )
+
+        if content:
+            output.append(
+                {
+                    "role": role,
+                    "content": content,
+                }
+            )
+
+    # Handle Responses API format
+    if hasattr(response, "output"):
+        content = []
+        role = "assistant"
+
+        for item in response.output:
+            if item.type == "message":
+                role = item.role
+
+                if hasattr(item, "content") and isinstance(item.content, list):
+                    for content_item in item.content:
+                        if (
+                            hasattr(content_item, "type")
+                            and content_item.type == "output_text"
+                            and hasattr(content_item, "text")
+                        ):
+                            content.append(
+                                {
+                                    "type": "text",
+                                    "text": content_item.text,
+                                }
+                            )
+
+                        elif hasattr(content_item, "text"):
+                            content.append({"type": "text", "text": content_item.text})
+
+                        elif (
+                            hasattr(content_item, "type")
+                            and content_item.type == "input_image"
+                            and hasattr(content_item, "image_url")
+                        ):
+                            image_content: FormattedImageContent = {
+                                "type": "image",
+                                "image": content_item.image_url,
+                            }
+                            content.append(image_content)
+
+                elif hasattr(item, "content"):
+                    text_content = {"type": "text", "text": str(item.content)}
+                    content.append(text_content)
+
+            elif hasattr(item, "type") and item.type == "function_call":
+                content.append(
+                    {
+                        "type": "function",
+                        "id": getattr(item, "call_id", getattr(item, "id", "")),
+                        "function": {
+                            "name": item.name,
+                            "arguments": getattr(item, "arguments", {}),
+                        },
+                    }
+                )
+
+        if content:
+            output.append(
+                {
+                    "role": role,
+                    "content": content,
+                }
+            )
+
+    return output
+
+
+def format_openai_input(
+    messages: Optional[List[Dict[str, Any]]] = None, input_data: Optional[Any] = None
+) -> List[FormattedMessage]:
+    """
+    Format OpenAI input messages.
+
+    Handles both messages parameter (Chat Completions) and input parameter (Responses API).
+
+    Args:
+        messages: List of message dictionaries for Chat Completions API
+        input_data: Input data for Responses API
+
+    Returns:
+        List of formatted messages
+    """
+
+    formatted_messages: List[FormattedMessage] = []
+
+    # Handle Chat Completions API format
+    if messages is not None:
+        for msg in messages:
+            formatted_messages.append(
+                {
+                    "role": msg.get("role", "user"),
+                    "content": msg.get("content", ""),
+                }
+            )
+
+    # Handle Responses API format
+    if input_data is not None:
+        if isinstance(input_data, list):
+            for item in input_data:
+                role = "user"
+                content = ""
+
+                if isinstance(item, dict):
+                    role = item.get("role", "user")
+                    content = item.get("content", "")
+
+                elif isinstance(item, str):
+                    content = item
+
+                else:
+                    content = str(item)
+
+                formatted_messages.append({"role": role, "content": content})
+
+        elif isinstance(input_data, str):
+            formatted_messages.append({"role": "user", "content": input_data})
+
+        else:
+            formatted_messages.append({"role": "user", "content": str(input_data)})
+
+    return formatted_messages
+
+
+def extract_openai_tools(kwargs: Dict[str, Any]) -> Optional[Any]:
+    """
+    Extract tool definitions from OpenAI API kwargs.
+
+    Args:
+        kwargs: Keyword arguments passed to OpenAI API
+
+    Returns:
+        Tool definitions if present, None otherwise
+    """
+
+    # Check for tools parameter (newer API)
+    if "tools" in kwargs:
+        return kwargs["tools"]
+
+    # Check for functions parameter (older API)
+    if "functions" in kwargs:
+        return kwargs["functions"]
+
+    return None
+
+
+def format_openai_streaming_content(
+    accumulated_content: str, tool_calls: Optional[List[Dict[str, Any]]] = None
+) -> List[FormattedContentItem]:
+    """
+    Format content from OpenAI streaming response.
+
+    Used by streaming handlers to format accumulated content.
+
+    Args:
+        accumulated_content: Accumulated text content from streaming
+        tool_calls: Optional list of tool calls accumulated during streaming
+
+    Returns:
+        List of formatted content items
+    """
+    formatted: List[FormattedContentItem] = []
+
+    # Add text content if present
+    if accumulated_content:
+        text_content: FormattedTextContent = {
+            "type": "text",
+            "text": accumulated_content,
+        }
+        formatted.append(text_content)
+
+    # Add tool calls if present
+    if tool_calls:
+        for tool_call in tool_calls:
+            function_call: FormattedFunctionCall = {
+                "type": "function",
+                "id": tool_call.get("id"),
+                "function": tool_call.get("function", {}),
+            }
+            formatted.append(function_call)
+
+    return formatted
+
+
+def extract_openai_web_search_count(response: Any) -> int:
+    """
+    Extract web search count from OpenAI response.
+
+    Uses a two-tier detection strategy:
+    1. Priority 1 (exact count): Check for output[].type == "web_search_call" (Responses API)
+    2. Priority 2 (binary detection): Check for various web search indicators:
+       - Root-level citations, search_results, or usage.search_context_size (Perplexity)
+       - Annotations with type "url_citation" in choices/output (including delta for streaming)
+
+    Args:
+        response: The response from OpenAI API
+
+    Returns:
+        Number of web search requests (exact count or binary 1/0)
+    """
+
+    # Priority 1: Check for exact count in Responses API output
+    if hasattr(response, "output"):
+        web_search_count = 0
+
+        for item in response.output:
+            if hasattr(item, "type") and item.type == "web_search_call":
+                web_search_count += 1
+
+        web_search_count = max(0, web_search_count)
+
+        if web_search_count > 0:
+            return web_search_count
+
+    # Priority 2: Binary detection (returns 1 or 0)
+
+    # Check root-level indicators (Perplexity)
+    if hasattr(response, "citations"):
+        citations = getattr(response, "citations")
+
+        if citations and len(citations) > 0:
+            return 1
+
+    if hasattr(response, "search_results"):
+        search_results = getattr(response, "search_results")
+
+        if search_results and len(search_results) > 0:
+            return 1
+
+    if hasattr(response, "usage") and hasattr(response.usage, "search_context_size"):
+        if response.usage.search_context_size:
+            return 1
+
+    # Check for url_citation annotations in choices (Chat Completions)
+    if hasattr(response, "choices"):
+        for choice in response.choices:
+            # Check message.annotations (non-streaming or final chunk)
+            if hasattr(choice, "message") and hasattr(choice.message, "annotations"):
+                annotations = choice.message.annotations
+
+                if annotations:
+                    for annotation in annotations:
+                        # Support both dict and object formats
+                        annotation_type = (
+                            annotation.get("type")
+                            if isinstance(annotation, dict)
+                            else getattr(annotation, "type", None)
+                        )
+
+                        if annotation_type == "url_citation":
+                            return 1
+
+            # Check delta.annotations (streaming chunks)
+            if hasattr(choice, "delta") and hasattr(choice.delta, "annotations"):
+                annotations = choice.delta.annotations
+
+                if annotations:
+                    for annotation in annotations:
+                        # Support both dict and object formats
+                        annotation_type = (
+                            annotation.get("type")
+                            if isinstance(annotation, dict)
+                            else getattr(annotation, "type", None)
+                        )
+
+                        if annotation_type == "url_citation":
+                            return 1
+
+    # Check for url_citation annotations in output (Responses API)
+    if hasattr(response, "output"):
+        for item in response.output:
+            if hasattr(item, "content") and isinstance(item.content, list):
+                for content_item in item.content:
+                    if hasattr(content_item, "annotations"):
+                        annotations = content_item.annotations
+
+                        if annotations:
+                            for annotation in annotations:
+                                # Support both dict and object formats
+                                annotation_type = (
+                                    annotation.get("type")
+                                    if isinstance(annotation, dict)
+                                    else getattr(annotation, "type", None)
+                                )
+
+                                if annotation_type == "url_citation":
+                                    return 1
+
+    return 0
+
+
+def extract_openai_usage_from_response(response: Any) -> TokenUsage:
+    """
+    Extract usage statistics from a full OpenAI response (non-streaming).
+    Handles both Chat Completions and Responses API.
+
+    Args:
+        response: The complete response from OpenAI API
+
+    Returns:
+        TokenUsage with standardized usage statistics
+    """
+    if not hasattr(response, "usage"):
+        return TokenUsage(input_tokens=0, output_tokens=0)
+
+    cached_tokens = 0
+    input_tokens = 0
+    output_tokens = 0
+    reasoning_tokens = 0
+
+    # Responses API format
+    if hasattr(response.usage, "input_tokens"):
+        input_tokens = response.usage.input_tokens
+    if hasattr(response.usage, "output_tokens"):
+        output_tokens = response.usage.output_tokens
+    if hasattr(response.usage, "input_tokens_details") and hasattr(
+        response.usage.input_tokens_details, "cached_tokens"
+    ):
+        cached_tokens = response.usage.input_tokens_details.cached_tokens
+    if hasattr(response.usage, "output_tokens_details") and hasattr(
+        response.usage.output_tokens_details, "reasoning_tokens"
+    ):
+        reasoning_tokens = response.usage.output_tokens_details.reasoning_tokens
+
+    # Chat Completions format
+    if hasattr(response.usage, "prompt_tokens"):
+        input_tokens = response.usage.prompt_tokens
+    if hasattr(response.usage, "completion_tokens"):
+        output_tokens = response.usage.completion_tokens
+    if hasattr(response.usage, "prompt_tokens_details") and hasattr(
+        response.usage.prompt_tokens_details, "cached_tokens"
+    ):
+        cached_tokens = response.usage.prompt_tokens_details.cached_tokens
+    if hasattr(response.usage, "completion_tokens_details") and hasattr(
+        response.usage.completion_tokens_details, "reasoning_tokens"
+    ):
+        reasoning_tokens = response.usage.completion_tokens_details.reasoning_tokens
+
+    result = TokenUsage(
+        input_tokens=input_tokens,
+        output_tokens=output_tokens,
+    )
+
+    if cached_tokens > 0:
+        result["cache_read_input_tokens"] = cached_tokens
+    if reasoning_tokens > 0:
+        result["reasoning_tokens"] = reasoning_tokens
+
+    web_search_count = extract_openai_web_search_count(response)
+    if web_search_count > 0:
+        result["web_search_count"] = web_search_count
+
+    return result
+
+
+def extract_openai_usage_from_chunk(
+    chunk: Any, provider_type: str = "chat"
+) -> TokenUsage:
+    """
+    Extract usage statistics from an OpenAI streaming chunk.
+
+    Handles both Chat Completions and Responses API formats.
+
+    Args:
+        chunk: Streaming chunk from OpenAI API
+        provider_type: Either "chat" or "responses" to handle different API formats
+
+    Returns:
+        Dictionary of usage statistics
+    """
+
+    usage: TokenUsage = TokenUsage()
+
+    if provider_type == "chat":
+        # Extract web search count from the chunk before checking for usage
+        # Web search indicators (citations, annotations) can appear on any chunk,
+        # not just those with usage data
+        web_search_count = extract_openai_web_search_count(chunk)
+        if web_search_count > 0:
+            usage["web_search_count"] = web_search_count
+
+        if not hasattr(chunk, "usage") or not chunk.usage:
+            return usage
+
+        # Chat Completions API uses prompt_tokens and completion_tokens
+        # Standardize to input_tokens and output_tokens
+        usage["input_tokens"] = getattr(chunk.usage, "prompt_tokens", 0)
+        usage["output_tokens"] = getattr(chunk.usage, "completion_tokens", 0)
+
+        # Handle cached tokens
+        if hasattr(chunk.usage, "prompt_tokens_details") and hasattr(
+            chunk.usage.prompt_tokens_details, "cached_tokens"
+        ):
+            usage["cache_read_input_tokens"] = (
+                chunk.usage.prompt_tokens_details.cached_tokens
+            )
+
+        # Handle reasoning tokens
+        if hasattr(chunk.usage, "completion_tokens_details") and hasattr(
+            chunk.usage.completion_tokens_details, "reasoning_tokens"
+        ):
+            usage["reasoning_tokens"] = (
+                chunk.usage.completion_tokens_details.reasoning_tokens
+            )
+
+    elif provider_type == "responses":
+        # For Responses API, usage is only in chunk.response.usage for completed events
+        if hasattr(chunk, "type") and chunk.type == "response.completed":
+            if (
+                hasattr(chunk, "response")
+                and hasattr(chunk.response, "usage")
+                and chunk.response.usage
+            ):
+                response_usage = chunk.response.usage
+                usage["input_tokens"] = getattr(response_usage, "input_tokens", 0)
+                usage["output_tokens"] = getattr(response_usage, "output_tokens", 0)
+
+                # Handle cached tokens
+                if hasattr(response_usage, "input_tokens_details") and hasattr(
+                    response_usage.input_tokens_details, "cached_tokens"
+                ):
+                    usage["cache_read_input_tokens"] = (
+                        response_usage.input_tokens_details.cached_tokens
+                    )
+
+                # Handle reasoning tokens
+                if hasattr(response_usage, "output_tokens_details") and hasattr(
+                    response_usage.output_tokens_details, "reasoning_tokens"
+                ):
+                    usage["reasoning_tokens"] = (
+                        response_usage.output_tokens_details.reasoning_tokens
+                    )
+
+                # Extract web search count from the complete response
+                if hasattr(chunk, "response"):
+                    web_search_count = extract_openai_web_search_count(chunk.response)
+                    if web_search_count > 0:
+                        usage["web_search_count"] = web_search_count
+
+    return usage
+
+
+def extract_openai_content_from_chunk(
+    chunk: Any, provider_type: str = "chat"
+) -> Optional[str]:
+    """
+    Extract content from an OpenAI streaming chunk.
+
+    Handles both Chat Completions and Responses API formats.
+
+    Args:
+        chunk: Streaming chunk from OpenAI API
+        provider_type: Either "chat" or "responses" to handle different API formats
+
+    Returns:
+        Text content if present, None otherwise
+    """
+
+    if provider_type == "chat":
+        # Chat Completions API format
+        if (
+            hasattr(chunk, "choices")
+            and chunk.choices
+            and len(chunk.choices) > 0
+            and chunk.choices[0].delta
+            and chunk.choices[0].delta.content
+        ):
+            return chunk.choices[0].delta.content
+
+    elif provider_type == "responses":
+        # Responses API format
+        if hasattr(chunk, "type") and chunk.type == "response.completed":
+            if hasattr(chunk, "response") and chunk.response:
+                res = chunk.response
+                if res.output and len(res.output) > 0:
+                    # Return the full output for responses
+                    return res.output[0]
+
+    return None
+
+
+def extract_openai_tool_calls_from_chunk(chunk: Any) -> Optional[List[Dict[str, Any]]]:
+    """
+    Extract tool calls from an OpenAI streaming chunk.
+
+    Args:
+        chunk: Streaming chunk from OpenAI API
+
+    Returns:
+        List of tool call deltas if present, None otherwise
+    """
+    if (
+        hasattr(chunk, "choices")
+        and chunk.choices
+        and len(chunk.choices) > 0
+        and chunk.choices[0].delta
+        and hasattr(chunk.choices[0].delta, "tool_calls")
+        and chunk.choices[0].delta.tool_calls
+    ):
+        tool_calls = []
+        for tool_call in chunk.choices[0].delta.tool_calls:
+            tc_dict = {
+                "index": getattr(tool_call, "index", None),
+            }
+
+            if hasattr(tool_call, "id") and tool_call.id:
+                tc_dict["id"] = tool_call.id
+
+            if hasattr(tool_call, "type") and tool_call.type:
+                tc_dict["type"] = tool_call.type
+
+            if hasattr(tool_call, "function") and tool_call.function:
+                function_dict = {}
+                if hasattr(tool_call.function, "name") and tool_call.function.name:
+                    function_dict["name"] = tool_call.function.name
+                if (
+                    hasattr(tool_call.function, "arguments")
+                    and tool_call.function.arguments
+                ):
+                    function_dict["arguments"] = tool_call.function.arguments
+                tc_dict["function"] = function_dict
+
+            tool_calls.append(tc_dict)
+        return tool_calls
+
+    return None
+
+
+def accumulate_openai_tool_calls(
+    accumulated_tool_calls: Dict[int, Dict[str, Any]],
+    chunk_tool_calls: List[Dict[str, Any]],
+) -> None:
+    """
+    Accumulate tool calls from streaming chunks.
+
+    OpenAI sends tool calls incrementally:
+    - First chunk has id, type, function.name and partial function.arguments
+    - Subsequent chunks have more function.arguments
+
+    Args:
+        accumulated_tool_calls: Dictionary mapping index to accumulated tool call data
+        chunk_tool_calls: List of tool call deltas from current chunk
+    """
+    for tool_call_delta in chunk_tool_calls:
+        index = tool_call_delta.get("index")
+        if index is None:
+            continue
+
+        # Initialize tool call if first time seeing this index
+        if index not in accumulated_tool_calls:
+            accumulated_tool_calls[index] = {
+                "id": "",
+                "type": "function",
+                "function": {
+                    "name": "",
+                    "arguments": "",
+                },
+            }
+
+        # Update with new data from delta
+        tc = accumulated_tool_calls[index]
+
+        if "id" in tool_call_delta and tool_call_delta["id"]:
+            tc["id"] = tool_call_delta["id"]
+
+        if "type" in tool_call_delta and tool_call_delta["type"]:
+            tc["type"] = tool_call_delta["type"]
+
+        if "function" in tool_call_delta:
+            func_delta = tool_call_delta["function"]
+            if "name" in func_delta and func_delta["name"]:
+                tc["function"]["name"] = func_delta["name"]
+            if "arguments" in func_delta and func_delta["arguments"]:
+                # Arguments are sent incrementally, concatenate them
+                tc["function"]["arguments"] += func_delta["arguments"]
+
+
+def format_openai_streaming_output(
+    accumulated_content: Any,
+    provider_type: str = "chat",
+    tool_calls: Optional[List[Dict[str, Any]]] = None,
+) -> List[FormattedMessage]:
+    """
+    Format the final output from OpenAI streaming.
+
+    Args:
+        accumulated_content: Accumulated content from streaming (string for chat, list for responses)
+        provider_type: Either "chat" or "responses" to handle different API formats
+        tool_calls: Optional list of accumulated tool calls
+
+    Returns:
+        List of formatted messages
+    """
+
+    if provider_type == "chat":
+        content_items: List[FormattedContentItem] = []
+
+        # Add text content if present
+        if isinstance(accumulated_content, str) and accumulated_content:
+            content_items.append({"type": "text", "text": accumulated_content})
+        elif isinstance(accumulated_content, list):
+            # If it's a list of strings, join them
+            text = "".join(str(item) for item in accumulated_content if item)
+            if text:
+                content_items.append({"type": "text", "text": text})
+
+        # Add tool calls if present
+        if tool_calls:
+            for tool_call in tool_calls:
+                if "function" in tool_call:
+                    function_call: FormattedFunctionCall = {
+                        "type": "function",
+                        "id": tool_call.get("id", ""),
+                        "function": tool_call["function"],
+                    }
+                    content_items.append(function_call)
+
+        # Return formatted message with content
+        if content_items:
+            return [{"role": "assistant", "content": content_items}]
+        else:
+            # Empty response
+            return [{"role": "assistant", "content": []}]
+
+    elif provider_type == "responses":
+        # Responses API: accumulated_content is a list of output items
+        if isinstance(accumulated_content, list) and accumulated_content:
+            # The output is already formatted, just return it
+            return accumulated_content
+        elif isinstance(accumulated_content, str):
+            return [
+                {
+                    "role": "assistant",
+                    "content": [{"type": "text", "text": accumulated_content}],
+                }
+            ]
+
+    # Fallback for any other format
+    return [
+        {
+            "role": "assistant",
+            "content": [{"type": "text", "text": str(accumulated_content)}],
+        }
+    ]
+
+
+def format_openai_streaming_input(
+    kwargs: Dict[str, Any], api_type: str = "chat"
+) -> Any:
+    """
+    Format OpenAI streaming input based on API type.
+
+    Args:
+        kwargs: Keyword arguments passed to OpenAI API
+        api_type: Either "chat" or "responses"
+
+    Returns:
+        Formatted input ready for PostHog tracking
+    """
+    from posthog.ai.utils import merge_system_prompt
+
+    return merge_system_prompt(kwargs, "openai")

posthog/ai/sanitization.py

@@ -0,0 +1,226 @@
+import re
+from typing import Any
+from urllib.parse import urlparse
+
+REDACTED_IMAGE_PLACEHOLDER = "[base64 image redacted]"
+
+
+def is_base64_data_url(text: str) -> bool:
+    return re.match(r"^data:([^;]+);base64,", text) is not None
+
+
+def is_valid_url(text: str) -> bool:
+    try:
+        result = urlparse(text)
+        return bool(result.scheme and result.netloc)
+    except Exception:
+        pass
+
+    return text.startswith(("/", "./", "../"))
+
+
+def is_raw_base64(text: str) -> bool:
+    if is_valid_url(text):
+        return False
+
+    return len(text) > 20 and re.match(r"^[A-Za-z0-9+/]+=*$", text) is not None
+
+
+def redact_base64_data_url(value: Any) -> Any:
+    if not isinstance(value, str):
+        return value
+
+    if is_base64_data_url(value):
+        return REDACTED_IMAGE_PLACEHOLDER
+
+    if is_raw_base64(value):
+        return REDACTED_IMAGE_PLACEHOLDER
+
+    return value
+
+
+def process_messages(messages: Any, transform_content_func) -> Any:
+    if not messages:
+        return messages
+
+    def process_content(content: Any) -> Any:
+        if isinstance(content, str):
+            return content
+
+        if not content:
+            return content
+
+        if isinstance(content, list):
+            return [transform_content_func(item) for item in content]
+
+        return transform_content_func(content)
+
+    def process_message(msg: Any) -> Any:
+        if not isinstance(msg, dict) or "content" not in msg:
+            return msg
+        return {**msg, "content": process_content(msg["content"])}
+
+    if isinstance(messages, list):
+        return [process_message(msg) for msg in messages]
+
+    return process_message(messages)
+
+
+def sanitize_openai_image(item: Any) -> Any:
+    if not isinstance(item, dict):
+        return item
+
+    if (
+        item.get("type") == "image_url"
+        and isinstance(item.get("image_url"), dict)
+        and "url" in item["image_url"]
+    ):
+        return {
+            **item,
+            "image_url": {
+                **item["image_url"],
+                "url": redact_base64_data_url(item["image_url"]["url"]),
+            },
+        }
+
+    return item
+
+
+def sanitize_openai_response_image(item: Any) -> Any:
+    if not isinstance(item, dict):
+        return item
+
+    if item.get("type") == "input_image" and "image_url" in item:
+        return {
+            **item,
+            "image_url": redact_base64_data_url(item["image_url"]),
+        }
+
+    return item
+
+
+def sanitize_anthropic_image(item: Any) -> Any:
+    if not isinstance(item, dict):
+        return item
+
+    if (
+        item.get("type") == "image"
+        and isinstance(item.get("source"), dict)
+        and item["source"].get("type") == "base64"
+        and "data" in item["source"]
+    ):
+        # For Anthropic, if the source type is "base64", we should always redact the data
+        # The provider is explicitly telling us this is base64 data
+        return {
+            **item,
+            "source": {
+                **item["source"],
+                "data": REDACTED_IMAGE_PLACEHOLDER,
+            },
+        }
+
+    return item
+
+
+def sanitize_gemini_part(part: Any) -> Any:
+    if not isinstance(part, dict):
+        return part
+
+    if (
+        "inline_data" in part
+        and isinstance(part["inline_data"], dict)
+        and "data" in part["inline_data"]
+    ):
+        # For Gemini, the inline_data structure indicates base64 data
+        # We should redact any string data in this context
+        return {
+            **part,
+            "inline_data": {
+                **part["inline_data"],
+                "data": REDACTED_IMAGE_PLACEHOLDER,
+            },
+        }
+
+    return part
+
+
+def process_gemini_item(item: Any) -> Any:
+    if not isinstance(item, dict):
+        return item
+
+    if "parts" in item and item["parts"]:
+        parts = item["parts"]
+        if isinstance(parts, list):
+            parts = [sanitize_gemini_part(part) for part in parts]
+        else:
+            parts = sanitize_gemini_part(parts)
+
+        return {**item, "parts": parts}
+
+    return item
+
+
+def sanitize_langchain_image(item: Any) -> Any:
+    if not isinstance(item, dict):
+        return item
+
+    if (
+        item.get("type") == "image_url"
+        and isinstance(item.get("image_url"), dict)
+        and "url" in item["image_url"]
+    ):
+        return {
+            **item,
+            "image_url": {
+                **item["image_url"],
+                "url": redact_base64_data_url(item["image_url"]["url"]),
+            },
+        }
+
+    if item.get("type") == "image" and "data" in item:
+        return {**item, "data": redact_base64_data_url(item["data"])}
+
+    if (
+        item.get("type") == "image"
+        and isinstance(item.get("source"), dict)
+        and "data" in item["source"]
+    ):
+        # Anthropic style - raw base64 in structured format, always redact
+        return {
+            **item,
+            "source": {
+                **item["source"],
+                "data": REDACTED_IMAGE_PLACEHOLDER,
+            },
+        }
+
+    if item.get("type") == "media" and "data" in item:
+        return {**item, "data": redact_base64_data_url(item["data"])}
+
+    return item
+
+
+def sanitize_openai(data: Any) -> Any:
+    return process_messages(data, sanitize_openai_image)
+
+
+def sanitize_openai_response(data: Any) -> Any:
+    return process_messages(data, sanitize_openai_response_image)
+
+
+def sanitize_anthropic(data: Any) -> Any:
+    return process_messages(data, sanitize_anthropic_image)
+
+
+def sanitize_gemini(data: Any) -> Any:
+    if not data:
+        return data
+
+    if isinstance(data, list):
+        return [process_gemini_item(item) for item in data]
+
+    return process_gemini_item(data)
+
+
+def sanitize_langchain(data: Any) -> Any:
+    return process_messages(data, sanitize_langchain_image)

posthog/ai/types.py

@@ -0,0 +1,125 @@
+"""
+Common type definitions for PostHog AI SDK.
+
+These types are used for formatting messages and responses across different AI providers
+(Anthropic, OpenAI, Gemini, etc.) to ensure consistency in tracking and data structure.
+"""
+
+from typing import Any, Dict, List, Optional, TypedDict, Union
+
+
+class FormattedTextContent(TypedDict):
+    """Formatted text content item."""
+
+    type: str  # Literal["text"]
+    text: str
+
+
+class FormattedFunctionCall(TypedDict, total=False):
+    """Formatted function/tool call content item."""
+
+    type: str  # Literal["function"]
+    id: Optional[str]
+    function: Dict[str, Any]  # Contains 'name' and 'arguments'
+
+
+class FormattedImageContent(TypedDict):
+    """Formatted image content item."""
+
+    type: str  # Literal["image"]
+    image: str
+
+
+# Union type for all formatted content items
+FormattedContentItem = Union[
+    FormattedTextContent,
+    FormattedFunctionCall,
+    FormattedImageContent,
+    Dict[str, Any],  # Fallback for unknown content types
+]
+
+
+class FormattedMessage(TypedDict):
+    """
+    Standardized message format for PostHog tracking.
+
+    Used across all providers to ensure consistent message structure
+    when sending events to PostHog.
+    """
+
+    role: str
+    content: Union[str, List[FormattedContentItem], Any]
+
+
+class TokenUsage(TypedDict, total=False):
+    """
+    Token usage information for AI model responses.
+
+    Different providers may populate different fields.
+    """
+
+    input_tokens: int
+    output_tokens: int
+    cache_read_input_tokens: Optional[int]
+    cache_creation_input_tokens: Optional[int]
+    reasoning_tokens: Optional[int]
+    web_search_count: Optional[int]
+
+
+class ProviderResponse(TypedDict, total=False):
+    """
+    Standardized provider response format.
+
+    Used for consistent response formatting across all providers.
+    """
+
+    messages: List[FormattedMessage]
+    usage: TokenUsage
+    error: Optional[str]
+
+
+class StreamingContentBlock(TypedDict, total=False):
+    """
+    Content block used during streaming to accumulate content.
+
+    Used for tracking text and function calls as they stream in.
+    """
+
+    type: str  # "text" or "function"
+    text: Optional[str]
+    id: Optional[str]
+    function: Optional[Dict[str, Any]]
+
+
+class ToolInProgress(TypedDict):
+    """
+    Tracks a tool/function call being accumulated during streaming.
+
+    Used by Anthropic to accumulate JSON input for tools.
+    """
+
+    block: StreamingContentBlock
+    input_string: str
+
+
+class StreamingEventData(TypedDict):
+    """
+    Standardized data for streaming events across all providers.
+
+    This type ensures consistent data structure when capturing streaming events,
+    with all provider-specific formatting already completed.
+    """
+
+    provider: str  # "openai", "anthropic", "gemini"
+    model: str
+    base_url: str
+    kwargs: Dict[str, Any]  # Original kwargs for tool extraction and special handling
+    formatted_input: Any  # Provider-formatted input ready for tracking
+    formatted_output: Any  # Provider-formatted output ready for tracking
+    usage_stats: TokenUsage
+    latency: float
+    distinct_id: Optional[str]
+    trace_id: Optional[str]
+    properties: Optional[Dict[str, Any]]
+    privacy_mode: bool
+    groups: Optional[Dict[str, Any]]

posthog/ai/utils.py

@@ -1,10 +1,83 @@
 import time
 import uuid
-from typing import Any, Callable, Dict, List, Optional
-
-from httpx import URL
+from typing import Any, Callable, Dict, List, Optional, cast
 
 from posthog.client import Client as PostHogClient
+from posthog.ai.types import FormattedMessage, StreamingEventData, TokenUsage
+from posthog.ai.sanitization import (
+    sanitize_openai,
+    sanitize_anthropic,
+    sanitize_gemini,
+    sanitize_langchain,
+)
+
+
+def merge_usage_stats(
+    target: TokenUsage, source: TokenUsage, mode: str = "incremental"
+) -> None:
+    """
+    Merge streaming usage statistics into target dict, handling None values.
+
+    Supports two modes:
+    - "incremental": Add source values to target (for APIs that report new tokens)
+    - "cumulative": Replace target with source values (for APIs that report totals)
+
+    Args:
+        target: Dictionary to update with usage stats
+        source: TokenUsage that may contain None values
+        mode: Either "incremental" or "cumulative"
+    """
+    if mode == "incremental":
+        # Add new values to existing totals
+        source_input = source.get("input_tokens")
+        if source_input is not None:
+            current = target.get("input_tokens") or 0
+            target["input_tokens"] = current + source_input
+
+        source_output = source.get("output_tokens")
+        if source_output is not None:
+            current = target.get("output_tokens") or 0
+            target["output_tokens"] = current + source_output
+
+        source_cache_read = source.get("cache_read_input_tokens")
+        if source_cache_read is not None:
+            current = target.get("cache_read_input_tokens") or 0
+            target["cache_read_input_tokens"] = current + source_cache_read
+
+        source_cache_creation = source.get("cache_creation_input_tokens")
+        if source_cache_creation is not None:
+            current = target.get("cache_creation_input_tokens") or 0
+            target["cache_creation_input_tokens"] = current + source_cache_creation
+
+        source_reasoning = source.get("reasoning_tokens")
+        if source_reasoning is not None:
+            current = target.get("reasoning_tokens") or 0
+            target["reasoning_tokens"] = current + source_reasoning
+
+        source_web_search = source.get("web_search_count")
+        if source_web_search is not None:
+            current = target.get("web_search_count") or 0
+            target["web_search_count"] = max(current, source_web_search)
+
+    elif mode == "cumulative":
+        # Replace with latest values (already cumulative)
+        if source.get("input_tokens") is not None:
+            target["input_tokens"] = source["input_tokens"]
+        if source.get("output_tokens") is not None:
+            target["output_tokens"] = source["output_tokens"]
+        if source.get("cache_read_input_tokens") is not None:
+            target["cache_read_input_tokens"] = source["cache_read_input_tokens"]
+        if source.get("cache_creation_input_tokens") is not None:
+            target["cache_creation_input_tokens"] = source[
+                "cache_creation_input_tokens"
+            ]
+        if source.get("reasoning_tokens") is not None:
+            target["reasoning_tokens"] = source["reasoning_tokens"]
+        if source.get("web_search_count") is not None:
+            target["web_search_count"] = source["web_search_count"]
+
+    else:
+        raise ValueError(f"Invalid mode: {mode}. Must be 'incremental' or 'cumulative'")
 
 
 def get_model_params(kwargs: Dict[str, Any]) -> Dict[str, Any]:
@@ -29,349 +102,135 @@ def get_model_params(kwargs: Dict[str, Any]) -> Dict[str, Any]:
     return model_params
 
 
-def get_usage(response, provider: str) -> Dict[str, Any]:
+def get_usage(response, provider: str) -> TokenUsage:
+    """
+    Extract usage statistics from response based on provider.
+    Delegates to provider-specific converter functions.
+    """
     if provider == "anthropic":
-        return {
-            "input_tokens": response.usage.input_tokens,
-            "output_tokens": response.usage.output_tokens,
-            "cache_read_input_tokens": response.usage.cache_read_input_tokens,
-            "cache_creation_input_tokens": response.usage.cache_creation_input_tokens,
-        }
+        from posthog.ai.anthropic.anthropic_converter import (
+            extract_anthropic_usage_from_response,
+        )
+
+        return extract_anthropic_usage_from_response(response)
     elif provider == "openai":
-        cached_tokens = 0
-        input_tokens = 0
-        output_tokens = 0
-        reasoning_tokens = 0
+        from posthog.ai.openai.openai_converter import (
+            extract_openai_usage_from_response,
+        )
 
-        # responses api
-        if hasattr(response.usage, "input_tokens"):
-            input_tokens = response.usage.input_tokens
-        if hasattr(response.usage, "output_tokens"):
-            output_tokens = response.usage.output_tokens
-        if hasattr(response.usage, "input_tokens_details") and hasattr(
-            response.usage.input_tokens_details, "cached_tokens"
-        ):
-            cached_tokens = response.usage.input_tokens_details.cached_tokens
-        if hasattr(response.usage, "output_tokens_details") and hasattr(
-            response.usage.output_tokens_details, "reasoning_tokens"
-        ):
-            reasoning_tokens = response.usage.output_tokens_details.reasoning_tokens
-
-        # chat completions
-        if hasattr(response.usage, "prompt_tokens"):
-            input_tokens = response.usage.prompt_tokens
-        if hasattr(response.usage, "completion_tokens"):
-            output_tokens = response.usage.completion_tokens
-        if hasattr(response.usage, "prompt_tokens_details") and hasattr(
-            response.usage.prompt_tokens_details, "cached_tokens"
-        ):
-            cached_tokens = response.usage.prompt_tokens_details.cached_tokens
-
-        return {
-            "input_tokens": input_tokens,
-            "output_tokens": output_tokens,
-            "cache_read_input_tokens": cached_tokens,
-            "reasoning_tokens": reasoning_tokens,
-        }
+        return extract_openai_usage_from_response(response)
     elif provider == "gemini":
-        input_tokens = 0
-        output_tokens = 0
+        from posthog.ai.gemini.gemini_converter import (
+            extract_gemini_usage_from_response,
+        )
 
-        if hasattr(response, "usage_metadata") and response.usage_metadata:
-            input_tokens = getattr(response.usage_metadata, "prompt_token_count", 0)
-            output_tokens = getattr(
-                response.usage_metadata, "candidates_token_count", 0
-            )
+        return extract_gemini_usage_from_response(response)
 
-        return {
-            "input_tokens": input_tokens,
-            "output_tokens": output_tokens,
-            "cache_read_input_tokens": 0,
-            "cache_creation_input_tokens": 0,
-            "reasoning_tokens": 0,
-        }
-    return {
-        "input_tokens": 0,
-        "output_tokens": 0,
-        "cache_read_input_tokens": 0,
-        "cache_creation_input_tokens": 0,
-        "reasoning_tokens": 0,
-    }
+    return TokenUsage(input_tokens=0, output_tokens=0)
 
 
 def format_response(response, provider: str):
     """
     Format a regular (non-streaming) response.
     """
-    output = []
-    if response is None:
-        return output
     if provider == "anthropic":
-        return format_response_anthropic(response)
+        from posthog.ai.anthropic.anthropic_converter import format_anthropic_response
+
+        return format_anthropic_response(response)
     elif provider == "openai":
-        return format_response_openai(response)
+        from posthog.ai.openai.openai_converter import format_openai_response
+
+        return format_openai_response(response)
     elif provider == "gemini":
-        return format_response_gemini(response)
-    return output
+        from posthog.ai.gemini.gemini_converter import format_gemini_response
 
-
-def format_response_anthropic(response):
-    output = []
-    content = []
-
-    for choice in response.content:
-        if (
-            hasattr(choice, "type")
-            and choice.type == "text"
-            and hasattr(choice, "text")
-            and choice.text
-        ):
-            content.append({"type": "text", "text": choice.text})
-        elif (
-            hasattr(choice, "type")
-            and choice.type == "tool_use"
-            and hasattr(choice, "name")
-            and hasattr(choice, "id")
-        ):
-            tool_call = {
-                "type": "function",
-                "id": choice.id,
-                "function": {
-                    "name": choice.name,
-                    "arguments": getattr(choice, "input", {}),
-                },
-            }
-            content.append(tool_call)
-
-    if content:
-        message = {
-            "role": "assistant",
-            "content": content,
-        }
-        output.append(message)
-
-    return output
-
-
-def format_response_openai(response):
-    output = []
-
-    if hasattr(response, "choices"):
-        content = []
-        role = "assistant"
-
-        for choice in response.choices:
-            # Handle Chat Completions response format
-            if hasattr(choice, "message") and choice.message:
-                if choice.message.role:
-                    role = choice.message.role
-
-                if choice.message.content:
-                    content.append({"type": "text", "text": choice.message.content})
-
-                if hasattr(choice.message, "tool_calls") and choice.message.tool_calls:
-                    for tool_call in choice.message.tool_calls:
-                        content.append(
-                            {
-                                "type": "function",
-                                "id": tool_call.id,
-                                "function": {
-                                    "name": tool_call.function.name,
-                                    "arguments": tool_call.function.arguments,
-                                },
-                            }
-                        )
-
-        if content:
-            message = {
-                "role": role,
-                "content": content,
-            }
-            output.append(message)
-
-    # Handle Responses API format
-    if hasattr(response, "output"):
-        content = []
-        role = "assistant"
-
-        for item in response.output:
-            if item.type == "message":
-                role = item.role
-
-                if hasattr(item, "content") and isinstance(item.content, list):
-                    for content_item in item.content:
-                        if (
-                            hasattr(content_item, "type")
-                            and content_item.type == "output_text"
-                            and hasattr(content_item, "text")
-                        ):
-                            content.append({"type": "text", "text": content_item.text})
-                        elif hasattr(content_item, "text"):
-                            content.append({"type": "text", "text": content_item.text})
-                        elif (
-                            hasattr(content_item, "type")
-                            and content_item.type == "input_image"
-                            and hasattr(content_item, "image_url")
-                        ):
-                            content.append(
-                                {
-                                    "type": "image",
-                                    "image": content_item.image_url,
-                                }
-                            )
-                elif hasattr(item, "content"):
-                    content.append({"type": "text", "text": str(item.content)})
-
-            elif hasattr(item, "type") and item.type == "function_call":
-                content.append(
-                    {
-                        "type": "function",
-                        "id": getattr(item, "call_id", getattr(item, "id", "")),
-                        "function": {
-                            "name": item.name,
-                            "arguments": getattr(item, "arguments", {}),
-                        },
-                    }
-                )
-
-        if content:
-            message = {
-                "role": role,
-                "content": content,
-            }
-            output.append(message)
-
-    return output
-
-
-def format_response_gemini(response):
-    output = []
-
-    if hasattr(response, "candidates") and response.candidates:
-        for candidate in response.candidates:
-            if hasattr(candidate, "content") and candidate.content:
-                content = []
-
-                if hasattr(candidate.content, "parts") and candidate.content.parts:
-                    for part in candidate.content.parts:
-                        if hasattr(part, "text") and part.text:
-                            content.append({"type": "text", "text": part.text})
-                        elif hasattr(part, "function_call") and part.function_call:
-                            function_call = part.function_call
-                            content.append(
-                                {
-                                    "type": "function",
-                                    "function": {
-                                        "name": function_call.name,
-                                        "arguments": function_call.args,
-                                    },
-                                }
-                            )
-
-                if content:
-                    message = {
-                        "role": "assistant",
-                        "content": content,
-                    }
-                    output.append(message)
-
-            elif hasattr(candidate, "text") and candidate.text:
-                output.append(
-                    {
-                        "role": "assistant",
-                        "content": [{"type": "text", "text": candidate.text}],
-                    }
-                )
-    elif hasattr(response, "text") and response.text:
-        output.append(
-            {
-                "role": "assistant",
-                "content": [{"type": "text", "text": response.text}],
-            }
-        )
-
-    return output
+        return format_gemini_response(response)
+    return []
 
 
 def extract_available_tool_calls(provider: str, kwargs: Dict[str, Any]):
+    """
+    Extract available tool calls for the given provider.
+    """
     if provider == "anthropic":
-        if "tools" in kwargs:
-            return kwargs["tools"]
+        from posthog.ai.anthropic.anthropic_converter import extract_anthropic_tools
 
-        return None
+        return extract_anthropic_tools(kwargs)
     elif provider == "gemini":
-        if "config" in kwargs and hasattr(kwargs["config"], "tools"):
-            return kwargs["config"].tools
+        from posthog.ai.gemini.gemini_converter import extract_gemini_tools
 
-        return None
+        return extract_gemini_tools(kwargs)
     elif provider == "openai":
-        if "tools" in kwargs:
-            return kwargs["tools"]
+        from posthog.ai.openai.openai_converter import extract_openai_tools
 
-        return None
+        return extract_openai_tools(kwargs)
+    return None
 
 
-def merge_system_prompt(kwargs: Dict[str, Any], provider: str):
-    messages: List[Dict[str, Any]] = []
+def merge_system_prompt(
+    kwargs: Dict[str, Any], provider: str
+) -> List[FormattedMessage]:
+    """
+    Merge system prompts and format messages for the given provider.
+    """
     if provider == "anthropic":
+        from posthog.ai.anthropic.anthropic_converter import format_anthropic_input
+
         messages = kwargs.get("messages") or []
-        if kwargs.get("system") is None:
-            return messages
-        return [{"role": "system", "content": kwargs.get("system")}] + messages
+        system = kwargs.get("system")
+        return format_anthropic_input(messages, system)
     elif provider == "gemini":
+        from posthog.ai.gemini.gemini_converter import format_gemini_input_with_system
+
         contents = kwargs.get("contents", [])
-        if isinstance(contents, str):
-            return [{"role": "user", "content": contents}]
-        elif isinstance(contents, list):
-            formatted = []
-            for item in contents:
-                if isinstance(item, str):
-                    formatted.append({"role": "user", "content": item})
-                elif hasattr(item, "text"):
-                    formatted.append({"role": "user", "content": item.text})
-                else:
-                    formatted.append({"role": "user", "content": str(item)})
-            return formatted
-        else:
-            return [{"role": "user", "content": str(contents)}]
+        config = kwargs.get("config")
+        return format_gemini_input_with_system(contents, config)
+    elif provider == "openai":
+        from posthog.ai.openai.openai_converter import format_openai_input
 
-    # For OpenAI, handle both Chat Completions and Responses API
-    if kwargs.get("messages") is not None:
-        messages = list(kwargs.get("messages", []))
+        # For OpenAI, handle both Chat Completions and Responses API
+        messages_param = kwargs.get("messages")
+        input_param = kwargs.get("input")
 
-    if kwargs.get("input") is not None:
-        input_data = kwargs.get("input")
-        if isinstance(input_data, list):
-            messages.extend(input_data)
-        else:
-            messages.append({"role": "user", "content": input_data})
+        # Get base formatted messages
+        messages = format_openai_input(messages_param, input_param)
 
-    # Check if system prompt is provided as a separate parameter
-    if kwargs.get("system") is not None:
-        has_system = any(msg.get("role") == "system" for msg in messages)
-        if not has_system:
-            messages = [{"role": "system", "content": kwargs.get("system")}] + messages
+        # Check if system prompt is provided as a separate parameter
+        if kwargs.get("system") is not None:
+            has_system = any(msg.get("role") == "system" for msg in messages)
+            if not has_system:
+                system_msg = cast(
+                    FormattedMessage,
+                    {"role": "system", "content": kwargs.get("system")},
+                )
+                messages = [system_msg] + messages
 
-    # For Responses API, add instructions to the system prompt if provided
-    if kwargs.get("instructions") is not None:
-        # Find the system message if it exists
-        system_idx = next(
-            (i for i, msg in enumerate(messages) if msg.get("role") == "system"), None
-        )
-
-        if system_idx is not None:
-            # Append instructions to existing system message
-            system_content = messages[system_idx].get("content", "")
-            messages[system_idx]["content"] = (
-                f"{system_content}\n\n{kwargs.get('instructions')}"
+        # For Responses API, add instructions to the system prompt if provided
+        if kwargs.get("instructions") is not None:
+            # Find the system message if it exists
+            system_idx = next(
+                (i for i, msg in enumerate(messages) if msg.get("role") == "system"),
+                None,
             )
-        else:
-            # Create a new system message with instructions
-            messages = [
-                {"role": "system", "content": kwargs.get("instructions")}
-            ] + messages
 
-    return messages
+            if system_idx is not None:
+                # Append instructions to existing system message
+                system_content = messages[system_idx].get("content", "")
+                messages[system_idx]["content"] = (
+                    f"{system_content}\n\n{kwargs.get('instructions')}"
+                )
+            else:
+                # Create a new system message with instructions
+                instruction_msg = cast(
+                    FormattedMessage,
+                    {"role": "system", "content": kwargs.get("instructions")},
+                )
+                messages = [instruction_msg] + messages
+
+        return messages
+
+    # Default case - return empty list
+    return []
 
 
 def call_llm_and_track_usage(
@@ -382,7 +241,7 @@ def call_llm_and_track_usage(
     posthog_properties: Optional[Dict[str, Any]],
     posthog_privacy_mode: bool,
     posthog_groups: Optional[Dict[str, Any]],
-    base_url: URL,
+    base_url: str,
     call_method: Callable[..., Any],
     **kwargs: Any,
 ) -> Any:
@@ -394,8 +253,8 @@ def call_llm_and_track_usage(
     response = None
     error = None
     http_status = 200
-    usage: Dict[str, Any] = {}
-    error_params: Dict[str, any] = {}
+    usage: TokenUsage = TokenUsage()
+    error_params: Dict[str, Any] = {}
 
     try:
         response = call_method(**kwargs)
@@ -422,12 +281,15 @@ def call_llm_and_track_usage(
             usage = get_usage(response, provider)
 
         messages = merge_system_prompt(kwargs, provider)
+        sanitized_messages = sanitize_messages(messages, provider)
 
         event_properties = {
             "$ai_provider": provider,
             "$ai_model": kwargs.get("model"),
             "$ai_model_parameters": get_model_params(kwargs),
-            "$ai_input": with_privacy_mode(ph_client, posthog_privacy_mode, messages),
+            "$ai_input": with_privacy_mode(
+                ph_client, posthog_privacy_mode, sanitized_messages
+            ),
             "$ai_output_choices": with_privacy_mode(
                 ph_client, posthog_privacy_mode, format_response(response, provider)
             ),
@@ -446,27 +308,21 @@ def call_llm_and_track_usage(
         if available_tool_calls:
             event_properties["$ai_tools"] = available_tool_calls
 
-        if (
-            usage.get("cache_read_input_tokens") is not None
-            and usage.get("cache_read_input_tokens", 0) > 0
-        ):
-            event_properties["$ai_cache_read_input_tokens"] = usage.get(
-                "cache_read_input_tokens", 0
-            )
+        cache_read = usage.get("cache_read_input_tokens")
+        if cache_read is not None and cache_read > 0:
+            event_properties["$ai_cache_read_input_tokens"] = cache_read
 
-        if (
-            usage.get("cache_creation_input_tokens") is not None
-            and usage.get("cache_creation_input_tokens", 0) > 0
-        ):
-            event_properties["$ai_cache_creation_input_tokens"] = usage.get(
-                "cache_creation_input_tokens", 0
-            )
+        cache_creation = usage.get("cache_creation_input_tokens")
+        if cache_creation is not None and cache_creation > 0:
+            event_properties["$ai_cache_creation_input_tokens"] = cache_creation
 
-        if (
-            usage.get("reasoning_tokens") is not None
-            and usage.get("reasoning_tokens", 0) > 0
-        ):
-            event_properties["$ai_reasoning_tokens"] = usage.get("reasoning_tokens", 0)
+        reasoning = usage.get("reasoning_tokens")
+        if reasoning is not None and reasoning > 0:
+            event_properties["$ai_reasoning_tokens"] = reasoning
+
+        web_search_count = usage.get("web_search_count")
+        if web_search_count is not None and web_search_count > 0:
+            event_properties["$ai_web_search_count"] = web_search_count
 
         if posthog_distinct_id is None:
             event_properties["$process_person_profile"] = False
@@ -500,7 +356,7 @@ async def call_llm_and_track_usage_async(
     posthog_properties: Optional[Dict[str, Any]],
     posthog_privacy_mode: bool,
     posthog_groups: Optional[Dict[str, Any]],
-    base_url: URL,
+    base_url: str,
     call_async_method: Callable[..., Any],
     **kwargs: Any,
 ) -> Any:
@@ -508,8 +364,8 @@ async def call_llm_and_track_usage_async(
     response = None
     error = None
     http_status = 200
-    usage: Dict[str, Any] = {}
-    error_params: Dict[str, any] = {}
+    usage: TokenUsage = TokenUsage()
+    error_params: Dict[str, Any] = {}
 
     try:
         response = await call_async_method(**kwargs)
@@ -536,12 +392,15 @@ async def call_llm_and_track_usage_async(
             usage = get_usage(response, provider)
 
         messages = merge_system_prompt(kwargs, provider)
+        sanitized_messages = sanitize_messages(messages, provider)
 
         event_properties = {
             "$ai_provider": provider,
             "$ai_model": kwargs.get("model"),
             "$ai_model_parameters": get_model_params(kwargs),
-            "$ai_input": with_privacy_mode(ph_client, posthog_privacy_mode, messages),
+            "$ai_input": with_privacy_mode(
+                ph_client, posthog_privacy_mode, sanitized_messages
+            ),
             "$ai_output_choices": with_privacy_mode(
                 ph_client, posthog_privacy_mode, format_response(response, provider)
             ),
@@ -560,21 +419,21 @@ async def call_llm_and_track_usage_async(
         if available_tool_calls:
             event_properties["$ai_tools"] = available_tool_calls
 
-        if (
-            usage.get("cache_read_input_tokens") is not None
-            and usage.get("cache_read_input_tokens", 0) > 0
-        ):
-            event_properties["$ai_cache_read_input_tokens"] = usage.get(
-                "cache_read_input_tokens", 0
-            )
+        cache_read = usage.get("cache_read_input_tokens")
+        if cache_read is not None and cache_read > 0:
+            event_properties["$ai_cache_read_input_tokens"] = cache_read
 
-        if (
-            usage.get("cache_creation_input_tokens") is not None
-            and usage.get("cache_creation_input_tokens", 0) > 0
-        ):
-            event_properties["$ai_cache_creation_input_tokens"] = usage.get(
-                "cache_creation_input_tokens", 0
-            )
+        cache_creation = usage.get("cache_creation_input_tokens")
+        if cache_creation is not None and cache_creation > 0:
+            event_properties["$ai_cache_creation_input_tokens"] = cache_creation
+
+        reasoning = usage.get("reasoning_tokens")
+        if reasoning is not None and reasoning > 0:
+            event_properties["$ai_reasoning_tokens"] = reasoning
+
+        web_search_count = usage.get("web_search_count")
+        if web_search_count is not None and web_search_count > 0:
+            event_properties["$ai_web_search_count"] = web_search_count
 
         if posthog_distinct_id is None:
             event_properties["$process_person_profile"] = False
@@ -600,7 +459,131 @@ async def call_llm_and_track_usage_async(
     return response
 
 
+def sanitize_messages(data: Any, provider: str) -> Any:
+    """Sanitize messages using provider-specific sanitization functions."""
+    if provider == "anthropic":
+        return sanitize_anthropic(data)
+    elif provider == "openai":
+        return sanitize_openai(data)
+    elif provider == "gemini":
+        return sanitize_gemini(data)
+    elif provider == "langchain":
+        return sanitize_langchain(data)
+    return data
+
+
 def with_privacy_mode(ph_client: PostHogClient, privacy_mode: bool, value: Any):
     if ph_client.privacy_mode or privacy_mode:
         return None
     return value
+
+
+def capture_streaming_event(
+    ph_client: PostHogClient,
+    event_data: StreamingEventData,
+):
+    """
+    Unified streaming event capture for all LLM providers.
+
+    This function handles the common logic for capturing streaming events across all providers.
+    All provider-specific formatting should be done BEFORE calling this function.
+
+    The function handles:
+    - Building PostHog event properties
+    - Extracting and adding tools based on provider
+    - Applying privacy mode
+    - Adding special token fields (cache, reasoning)
+    - Provider-specific fields (e.g., OpenAI instructions)
+    - Sending the event to PostHog
+
+    Args:
+        ph_client: PostHog client instance
+        event_data: Standardized streaming event data containing all necessary information
+    """
+    trace_id = event_data.get("trace_id") or str(uuid.uuid4())
+
+    # Build base event properties
+    event_properties = {
+        "$ai_provider": event_data["provider"],
+        "$ai_model": event_data["model"],
+        "$ai_model_parameters": get_model_params(event_data["kwargs"]),
+        "$ai_input": with_privacy_mode(
+            ph_client,
+            event_data["privacy_mode"],
+            event_data["formatted_input"],
+        ),
+        "$ai_output_choices": with_privacy_mode(
+            ph_client,
+            event_data["privacy_mode"],
+            event_data["formatted_output"],
+        ),
+        "$ai_http_status": 200,
+        "$ai_input_tokens": event_data["usage_stats"].get("input_tokens", 0),
+        "$ai_output_tokens": event_data["usage_stats"].get("output_tokens", 0),
+        "$ai_latency": event_data["latency"],
+        "$ai_trace_id": trace_id,
+        "$ai_base_url": str(event_data["base_url"]),
+        **(event_data.get("properties") or {}),
+    }
+
+    # Extract and add tools based on provider
+    available_tools = extract_available_tool_calls(
+        event_data["provider"],
+        event_data["kwargs"],
+    )
+    if available_tools:
+        event_properties["$ai_tools"] = available_tools
+
+    # Add optional token fields
+    # For Anthropic, always include cache fields even if 0 (backward compatibility)
+    # For others, only include if present and non-zero
+    if event_data["provider"] == "anthropic":
+        # Anthropic always includes cache fields
+        cache_read = event_data["usage_stats"].get("cache_read_input_tokens", 0)
+        cache_creation = event_data["usage_stats"].get("cache_creation_input_tokens", 0)
+        event_properties["$ai_cache_read_input_tokens"] = cache_read
+        event_properties["$ai_cache_creation_input_tokens"] = cache_creation
+    else:
+        # Other providers only include if non-zero
+        optional_token_fields = [
+            "cache_read_input_tokens",
+            "cache_creation_input_tokens",
+            "reasoning_tokens",
+        ]
+
+        for field in optional_token_fields:
+            value = event_data["usage_stats"].get(field)
+            if value is not None and isinstance(value, int) and value > 0:
+                event_properties[f"$ai_{field}"] = value
+
+    # Add web search count if present (all providers)
+    web_search_count = event_data["usage_stats"].get("web_search_count")
+    if (
+        web_search_count is not None
+        and isinstance(web_search_count, int)
+        and web_search_count > 0
+    ):
+        event_properties["$ai_web_search_count"] = web_search_count
+
+    # Handle provider-specific fields
+    if (
+        event_data["provider"] == "openai"
+        and event_data["kwargs"].get("instructions") is not None
+    ):
+        event_properties["$ai_instructions"] = with_privacy_mode(
+            ph_client,
+            event_data["privacy_mode"],
+            event_data["kwargs"]["instructions"],
+        )
+
+    if event_data.get("distinct_id") is None:
+        event_properties["$process_person_profile"] = False
+
+    # Send event to PostHog
+    if hasattr(ph_client, "capture"):
+        ph_client.capture(
+            distinct_id=event_data.get("distinct_id") or trace_id,
+            event="$ai_generation",
+            properties=event_properties,
+            groups=event_data.get("groups"),
+        )

posthog/client.py

@@ -19,8 +19,15 @@ from posthog.exception_utils import (
     handle_in_app,
     exception_is_already_captured,
     mark_exception_as_captured,
+    try_attach_code_variables_to_frames,
+    DEFAULT_CODE_VARIABLES_MASK_PATTERNS,
+    DEFAULT_CODE_VARIABLES_IGNORE_PATTERNS,
+)
+from posthog.feature_flags import (
+    InconclusiveMatchError,
+    RequiresServerEvaluation,
+    match_feature_flag_properties,
 )
-from posthog.feature_flags import InconclusiveMatchError, match_feature_flag_properties
 from posthog.poller import Poller
 from posthog.request import (
     DEFAULT_HOST,
@@ -35,6 +42,9 @@ from posthog.contexts import (
     _get_current_context,
     get_context_distinct_id,
     get_context_session_id,
+    get_capture_exception_code_variables_context,
+    get_code_variables_mask_patterns_context,
+    get_code_variables_ignore_patterns_context,
     new_context,
 )
 from posthog.types import (
@@ -44,6 +54,7 @@ from posthog.types import (
     FlagsAndPayloads,
     FlagsResponse,
     FlagValue,
+    SendFeatureFlagsOptions,
     normalize_flags_response,
     to_flags_and_payloads,
     to_payloads,
@@ -55,7 +66,6 @@ from posthog.utils import (
     SizeLimitedDict,
     clean,
     guess_timezone,
-    remove_trailing_slash,
     system_context,
 )
 from posthog.version import VERSION
@@ -87,6 +97,7 @@ def add_context_tags(properties):
     current_context = _get_current_context()
     if current_context:
         context_tags = current_context.collect_tags()
+        properties["$context_tags"] = set(context_tags.keys())
         # We want explicitly passed properties to override context tags
         context_tags.update(properties)
         properties = context_tags
@@ -97,6 +108,34 @@ def add_context_tags(properties):
     return properties
 
 
+def no_throw(default_return=None):
+    """
+    Decorator to prevent raising exceptions from public API methods.
+    Note that this doesn't prevent errors from propagating via `on_error`.
+    Exceptions will still be raised if the debug flag is enabled.
+
+    Args:
+        default_return: Value to return on exception (default: None)
+    """
+
+    def decorator(func):
+        from functools import wraps
+
+        @wraps(func)
+        def wrapper(self, *args, **kwargs):
+            try:
+                return func(self, *args, **kwargs)
+            except Exception as e:
+                if self.debug:
+                    raise e
+                self.log.exception(f"Error in {func.__name__}: {e}")
+                return default_return
+
+        return wrapper
+
+    return decorator
+
+
 class Client(object):
     """
     This is the SDK reference for the PostHog Python SDK.
@@ -145,6 +184,9 @@ class Client(object):
         before_send=None,
         flag_fallback_cache_url=None,
         enable_local_evaluation=True,
+        capture_exception_code_variables=False,
+        code_variables_mask_patterns=None,
+        code_variables_ignore_patterns=None,
     ):
         """
         Initialize a new PostHog client instance.
@@ -190,6 +232,7 @@ class Client(object):
         self.distinct_ids_feature_flags_reported = SizeLimitedDict(MAX_DICT_SIZE, set)
         self.flag_cache = self._initialize_flag_cache(flag_fallback_cache_url)
         self.flag_definition_version = 0
+        self._flags_etag: Optional[str] = None
         self.disabled = disabled
         self.disable_geoip = disable_geoip
         self.historical_migration = historical_migration
@@ -200,6 +243,18 @@ class Client(object):
         self.privacy_mode = privacy_mode
         self.enable_local_evaluation = enable_local_evaluation
 
+        self.capture_exception_code_variables = capture_exception_code_variables
+        self.code_variables_mask_patterns = (
+            code_variables_mask_patterns
+            if code_variables_mask_patterns is not None
+            else DEFAULT_CODE_VARIABLES_MASK_PATTERNS
+        )
+        self.code_variables_ignore_patterns = (
+            code_variables_ignore_patterns
+            if code_variables_ignore_patterns is not None
+            else DEFAULT_CODE_VARIABLES_IGNORE_PATTERNS
+        )
+
         if project_root is None:
             try:
                 project_root = os.getcwd()
@@ -241,8 +296,9 @@ class Client(object):
             # to call flush().
             if send:
                 atexit.register(self.join)
-            for n in range(thread):
-                self.consumers = []
+
+            self.consumers = []
+            for _ in range(thread):
                 consumer = Consumer(
                     self.queue,
                     self.api_key,
@@ -312,6 +368,7 @@ class Client(object):
         person_properties=None,
         group_properties=None,
         disable_geoip=None,
+        flag_keys_to_evaluate: Optional[list[str]] = None,
     ) -> dict[str, Union[bool, str]]:
         """
         Get feature flag variants for a user by calling decide.
@@ -322,12 +379,19 @@ class Client(object):
             person_properties: A dictionary of person properties.
             group_properties: A dictionary of group properties.
             disable_geoip: Whether to disable GeoIP for this request.
+            flag_keys_to_evaluate: A list of specific flag keys to evaluate. If provided,
+                only these flags will be evaluated, improving performance.
 
         Category:
-            Feature Flags
+            Feature flags
         """
         resp_data = self.get_flags_decision(
-            distinct_id, groups, person_properties, group_properties, disable_geoip
+            distinct_id,
+            groups,
+            person_properties,
+            group_properties,
+            disable_geoip,
+            flag_keys_to_evaluate,
         )
         return to_values(resp_data) or {}
 
@@ -338,6 +402,7 @@ class Client(object):
         person_properties=None,
         group_properties=None,
         disable_geoip=None,
+        flag_keys_to_evaluate: Optional[list[str]] = None,
     ) -> dict[str, str]:
         """
         Get feature flag payloads for a user by calling decide.
@@ -348,6 +413,8 @@ class Client(object):
             person_properties: A dictionary of person properties.
             group_properties: A dictionary of group properties.
             disable_geoip: Whether to disable GeoIP for this request.
+            flag_keys_to_evaluate: A list of specific flag keys to evaluate. If provided,
+                only these flags will be evaluated, improving performance.
 
         Examples:
             ```python
@@ -355,10 +422,15 @@ class Client(object):
             ```
 
         Category:
-            Feature Flags
+            Feature flags
         """
         resp_data = self.get_flags_decision(
-            distinct_id, groups, person_properties, group_properties, disable_geoip
+            distinct_id,
+            groups,
+            person_properties,
+            group_properties,
+            disable_geoip,
+            flag_keys_to_evaluate,
         )
         return to_payloads(resp_data) or {}
 
@@ -369,6 +441,7 @@ class Client(object):
         person_properties=None,
         group_properties=None,
         disable_geoip=None,
+        flag_keys_to_evaluate: Optional[list[str]] = None,
     ) -> FlagsAndPayloads:
         """
         Get feature flags and payloads for a user by calling decide.
@@ -379,6 +452,8 @@ class Client(object):
             person_properties: A dictionary of person properties.
             group_properties: A dictionary of group properties.
             disable_geoip: Whether to disable GeoIP for this request.
+            flag_keys_to_evaluate: A list of specific flag keys to evaluate. If provided,
+                only these flags will be evaluated, improving performance.
 
         Examples:
             ```python
@@ -386,10 +461,15 @@ class Client(object):
             ```
 
         Category:
-            Feature Flags
+            Feature flags
         """
         resp = self.get_flags_decision(
-            distinct_id, groups, person_properties, group_properties, disable_geoip
+            distinct_id,
+            groups,
+            person_properties,
+            group_properties,
+            disable_geoip,
+            flag_keys_to_evaluate,
         )
         return to_flags_and_payloads(resp)
 
@@ -400,6 +480,7 @@ class Client(object):
         person_properties=None,
         group_properties=None,
         disable_geoip=None,
+        flag_keys_to_evaluate: Optional[list[str]] = None,
     ) -> FlagsResponse:
         """
         Get feature flags decision.
@@ -410,6 +491,8 @@ class Client(object):
             person_properties: A dictionary of person properties.
             group_properties: A dictionary of group properties.
             disable_geoip: Whether to disable GeoIP for this request.
+            flag_keys_to_evaluate: A list of specific flag keys to evaluate. If provided,
+                only these flags will be evaluated, improving performance.
 
         Examples:
             ```python
@@ -417,7 +500,7 @@ class Client(object):
             ```
 
         Category:
-            Feature Flags
+            Feature flags
         """
         groups = groups or {}
         person_properties = person_properties or {}
@@ -440,6 +523,9 @@ class Client(object):
             "geoip_disable": disable_geoip,
         }
 
+        if flag_keys_to_evaluate:
+            request_data["flag_keys_to_evaluate"] = flag_keys_to_evaluate
+
         resp_data = flags(
             self.api_key,
             self.host,
@@ -449,6 +535,7 @@ class Client(object):
 
         return normalize_flags_response(resp_data)
 
+    @no_throw()
     def capture(
         self, event: str, **kwargs: Unpack[OptionalCaptureArgs]
     ) -> Optional[str]:
@@ -536,7 +623,7 @@ class Client(object):
         if flag_options["should_send"]:
             try:
                 if flag_options["only_evaluate_locally"] is True:
-                    # Only use local evaluation
+                    # Local evaluation explicitly requested
                     feature_variants = self.get_all_flags(
                         distinct_id,
                         groups=(groups or {}),
@@ -544,15 +631,38 @@ class Client(object):
                         group_properties=flag_options["group_properties"],
                         disable_geoip=disable_geoip,
                         only_evaluate_locally=True,
+                        flag_keys_to_evaluate=flag_options["flag_keys_filter"],
                     )
-                else:
-                    # Default behavior - use remote evaluation
+                elif flag_options["only_evaluate_locally"] is False:
+                    # Remote evaluation explicitly requested
                     feature_variants = self.get_feature_variants(
                         distinct_id,
                         groups,
                         person_properties=flag_options["person_properties"],
                         group_properties=flag_options["group_properties"],
                         disable_geoip=disable_geoip,
+                        flag_keys_to_evaluate=flag_options["flag_keys_filter"],
+                    )
+                elif self.feature_flags:
+                    # Local flags available, prefer local evaluation
+                    feature_variants = self.get_all_flags(
+                        distinct_id,
+                        groups=(groups or {}),
+                        person_properties=flag_options["person_properties"],
+                        group_properties=flag_options["group_properties"],
+                        disable_geoip=disable_geoip,
+                        only_evaluate_locally=True,
+                        flag_keys_to_evaluate=flag_options["flag_keys_filter"],
+                    )
+                else:
+                    # Fall back to remote evaluation
+                    feature_variants = self.get_feature_variants(
+                        distinct_id,
+                        groups,
+                        person_properties=flag_options["person_properties"],
+                        group_properties=flag_options["group_properties"],
+                        disable_geoip=disable_geoip,
+                        flag_keys_to_evaluate=flag_options["flag_keys_filter"],
                     )
             except Exception as e:
                 self.log.exception(
@@ -585,7 +695,7 @@ class Client(object):
 
         return self._enqueue(msg, disable_geoip)
 
-    def _parse_send_feature_flags(self, send_feature_flags) -> dict:
+    def _parse_send_feature_flags(self, send_feature_flags) -> SendFeatureFlagsOptions:
         """
         Parse and normalize send_feature_flags parameter into a standard format.
 
@@ -593,8 +703,8 @@ class Client(object):
             send_feature_flags: Either bool or SendFeatureFlagsOptions dict
 
         Returns:
-            dict: Normalized options with keys: should_send, only_evaluate_locally,
-                  person_properties, group_properties
+            SendFeatureFlagsOptions: Normalized options with keys: should_send, only_evaluate_locally,
+                  person_properties, group_properties, flag_keys_filter
 
         Raises:
             TypeError: If send_feature_flags is not bool or dict
@@ -607,6 +717,7 @@ class Client(object):
                 ),
                 "person_properties": send_feature_flags.get("person_properties"),
                 "group_properties": send_feature_flags.get("group_properties"),
+                "flag_keys_filter": send_feature_flags.get("flag_keys_filter"),
             }
         elif isinstance(send_feature_flags, bool):
             return {
@@ -614,6 +725,7 @@ class Client(object):
                 "only_evaluate_locally": None,
                 "person_properties": None,
                 "group_properties": None,
+                "flag_keys_filter": None,
             }
         else:
             raise TypeError(
@@ -621,6 +733,7 @@ class Client(object):
                 f"Expected bool or dict."
             )
 
+    @no_throw()
     def set(self, **kwargs: Unpack[OptionalSetArgs]) -> Optional[str]:
         """
         Set properties on a person profile.
@@ -635,25 +748,13 @@ class Client(object):
         Examples:
             ```python
             # Set with distinct id
-            posthog.capture(
-                'event_name',
-                distinct_id='user-distinct-id',
-                properties={
-                    '$set': {'name': 'Max Hedgehog'},
-                    '$set_once': {'initial_url': '/blog'}
-                }
-            )
-            ```
-            ```python
-            # Set using context
-            from posthog import new_context, identify_context
-            with new_context():
-                identify_context('user-distinct-id')
-                posthog.capture('event_name')
+            posthog.set(distinct_id='user123', properties={'name': 'Max Hedgehog'})
             ```
 
         Category:
             Identification
+
+        Note: This method will not raise exceptions. Errors are logged.
         """
         distinct_id = kwargs.get("distinct_id", None)
         properties = kwargs.get("properties", None)
@@ -680,6 +781,7 @@ class Client(object):
 
         return self._enqueue(msg, disable_geoip)
 
+    @no_throw()
     def set_once(self, **kwargs: Unpack[OptionalSetArgs]) -> Optional[str]:
         """
         Set properties on a person profile only if they haven't been set before.
@@ -698,6 +800,8 @@ class Client(object):
 
         Category:
             Identification
+
+        Note: This method will not raise exceptions. Errors are logged.
         """
         distinct_id = kwargs.get("distinct_id", None)
         properties = kwargs.get("properties", None)
@@ -723,6 +827,7 @@ class Client(object):
 
         return self._enqueue(msg, disable_geoip)
 
+    @no_throw()
     def group_identify(
         self,
         group_type: str,
@@ -755,6 +860,8 @@ class Client(object):
 
         Category:
             Identification
+
+        Note: This method will not raise exceptions. Errors are logged.
         """
         properties = properties or {}
 
@@ -779,6 +886,7 @@ class Client(object):
 
         return self._enqueue(msg, disable_geoip)
 
+    @no_throw()
     def alias(
         self,
         previous_id: str,
@@ -804,6 +912,8 @@ class Client(object):
 
         Category:
             Identification
+
+        Note: This method will not raise exceptions. Errors are logged.
         """
         (distinct_id, personless) = get_identity_state(distinct_id)
 
@@ -891,15 +1001,38 @@ class Client(object):
             all_exceptions_with_trace_and_in_app = event["exception"]["values"]
 
             properties = {
-                "$exception_type": all_exceptions_with_trace_and_in_app[0].get("type"),
-                "$exception_message": all_exceptions_with_trace_and_in_app[0].get(
-                    "value"
-                ),
                 "$exception_list": all_exceptions_with_trace_and_in_app,
-                "$exception_personURL": f"{remove_trailing_slash(self.raw_host)}/project/{self.api_key}/person/{distinct_id}",
                 **properties,
             }
 
+            context_enabled = get_capture_exception_code_variables_context()
+            context_mask = get_code_variables_mask_patterns_context()
+            context_ignore = get_code_variables_ignore_patterns_context()
+
+            enabled = (
+                context_enabled
+                if context_enabled is not None
+                else self.capture_exception_code_variables
+            )
+            mask_patterns = (
+                context_mask
+                if context_mask is not None
+                else self.code_variables_mask_patterns
+            )
+            ignore_patterns = (
+                context_ignore
+                if context_ignore is not None
+                else self.code_variables_ignore_patterns
+            )
+
+            if enabled:
+                try_attach_code_variables_to_frames(
+                    all_exceptions_with_trace_and_in_app,
+                    exc_info,
+                    mask_patterns=mask_patterns,
+                    ignore_patterns=ignore_patterns,
+                )
+
             if self.log_captured_exceptions:
                 self.log.exception(exception, extra=kwargs)
 
@@ -1068,11 +1201,29 @@ class Client(object):
                 f"/api/feature_flag/local_evaluation/?token={self.api_key}&send_cohorts",
                 self.host,
                 timeout=10,
+                etag=self._flags_etag,
             )
 
-            self.feature_flags = response["flags"] or []
-            self.group_type_mapping = response["group_type_mapping"] or {}
-            self.cohorts = response["cohorts"] or {}
+            # Update stored ETag (clear if server stops sending one)
+            self._flags_etag = response.etag
+
+            # If 304 Not Modified, flags haven't changed - skip processing
+            if response.not_modified:
+                self.log.debug(
+                    "[FEATURE FLAGS] Flags not modified (304), using cached data"
+                )
+                self._last_feature_flag_poll = datetime.now(tz=tzutc())
+                return
+
+            if response.data is None:
+                self.log.error(
+                    "[FEATURE FLAGS] Unexpected empty response data in non-304 response"
+                )
+                return
+
+            self.feature_flags = response.data["flags"] or []
+            self.group_type_mapping = response.data["group_type_mapping"] or {}
+            self.cohorts = response.data["cohorts"] or {}
 
             # Check if flag definitions changed and update version
             if self.flag_cache and old_flags_by_key != (
@@ -1133,7 +1284,7 @@ class Client(object):
             ```
 
         Category:
-            Feature Flags
+            Feature flags
         """
         if not self.personal_api_key:
             self.log.warning(
@@ -1168,6 +1319,9 @@ class Client(object):
         person_properties = person_properties or {}
         group_properties = group_properties or {}
 
+        # Create evaluation cache for flag dependencies
+        evaluation_cache: dict[str, Optional[FlagValue]] = {}
+
         if feature_flag.get("ensure_experience_continuity", False):
             raise InconclusiveMatchError("Flag has experience continuity enabled")
 
@@ -1183,12 +1337,12 @@ class Client(object):
                 self.log.warning(
                     f"[FEATURE FLAGS] Unknown group type index {aggregation_group_type_index} for feature flag {feature_flag['key']}"
                 )
-                # failover to `/decide/`
+                # failover to `/flags`
                 raise InconclusiveMatchError("Flag has unknown group type index")
 
             if group_name not in groups:
                 # Group flags are never enabled in `groups` aren't passed in
-                # don't failover to `/decide/`, since response will be the same
+                # don't failover to `/flags`, since response will be the same
                 if warn_on_unknown_groups:
                     self.log.warning(
                         f"[FEATURE FLAGS] Can't compute group feature flag: {feature_flag['key']} without group names passed in"
@@ -1201,11 +1355,20 @@ class Client(object):
 
             focused_group_properties = group_properties[group_name]
             return match_feature_flag_properties(
-                feature_flag, groups[group_name], focused_group_properties
+                feature_flag,
+                groups[group_name],
+                focused_group_properties,
+                self.feature_flags_by_key,
+                evaluation_cache,
             )
         else:
             return match_feature_flag_properties(
-                feature_flag, distinct_id, person_properties, self.cohorts
+                feature_flag,
+                distinct_id,
+                person_properties,
+                self.cohorts,
+                self.feature_flags_by_key,
+                evaluation_cache,
             )
 
     def feature_enabled(
@@ -1243,7 +1406,7 @@ class Client(object):
             ```
 
         Category:
-            Feature Flags
+            Feature flags
         """
         response = self.get_feature_flag(
             key,
@@ -1292,6 +1455,7 @@ class Client(object):
         flag_result = None
         flag_details = None
         request_id = None
+        evaluated_at = None
 
         flag_value = self._locally_evaluate_flag(
             key, distinct_id, groups, person_properties, group_properties
@@ -1316,13 +1480,15 @@ class Client(object):
                 )
         elif not only_evaluate_locally:
             try:
-                flag_details, request_id = self._get_feature_flag_details_from_decide(
-                    key,
-                    distinct_id,
-                    groups,
-                    person_properties,
-                    group_properties,
-                    disable_geoip,
+                flag_details, request_id, evaluated_at = (
+                    self._get_feature_flag_details_from_server(
+                        key,
+                        distinct_id,
+                        groups,
+                        person_properties,
+                        group_properties,
+                        disable_geoip,
+                    )
                 )
                 flag_result = FeatureFlagResult.from_flag_details(
                     flag_details, override_match_value
@@ -1361,6 +1527,7 @@ class Client(object):
                 groups,
                 disable_geoip,
                 request_id,
+                evaluated_at,
                 flag_details,
             )
 
@@ -1451,7 +1618,7 @@ class Client(object):
             ```
 
         Category:
-            Feature Flags
+            Feature flags
         """
         feature_flag_result = self.get_feature_flag_result(
             key,
@@ -1495,7 +1662,7 @@ class Client(object):
                     self.log.debug(
                         f"Successfully computed flag locally: {key} -> {response}"
                     )
-                except InconclusiveMatchError as e:
+                except (RequiresServerEvaluation, InconclusiveMatchError) as e:
                     self.log.debug(f"Failed to compute flag {key} locally: {e}")
                 except Exception as e:
                     self.log.exception(
@@ -1541,7 +1708,7 @@ class Client(object):
             ```
 
         Category:
-            Feature Flags
+            Feature flags
         """
         feature_flag_result = self._get_feature_flag_result(
             key,
@@ -1556,7 +1723,7 @@ class Client(object):
         )
         return feature_flag_result.payload if feature_flag_result else None
 
-    def _get_feature_flag_details_from_decide(
+    def _get_feature_flag_details_from_server(
         self,
         key: str,
         distinct_id: ID_TYPES,
@@ -1564,17 +1731,23 @@ class Client(object):
         person_properties: dict[str, str],
         group_properties: dict[str, str],
         disable_geoip: Optional[bool],
-    ) -> tuple[Optional[FeatureFlag], Optional[str]]:
+    ) -> tuple[Optional[FeatureFlag], Optional[str], Optional[int]]:
         """
-        Calls /decide and returns the flag details and request id
+        Calls /flags and returns the flag details, request id, and evaluated at timestamp
         """
         resp_data = self.get_flags_decision(
-            distinct_id, groups, person_properties, group_properties, disable_geoip
+            distinct_id,
+            groups,
+            person_properties,
+            group_properties,
+            disable_geoip,
+            flag_keys_to_evaluate=[key],
         )
         request_id = resp_data.get("requestId")
+        evaluated_at = resp_data.get("evaluatedAt")
         flags = resp_data.get("flags")
         flag_details = flags.get(key) if flags else None
-        return flag_details, request_id
+        return flag_details, request_id, evaluated_at
 
     def _capture_feature_flag_called(
         self,
@@ -1586,6 +1759,7 @@ class Client(object):
         groups: Dict[str, str],
         disable_geoip: Optional[bool],
         request_id: Optional[str],
+        evaluated_at: Optional[int],
         flag_details: Optional[FeatureFlag],
     ):
         feature_flag_reported_key = (
@@ -1609,6 +1783,8 @@ class Client(object):
 
             if request_id:
                 properties["$feature_flag_request_id"] = request_id
+            if evaluated_at:
+                properties["$feature_flag_evaluated_at"] = evaluated_at
             if isinstance(flag_details, FeatureFlag):
                 if flag_details.reason and flag_details.reason.description:
                     properties["$feature_flag_reason"] = flag_details.reason.description
@@ -1644,6 +1820,7 @@ class Client(object):
         try:
             return remote_config(
                 self.personal_api_key,
+                self.api_key,
                 self.host,
                 key,
                 timeout=self.feature_flags_request_timeout_seconds,
@@ -1684,6 +1861,7 @@ class Client(object):
         group_properties=None,
         only_evaluate_locally=False,
         disable_geoip=None,
+        flag_keys_to_evaluate: Optional[list[str]] = None,
     ) -> Optional[dict[str, Union[bool, str]]]:
         """
         Get all feature flags for a user.
@@ -1695,6 +1873,8 @@ class Client(object):
             group_properties: A dictionary of group properties.
             only_evaluate_locally: Whether to only evaluate locally.
             disable_geoip: Whether to disable GeoIP for this request.
+            flag_keys_to_evaluate: A list of specific flag keys to evaluate. If provided,
+                only these flags will be evaluated, improving performance.
 
         Examples:
             ```python
@@ -1702,7 +1882,7 @@ class Client(object):
             ```
 
         Category:
-            Feature Flags
+            Feature flags
         """
         response = self.get_all_flags_and_payloads(
             distinct_id,
@@ -1711,6 +1891,7 @@ class Client(object):
             group_properties=group_properties,
             only_evaluate_locally=only_evaluate_locally,
             disable_geoip=disable_geoip,
+            flag_keys_to_evaluate=flag_keys_to_evaluate,
         )
 
         return response["featureFlags"]
@@ -1724,6 +1905,7 @@ class Client(object):
         group_properties=None,
         only_evaluate_locally=False,
         disable_geoip=None,
+        flag_keys_to_evaluate: Optional[list[str]] = None,
     ) -> FlagsAndPayloads:
         """
         Get all feature flags and their payloads for a user.
@@ -1735,6 +1917,8 @@ class Client(object):
             group_properties: A dictionary of group properties.
             only_evaluate_locally: Whether to only evaluate locally.
             disable_geoip: Whether to disable GeoIP for this request.
+            flag_keys_to_evaluate: A list of specific flag keys to evaluate. If provided,
+                only these flags will be evaluated, improving performance.
 
         Examples:
             ```python
@@ -1742,7 +1926,7 @@ class Client(object):
             ```
 
         Category:
-            Feature Flags
+            Feature flags
         """
         if self.disabled:
             return {"featureFlags": None, "featureFlagPayloads": None}
@@ -1753,14 +1937,15 @@ class Client(object):
             )
         )
 
-        response, fallback_to_decide = self._get_all_flags_and_payloads_locally(
+        response, fallback_to_flags = self._get_all_flags_and_payloads_locally(
             distinct_id,
             groups=groups,
             person_properties=person_properties,
             group_properties=group_properties,
+            flag_keys_to_evaluate=flag_keys_to_evaluate,
         )
 
-        if fallback_to_decide and not only_evaluate_locally:
+        if fallback_to_flags and not only_evaluate_locally:
             try:
                 decide_response = self.get_flags_decision(
                     distinct_id,
@@ -1768,6 +1953,7 @@ class Client(object):
                     person_properties=person_properties,
                     group_properties=group_properties,
                     disable_geoip=disable_geoip,
+                    flag_keys_to_evaluate=flag_keys_to_evaluate,
                 )
                 return to_flags_and_payloads(decide_response)
             except Exception as e:
@@ -1785,6 +1971,7 @@ class Client(object):
         person_properties=None,
         group_properties=None,
         warn_on_unknown_groups=False,
+        flag_keys_to_evaluate: Optional[list[str]] = None,
     ) -> tuple[FlagsAndPayloads, bool]:
         person_properties = person_properties or {}
         group_properties = group_properties or {}
@@ -1794,10 +1981,18 @@ class Client(object):
 
         flags: dict[str, FlagValue] = {}
         payloads: dict[str, str] = {}
-        fallback_to_decide = False
+        fallback_to_flags = False
         # If loading in previous line failed
         if self.feature_flags:
-            for flag in self.feature_flags:
+            # Filter flags based on flag_keys_to_evaluate if provided
+            flags_to_process = self.feature_flags
+            if flag_keys_to_evaluate:
+                flag_keys_set = set(flag_keys_to_evaluate)
+                flags_to_process = [
+                    flag for flag in self.feature_flags if flag["key"] in flag_keys_set
+                ]
+
+            for flag in flags_to_process:
                 try:
                     flags[flag["key"]] = self._compute_flag_locally(
                         flag,
@@ -1813,20 +2008,20 @@ class Client(object):
                     if matched_payload is not None:
                         payloads[flag["key"]] = matched_payload
                 except InconclusiveMatchError:
-                    # No need to log this, since it's just telling us to fall back to `/decide`
-                    fallback_to_decide = True
+                    # No need to log this, since it's just telling us to fall back to `/flags`
+                    fallback_to_flags = True
                 except Exception as e:
                     self.log.exception(
                         f"[FEATURE FLAGS] Error while computing variant and payload: {e}"
                     )
-                    fallback_to_decide = True
+                    fallback_to_flags = True
         else:
-            fallback_to_decide = True
+            fallback_to_flags = True
 
         return {
             "featureFlags": flags,
             "featureFlagPayloads": payloads,
-        }, fallback_to_decide
+        }, fallback_to_flags
 
     def _initialize_flag_cache(self, cache_url):
         """Initialize feature flag cache for graceful degradation during service outages.
@@ -1937,7 +2132,7 @@ class Client(object):
             for group_name in groups:
                 all_group_properties[group_name] = {
                     "$group_key": groups[group_name],
-                    **(group_properties.get(group_name) or {}),
+                    **((group_properties or {}).get(group_name) or {}),
                 }
 
         return all_person_properties, all_group_properties

posthog/contexts.py

@@ -22,6 +22,9 @@ class ContextScope:
         self.session_id: Optional[str] = None
         self.distinct_id: Optional[str] = None
         self.tags: Dict[str, Any] = {}
+        self.capture_exception_code_variables: Optional[bool] = None
+        self.code_variables_mask_patterns: Optional[list] = None
+        self.code_variables_ignore_patterns: Optional[list] = None
 
     def set_session_id(self, session_id: str):
         self.session_id = session_id
@@ -32,6 +35,15 @@ class ContextScope:
     def add_tag(self, key: str, value: Any):
         self.tags[key] = value
 
+    def set_capture_exception_code_variables(self, enabled: bool):
+        self.capture_exception_code_variables = enabled
+
+    def set_code_variables_mask_patterns(self, mask_patterns: list):
+        self.code_variables_mask_patterns = mask_patterns
+
+    def set_code_variables_ignore_patterns(self, ignore_patterns: list):
+        self.code_variables_ignore_patterns = ignore_patterns
+
     def get_parent(self):
         return self.parent
 
@@ -59,6 +71,27 @@ class ContextScope:
             tags.update(new_tags)
         return tags
 
+    def get_capture_exception_code_variables(self) -> Optional[bool]:
+        if self.capture_exception_code_variables is not None:
+            return self.capture_exception_code_variables
+        if self.parent is not None and not self.fresh:
+            return self.parent.get_capture_exception_code_variables()
+        return None
+
+    def get_code_variables_mask_patterns(self) -> Optional[list]:
+        if self.code_variables_mask_patterns is not None:
+            return self.code_variables_mask_patterns
+        if self.parent is not None and not self.fresh:
+            return self.parent.get_code_variables_mask_patterns()
+        return None
+
+    def get_code_variables_ignore_patterns(self) -> Optional[list]:
+        if self.code_variables_ignore_patterns is not None:
+            return self.code_variables_ignore_patterns
+        if self.parent is not None and not self.fresh:
+            return self.parent.get_code_variables_ignore_patterns()
+        return None
+
 
 _context_stack: contextvars.ContextVar[Optional[ContextScope]] = contextvars.ContextVar(
     "posthog_context_stack", default=None
@@ -243,6 +276,54 @@ def get_context_distinct_id() -> Optional[str]:
     return None
 
 
+def set_capture_exception_code_variables_context(enabled: bool) -> None:
+    """
+    Set whether code variables are captured for the current context.
+    """
+    current_context = _get_current_context()
+    if current_context:
+        current_context.set_capture_exception_code_variables(enabled)
+
+
+def set_code_variables_mask_patterns_context(mask_patterns: list) -> None:
+    """
+    Variable names matching these patterns will be masked with *** when capturing code variables.
+    """
+    current_context = _get_current_context()
+    if current_context:
+        current_context.set_code_variables_mask_patterns(mask_patterns)
+
+
+def set_code_variables_ignore_patterns_context(ignore_patterns: list) -> None:
+    """
+    Variable names matching these patterns will be ignored completely when capturing code variables.
+    """
+    current_context = _get_current_context()
+    if current_context:
+        current_context.set_code_variables_ignore_patterns(ignore_patterns)
+
+
+def get_capture_exception_code_variables_context() -> Optional[bool]:
+    current_context = _get_current_context()
+    if current_context:
+        return current_context.get_capture_exception_code_variables()
+    return None
+
+
+def get_code_variables_mask_patterns_context() -> Optional[list]:
+    current_context = _get_current_context()
+    if current_context:
+        return current_context.get_code_variables_mask_patterns()
+    return None
+
+
+def get_code_variables_ignore_patterns_context() -> Optional[list]:
+    current_context = _get_current_context()
+    if current_context:
+        return current_context.get_code_variables_ignore_patterns()
+    return None
+
+
 F = TypeVar("F", bound=Callable[..., Any])
 
 

posthog/exception_utils.py

@@ -5,6 +5,7 @@
 # 💖open source (under MIT License)
 # We want to keep payloads as similar to Sentry as possible for easy interoperability
 
+import json
 import linecache
 import os
 import re
@@ -26,6 +27,7 @@ from typing import (  # noqa: F401
     Union,
     cast,
     TYPE_CHECKING,
+    Pattern,
 )
 
 from posthog.args import ExcInfo, ExceptionArg  # noqa: F401
@@ -40,6 +42,46 @@ except ImportError:
 
 DEFAULT_MAX_VALUE_LENGTH = 1024
 
+DEFAULT_CODE_VARIABLES_MASK_PATTERNS = [
+    r"(?i).*password.*",
+    r"(?i).*secret.*",
+    r"(?i).*passwd.*",
+    r"(?i).*pwd.*",
+    r"(?i).*api_key.*",
+    r"(?i).*apikey.*",
+    r"(?i).*auth.*",
+    r"(?i).*credentials.*",
+    r"(?i).*privatekey.*",
+    r"(?i).*private_key.*",
+    r"(?i).*token.*",
+    r"(?i).*aws_access_key_id.*",
+    r"(?i).*_pass",
+    r"(?i)sk_.*",
+    r"(?i).*jwt.*",
+]
+
+DEFAULT_CODE_VARIABLES_IGNORE_PATTERNS = [r"^__.*"]
+
+CODE_VARIABLES_REDACTED_VALUE = "$$_posthog_redacted_based_on_masking_rules_$$"
+
+DEFAULT_TOTAL_VARIABLES_SIZE_LIMIT = 20 * 1024
+
+
+class VariableSizeLimiter:
+    def __init__(self, max_size=DEFAULT_TOTAL_VARIABLES_SIZE_LIMIT):
+        self.max_size = max_size
+        self.current_size = 0
+
+    def can_add(self, size):
+        return self.current_size + size <= self.max_size
+
+    def add(self, size):
+        self.current_size += size
+
+    def get_remaining_space(self):
+        return self.max_size - self.current_size
+
+
 LogLevelStr = Literal["fatal", "critical", "error", "warning", "info", "debug"]
 
 Event = TypedDict(
@@ -884,3 +926,209 @@ def strip_string(value, max_length=None):
             "rem": [["!limit", "x", max_length - 3, max_length]],
         },
     )
+
+
+def _compile_patterns(patterns):
+    compiled = []
+    for pattern in patterns:
+        try:
+            compiled.append(re.compile(pattern))
+        except Exception:
+            pass
+    return compiled
+
+
+def _pattern_matches(name, patterns):
+    for pattern in patterns:
+        if pattern.search(name):
+            return True
+    return False
+
+
+def _mask_sensitive_data(value, compiled_mask):
+    if not compiled_mask:
+        return value
+
+    if isinstance(value, dict):
+        result = {}
+        for k, v in value.items():
+            key_str = str(k) if not isinstance(k, str) else k
+            if _pattern_matches(key_str, compiled_mask):
+                result[k] = CODE_VARIABLES_REDACTED_VALUE
+            else:
+                result[k] = _mask_sensitive_data(v, compiled_mask)
+        return result
+    elif isinstance(value, (list, tuple)):
+        masked_items = [_mask_sensitive_data(item, compiled_mask) for item in value]
+        return type(value)(masked_items)
+    elif isinstance(value, str):
+        if _pattern_matches(value, compiled_mask):
+            return CODE_VARIABLES_REDACTED_VALUE
+        return value
+    else:
+        return value
+
+
+def _serialize_variable_value(value, limiter, max_length=1024, compiled_mask=None):
+    try:
+        if value is None:
+            result = "None"
+        elif isinstance(value, bool):
+            result = str(value)
+        elif isinstance(value, (int, float)):
+            result_size = len(str(value))
+            if not limiter.can_add(result_size):
+                return None
+            limiter.add(result_size)
+            return value
+        elif isinstance(value, str):
+            if compiled_mask and _pattern_matches(value, compiled_mask):
+                result = CODE_VARIABLES_REDACTED_VALUE
+            else:
+                result = value
+        else:
+            masked_value = _mask_sensitive_data(value, compiled_mask)
+            result = json.dumps(masked_value)
+
+        if len(result) > max_length:
+            result = result[: max_length - 3] + "..."
+
+        result_size = len(result)
+        if not limiter.can_add(result_size):
+            return None
+        limiter.add(result_size)
+
+        return result
+    except Exception:
+        try:
+            result = repr(value)
+            if len(result) > max_length:
+                result = result[: max_length - 3] + "..."
+
+            result_size = len(result)
+            if not limiter.can_add(result_size):
+                return None
+            limiter.add(result_size)
+            return result
+        except Exception:
+            try:
+                fallback = f"<{type(value).__name__}>"
+                fallback_size = len(fallback)
+                if not limiter.can_add(fallback_size):
+                    return None
+                limiter.add(fallback_size)
+                return fallback
+            except Exception:
+                fallback = "<unserializable object>"
+                fallback_size = len(fallback)
+                if not limiter.can_add(fallback_size):
+                    return None
+                limiter.add(fallback_size)
+                return fallback
+
+
+def _is_simple_type(value):
+    return isinstance(value, (type(None), bool, int, float, str))
+
+
+def serialize_code_variables(
+    frame, limiter, mask_patterns=None, ignore_patterns=None, max_length=1024
+):
+    if mask_patterns is None:
+        mask_patterns = []
+    if ignore_patterns is None:
+        ignore_patterns = []
+
+    compiled_mask = _compile_patterns(mask_patterns)
+    compiled_ignore = _compile_patterns(ignore_patterns)
+
+    try:
+        local_vars = frame.f_locals.copy()
+    except Exception:
+        return {}
+
+    simple_vars = {}
+    complex_vars = {}
+
+    for name, value in local_vars.items():
+        if _pattern_matches(name, compiled_ignore):
+            continue
+
+        if _is_simple_type(value):
+            simple_vars[name] = value
+        else:
+            complex_vars[name] = value
+
+    result = {}
+
+    all_vars = {**simple_vars, **complex_vars}
+    ordered_names = list(sorted(simple_vars.keys())) + list(sorted(complex_vars.keys()))
+
+    for name in ordered_names:
+        value = all_vars[name]
+
+        if _pattern_matches(name, compiled_mask):
+            redacted_value = CODE_VARIABLES_REDACTED_VALUE
+            redacted_size = len(redacted_value)
+            if not limiter.can_add(redacted_size):
+                break
+            limiter.add(redacted_size)
+            result[name] = redacted_value
+        else:
+            serialized = _serialize_variable_value(
+                value, limiter, max_length, compiled_mask
+            )
+            if serialized is None:
+                break
+            result[name] = serialized
+
+    return result
+
+
+def try_attach_code_variables_to_frames(
+    all_exceptions, exc_info, mask_patterns, ignore_patterns
+):
+    try:
+        attach_code_variables_to_frames(
+            all_exceptions, exc_info, mask_patterns, ignore_patterns
+        )
+    except Exception:
+        pass
+
+
+def attach_code_variables_to_frames(
+    all_exceptions, exc_info, mask_patterns, ignore_patterns
+):
+    exc_type, exc_value, traceback = exc_info
+
+    if traceback is None:
+        return
+
+    tb_frames = list(iter_stacks(traceback))
+
+    if not tb_frames:
+        return
+
+    limiter = VariableSizeLimiter()
+
+    for exception in all_exceptions:
+        stacktrace = exception.get("stacktrace")
+        if not stacktrace or "frames" not in stacktrace:
+            continue
+
+        serialized_frames = stacktrace["frames"]
+
+        for serialized_frame, tb_item in zip(serialized_frames, tb_frames):
+            if not serialized_frame.get("in_app"):
+                continue
+
+            variables = serialize_code_variables(
+                tb_item.tb_frame,
+                limiter,
+                mask_patterns=mask_patterns,
+                ignore_patterns=ignore_patterns,
+                max_length=1024,
+            )
+
+            if variables:
+                serialized_frame["code_variables"] = variables

posthog/feature_flags.py

@@ -22,6 +22,18 @@ class InconclusiveMatchError(Exception):
     pass
 
 
+class RequiresServerEvaluation(Exception):
+    """
+    Raised when feature flag evaluation requires server-side data that is not
+    available locally (e.g., static cohorts, experience continuity).
+
+    This error should propagate immediately to trigger API fallback, unlike
+    InconclusiveMatchError which allows trying other conditions.
+    """
+
+    pass
+
+
 # This function takes a distinct_id and a feature flag key and returns a float between 0 and 1.
 # Given the same distinct_id and key, it'll always return the same float. These floats are
 # uniformly distributed between 0 and 1, so if we want to show this feature to 20% of traffic
@@ -55,8 +67,161 @@ def variant_lookup_table(feature_flag):
     return lookup_table
 
 
+def evaluate_flag_dependency(
+    property, flags_by_key, evaluation_cache, distinct_id, properties, cohort_properties
+):
+    """
+    Evaluate a flag dependency property according to the dependency chain algorithm.
+
+    Args:
+        property: Flag property with type="flag" and dependency_chain
+        flags_by_key: Dictionary of all flags by their key
+        evaluation_cache: Cache for storing evaluation results
+        distinct_id: The distinct ID being evaluated
+        properties: Person properties for evaluation
+        cohort_properties: Cohort properties for evaluation
+
+    Returns:
+        bool: True if all dependencies in the chain evaluate to True, False otherwise
+    """
+    if flags_by_key is None or evaluation_cache is None:
+        # Cannot evaluate flag dependencies without required context
+        raise InconclusiveMatchError(
+            f"Cannot evaluate flag dependency on '{property.get('key', 'unknown')}' without flags_by_key and evaluation_cache"
+        )
+
+    # Check if dependency_chain is present - it should always be provided for flag dependencies
+    if "dependency_chain" not in property:
+        # Missing dependency_chain indicates malformed server data
+        raise InconclusiveMatchError(
+            f"Flag dependency property for '{property.get('key', 'unknown')}' is missing required 'dependency_chain' field"
+        )
+
+    dependency_chain = property["dependency_chain"]
+
+    # Handle circular dependency (empty chain means circular)
+    if len(dependency_chain) == 0:
+        log.debug(f"Circular dependency detected for flag: {property.get('key')}")
+        raise InconclusiveMatchError(
+            f"Circular dependency detected for flag '{property.get('key', 'unknown')}'"
+        )
+
+    # Evaluate all dependencies in the chain order
+    for dep_flag_key in dependency_chain:
+        if dep_flag_key not in evaluation_cache:
+            # Need to evaluate this dependency first
+            dep_flag = flags_by_key.get(dep_flag_key)
+            if not dep_flag:
+                # Missing flag dependency - cannot evaluate locally
+                evaluation_cache[dep_flag_key] = None
+                raise InconclusiveMatchError(
+                    f"Cannot evaluate flag dependency '{dep_flag_key}' - flag not found in local flags"
+                )
+            else:
+                # Check if the flag is active (same check as in client._compute_flag_locally)
+                if not dep_flag.get("active"):
+                    evaluation_cache[dep_flag_key] = False
+                else:
+                    # Recursively evaluate the dependency
+                    try:
+                        dep_result = match_feature_flag_properties(
+                            dep_flag,
+                            distinct_id,
+                            properties,
+                            cohort_properties,
+                            flags_by_key,
+                            evaluation_cache,
+                        )
+                        evaluation_cache[dep_flag_key] = dep_result
+                    except InconclusiveMatchError as e:
+                        # If we can't evaluate a dependency, store None and propagate the error
+                        evaluation_cache[dep_flag_key] = None
+                        raise InconclusiveMatchError(
+                            f"Cannot evaluate flag dependency '{dep_flag_key}': {e}"
+                        ) from e
+
+        # Check the cached result
+        cached_result = evaluation_cache[dep_flag_key]
+        if cached_result is None:
+            # Previously inconclusive - raise error again
+            raise InconclusiveMatchError(
+                f"Flag dependency '{dep_flag_key}' was previously inconclusive"
+            )
+        elif not cached_result:
+            # Definitive False result - dependency failed
+            return False
+
+    # All dependencies in the chain have been evaluated successfully
+    # Now check if the final flag value matches the expected value in the property
+    flag_key = property.get("key")
+    expected_value = property.get("value")
+    operator = property.get("operator", "exact")
+
+    if flag_key and expected_value is not None:
+        # Get the actual value of the flag we're checking
+        actual_value = evaluation_cache.get(flag_key)
+
+        if actual_value is None:
+            # Flag wasn't evaluated - this shouldn't happen if dependency chain is correct
+            raise InconclusiveMatchError(
+                f"Flag '{flag_key}' was not evaluated despite being in dependency chain"
+            )
+
+        # For flag dependencies, we need to compare the actual flag result with expected value
+        # using the flag_evaluates_to operator logic
+        if operator == "flag_evaluates_to":
+            return matches_dependency_value(expected_value, actual_value)
+        else:
+            # This should never happen, but just to be defensive.
+            raise InconclusiveMatchError(
+                f"Flag dependency property for '{property.get('key', 'unknown')}' has invalid operator '{operator}'"
+            )
+
+    # If no value check needed, return True (all dependencies passed)
+    return True
+
+
+def matches_dependency_value(expected_value, actual_value):
+    """
+    Check if the actual flag value matches the expected dependency value.
+
+    This follows the same logic as the C# MatchesDependencyValue function:
+    - String variant case: check for exact match or boolean true
+    - Boolean case: must match expected boolean value
+
+    Args:
+        expected_value: The expected value from the property
+        actual_value: The actual value returned by the flag evaluation
+
+    Returns:
+        bool: True if the values match according to flag dependency rules
+    """
+    # String variant case - check for exact match or boolean true
+    if isinstance(actual_value, str) and len(actual_value) > 0:
+        if isinstance(expected_value, bool):
+            # Any variant matches boolean true
+            return expected_value
+        elif isinstance(expected_value, str):
+            # variants are case-sensitive, hence our comparison is too
+            return actual_value == expected_value
+        else:
+            return False
+
+    # Boolean case - must match expected boolean value
+    elif isinstance(actual_value, bool) and isinstance(expected_value, bool):
+        return actual_value == expected_value
+
+    # Default case
+    return False
+
+
 def match_feature_flag_properties(
-    flag, distinct_id, properties, cohort_properties=None
+    flag,
+    distinct_id,
+    properties,
+    cohort_properties=None,
+    flags_by_key=None,
+    evaluation_cache=None,
 ) -> FlagValue:
     flag_conditions = (flag.get("filters") or {}).get("groups") or []
     is_inconclusive = False
@@ -67,19 +232,18 @@ def match_feature_flag_properties(
     ) or []
     valid_variant_keys = [variant["key"] for variant in flag_variants]
 
-    # Stable sort conditions with variant overrides to the top. This ensures that if overrides are present, they are
-    # evaluated first, and the variant override is applied to the first matching condition.
-    sorted_flag_conditions = sorted(
-        flag_conditions,
-        key=lambda condition: 0 if condition.get("variant") else 1,
-    )
-
-    for condition in sorted_flag_conditions:
+    for condition in flag_conditions:
         try:
             # if any one condition resolves to True, we can shortcircuit and return
             # the matching variant
             if is_condition_match(
-                flag, distinct_id, condition, properties, cohort_properties
+                flag,
+                distinct_id,
+                condition,
+                properties,
+                cohort_properties,
+                flags_by_key,
+                evaluation_cache,
             ):
                 variant_override = condition.get("variant")
                 if variant_override and variant_override in valid_variant_keys:
@@ -87,7 +251,12 @@ def match_feature_flag_properties(
                 else:
                     variant = get_matching_variant(flag, distinct_id)
                 return variant or True
+        except RequiresServerEvaluation:
+            # Static cohort or other missing server-side data - must fallback to API
+            raise
         except InconclusiveMatchError:
+            # Evaluation error (bad regex, invalid date, missing property, etc.)
+            # Track that we had an inconclusive match, but try other conditions
             is_inconclusive = True
 
     if is_inconclusive:
@@ -101,22 +270,36 @@ def match_feature_flag_properties(
 
 
 def is_condition_match(
-    feature_flag, distinct_id, condition, properties, cohort_properties
+    feature_flag,
+    distinct_id,
+    condition,
+    properties,
+    cohort_properties,
+    flags_by_key=None,
+    evaluation_cache=None,
 ) -> bool:
     rollout_percentage = condition.get("rollout_percentage")
     if len(condition.get("properties") or []) > 0:
         for prop in condition.get("properties"):
             property_type = prop.get("type")
             if property_type == "cohort":
-                matches = match_cohort(prop, properties, cohort_properties)
+                matches = match_cohort(
+                    prop,
+                    properties,
+                    cohort_properties,
+                    flags_by_key,
+                    evaluation_cache,
+                    distinct_id,
+                )
             elif property_type == "flag":
-                log.warning(
-                    "Flag dependency filters are not supported in local evaluation. "
-                    "Skipping condition for flag '%s' with dependency on flag '%s'",
-                    feature_flag.get("key", "unknown"),
-                    prop.get("key", "unknown"),
+                matches = evaluate_flag_dependency(
+                    prop,
+                    flags_by_key,
+                    evaluation_cache,
+                    distinct_id,
+                    properties,
+                    cohort_properties,
                 )
-                continue
             else:
                 matches = match_property(prop, properties)
             if not matches:
@@ -264,7 +447,14 @@ def match_property(property, property_values) -> bool:
     raise InconclusiveMatchError(f"Unknown operator {operator}")
 
 
-def match_cohort(property, property_values, cohort_properties) -> bool:
+def match_cohort(
+    property,
+    property_values,
+    cohort_properties,
+    flags_by_key=None,
+    evaluation_cache=None,
+    distinct_id=None,
+) -> bool:
     # Cohort properties are in the form of property groups like this:
     # {
     #     "cohort_id": {
@@ -276,15 +466,29 @@ def match_cohort(property, property_values, cohort_properties) -> bool:
     # }
     cohort_id = str(property.get("value"))
     if cohort_id not in cohort_properties:
-        raise InconclusiveMatchError(
-            "can't match cohort without a given cohort property value"
+        raise RequiresServerEvaluation(
+            f"cohort {cohort_id} not found in local cohorts - likely a static cohort that requires server evaluation"
         )
 
     property_group = cohort_properties[cohort_id]
-    return match_property_group(property_group, property_values, cohort_properties)
+    return match_property_group(
+        property_group,
+        property_values,
+        cohort_properties,
+        flags_by_key,
+        evaluation_cache,
+        distinct_id,
+    )
 
 
-def match_property_group(property_group, property_values, cohort_properties) -> bool:
+def match_property_group(
+    property_group,
+    property_values,
+    cohort_properties,
+    flags_by_key=None,
+    evaluation_cache=None,
+    distinct_id=None,
+) -> bool:
     if not property_group:
         return True
 
@@ -301,7 +505,14 @@ def match_property_group(property_group, property_values, cohort_properties) ->
         # a nested property group
         for prop in properties:
             try:
-                matches = match_property_group(prop, property_values, cohort_properties)
+                matches = match_property_group(
+                    prop,
+                    property_values,
+                    cohort_properties,
+                    flags_by_key,
+                    evaluation_cache,
+                    distinct_id,
+                )
                 if property_group_type == "AND":
                     if not matches:
                         return False
@@ -309,6 +520,9 @@ def match_property_group(property_group, property_values, cohort_properties) ->
                     # OR group
                     if matches:
                         return True
+            except RequiresServerEvaluation:
+                # Immediately propagate - this condition requires server-side data
+                raise
             except InconclusiveMatchError as e:
                 log.debug(f"Failed to compute property {prop} locally: {e}")
                 error_matching_locally = True
@@ -324,14 +538,23 @@ def match_property_group(property_group, property_values, cohort_properties) ->
         for prop in properties:
             try:
                 if prop.get("type") == "cohort":
-                    matches = match_cohort(prop, property_values, cohort_properties)
+                    matches = match_cohort(
+                        prop,
+                        property_values,
+                        cohort_properties,
+                        flags_by_key,
+                        evaluation_cache,
+                        distinct_id,
+                    )
                 elif prop.get("type") == "flag":
-                    log.warning(
-                        "Flag dependency filters are not supported in local evaluation. "
-                        "Skipping condition with dependency on flag '%s'",
-                        prop.get("key", "unknown"),
+                    matches = evaluate_flag_dependency(
+                        prop,
+                        flags_by_key,
+                        evaluation_cache,
+                        distinct_id,
+                        property_values,
+                        cohort_properties,
                     )
-                    continue
                 else:
                     matches = match_property(prop, property_values)
 
@@ -349,6 +572,9 @@ def match_property_group(property_group, property_values, cohort_properties) ->
                         return True
                     if not matches and negation:
                         return True
+            except RequiresServerEvaluation:
+                # Immediately propagate - this condition requires server-side data
+                raise
             except InconclusiveMatchError as e:
                 log.debug(f"Failed to compute property {prop} locally: {e}")
                 error_matching_locally = True

posthog/integrations/django.py

@@ -1,10 +1,24 @@
 from typing import TYPE_CHECKING, cast
-from posthog import contexts, capture_exception
+from posthog import contexts
 from posthog.client import Client
 
+try:
+    from asgiref.sync import iscoroutinefunction, markcoroutinefunction
+except ImportError:
+    # Fallback for older Django versions without asgiref
+    import asyncio
+
+    iscoroutinefunction = asyncio.iscoroutinefunction
+
+    # No-op fallback for markcoroutinefunction
+    # Older Django versions without asgiref typically don't support async middleware anyway
+    def markcoroutinefunction(func):
+        return func
+
+
 if TYPE_CHECKING:
     from django.http import HttpRequest, HttpResponse  # noqa: F401
-    from typing import Callable, Dict, Any, Optional  # noqa: F401
+    from typing import Callable, Dict, Any, Optional, Union, Awaitable  # noqa: F401
 
 
 class PosthogContextMiddleware:
@@ -31,11 +45,24 @@ class PosthogContextMiddleware:
     See the context documentation for more information. The extracted distinct ID and session ID, if found, are used to
     associate all events captured in the middleware context with the same distinct ID and session as currently active on the
     frontend. See the documentation for `set_context_session` and `identify_context` for more details.
+
+    This middleware is hybrid-capable: it supports both WSGI (sync) and ASGI (async) Django applications. The middleware
+    detects at initialization whether the next middleware in the chain is async or sync, and adapts its behavior accordingly.
+    This ensures compatibility with both pure sync and pure async middleware chains, as well as mixed chains in ASGI mode.
     """
 
+    sync_capable = True
+    async_capable = True
+
     def __init__(self, get_response):
-        # type: (Callable[[HttpRequest], HttpResponse]) -> None
+        # type: (Union[Callable[[HttpRequest], HttpResponse], Callable[[HttpRequest], Awaitable[HttpResponse]]]) -> None
         self.get_response = get_response
+        self._is_coroutine = iscoroutinefunction(get_response)
+
+        # Mark this instance as a coroutine function if get_response is async
+        # This is required for Django to correctly detect async middleware
+        if self._is_coroutine:
+            markcoroutinefunction(self)
 
         from django.conf import settings
 
@@ -85,9 +112,18 @@ class PosthogContextMiddleware:
 
     def extract_tags(self, request):
         # type: (HttpRequest) -> Dict[str, Any]
-        tags = {}
+        """Extract tags from request in sync context."""
+        user_id, user_email = self.extract_request_user(request)
+        return self._build_tags(request, user_id, user_email)
 
-        (user_id, user_email) = self.extract_request_user(request)
+    def _build_tags(self, request, user_id, user_email):
+        # type: (HttpRequest, Optional[str], Optional[str]) -> Dict[str, Any]
+        """
+        Build tags dict from request and user info.
+
+        Centralized tag extraction logic used by both sync and async paths.
+        """
+        tags = {}
 
         # Extract session ID from X-POSTHOG-SESSION-ID header
         session_id = request.headers.get("X-POSTHOG-SESSION-ID")
@@ -139,43 +175,145 @@ class PosthogContextMiddleware:
         return tags
 
     def extract_request_user(self, request):
+        # type: (HttpRequest) -> tuple[Optional[str], Optional[str]]
+        """Extract user ID and email from request in sync context."""
+        user = getattr(request, "user", None)
+        return self._resolve_user_details(user)
+
+    async def aextract_tags(self, request):
+        # type: (HttpRequest) -> Dict[str, Any]
+        """
+        Async version of extract_tags for use in async request handling.
+
+        Uses await request.auser() instead of request.user to avoid
+        SynchronousOnlyOperation in async context.
+
+        Follows Django's naming convention for async methods (auser, asave, etc.).
+        """
+        user_id, user_email = await self.aextract_request_user(request)
+        return self._build_tags(request, user_id, user_email)
+
+    async def aextract_request_user(self, request):
+        # type: (HttpRequest) -> tuple[Optional[str], Optional[str]]
+        """
+        Async version of extract_request_user for use in async request handling.
+
+        Uses await request.auser() instead of request.user to avoid
+        SynchronousOnlyOperation in async context.
+
+        Follows Django's naming convention for async methods (auser, asave, etc.).
+        """
+        auser = getattr(request, "auser", None)
+        if callable(auser):
+            try:
+                user = await auser()
+                return self._resolve_user_details(user)
+            except Exception:
+                # If auser() fails, return empty - don't break the request
+                # Real errors (permissions, broken auth) will be logged by Django
+                return None, None
+
+        # Fallback for test requests without auser
+        return None, None
+
+    def _resolve_user_details(self, user):
+        # type: (Any) -> tuple[Optional[str], Optional[str]]
+        """
+        Extract user ID and email from a user object.
+
+        Handles both authenticated and unauthenticated users, as well as
+        legacy Django where is_authenticated was a method.
+        """
         user_id = None
         email = None
 
-        user = getattr(request, "user", None)
+        if user is None:
+            return user_id, email
 
-        if user and getattr(user, "is_authenticated", False):
-            try:
-                user_id = str(user.pk)
-            except Exception:
-                pass
+        # Handle is_authenticated (property in modern Django, method in legacy)
+        is_authenticated = getattr(user, "is_authenticated", False)
+        if callable(is_authenticated):
+            is_authenticated = is_authenticated()
 
-            try:
-                email = str(user.email)
-            except Exception:
-                pass
+        if not is_authenticated:
+            return user_id, email
+
+        # Extract user primary key
+        user_pk = getattr(user, "pk", None)
+        if user_pk is not None:
+            user_id = str(user_pk)
+
+        # Extract user email
+        user_email = getattr(user, "email", None)
+        if user_email:
+            email = str(user_email)
 
         return user_id, email
 
     def __call__(self, request):
-        # type: (HttpRequest) -> HttpResponse
+        # type: (HttpRequest) -> Union[HttpResponse, Awaitable[HttpResponse]]
+        """
+        Unified entry point for both sync and async request handling.
+
+        When sync_capable and async_capable are both True, Django passes requests
+        without conversion. This method detects the mode and routes accordingly.
+        """
+        if self._is_coroutine:
+            return self.__acall__(request)
+        else:
+            # Synchronous path
+            if self.request_filter and not self.request_filter(request):
+                return self.get_response(request)
+
+            with contexts.new_context(self.capture_exceptions, client=self.client):
+                for k, v in self.extract_tags(request).items():
+                    contexts.tag(k, v)
+
+                return self.get_response(request)
+
+    async def __acall__(self, request):
+        # type: (HttpRequest) -> Awaitable[HttpResponse]
+        """
+        Asynchronous entry point for async request handling.
+
+        This method is called when the middleware chain is async.
+        Uses aextract_tags() which calls request.auser() to avoid
+        SynchronousOnlyOperation when accessing user in async context.
+        """
         if self.request_filter and not self.request_filter(request):
-            return self.get_response(request)
+            return await self.get_response(request)
 
         with contexts.new_context(self.capture_exceptions, client=self.client):
-            for k, v in self.extract_tags(request).items():
+            for k, v in (await self.aextract_tags(request)).items():
                 contexts.tag(k, v)
 
-            return self.get_response(request)
+            return await self.get_response(request)
 
     def process_exception(self, request, exception):
+        # type: (HttpRequest, Exception) -> None
+        """
+        Process exceptions from views and downstream middleware.
+
+        Django calls this WHILE still inside the context created by __call__,
+        so request tags have already been extracted and set. This method just
+        needs to capture the exception directly.
+
+        Django converts view exceptions into responses before they propagate through
+        the middleware stack, so the context manager in __call__/__acall__ never sees them.
+
+        Note: Django's process_exception is always synchronous, even for async views.
+        """
         if self.request_filter and not self.request_filter(request):
             return
 
         if not self.capture_exceptions:
             return
 
+        # Context and tags already set by __call__ or __acall__
+        # Just capture the exception
         if self.client:
             self.client.capture_exception(exception)
         else:
+            from posthog import capture_exception
+
             capture_exception(exception)

posthog/request.py

@@ -1,28 +1,125 @@
 import json
 import logging
+import re
+import socket
+from dataclasses import dataclass
 from datetime import date, datetime
 from gzip import GzipFile
 from io import BytesIO
-from typing import Any, Optional, Union
+from typing import Any, List, Optional, Tuple, Union
+
 
 import requests
 from dateutil.tz import tzutc
+from requests.adapters import HTTPAdapter  # type: ignore[import-untyped]
+from urllib3.connection import HTTPConnection
 from urllib3.util.retry import Retry
 
 from posthog.utils import remove_trailing_slash
 from posthog.version import VERSION
 
-# Retry on both connect and read errors
-# by default read errors will only retry idempotent HTTP methods (so not POST)
-adapter = requests.adapters.HTTPAdapter(
-    max_retries=Retry(
-        total=2,
-        connect=2,
-        read=2,
+SocketOptions = List[Tuple[int, int, Union[int, bytes]]]
+
+KEEPALIVE_IDLE_SECONDS = 60
+KEEPALIVE_INTERVAL_SECONDS = 60
+KEEPALIVE_PROBE_COUNT = 3
+
+# TCP keepalive probes idle connections to prevent them from being dropped.
+# SO_KEEPALIVE is cross-platform, but timing options vary:
+# - Linux: TCP_KEEPIDLE, TCP_KEEPINTVL, TCP_KEEPCNT
+# - macOS: only SO_KEEPALIVE (uses system defaults)
+# - Windows: TCP_KEEPIDLE, TCP_KEEPINTVL (since Windows 10 1709)
+KEEP_ALIVE_SOCKET_OPTIONS: SocketOptions = list(
+    HTTPConnection.default_socket_options
+) + [
+    (socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1),
+]
+for attr, value in [
+    ("TCP_KEEPIDLE", KEEPALIVE_IDLE_SECONDS),
+    ("TCP_KEEPINTVL", KEEPALIVE_INTERVAL_SECONDS),
+    ("TCP_KEEPCNT", KEEPALIVE_PROBE_COUNT),
+]:
+    if hasattr(socket, attr):
+        KEEP_ALIVE_SOCKET_OPTIONS.append((socket.SOL_TCP, getattr(socket, attr), value))
+
+
+def _mask_tokens_in_url(url: str) -> str:
+    """Mask token values in URLs for safe logging, keeping first 10 chars visible."""
+    return re.sub(r"(token=)([^&]{10})[^&]*", r"\1\2...", url)
+
+
+@dataclass
+class GetResponse:
+    """Response from a GET request with ETag support."""
+
+    data: Any
+    etag: Optional[str] = None
+    not_modified: bool = False
+
+
+class HTTPAdapterWithSocketOptions(HTTPAdapter):
+    """HTTPAdapter with configurable socket options."""
+
+    def __init__(self, *args, socket_options: Optional[SocketOptions] = None, **kwargs):
+        self.socket_options = socket_options
+        super().__init__(*args, **kwargs)
+
+    def init_poolmanager(self, *args, **kwargs):
+        if self.socket_options is not None:
+            kwargs["socket_options"] = self.socket_options
+        super().init_poolmanager(*args, **kwargs)
+
+
+def _build_session(socket_options: Optional[SocketOptions] = None) -> requests.Session:
+    adapter = HTTPAdapterWithSocketOptions(
+        max_retries=Retry(
+            total=2,
+            connect=2,
+            read=2,
+        ),
+        socket_options=socket_options,
     )
-)
-_session = requests.sessions.Session()
-_session.mount("https://", adapter)
+    session = requests.sessions.Session()
+    session.mount("https://", adapter)
+    return session
+
+
+_session = _build_session()
+_socket_options: Optional[SocketOptions] = None
+_pooling_enabled = True
+
+
+def _get_session() -> requests.Session:
+    if _pooling_enabled:
+        return _session
+    return _build_session(_socket_options)
+
+
+def set_socket_options(socket_options: Optional[SocketOptions]) -> None:
+    """
+    Configure socket options for all HTTP connections.
+
+    Example:
+        from posthog import set_socket_options
+        set_socket_options([(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1)])
+    """
+    global _session, _socket_options
+    if socket_options == _socket_options:
+        return
+    _socket_options = socket_options
+    _session = _build_session(socket_options)
+
+
+def enable_keep_alive() -> None:
+    """Enable TCP keepalive to prevent idle connections from being dropped."""
+    set_socket_options(KEEP_ALIVE_SOCKET_OPTIONS)
+
+
+def disable_connection_reuse() -> None:
+    """Disable connection reuse, creating a fresh connection for each request."""
+    global _pooling_enabled
+    _pooling_enabled = False
+
 
 US_INGESTION_ENDPOINT = "https://us.i.posthog.com"
 EU_INGESTION_ENDPOINT = "https://eu.i.posthog.com"
@@ -68,7 +165,7 @@ def post(
             gz.write(data.encode("utf-8"))
         data = buf.getvalue()
 
-    res = _session.post(url, data=data, headers=headers, timeout=timeout)
+    res = _get_session().post(url, data=data, headers=headers, timeout=timeout)
 
     if res.status_code == 200:
         log.debug("data uploaded successfully")
@@ -132,15 +229,20 @@ def flags(
 
 
 def remote_config(
-    personal_api_key: str, host: Optional[str] = None, key: str = "", timeout: int = 15
+    personal_api_key: str,
+    project_api_key: str,
+    host: Optional[str] = None,
+    key: str = "",
+    timeout: int = 15,
 ) -> Any:
     """Get remote config flag value from remote_config API endpoint"""
-    return get(
+    response = get(
         personal_api_key,
-        f"/api/projects/@current/feature_flags/{key}/remote_config/",
+        f"/api/projects/@current/feature_flags/{key}/remote_config?token={project_api_key}",
         host,
         timeout,
     )
+    return response.data
 
 
 def batch_post(
@@ -158,15 +260,42 @@ def batch_post(
 
 
 def get(
-    api_key: str, url: str, host: Optional[str] = None, timeout: Optional[int] = None
-) -> requests.Response:
-    url = remove_trailing_slash(host or DEFAULT_HOST) + url
-    res = requests.get(
-        url,
-        headers={"Authorization": "Bearer %s" % api_key, "User-Agent": USER_AGENT},
-        timeout=timeout,
+    api_key: str,
+    url: str,
+    host: Optional[str] = None,
+    timeout: Optional[int] = None,
+    etag: Optional[str] = None,
+) -> GetResponse:
+    """
+    Make a GET request with optional ETag support.
+
+    If an etag is provided, sends If-None-Match header. Returns GetResponse with:
+    - not_modified=True and data=None if server returns 304
+    - not_modified=False and data=response if server returns 200
+    """
+    log = logging.getLogger("posthog")
+    full_url = remove_trailing_slash(host or DEFAULT_HOST) + url
+    headers = {"Authorization": "Bearer %s" % api_key, "User-Agent": USER_AGENT}
+
+    if etag:
+        headers["If-None-Match"] = etag
+
+    res = _get_session().get(full_url, headers=headers, timeout=timeout)
+
+    masked_url = _mask_tokens_in_url(full_url)
+
+    # Handle 304 Not Modified
+    if res.status_code == 304:
+        log.debug(f"GET {masked_url} returned 304 Not Modified")
+        response_etag = res.headers.get("ETag")
+        return GetResponse(data=None, etag=response_etag or etag, not_modified=True)
+
+    # Handle normal response
+    data = _process_response(
+        res, success_message=f"GET {masked_url} completed successfully"
     )
-    return _process_response(res, success_message=f"GET {url} completed successfully")
+    response_etag = res.headers.get("ETag")
+    return GetResponse(data=data, etag=response_etag, not_modified=False)
 
 
 class APIError(Exception):

posthog/test/ai/anthropic/test_anthropic.py

@@ -1,5 +1,3 @@
-import os
-import time
 from unittest.mock import patch
 
 import pytest
@@ -13,14 +11,95 @@ try:
 except ImportError:
     ANTHROPIC_AVAILABLE = False
 
-ANTHROPIC_API_KEY = os.getenv("ANTHROPIC_API_KEY")
-
 # Skip all tests if Anthropic is not available
 pytestmark = pytest.mark.skipif(
     not ANTHROPIC_AVAILABLE, reason="Anthropic package is not available"
 )
 
 
+# =======================
+# Reusable Mock Helpers
+# =======================
+
+
+class MockContent:
+    """Reusable mock content class for Anthropic responses."""
+
+    def __init__(self, text="Bar", content_type="text"):
+        self.type = content_type
+        self.text = text
+
+
+class MockUsage:
+    """Reusable mock usage class for Anthropic responses."""
+
+    def __init__(
+        self,
+        input_tokens=18,
+        output_tokens=1,
+        cache_read_input_tokens=0,
+        cache_creation_input_tokens=0,
+    ):
+        self.input_tokens = input_tokens
+        self.output_tokens = output_tokens
+        self.cache_read_input_tokens = cache_read_input_tokens
+        self.cache_creation_input_tokens = cache_creation_input_tokens
+
+
+class MockResponse:
+    """Reusable mock response class for Anthropic messages."""
+
+    def __init__(
+        self,
+        content_text="Bar",
+        model="claude-3-opus-20240229",
+        input_tokens=18,
+        output_tokens=1,
+        cache_read=0,
+        cache_creation=0,
+    ):
+        self.content = [MockContent(text=content_text)]
+        self.model = model
+        self.usage = MockUsage(
+            input_tokens=input_tokens,
+            output_tokens=output_tokens,
+            cache_read_input_tokens=cache_read,
+            cache_creation_input_tokens=cache_creation,
+        )
+
+
+def create_mock_response(**kwargs):
+    """Factory function to create mock responses with custom parameters."""
+    return MockResponse(**kwargs)
+
+
+# Streaming mock helpers
+class MockStreamEvent:
+    """Reusable mock event class for streaming responses."""
+
+    def __init__(self, event_type=None, **kwargs):
+        self.type = event_type
+        for key, value in kwargs.items():
+            setattr(self, key, value)
+
+
+class MockContentBlock:
+    """Reusable mock content block for streaming."""
+
+    def __init__(self, block_type, **kwargs):
+        self.type = block_type
+        for key, value in kwargs.items():
+            setattr(self, key, value)
+
+
+class MockDelta:
+    """Reusable mock delta for streaming events."""
+
+    def __init__(self, **kwargs):
+        for key, value in kwargs.items():
+            setattr(self, key, value)
+
+
 @pytest.fixture
 def mock_client():
     with patch("posthog.client.Client") as mock_client:
@@ -46,22 +125,77 @@ def mock_anthropic_response():
 
 
 @pytest.fixture
-def mock_anthropic_stream():
-    class MockStreamEvent:
-        def __init__(self, content, usage=None):
-            self.content = content
-            self.usage = usage
+def mock_anthropic_stream_with_tools():
+    """Mock stream events for tool calls."""
+
+    class MockMessage:
+        def __init__(self):
+            self.usage = MockUsage(
+                input_tokens=50,
+                cache_creation_input_tokens=0,
+                cache_read_input_tokens=5,
+            )
 
     def stream_generator():
-        yield MockStreamEvent("A")
-        yield MockStreamEvent("B")
-        yield MockStreamEvent(
-            "C",
-            usage=Usage(
-                input_tokens=20,
-                output_tokens=10,
-            ),
+        # Message start with usage
+        event = MockStreamEvent("message_start")
+        event.message = MockMessage()
+        yield event
+
+        # Text block start
+        event = MockStreamEvent("content_block_start")
+        event.content_block = MockContentBlock("text")
+        event.index = 0
+        yield event
+
+        # Text delta
+        event = MockStreamEvent("content_block_delta")
+        event.delta = MockDelta(text="I'll check the weather for you.")
+        event.index = 0
+        yield event
+
+        # Text block stop
+        event = MockStreamEvent("content_block_stop")
+        event.index = 0
+        yield event
+
+        # Tool use block start
+        event = MockStreamEvent("content_block_start")
+        event.content_block = MockContentBlock(
+            "tool_use", id="toolu_stream123", name="get_weather"
         )
+        event.index = 1
+        yield event
+
+        # Tool input delta 1
+        event = MockStreamEvent("content_block_delta")
+        event.delta = MockDelta(
+            type="input_json_delta", partial_json='{"location": "San'
+        )
+        event.index = 1
+        yield event
+
+        # Tool input delta 2
+        event = MockStreamEvent("content_block_delta")
+        event.delta = MockDelta(
+            type="input_json_delta", partial_json=' Francisco", "unit": "celsius"}'
+        )
+        event.index = 1
+        yield event
+
+        # Tool block stop
+        event = MockStreamEvent("content_block_stop")
+        event.index = 1
+        yield event
+
+        # Message delta with final usage
+        event = MockStreamEvent("message_delta")
+        event.usage = MockUsage(output_tokens=25)
+        yield event
+
+        # Message stop
+        event = MockStreamEvent("message_stop")
+        yield event
 
     return stream_generator()
 
@@ -174,83 +308,6 @@ def test_basic_completion(mock_client, mock_anthropic_response):
         assert isinstance(props["$ai_latency"], float)
 
 
-def test_streaming(mock_client, mock_anthropic_stream):
-    with patch(
-        "anthropic.resources.Messages.create", return_value=mock_anthropic_stream
-    ):
-        client = Anthropic(api_key="test-key", posthog_client=mock_client)
-        response = client.messages.create(
-            model="claude-3-opus-20240229",
-            messages=[{"role": "user", "content": "Hello"}],
-            stream=True,
-            posthog_distinct_id="test-id",
-            posthog_properties={"foo": "bar"},
-        )
-
-        # Consume the stream
-        chunks = list(response)
-        assert len(chunks) == 3
-        assert chunks[0].content == "A"
-        assert chunks[1].content == "B"
-        assert chunks[2].content == "C"
-
-        # Wait a bit to ensure the capture is called
-        time.sleep(0.1)
-        assert mock_client.capture.call_count == 1
-
-        call_args = mock_client.capture.call_args[1]
-        props = call_args["properties"]
-
-        assert call_args["distinct_id"] == "test-id"
-        assert call_args["event"] == "$ai_generation"
-        assert props["$ai_provider"] == "anthropic"
-        assert props["$ai_model"] == "claude-3-opus-20240229"
-        assert props["$ai_input"] == [{"role": "user", "content": "Hello"}]
-        assert props["$ai_output_choices"] == [{"role": "assistant", "content": "ABC"}]
-        assert props["$ai_input_tokens"] == 20
-        assert props["$ai_output_tokens"] == 10
-        assert isinstance(props["$ai_latency"], float)
-        assert props["foo"] == "bar"
-
-
-def test_streaming_with_stream_endpoint(mock_client, mock_anthropic_stream):
-    with patch(
-        "anthropic.resources.Messages.create", return_value=mock_anthropic_stream
-    ):
-        client = Anthropic(api_key="test-key", posthog_client=mock_client)
-        response = client.messages.stream(
-            model="claude-3-opus-20240229",
-            messages=[{"role": "user", "content": "Hello"}],
-            posthog_distinct_id="test-id",
-            posthog_properties={"foo": "bar"},
-        )
-
-        # Consume the stream
-        chunks = list(response)
-        assert len(chunks) == 3
-        assert chunks[0].content == "A"
-        assert chunks[1].content == "B"
-        assert chunks[2].content == "C"
-
-        # Wait a bit to ensure the capture is called
-        time.sleep(0.1)
-        assert mock_client.capture.call_count == 1
-
-        call_args = mock_client.capture.call_args[1]
-        props = call_args["properties"]
-
-        assert call_args["distinct_id"] == "test-id"
-        assert call_args["event"] == "$ai_generation"
-        assert props["$ai_provider"] == "anthropic"
-        assert props["$ai_model"] == "claude-3-opus-20240229"
-        assert props["$ai_input"] == [{"role": "user", "content": "Hello"}]
-        assert props["$ai_output_choices"] == [{"role": "assistant", "content": "ABC"}]
-        assert props["$ai_input_tokens"] == 20
-        assert props["$ai_output_tokens"] == 10
-        assert isinstance(props["$ai_latency"], float)
-        assert props["foo"] == "bar"
-
-
 def test_groups(mock_client, mock_anthropic_response):
     with patch(
         "anthropic.resources.Messages.create", return_value=mock_anthropic_response
@@ -313,18 +370,23 @@ def test_privacy_mode_global(mock_client, mock_anthropic_response):
         assert props["$ai_output_choices"] is None
 
 
-@pytest.mark.skipif(not ANTHROPIC_API_KEY, reason="ANTHROPIC_API_KEY is not set")
 def test_basic_integration(mock_client):
-    client = Anthropic(posthog_client=mock_client)
-    client.messages.create(
-        model="claude-3-opus-20240229",
-        messages=[{"role": "user", "content": "Foo"}],
-        max_tokens=1,
-        temperature=0,
-        posthog_distinct_id="test-id",
-        posthog_properties={"foo": "bar"},
-        system="You must always answer with 'Bar'.",
-    )
+    """Test basic non-streaming integration."""
+
+    with patch(
+        "anthropic.resources.Messages.create",
+        return_value=create_mock_response(),
+    ):
+        client = Anthropic(posthog_client=mock_client)
+        client.messages.create(
+            model="claude-3-opus-20240229",
+            messages=[{"role": "user", "content": "Foo"}],
+            max_tokens=1,
+            temperature=0,
+            posthog_distinct_id="test-id",
+            posthog_properties={"foo": "bar"},
+            system="You must always answer with 'Bar'.",
+        )
 
     assert mock_client.capture.call_count == 1
 
@@ -349,17 +411,28 @@ def test_basic_integration(mock_client):
     assert isinstance(props["$ai_latency"], float)
 
 
-@pytest.mark.skipif(not ANTHROPIC_API_KEY, reason="ANTHROPIC_API_KEY is not set")
 async def test_basic_async_integration(mock_client):
-    client = AsyncAnthropic(posthog_client=mock_client)
-    await client.messages.create(
-        model="claude-3-opus-20240229",
-        messages=[{"role": "user", "content": "You must always answer with 'Bar'."}],
-        max_tokens=1,
-        temperature=0,
-        posthog_distinct_id="test-id",
-        posthog_properties={"foo": "bar"},
-    )
+    """Test async non-streaming integration."""
+
+    # Make the mock async
+    async def mock_async_create(**kwargs):
+        return create_mock_response(input_tokens=16)
+
+    with patch(
+        "anthropic.resources.messages.AsyncMessages.create",
+        side_effect=mock_async_create,
+    ):
+        client = AsyncAnthropic(posthog_client=mock_client)
+        await client.messages.create(
+            model="claude-3-opus-20240229",
+            messages=[
+                {"role": "user", "content": "You must always answer with 'Bar'."}
+            ],
+            max_tokens=1,
+            temperature=0,
+            posthog_distinct_id="test-id",
+            posthog_properties={"foo": "bar"},
+        )
 
     assert mock_client.capture.call_count == 1
 
@@ -381,52 +454,50 @@ async def test_basic_async_integration(mock_client):
     assert isinstance(props["$ai_latency"], float)
 
 
-def test_streaming_system_prompt(mock_client, mock_anthropic_stream):
+async def test_async_streaming_system_prompt(mock_client):
+    """Test async streaming with system prompt."""
+
+    # Create a simple mock async stream using reusable helpers
+    async def mock_async_stream():
+        # Yield some events
+        yield MockStreamEvent(type="message_start")
+        yield MockStreamEvent(type="content_block_start")
+        yield MockStreamEvent(type="content_block_delta", text="Bar")
+
+        # Final message with usage
+        final_msg = MockStreamEvent(type="message_delta")
+        final_msg.usage = MockUsage(
+            input_tokens=10,
+            output_tokens=5,
+            cache_read_input_tokens=0,
+            cache_creation_input_tokens=0,
+        )
+        yield final_msg
+
+    # Mock create to return a coroutine that yields the async generator
+    # This matches the actual behavior when stream=True with await
+    async def async_create_wrapper(**kwargs):
+        return mock_async_stream()
+
     with patch(
-        "anthropic.resources.Messages.create", return_value=mock_anthropic_stream
+        "anthropic.resources.messages.AsyncMessages.create",
+        side_effect=async_create_wrapper,
     ):
-        client = Anthropic(api_key="test-key", posthog_client=mock_client)
-        response = client.messages.create(
+        client = AsyncAnthropic(posthog_client=mock_client)
+        response = await client.messages.create(
             model="claude-3-opus-20240229",
-            system="Foo",
-            messages=[{"role": "user", "content": "Bar"}],
+            system="You must always answer with 'Bar'.",
+            messages=[{"role": "user", "content": "Foo"}],
             stream=True,
+            max_tokens=1,
         )
 
-        # Consume the stream
-        list(response)
+        # Consume the stream - async finally block completes before this returns
+        [c async for c in response]
 
-        # Wait a bit to ensure the capture is called
-        time.sleep(0.1)
+        # Capture happens in the async finally block before generator completes
         assert mock_client.capture.call_count == 1
 
-        call_args = mock_client.capture.call_args[1]
-        props = call_args["properties"]
-
-        assert props["$ai_input"] == [
-            {"role": "system", "content": "Foo"},
-            {"role": "user", "content": "Bar"},
-        ]
-
-
-@pytest.mark.skipif(not ANTHROPIC_API_KEY, reason="ANTHROPIC_API_KEY is not set")
-async def test_async_streaming_system_prompt(mock_client, mock_anthropic_stream):
-    client = AsyncAnthropic(posthog_client=mock_client)
-    response = await client.messages.create(
-        model="claude-3-opus-20240229",
-        system="You must always answer with 'Bar'.",
-        messages=[{"role": "user", "content": "Foo"}],
-        stream=True,
-        max_tokens=1,
-    )
-
-    # Consume the stream
-    [c async for c in response]
-
-    # Wait a bit to ensure the capture is called
-    time.sleep(0.1)
-    assert mock_client.capture.call_count == 1
-
     call_args = mock_client.capture.call_args[1]
     props = call_args["properties"]
 
@@ -746,3 +817,449 @@ def test_async_tool_calls_in_output_choices(
         assert props["$ai_input_tokens"] == 25
         assert props["$ai_output_tokens"] == 15
         assert props["$ai_http_status"] == 200
+
+
+def test_streaming_with_tool_calls(mock_client, mock_anthropic_stream_with_tools):
+    """Test that tool calls are properly captured in streaming mode."""
+    with patch(
+        "anthropic.resources.Messages.create",
+        return_value=mock_anthropic_stream_with_tools,
+    ):
+        client = Anthropic(api_key="test-key", posthog_client=mock_client)
+        response = client.messages.create(
+            model="claude-3-5-sonnet-20241022",
+            system="You are a helpful weather assistant.",
+            messages=[
+                {"role": "user", "content": "What's the weather in San Francisco?"}
+            ],
+            tools=[
+                {
+                    "name": "get_weather",
+                    "description": "Get weather information",
+                    "input_schema": {
+                        "type": "object",
+                        "properties": {
+                            "location": {"type": "string"},
+                            "unit": {"type": "string"},
+                        },
+                        "required": ["location"],
+                    },
+                }
+            ],
+            stream=True,
+            posthog_distinct_id="test-id",
+        )
+
+        # Consume the stream - this triggers the finally block synchronously
+        list(response)
+
+        # Capture happens synchronously when generator is exhausted
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        assert call_args["distinct_id"] == "test-id"
+        assert call_args["event"] == "$ai_generation"
+        assert props["$ai_provider"] == "anthropic"
+        assert props["$ai_model"] == "claude-3-5-sonnet-20241022"
+
+        # Verify system prompt is included in input
+        assert props["$ai_input"] == [
+            {"role": "system", "content": "You are a helpful weather assistant."},
+            {"role": "user", "content": "What's the weather in San Francisco?"},
+        ]
+
+        # Verify that tools are captured in the properties
+        assert props["$ai_tools"] == [
+            {
+                "name": "get_weather",
+                "description": "Get weather information",
+                "input_schema": {
+                    "type": "object",
+                    "properties": {
+                        "location": {"type": "string"},
+                        "unit": {"type": "string"},
+                    },
+                    "required": ["location"],
+                },
+            }
+        ]
+
+        # Verify output contains both text and tool call
+        output_choices = props["$ai_output_choices"]
+        assert len(output_choices) == 1
+
+        assistant_message = output_choices[0]
+        assert assistant_message["role"] == "assistant"
+
+        content = assistant_message["content"]
+        assert isinstance(content, list)
+        assert len(content) == 2
+
+        # Verify text block
+        text_block = content[0]
+        assert text_block["type"] == "text"
+        assert text_block["text"] == "I'll check the weather for you."
+
+        # Verify tool call block
+        tool_block = content[1]
+        assert tool_block["type"] == "function"
+        assert tool_block["id"] == "toolu_stream123"
+        assert tool_block["function"]["name"] == "get_weather"
+        assert tool_block["function"]["arguments"] == {
+            "location": "San Francisco",
+            "unit": "celsius",
+        }
+
+        # Check token usage
+        assert props["$ai_input_tokens"] == 50
+        assert props["$ai_output_tokens"] == 25
+        assert props["$ai_cache_read_input_tokens"] == 5
+        assert props["$ai_cache_creation_input_tokens"] == 0
+
+
+def test_async_streaming_with_tool_calls(mock_client, mock_anthropic_stream_with_tools):
+    """Test that tool calls are properly captured in async streaming mode."""
+    import asyncio
+
+    async def mock_async_generator():
+        # Convert regular generator to async generator
+        for event in mock_anthropic_stream_with_tools:
+            yield event
+
+    async def mock_async_create(**kwargs):
+        # Return the async generator (to be awaited by the implementation)
+        return mock_async_generator()
+
+    with patch(
+        "anthropic.resources.AsyncMessages.create",
+        side_effect=mock_async_create,
+    ):
+        async_client = AsyncAnthropic(api_key="test-key", posthog_client=mock_client)
+
+        async def run_test():
+            response = await async_client.messages.create(
+                model="claude-3-5-sonnet-20241022",
+                system="You are a helpful weather assistant.",
+                messages=[
+                    {"role": "user", "content": "What's the weather in San Francisco?"}
+                ],
+                tools=[
+                    {
+                        "name": "get_weather",
+                        "description": "Get weather information",
+                        "input_schema": {
+                            "type": "object",
+                            "properties": {
+                                "location": {"type": "string"},
+                                "unit": {"type": "string"},
+                            },
+                            "required": ["location"],
+                        },
+                    }
+                ],
+                stream=True,
+                posthog_distinct_id="test-id",
+            )
+
+            # Consume the async stream
+            [event async for event in response]
+
+        # asyncio.run() waits for all async operations to complete
+        asyncio.run(run_test())
+
+        # Capture completes before asyncio.run() returns
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        assert call_args["distinct_id"] == "test-id"
+        assert call_args["event"] == "$ai_generation"
+        assert props["$ai_provider"] == "anthropic"
+        assert props["$ai_model"] == "claude-3-5-sonnet-20241022"
+
+        # Verify system prompt is included in input
+        assert props["$ai_input"] == [
+            {"role": "system", "content": "You are a helpful weather assistant."},
+            {"role": "user", "content": "What's the weather in San Francisco?"},
+        ]
+
+        # Verify that tools are captured in the properties
+        assert props["$ai_tools"] == [
+            {
+                "name": "get_weather",
+                "description": "Get weather information",
+                "input_schema": {
+                    "type": "object",
+                    "properties": {
+                        "location": {"type": "string"},
+                        "unit": {"type": "string"},
+                    },
+                    "required": ["location"],
+                },
+            }
+        ]
+
+        # Verify output contains both text and tool call
+        output_choices = props["$ai_output_choices"]
+        assert len(output_choices) == 1
+
+        assistant_message = output_choices[0]
+        assert assistant_message["role"] == "assistant"
+
+        content = assistant_message["content"]
+        assert isinstance(content, list)
+        assert len(content) == 2
+
+        # Verify text block
+        text_block = content[0]
+        assert text_block["type"] == "text"
+        assert text_block["text"] == "I'll check the weather for you."
+
+        # Verify tool call block
+        tool_block = content[1]
+        assert tool_block["type"] == "function"
+        assert tool_block["id"] == "toolu_stream123"
+        assert tool_block["function"]["name"] == "get_weather"
+        assert tool_block["function"]["arguments"] == {
+            "location": "San Francisco",
+            "unit": "celsius",
+        }
+
+        # Check token usage
+        assert props["$ai_input_tokens"] == 50
+        assert props["$ai_output_tokens"] == 25
+        assert props["$ai_cache_read_input_tokens"] == 5
+        assert props["$ai_cache_creation_input_tokens"] == 0
+
+
+def test_web_search_count(mock_client):
+    """Test that web search count is properly tracked from Anthropic responses."""
+
+    # Create a mock usage with web search
+    class MockServerToolUse:
+        def __init__(self):
+            self.web_search_requests = 3
+
+    class MockUsageWithWebSearch:
+        def __init__(self):
+            self.input_tokens = 100
+            self.output_tokens = 50
+            self.cache_read_input_tokens = 0
+            self.cache_creation_input_tokens = 0
+            self.server_tool_use = MockServerToolUse()
+
+    class MockResponseWithWebSearch:
+        def __init__(self):
+            self.content = [MockContent(text="Search results show...")]
+            self.model = "claude-3-opus-20240229"
+            self.usage = MockUsageWithWebSearch()
+
+    mock_response = MockResponseWithWebSearch()
+
+    with patch("anthropic.resources.Messages.create", return_value=mock_response):
+        client = Anthropic(api_key="test-key", posthog_client=mock_client)
+        response = client.messages.create(
+            model="claude-3-opus-20240229",
+            messages=[{"role": "user", "content": "Search for recent news"}],
+            posthog_distinct_id="test-id",
+        )
+
+        assert response == mock_response
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        # Verify web search count is captured
+        assert props["$ai_web_search_count"] == 3
+        assert props["$ai_input_tokens"] == 100
+        assert props["$ai_output_tokens"] == 50
+
+
+@pytest.fixture
+def mock_anthropic_stream_with_web_search():
+    """Mock stream events for web search."""
+
+    class MockServerToolUse:
+        def __init__(self):
+            self.web_search_requests = 2
+
+    class MockMessage:
+        def __init__(self):
+            self.usage = MockUsage(
+                input_tokens=50,
+                cache_creation_input_tokens=0,
+                cache_read_input_tokens=5,
+            )
+
+    def stream_generator():
+        # Message start with usage
+        event = MockStreamEvent("message_start")
+        event.message = MockMessage()
+        yield event
+
+        # Text block start
+        event = MockStreamEvent("content_block_start")
+        event.content_block = MockContentBlock("text")
+        event.index = 0
+        yield event
+
+        # Text delta
+        event = MockStreamEvent("content_block_delta")
+        event.delta = MockDelta(text="Here are the search results...")
+        event.index = 0
+        yield event
+
+        # Text block stop
+        event = MockStreamEvent("content_block_stop")
+        event.index = 0
+        yield event
+
+        # Message delta with final usage including web search
+        event = MockStreamEvent("message_delta")
+        usage = MockUsage(output_tokens=25)
+        usage.server_tool_use = MockServerToolUse()
+        event.usage = usage
+        yield event
+
+        # Message stop
+        event = MockStreamEvent("message_stop")
+        yield event
+
+    return stream_generator()
+
+
+def test_streaming_with_web_search(mock_client, mock_anthropic_stream_with_web_search):
+    """Test that web search count is properly captured in streaming mode."""
+    with patch(
+        "anthropic.resources.Messages.create",
+        return_value=mock_anthropic_stream_with_web_search,
+    ):
+        client = Anthropic(api_key="test-key", posthog_client=mock_client)
+        response = client.messages.create(
+            model="claude-3-opus-20240229",
+            messages=[{"role": "user", "content": "Search for recent news"}],
+            stream=True,
+            posthog_distinct_id="test-id",
+        )
+
+        # Consume the stream - this triggers the finally block synchronously
+        list(response)
+
+        # Capture happens synchronously when generator is exhausted
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        # Verify web search count is captured
+        assert props["$ai_web_search_count"] == 2
+        assert props["$ai_input_tokens"] == 50
+        assert props["$ai_output_tokens"] == 25
+
+
+def test_async_with_web_search(mock_client):
+    """Test that web search count is properly tracked in async non-streaming mode."""
+    import asyncio
+
+    # Create a mock usage with web search
+    class MockServerToolUse:
+        def __init__(self):
+            self.web_search_requests = 3
+
+    class MockUsageWithWebSearch:
+        def __init__(self):
+            self.input_tokens = 100
+            self.output_tokens = 50
+            self.cache_read_input_tokens = 0
+            self.cache_creation_input_tokens = 0
+            self.server_tool_use = MockServerToolUse()
+
+    class MockResponseWithWebSearch:
+        def __init__(self):
+            self.content = [MockContent(text="Search results show...")]
+            self.model = "claude-3-opus-20240229"
+            self.usage = MockUsageWithWebSearch()
+
+    mock_response = MockResponseWithWebSearch()
+
+    async def mock_async_create(**kwargs):
+        return mock_response
+
+    with patch(
+        "anthropic.resources.AsyncMessages.create",
+        side_effect=mock_async_create,
+    ):
+        async_client = AsyncAnthropic(api_key="test-key", posthog_client=mock_client)
+
+        async def run_test():
+            response = await async_client.messages.create(
+                model="claude-3-opus-20240229",
+                messages=[{"role": "user", "content": "Search for recent news"}],
+                posthog_distinct_id="test-id",
+            )
+            return response
+
+        # asyncio.run() waits for all async operations to complete
+        response = asyncio.run(run_test())
+
+        assert response == mock_response
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        # Verify web search count is captured
+        assert props["$ai_web_search_count"] == 3
+        assert props["$ai_input_tokens"] == 100
+        assert props["$ai_output_tokens"] == 50
+
+
+def test_async_streaming_with_web_search(
+    mock_client, mock_anthropic_stream_with_web_search
+):
+    """Test that web search count is properly captured in async streaming mode."""
+    import asyncio
+
+    async def mock_async_generator():
+        # Convert regular generator to async generator
+        for event in mock_anthropic_stream_with_web_search:
+            yield event
+
+    async def mock_async_create(**kwargs):
+        # Return the async generator (to be awaited by the implementation)
+        return mock_async_generator()
+
+    with patch(
+        "anthropic.resources.AsyncMessages.create",
+        side_effect=mock_async_create,
+    ):
+        async_client = AsyncAnthropic(api_key="test-key", posthog_client=mock_client)
+
+        async def run_test():
+            response = await async_client.messages.create(
+                model="claude-3-opus-20240229",
+                messages=[{"role": "user", "content": "Search for recent news"}],
+                stream=True,
+                posthog_distinct_id="test-id",
+            )
+
+            # Consume the async stream
+            [event async for event in response]
+
+        # asyncio.run() waits for all async operations to complete
+        asyncio.run(run_test())
+
+        # Capture completes before asyncio.run() returns
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        # Verify web search count is captured
+        assert props["$ai_web_search_count"] == 2
+        assert props["$ai_input_tokens"] == 50
+        assert props["$ai_output_tokens"] == 25

posthog/test/ai/gemini/test_gemini.py

@@ -31,6 +31,9 @@ def mock_gemini_response():
     mock_usage = MagicMock()
     mock_usage.prompt_token_count = 20
     mock_usage.candidates_token_count = 10
+    # Ensure cache and reasoning tokens are not present (not MagicMock)
+    mock_usage.cached_content_token_count = 0
+    mock_usage.thoughts_token_count = 0
     mock_response.usage_metadata = mock_usage
 
     mock_candidate = MagicMock()
@@ -64,6 +67,8 @@ def mock_gemini_response_with_function_calls():
     mock_usage = MagicMock()
     mock_usage.prompt_token_count = 25
     mock_usage.candidates_token_count = 15
+    mock_usage.cached_content_token_count = 0
+    mock_usage.thoughts_token_count = 0
     mock_response.usage_metadata = mock_usage
 
     # Mock function call
@@ -110,6 +115,8 @@ def mock_gemini_response_function_calls_only():
     mock_usage = MagicMock()
     mock_usage.prompt_token_count = 30
     mock_usage.candidates_token_count = 12
+    mock_usage.cached_content_token_count = 0
+    mock_usage.thoughts_token_count = 0
     mock_response.usage_metadata = mock_usage
 
     # Mock function call
@@ -180,6 +187,8 @@ def test_new_client_streaming_with_generate_content_stream(
         mock_usage1 = MagicMock()
         mock_usage1.prompt_token_count = 10
         mock_usage1.candidates_token_count = 5
+        mock_usage1.cached_content_token_count = 0
+        mock_usage1.thoughts_token_count = 0
         mock_chunk1.usage_metadata = mock_usage1
 
         mock_chunk2 = MagicMock()
@@ -187,6 +196,8 @@ def test_new_client_streaming_with_generate_content_stream(
         mock_usage2 = MagicMock()
         mock_usage2.prompt_token_count = 10
         mock_usage2.candidates_token_count = 10
+        mock_usage2.cached_content_token_count = 0
+        mock_usage2.thoughts_token_count = 0
         mock_chunk2.usage_metadata = mock_usage2
 
         yield mock_chunk1
@@ -226,6 +237,91 @@ def test_new_client_streaming_with_generate_content_stream(
     assert isinstance(props["$ai_latency"], float)
 
 
+def test_new_client_streaming_with_tools(mock_client, mock_google_genai_client):
+    """Test that tools are captured in streaming mode"""
+
+    def mock_streaming_response():
+        mock_chunk1 = MagicMock()
+        mock_chunk1.text = "I'll check "
+        mock_usage1 = MagicMock()
+        mock_usage1.prompt_token_count = 15
+        mock_usage1.candidates_token_count = 5
+        mock_usage1.cached_content_token_count = 0
+        mock_usage1.thoughts_token_count = 0
+        mock_chunk1.usage_metadata = mock_usage1
+
+        mock_chunk2 = MagicMock()
+        mock_chunk2.text = "the weather"
+        mock_usage2 = MagicMock()
+        mock_usage2.prompt_token_count = 15
+        mock_usage2.candidates_token_count = 10
+        mock_usage2.cached_content_token_count = 0
+        mock_usage2.thoughts_token_count = 0
+        mock_chunk2.usage_metadata = mock_usage2
+
+        yield mock_chunk1
+        yield mock_chunk2
+
+    # Mock the generate_content_stream method
+    mock_google_genai_client.models.generate_content_stream.return_value = (
+        mock_streaming_response()
+    )
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+
+    # Create mock tools configuration
+    mock_tool = MagicMock()
+    mock_tool.function_declarations = [
+        MagicMock(
+            name="get_current_weather",
+            description="Gets the current weather for a given location.",
+            parameters=MagicMock(
+                type="OBJECT",
+                properties={
+                    "location": MagicMock(
+                        type="STRING",
+                        description="The city and state, e.g. San Francisco, CA",
+                    )
+                },
+                required=["location"],
+            ),
+        )
+    ]
+
+    mock_config = MagicMock()
+    mock_config.tools = [mock_tool]
+
+    response = client.models.generate_content_stream(
+        model="gemini-2.0-flash",
+        contents=["What's the weather in SF?"],
+        config=mock_config,
+        posthog_distinct_id="test-id",
+        posthog_properties={"feature": "streaming_with_tools"},
+    )
+
+    chunks = list(response)
+    assert len(chunks) == 2
+    assert chunks[0].text == "I'll check "
+    assert chunks[1].text == "the weather"
+
+    # Check that the streaming event was captured with tools
+    assert mock_client.capture.call_count == 1
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "gemini"
+    assert props["$ai_model"] == "gemini-2.0-flash"
+    assert props["$ai_input_tokens"] == 15
+    assert props["$ai_output_tokens"] == 10
+    assert props["feature"] == "streaming_with_tools"
+    assert isinstance(props["$ai_latency"], float)
+
+    # Verify that tools are captured in the $ai_tools property in streaming mode
+    assert props["$ai_tools"] == [mock_tool]
+
+
 def test_new_client_groups(mock_client, mock_google_genai_client, mock_gemini_response):
     """Test groups functionality with new Client API"""
     mock_google_genai_client.models.generate_content.return_value = mock_gemini_response
@@ -302,12 +398,32 @@ def test_new_client_different_input_formats(
     props = call_args["properties"]
     assert props["$ai_input"] == [{"role": "user", "content": "Hello"}]
 
-    # Test list input
-    mock_client.capture.reset_mock()
-    mock_part = MagicMock()
-    mock_part.text = "List item"
+    # Test Gemini-specific format with parts array (like in the screenshot)
+    mock_client.reset_mock()
     client.models.generate_content(
-        model="gemini-2.0-flash", contents=[mock_part], posthog_distinct_id="test-id"
+        model="gemini-2.0-flash",
+        contents=[{"role": "user", "parts": [{"text": "hey"}]}],
+        posthog_distinct_id="test-id",
+    )
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+    assert props["$ai_input"] == [{"role": "user", "content": "hey"}]
+
+    # Test multiple parts in the parts array
+    mock_client.reset_mock()
+    client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=[{"role": "user", "parts": [{"text": "Hello "}, {"text": "world"}]}],
+        posthog_distinct_id="test-id",
+    )
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+    assert props["$ai_input"] == [{"role": "user", "content": "Hello world"}]
+
+    # Test list input with string
+    mock_client.capture.reset_mock()
+    client.models.generate_content(
+        model="gemini-2.0-flash", contents=["List item"], posthog_distinct_id="test-id"
     )
     call_args = mock_client.capture.call_args[1]
     props = call_args["properties"]
@@ -401,6 +517,78 @@ def test_new_client_override_defaults(
     assert props["urgent"] is True  # from call
 
 
+def test_vertex_ai_parameters_passed_through(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test that Vertex AI parameters are properly passed to genai.Client"""
+    mock_google_genai_client.models.generate_content.return_value = mock_gemini_response
+
+    # Mock credentials object
+    mock_credentials = MagicMock()
+    mock_debug_config = MagicMock()
+    mock_http_options = MagicMock()
+
+    # Create client with Vertex AI parameters
+    Client(
+        vertexai=True,
+        credentials=mock_credentials,
+        project="test-project",
+        location="us-central1",
+        debug_config=mock_debug_config,
+        http_options=mock_http_options,
+        posthog_client=mock_client,
+    )
+
+    # Verify genai.Client was called with correct parameters
+    google_genai.Client.assert_called_once_with(
+        vertexai=True,
+        credentials=mock_credentials,
+        project="test-project",
+        location="us-central1",
+        debug_config=mock_debug_config,
+        http_options=mock_http_options,
+    )
+
+
+def test_api_key_mode(mock_client, mock_google_genai_client):
+    """Test API key authentication mode"""
+
+    # Create client with just API key (traditional mode)
+    Client(
+        api_key="test-api-key",
+        posthog_client=mock_client,
+    )
+
+    # Verify genai.Client was called with only api_key
+    google_genai.Client.assert_called_once_with(api_key="test-api-key")
+
+
+def test_vertex_ai_mode_with_optional_api_key(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test Vertex AI mode with optional API key"""
+    mock_google_genai_client.models.generate_content.return_value = mock_gemini_response
+
+    mock_credentials = MagicMock()
+
+    # Create client with Vertex AI + API key
+    Client(
+        vertexai=True,
+        api_key="test-api-key",
+        credentials=mock_credentials,
+        project="test-project",
+        posthog_client=mock_client,
+    )
+
+    # Verify genai.Client was called with both Vertex AI params and API key
+    google_genai.Client.assert_called_once_with(
+        vertexai=True,
+        api_key="test-api-key",
+        credentials=mock_credentials,
+        project="test-project",
+    )
+
+
 def test_tool_use_response(mock_client, mock_google_genai_client, mock_gemini_response):
     """Test that tools defined in config are captured in $ai_tools property"""
     mock_google_genai_client.models.generate_content.return_value = mock_gemini_response
@@ -428,6 +616,8 @@ def test_tool_use_response(mock_client, mock_google_genai_client, mock_gemini_re
 
     mock_config = MagicMock()
     mock_config.tools = [mock_tool]
+    # Explicitly specify this config doesn't have system_instruction
+    del mock_config.system_instruction
 
     response = client.models.generate_content(
         model="gemini-2.5-flash",
@@ -557,3 +747,352 @@ def test_function_calls_only_no_content(
     assert props["$ai_input_tokens"] == 30
     assert props["$ai_output_tokens"] == 12
     assert props["$ai_http_status"] == 200
+
+
+def test_cache_and_reasoning_tokens(mock_client, mock_google_genai_client):
+    """Test that cache and reasoning tokens are properly extracted"""
+    # Create a mock response with cache and reasoning tokens
+    mock_response = MagicMock()
+    mock_response.text = "Test response with cache"
+
+    mock_usage = MagicMock()
+    mock_usage.prompt_token_count = 100
+    mock_usage.candidates_token_count = 50
+    mock_usage.cached_content_token_count = 30  # Cache tokens
+    mock_usage.thoughts_token_count = 10  # Reasoning tokens
+    mock_response.usage_metadata = mock_usage
+
+    # Mock candidates
+    mock_candidate = MagicMock()
+    mock_candidate.text = "Test response with cache"
+    mock_response.candidates = [mock_candidate]
+
+    mock_google_genai_client.models.generate_content.return_value = mock_response
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+
+    response = client.models.generate_content(
+        model="gemini-2.5-pro",
+        contents="Test with cache",
+        posthog_distinct_id="test-id",
+    )
+
+    assert response == mock_response
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Check that all token types are present
+    assert props["$ai_input_tokens"] == 100
+    assert props["$ai_output_tokens"] == 50
+    assert props["$ai_cache_read_input_tokens"] == 30
+    assert props["$ai_reasoning_tokens"] == 10
+
+
+def test_streaming_cache_and_reasoning_tokens(mock_client, mock_google_genai_client):
+    """Test that cache and reasoning tokens are properly extracted in streaming"""
+    # Create mock chunks with cache and reasoning tokens
+    chunk1 = MagicMock()
+    chunk1.text = "Hello "
+    chunk1_usage = MagicMock()
+    chunk1_usage.prompt_token_count = 100
+    chunk1_usage.candidates_token_count = 5
+    chunk1_usage.cached_content_token_count = 30  # Cache tokens
+    chunk1_usage.thoughts_token_count = 0
+    chunk1.usage_metadata = chunk1_usage
+
+    chunk2 = MagicMock()
+    chunk2.text = "world!"
+    chunk2_usage = MagicMock()
+    chunk2_usage.prompt_token_count = 100
+    chunk2_usage.candidates_token_count = 10
+    chunk2_usage.cached_content_token_count = 30  # Same cache tokens
+    chunk2_usage.thoughts_token_count = 5  # Reasoning tokens
+    chunk2.usage_metadata = chunk2_usage
+
+    mock_stream = iter([chunk1, chunk2])
+    mock_google_genai_client.models.generate_content_stream.return_value = mock_stream
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+
+    response = client.models.generate_content_stream(
+        model="gemini-2.5-pro",
+        contents="Test streaming with cache",
+        posthog_distinct_id="test-id",
+    )
+
+    # Consume the stream
+    result = list(response)
+    assert len(result) == 2
+
+    # Check PostHog capture was called
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Check that all token types are present (should use final chunk's usage)
+    assert props["$ai_input_tokens"] == 100
+    assert props["$ai_output_tokens"] == 10
+    assert props["$ai_cache_read_input_tokens"] == 30
+    assert props["$ai_reasoning_tokens"] == 5
+
+
+def test_web_search_grounding(mock_client, mock_google_genai_client):
+    """Test web search detection via grounding_metadata."""
+
+    # Create mock response with grounding metadata
+    mock_response = MagicMock()
+
+    # Mock usage metadata
+    mock_usage = MagicMock()
+    mock_usage.prompt_token_count = 60
+    mock_usage.candidates_token_count = 40
+    mock_usage.cached_content_token_count = 0
+    mock_usage.thoughts_token_count = 0
+    mock_response.usage_metadata = mock_usage
+
+    # Mock grounding metadata
+    mock_grounding_chunk = MagicMock()
+    mock_grounding_chunk.uri = "https://example.com"
+
+    mock_grounding_metadata = MagicMock()
+    mock_grounding_metadata.grounding_chunks = [mock_grounding_chunk]
+
+    # Mock text part
+    mock_text_part = MagicMock()
+    mock_text_part.text = "According to search results..."
+    type(mock_text_part).text = mock_text_part.text
+
+    # Mock content with parts
+    mock_content = MagicMock()
+    mock_content.parts = [mock_text_part]
+
+    # Mock candidate with grounding metadata
+    mock_candidate = MagicMock()
+    mock_candidate.content = mock_content
+    mock_candidate.grounding_metadata = mock_grounding_metadata
+    type(mock_candidate).grounding_metadata = mock_candidate.grounding_metadata
+
+    mock_response.candidates = [mock_candidate]
+    mock_response.text = "According to search results..."
+
+    # Mock the generate_content method
+    mock_google_genai_client.models.generate_content.return_value = mock_response
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+    response = client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents="What's the latest news?",
+        posthog_distinct_id="test-id",
+    )
+
+    assert response == mock_response
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Verify web search count is detected (binary for grounding)
+    assert props["$ai_web_search_count"] == 1
+    assert props["$ai_input_tokens"] == 60
+    assert props["$ai_output_tokens"] == 40
+
+
+def test_streaming_with_web_search(mock_client, mock_google_genai_client):
+    """Test that web search count is properly captured in streaming mode."""
+
+    def mock_streaming_response():
+        # Create chunk 1 with grounding metadata
+        mock_chunk1 = MagicMock()
+        mock_chunk1.text = "According to "
+
+        mock_usage1 = MagicMock()
+        mock_usage1.prompt_token_count = 30
+        mock_usage1.candidates_token_count = 5
+        mock_usage1.cached_content_token_count = 0
+        mock_usage1.thoughts_token_count = 0
+        mock_chunk1.usage_metadata = mock_usage1
+
+        # Add grounding metadata to first chunk
+        mock_grounding_chunk = MagicMock()
+        mock_grounding_chunk.uri = "https://example.com"
+
+        mock_grounding_metadata = MagicMock()
+        mock_grounding_metadata.grounding_chunks = [mock_grounding_chunk]
+
+        mock_candidate1 = MagicMock()
+        mock_candidate1.grounding_metadata = mock_grounding_metadata
+        type(mock_candidate1).grounding_metadata = mock_candidate1.grounding_metadata
+
+        mock_chunk1.candidates = [mock_candidate1]
+
+        # Create chunk 2
+        mock_chunk2 = MagicMock()
+        mock_chunk2.text = "search results..."
+
+        mock_usage2 = MagicMock()
+        mock_usage2.prompt_token_count = 30
+        mock_usage2.candidates_token_count = 15
+        mock_usage2.cached_content_token_count = 0
+        mock_usage2.thoughts_token_count = 0
+        mock_chunk2.usage_metadata = mock_usage2
+
+        mock_candidate2 = MagicMock()
+        mock_chunk2.candidates = [mock_candidate2]
+
+        yield mock_chunk1
+        yield mock_chunk2
+
+    # Mock the generate_content_stream method
+    mock_google_genai_client.models.generate_content_stream.return_value = (
+        mock_streaming_response()
+    )
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+
+    response = client.models.generate_content_stream(
+        model="gemini-2.5-flash",
+        contents="What's the latest news?",
+        posthog_distinct_id="test-id",
+    )
+
+    chunks = list(response)
+    assert len(chunks) == 2
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Verify web search count is detected (binary for grounding)
+    assert props["$ai_web_search_count"] == 1
+    assert props["$ai_input_tokens"] == 30
+    assert props["$ai_output_tokens"] == 15
+
+
+def test_empty_grounding_metadata_no_web_search(mock_client, mock_google_genai_client):
+    """Test that empty grounding_metadata (all null fields) does not count as web search."""
+
+    # Create mock response with empty grounding metadata (all null fields)
+    mock_response = MagicMock()
+
+    # Mock usage metadata
+    mock_usage = MagicMock()
+    mock_usage.prompt_token_count = 10
+    mock_usage.candidates_token_count = 10
+    mock_usage.cached_content_token_count = 0
+    mock_usage.thoughts_token_count = 0
+    mock_response.usage_metadata = mock_usage
+
+    # Mock empty grounding metadata (all fields are None)
+    mock_grounding_metadata = MagicMock()
+    mock_grounding_metadata.web_search_queries = None
+    mock_grounding_metadata.grounding_chunks = None
+    mock_grounding_metadata.grounding_supports = None
+    mock_grounding_metadata.retrieval_metadata = None
+    mock_grounding_metadata.retrieval_queries = None
+    mock_grounding_metadata.search_entry_point = None
+
+    # Mock text part
+    mock_text_part = MagicMock()
+    mock_text_part.text = "Hey there! How can I help you today?"
+    type(mock_text_part).text = mock_text_part.text
+
+    # Mock content with parts
+    mock_content = MagicMock()
+    mock_content.parts = [mock_text_part]
+
+    # Mock candidate with empty grounding metadata
+    mock_candidate = MagicMock()
+    mock_candidate.content = mock_content
+    mock_candidate.grounding_metadata = mock_grounding_metadata
+    type(mock_candidate).grounding_metadata = mock_candidate.grounding_metadata
+
+    mock_response.candidates = [mock_candidate]
+    mock_response.text = "Hey there! How can I help you today?"
+
+    # Mock the generate_content method
+    mock_google_genai_client.models.generate_content.return_value = mock_response
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+
+    response = client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents="Hello",
+        posthog_distinct_id="test-id",
+    )
+
+    assert response == mock_response
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Verify web search count is 0 (not present in properties when 0)
+    assert "$ai_web_search_count" not in props
+    assert props["$ai_input_tokens"] == 10
+    assert props["$ai_output_tokens"] == 10
+
+
+def test_empty_array_grounding_metadata_no_web_search(
+    mock_client, mock_google_genai_client
+):
+    """Test that grounding_metadata with empty arrays does not count as web search."""
+
+    # Create mock response with grounding metadata having empty arrays
+    mock_response = MagicMock()
+
+    # Mock usage metadata
+    mock_usage = MagicMock()
+    mock_usage.prompt_token_count = 15
+    mock_usage.candidates_token_count = 12
+    mock_usage.cached_content_token_count = 0
+    mock_usage.thoughts_token_count = 0
+    mock_response.usage_metadata = mock_usage
+
+    # Mock grounding metadata with empty arrays
+    mock_grounding_metadata = MagicMock()
+    mock_grounding_metadata.web_search_queries = []
+    mock_grounding_metadata.grounding_chunks = []
+    mock_grounding_metadata.grounding_supports = []
+
+    # Mock text part
+    mock_text_part = MagicMock()
+    mock_text_part.text = "I can help with that."
+    type(mock_text_part).text = mock_text_part.text
+
+    # Mock content with parts
+    mock_content = MagicMock()
+    mock_content.parts = [mock_text_part]
+
+    # Mock candidate with grounding metadata containing empty arrays
+    mock_candidate = MagicMock()
+    mock_candidate.content = mock_content
+    mock_candidate.grounding_metadata = mock_grounding_metadata
+    type(mock_candidate).grounding_metadata = mock_candidate.grounding_metadata
+
+    mock_response.candidates = [mock_candidate]
+    mock_response.text = "I can help with that."
+
+    # Mock the generate_content method
+    mock_google_genai_client.models.generate_content.return_value = mock_response
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+
+    response = client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents="What can you do?",
+        posthog_distinct_id="test-id",
+    )
+
+    assert response == mock_response
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Verify web search count is 0 (not present in properties when 0)
+    assert "$ai_web_search_count" not in props
+    assert props["$ai_input_tokens"] == 15
+    assert props["$ai_output_tokens"] == 12

posthog/test/ai/gemini/test_gemini_async.py

@@ -0,0 +1,843 @@
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+try:
+    from google import genai as google_genai
+
+    from posthog.ai.gemini import AsyncClient
+
+    GEMINI_AVAILABLE = True
+except ImportError:
+    GEMINI_AVAILABLE = False
+
+pytestmark = [
+    pytest.mark.skipif(
+        not GEMINI_AVAILABLE, reason="Google Gemini package is not available"
+    ),
+    pytest.mark.asyncio,
+]
+
+
+@pytest.fixture
+def mock_client():
+    with patch("posthog.client.Client") as mock_client:
+        mock_client.privacy_mode = False
+        yield mock_client
+
+
+@pytest.fixture
+def mock_gemini_response():
+    mock_response = MagicMock()
+    mock_response.text = "Test response from Gemini"
+
+    mock_usage = MagicMock()
+    mock_usage.prompt_token_count = 20
+    mock_usage.candidates_token_count = 10
+    # Ensure cache and reasoning tokens are not present (not MagicMock)
+    mock_usage.cached_content_token_count = 0
+    mock_usage.thoughts_token_count = 0
+    mock_response.usage_metadata = mock_usage
+
+    mock_candidate = MagicMock()
+    mock_candidate.text = "Test response from Gemini"
+    mock_content = MagicMock()
+    mock_part = MagicMock()
+    mock_part.text = "Test response from Gemini"
+    mock_content.parts = [mock_part]
+    mock_candidate.content = mock_content
+    mock_response.candidates = [mock_candidate]
+
+    return mock_response
+
+
+@pytest.fixture
+def mock_google_genai_client():
+    """Mock for the google-genai Client with async support"""
+    with patch.object(google_genai, "Client") as mock_client_class:
+        mock_client_instance = MagicMock()
+        mock_models = MagicMock()
+        mock_aio = MagicMock()
+        mock_aio_models = MagicMock()
+
+        mock_client_instance.models = mock_models
+        mock_client_instance.aio = mock_aio
+        mock_aio.models = mock_aio_models
+
+        mock_client_class.return_value = mock_client_instance
+        yield mock_client_instance
+
+
+@pytest.fixture
+def mock_gemini_response_with_function_calls():
+    mock_response = MagicMock()
+
+    # Mock usage metadata
+    mock_usage = MagicMock()
+    mock_usage.prompt_token_count = 25
+    mock_usage.candidates_token_count = 15
+    mock_usage.cached_content_token_count = 0
+    mock_usage.thoughts_token_count = 0
+    mock_response.usage_metadata = mock_usage
+
+    # Mock function call
+    mock_function_call = MagicMock()
+    mock_function_call.name = "get_current_weather"
+    mock_function_call.args = {"location": "San Francisco"}
+
+    # Mock text part 1
+    mock_text_part1 = MagicMock()
+    mock_text_part1.text = "I'll check the weather for you."
+    type(mock_text_part1).text = mock_text_part1.text
+
+    # Mock text part 2
+    mock_text_part2 = MagicMock()
+    mock_text_part2.text = " Let me look that up."
+    type(mock_text_part2).text = mock_text_part2.text
+
+    # Mock function call part
+    mock_function_part = MagicMock()
+    mock_function_part.function_call = mock_function_call
+    type(mock_function_part).function_call = mock_function_part.function_call
+    del mock_function_part.text
+
+    # Mock content with 2 text parts and 1 function call part
+    mock_content = MagicMock()
+    mock_content.parts = [mock_text_part1, mock_text_part2, mock_function_part]
+
+    # Mock candidate
+    mock_candidate = MagicMock()
+    mock_candidate.content = mock_content
+    mock_response.candidates = [mock_candidate]
+
+    return mock_response
+
+
+async def test_async_client_basic_generation(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test the async Client/AsyncModels API structure"""
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    response = await client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=["Tell me a fun fact about hedgehogs"],
+        posthog_distinct_id="test-id",
+        posthog_properties={"foo": "bar"},
+    )
+
+    assert response == mock_gemini_response
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "gemini"
+    assert props["$ai_model"] == "gemini-2.0-flash"
+    assert props["$ai_input_tokens"] == 20
+    assert props["$ai_output_tokens"] == 10
+    assert props["foo"] == "bar"
+    assert "$ai_trace_id" in props
+    assert props["$ai_latency"] > 0
+
+
+async def test_async_client_streaming_with_generate_content_stream(
+    mock_client, mock_google_genai_client
+):
+    """Test the async generate_content_stream method"""
+
+    async def mock_streaming_response():
+        mock_chunk1 = MagicMock()
+        mock_chunk1.text = "Hello "
+        mock_usage1 = MagicMock()
+        mock_usage1.prompt_token_count = 10
+        mock_usage1.candidates_token_count = 5
+        mock_usage1.cached_content_token_count = 0
+        mock_usage1.thoughts_token_count = 0
+        mock_chunk1.usage_metadata = mock_usage1
+        yield mock_chunk1
+
+        mock_chunk2 = MagicMock()
+        mock_chunk2.text = "world!"
+        mock_usage2 = MagicMock()
+        mock_usage2.prompt_token_count = 10
+        mock_usage2.candidates_token_count = 10
+        mock_usage2.cached_content_token_count = 0
+        mock_usage2.thoughts_token_count = 0
+        mock_chunk2.usage_metadata = mock_usage2
+        yield mock_chunk2
+
+    # Mock the async generate_content_stream method
+    mock_google_genai_client.aio.models.generate_content_stream = AsyncMock(
+        return_value=mock_streaming_response()
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    response = await client.models.generate_content_stream(
+        model="gemini-2.0-flash",
+        contents=["Write a short story"],
+        posthog_distinct_id="test-id",
+        posthog_properties={"feature": "streaming"},
+    )
+
+    chunks = []
+    async for chunk in response:
+        chunks.append(chunk)
+
+    assert len(chunks) == 2
+    assert chunks[0].text == "Hello "
+    assert chunks[1].text == "world!"
+
+    # Check that the streaming event was captured
+    assert mock_client.capture.call_count == 1
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "gemini"
+    assert props["$ai_model"] == "gemini-2.0-flash"
+    assert props["$ai_input_tokens"] == 10
+    assert props["$ai_output_tokens"] == 10
+    assert props["feature"] == "streaming"
+    assert isinstance(props["$ai_latency"], float)
+
+
+async def test_async_client_streaming_with_tools(mock_client, mock_google_genai_client):
+    """Test that tools are captured in async streaming mode"""
+
+    async def mock_streaming_response():
+        mock_chunk1 = MagicMock()
+        mock_chunk1.text = "I'll check "
+        mock_usage1 = MagicMock()
+        mock_usage1.prompt_token_count = 15
+        mock_usage1.candidates_token_count = 5
+        mock_usage1.cached_content_token_count = 0
+        mock_usage1.thoughts_token_count = 0
+        mock_chunk1.usage_metadata = mock_usage1
+        yield mock_chunk1
+
+        mock_chunk2 = MagicMock()
+        mock_chunk2.text = "the weather"
+        mock_usage2 = MagicMock()
+        mock_usage2.prompt_token_count = 15
+        mock_usage2.candidates_token_count = 10
+        mock_usage2.cached_content_token_count = 0
+        mock_usage2.thoughts_token_count = 0
+        mock_chunk2.usage_metadata = mock_usage2
+        yield mock_chunk2
+
+    # Mock the async generate_content_stream method
+    mock_google_genai_client.aio.models.generate_content_stream = AsyncMock(
+        return_value=mock_streaming_response()
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    # Create mock tools configuration
+    mock_tool = MagicMock()
+    mock_tool.function_declarations = [
+        MagicMock(
+            name="get_current_weather",
+            description="Gets the current weather for a given location.",
+            parameters=MagicMock(
+                type="OBJECT",
+                properties={
+                    "location": MagicMock(
+                        type="STRING",
+                        description="The city and state, e.g. San Francisco, CA",
+                    )
+                },
+                required=["location"],
+            ),
+        )
+    ]
+
+    mock_config = MagicMock()
+    mock_config.tools = [mock_tool]
+
+    response = await client.models.generate_content_stream(
+        model="gemini-2.0-flash",
+        contents=["What's the weather in SF?"],
+        config=mock_config,
+        posthog_distinct_id="test-id",
+        posthog_properties={"feature": "streaming_with_tools"},
+    )
+
+    chunks = []
+    async for chunk in response:
+        chunks.append(chunk)
+
+    assert len(chunks) == 2
+    assert chunks[0].text == "I'll check "
+    assert chunks[1].text == "the weather"
+
+    # Check that the streaming event was captured with tools
+    assert mock_client.capture.call_count == 1
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "gemini"
+    assert props["$ai_model"] == "gemini-2.0-flash"
+    assert props["$ai_input_tokens"] == 15
+    assert props["$ai_output_tokens"] == 10
+    assert props["feature"] == "streaming_with_tools"
+    assert isinstance(props["$ai_latency"], float)
+
+    # Verify that tools are captured in the $ai_tools property in streaming mode
+    assert props["$ai_tools"] == [mock_tool]
+
+
+async def test_async_client_groups(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test groups functionality with async Client API"""
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    await client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=["Hello"],
+        posthog_distinct_id="test-id",
+        posthog_groups={"company": "company_123"},
+    )
+
+    call_args = mock_client.capture.call_args[1]
+    assert call_args["groups"] == {"company": "company_123"}
+
+
+async def test_async_client_privacy_mode_local(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test local privacy mode with async Client API"""
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    await client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=["Hello"],
+        posthog_distinct_id="test-id",
+        posthog_privacy_mode=True,
+    )
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+    assert props["$ai_input"] is None
+    assert props["$ai_output_choices"] is None
+
+
+async def test_async_client_privacy_mode_global(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test global privacy mode with async Client API"""
+    mock_client.privacy_mode = True
+
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    await client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=["Hello"],
+        posthog_distinct_id="test-id",
+    )
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+    assert props["$ai_input"] is None
+    assert props["$ai_output_choices"] is None
+
+
+async def test_async_client_different_input_formats(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test different input formats with async Client API"""
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    # Test string input
+    await client.models.generate_content(
+        model="gemini-2.0-flash", contents="Hello", posthog_distinct_id="test-id"
+    )
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+    assert props["$ai_input"] == [{"role": "user", "content": "Hello"}]
+
+    # Test Gemini-specific format with parts array
+    mock_client.reset_mock()
+    await client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=[{"role": "user", "parts": [{"text": "hey"}]}],
+        posthog_distinct_id="test-id",
+    )
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+    assert props["$ai_input"] == [{"role": "user", "content": "hey"}]
+
+    # Test multiple parts in the parts array
+    mock_client.reset_mock()
+    await client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=[{"role": "user", "parts": [{"text": "Hello "}, {"text": "world"}]}],
+        posthog_distinct_id="test-id",
+    )
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+    assert props["$ai_input"] == [{"role": "user", "content": "Hello world"}]
+
+    # Test list input with string
+    mock_client.capture.reset_mock()
+    await client.models.generate_content(
+        model="gemini-2.0-flash", contents=["List item"], posthog_distinct_id="test-id"
+    )
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+    assert props["$ai_input"] == [{"role": "user", "content": "List item"}]
+
+
+async def test_async_client_model_parameters(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test model parameters with async Client API"""
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    await client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=["Hello"],
+        posthog_distinct_id="test-id",
+        temperature=0.7,
+        max_tokens=100,
+    )
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+    assert props["$ai_model_parameters"]["temperature"] == 0.7
+    assert props["$ai_model_parameters"]["max_tokens"] == 100
+
+
+async def test_async_client_default_settings(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test async client with default PostHog settings"""
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response
+    )
+
+    client = AsyncClient(
+        api_key="test-key",
+        posthog_client=mock_client,
+        posthog_distinct_id="default_user",
+        posthog_properties={"team": "ai"},
+        posthog_privacy_mode=False,
+        posthog_groups={"company": "acme_corp"},
+    )
+
+    # Call without overriding defaults
+    await client.models.generate_content(model="gemini-2.0-flash", contents=["Hello"])
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "default_user"
+    assert call_args["groups"] == {"company": "acme_corp"}
+    assert props["team"] == "ai"
+
+
+async def test_async_client_override_defaults(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test overriding async client defaults per call"""
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response
+    )
+
+    client = AsyncClient(
+        api_key="test-key",
+        posthog_client=mock_client,
+        posthog_distinct_id="default_user",
+        posthog_properties={"team": "ai"},
+        posthog_privacy_mode=False,
+        posthog_groups={"company": "acme_corp"},
+    )
+
+    # Override defaults in call
+    await client.models.generate_content(
+        model="gemini-2.0-flash",
+        contents=["Hello"],
+        posthog_distinct_id="specific_user",
+        posthog_properties={"feature": "chat", "urgent": True},
+        posthog_privacy_mode=True,
+        posthog_groups={"organization": "special_org"},
+    )
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Check overrides
+    assert call_args["distinct_id"] == "specific_user"
+    assert call_args["groups"] == {"organization": "special_org"}
+    assert props["$ai_input"] is None  # privacy mode was overridden
+
+    # Check merged properties (defaults + call-specific)
+    assert props["team"] == "ai"  # from defaults
+    assert props["feature"] == "chat"  # from call
+    assert props["urgent"] is True  # from call
+
+
+async def test_async_vertex_ai_parameters_passed_through(
+    mock_client, mock_google_genai_client, mock_gemini_response
+):
+    """Test that Vertex AI parameters are properly passed to genai.Client"""
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response
+    )
+
+    # Mock credentials object
+    mock_credentials = MagicMock()
+    mock_debug_config = MagicMock()
+    mock_http_options = MagicMock()
+
+    # Create client with Vertex AI parameters
+    AsyncClient(
+        vertexai=True,
+        credentials=mock_credentials,
+        project="test-project",
+        location="us-central1",
+        debug_config=mock_debug_config,
+        http_options=mock_http_options,
+        posthog_client=mock_client,
+    )
+
+    # Verify genai.Client was called with correct parameters
+    google_genai.Client.assert_called_once_with(
+        vertexai=True,
+        credentials=mock_credentials,
+        project="test-project",
+        location="us-central1",
+        debug_config=mock_debug_config,
+        http_options=mock_http_options,
+    )
+
+
+async def test_async_api_key_mode(mock_client, mock_google_genai_client):
+    """Test API key authentication mode with async client"""
+
+    # Create async client with just API key (traditional mode)
+    AsyncClient(
+        api_key="test-api-key",
+        posthog_client=mock_client,
+    )
+
+    # Verify genai.Client was called with only api_key
+    google_genai.Client.assert_called_once_with(api_key="test-api-key")
+
+
+async def test_async_function_calls_in_output_choices(
+    mock_client, mock_google_genai_client, mock_gemini_response_with_function_calls
+):
+    """Test that function calls are properly included in $ai_output_choices with async"""
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_gemini_response_with_function_calls
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    response = await client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents=["What's the weather in San Francisco?"],
+        posthog_distinct_id="test-id",
+    )
+
+    assert response == mock_gemini_response_with_function_calls
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "gemini"
+    assert props["$ai_model"] == "gemini-2.5-flash"
+    assert props["$ai_output_choices"] == [
+        {
+            "role": "assistant",
+            "content": [
+                {"type": "text", "text": "I'll check the weather for you."},
+                {"type": "text", "text": " Let me look that up."},
+                {
+                    "type": "function",
+                    "function": {
+                        "name": "get_current_weather",
+                        "arguments": {"location": "San Francisco"},
+                    },
+                },
+            ],
+        }
+    ]
+
+    # Check token usage
+    assert props["$ai_input_tokens"] == 25
+    assert props["$ai_output_tokens"] == 15
+    assert props["$ai_http_status"] == 200
+
+
+async def test_async_cache_and_reasoning_tokens(mock_client, mock_google_genai_client):
+    """Test that cache and reasoning tokens are properly extracted with async"""
+    # Create a mock response with cache and reasoning tokens
+    mock_response = MagicMock()
+    mock_response.text = "Test response with cache"
+
+    mock_usage = MagicMock()
+    mock_usage.prompt_token_count = 100
+    mock_usage.candidates_token_count = 50
+    mock_usage.cached_content_token_count = 30  # Cache tokens
+    mock_usage.thoughts_token_count = 10  # Reasoning tokens
+    mock_response.usage_metadata = mock_usage
+
+    # Mock candidates
+    mock_candidate = MagicMock()
+    mock_candidate.text = "Test response with cache"
+    mock_response.candidates = [mock_candidate]
+
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_response
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    response = await client.models.generate_content(
+        model="gemini-2.5-pro",
+        contents="Test with cache",
+        posthog_distinct_id="test-id",
+    )
+
+    assert response == mock_response
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Check that all token types are present
+    assert props["$ai_input_tokens"] == 100
+    assert props["$ai_output_tokens"] == 50
+    assert props["$ai_cache_read_input_tokens"] == 30
+    assert props["$ai_reasoning_tokens"] == 10
+
+
+async def test_async_streaming_cache_and_reasoning_tokens(
+    mock_client, mock_google_genai_client
+):
+    """Test that cache and reasoning tokens are properly extracted in async streaming"""
+
+    async def mock_streaming_response():
+        # Create mock chunks with cache and reasoning tokens
+        chunk1 = MagicMock()
+        chunk1.text = "Hello "
+        chunk1_usage = MagicMock()
+        chunk1_usage.prompt_token_count = 100
+        chunk1_usage.candidates_token_count = 5
+        chunk1_usage.cached_content_token_count = 30  # Cache tokens
+        chunk1_usage.thoughts_token_count = 0
+        chunk1.usage_metadata = chunk1_usage
+        yield chunk1
+
+        chunk2 = MagicMock()
+        chunk2.text = "world!"
+        chunk2_usage = MagicMock()
+        chunk2_usage.prompt_token_count = 100
+        chunk2_usage.candidates_token_count = 10
+        chunk2_usage.cached_content_token_count = 30  # Same cache tokens
+        chunk2_usage.thoughts_token_count = 5  # Reasoning tokens
+        chunk2.usage_metadata = chunk2_usage
+        yield chunk2
+
+    mock_google_genai_client.aio.models.generate_content_stream = AsyncMock(
+        return_value=mock_streaming_response()
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    response = await client.models.generate_content_stream(
+        model="gemini-2.5-pro",
+        contents="Test streaming with cache",
+        posthog_distinct_id="test-id",
+    )
+
+    # Consume the stream
+    result = []
+    async for chunk in response:
+        result.append(chunk)
+
+    assert len(result) == 2
+
+    # Check PostHog capture was called
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Check that all token types are present (should use final chunk's usage)
+    assert props["$ai_input_tokens"] == 100
+    assert props["$ai_output_tokens"] == 10
+    assert props["$ai_cache_read_input_tokens"] == 30
+    assert props["$ai_reasoning_tokens"] == 5
+
+
+async def test_async_web_search_grounding(mock_client, mock_google_genai_client):
+    """Test async web search detection via grounding_metadata."""
+
+    # Create mock response with grounding metadata
+    mock_response = MagicMock()
+
+    # Mock usage metadata
+    mock_usage = MagicMock()
+    mock_usage.prompt_token_count = 60
+    mock_usage.candidates_token_count = 40
+    mock_usage.cached_content_token_count = 0
+    mock_usage.thoughts_token_count = 0
+    mock_response.usage_metadata = mock_usage
+
+    # Mock grounding metadata
+    mock_grounding_chunk = MagicMock()
+    mock_grounding_chunk.uri = "https://example.com"
+
+    mock_grounding_metadata = MagicMock()
+    mock_grounding_metadata.grounding_chunks = [mock_grounding_chunk]
+
+    # Mock text part
+    mock_text_part = MagicMock()
+    mock_text_part.text = "According to search results..."
+    type(mock_text_part).text = mock_text_part.text
+
+    # Mock content with parts
+    mock_content = MagicMock()
+    mock_content.parts = [mock_text_part]
+
+    # Mock candidate with grounding metadata
+    mock_candidate = MagicMock()
+    mock_candidate.content = mock_content
+    mock_candidate.grounding_metadata = mock_grounding_metadata
+    type(mock_candidate).grounding_metadata = mock_candidate.grounding_metadata
+
+    mock_response.candidates = [mock_candidate]
+    mock_response.text = "According to search results..."
+
+    # Mock the async generate_content method
+    mock_google_genai_client.aio.models.generate_content = AsyncMock(
+        return_value=mock_response
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+    response = await client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents="What's the latest news?",
+        posthog_distinct_id="test-id",
+    )
+
+    assert response == mock_response
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Verify web search count is detected (binary for grounding)
+    assert props["$ai_web_search_count"] == 1
+    assert props["$ai_input_tokens"] == 60
+    assert props["$ai_output_tokens"] == 40
+
+
+async def test_async_streaming_with_web_search(mock_client, mock_google_genai_client):
+    """Test that web search count is properly captured in async streaming mode."""
+
+    async def mock_streaming_response():
+        # Create chunk 1 with grounding metadata
+        mock_chunk1 = MagicMock()
+        mock_chunk1.text = "According to "
+
+        mock_usage1 = MagicMock()
+        mock_usage1.prompt_token_count = 30
+        mock_usage1.candidates_token_count = 5
+        mock_usage1.cached_content_token_count = 0
+        mock_usage1.thoughts_token_count = 0
+        mock_chunk1.usage_metadata = mock_usage1
+
+        # Add grounding metadata to first chunk
+        mock_grounding_chunk = MagicMock()
+        mock_grounding_chunk.uri = "https://example.com"
+
+        mock_grounding_metadata = MagicMock()
+        mock_grounding_metadata.grounding_chunks = [mock_grounding_chunk]
+
+        mock_candidate1 = MagicMock()
+        mock_candidate1.grounding_metadata = mock_grounding_metadata
+        type(mock_candidate1).grounding_metadata = mock_candidate1.grounding_metadata
+
+        mock_chunk1.candidates = [mock_candidate1]
+        yield mock_chunk1
+
+        # Create chunk 2
+        mock_chunk2 = MagicMock()
+        mock_chunk2.text = "search results..."
+
+        mock_usage2 = MagicMock()
+        mock_usage2.prompt_token_count = 30
+        mock_usage2.candidates_token_count = 15
+        mock_usage2.cached_content_token_count = 0
+        mock_usage2.thoughts_token_count = 0
+        mock_chunk2.usage_metadata = mock_usage2
+
+        mock_candidate2 = MagicMock()
+        mock_chunk2.candidates = [mock_candidate2]
+        yield mock_chunk2
+
+    # Mock the async generate_content_stream method
+    mock_google_genai_client.aio.models.generate_content_stream = AsyncMock(
+        return_value=mock_streaming_response()
+    )
+
+    client = AsyncClient(api_key="test-key", posthog_client=mock_client)
+
+    response = await client.models.generate_content_stream(
+        model="gemini-2.5-flash",
+        contents="What's the latest news?",
+        posthog_distinct_id="test-id",
+    )
+
+    chunks = []
+    async for chunk in response:
+        chunks.append(chunk)
+
+    assert len(chunks) == 2
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Verify web search count is detected (binary for grounding)
+    assert props["$ai_web_search_count"] == 1
+    assert props["$ai_input_tokens"] == 30
+    assert props["$ai_output_tokens"] == 15

posthog/test/ai/langchain/__init__.py

@@ -1,5 +1,5 @@
 import pytest
 
-pytest.importorskip("langchain")
+pytest.importorskip("langchain_core")
 pytest.importorskip("langchain_community")
 pytest.importorskip("langgraph")

posthog/test/ai/langchain/test_callbacks.py

@@ -113,6 +113,7 @@ def test_metadata_capture(mock_client):
         base_url="https://us.posthog.com",
         name="test",
         end_time=None,
+        posthog_properties=None,
     )
     assert callbacks._runs[run_id] == expected
     with patch("time.time", return_value=1234567891):
@@ -204,6 +205,7 @@ def test_basic_chat_chain(mock_client, stream):
     # Generation is second
     assert generation_args["event"] == "$ai_generation"
     assert "distinct_id" in generation_args
+    assert generation_props["$ai_framework"] == "langchain"
     assert "$ai_model" in generation_props
     assert "$ai_provider" in generation_props
     assert generation_props["$ai_input"] == [
@@ -1123,9 +1125,9 @@ def test_anthropic_chain(mock_client):
     )
     chain = prompt | ChatAnthropic(
         api_key=ANTHROPIC_API_KEY,
-        model="claude-3-opus-20240229",
+        model="claude-sonnet-4-5-20250929",
         temperature=0,
-        max_tokens=1,
+        max_tokens=1024,
     )
     callbacks = CallbackHandler(
         mock_client,
@@ -1148,12 +1150,12 @@ def test_anthropic_chain(mock_client):
     assert gen_args["event"] == "$ai_generation"
     assert gen_props["$ai_trace_id"] == "test-trace-id"
     assert gen_props["$ai_provider"] == "anthropic"
-    assert gen_props["$ai_model"] == "claude-3-opus-20240229"
+    assert gen_props["$ai_model"] == "claude-sonnet-4-5-20250929"
     assert gen_props["foo"] == "bar"
 
     assert gen_props["$ai_model_parameters"] == {
         "temperature": 0.0,
-        "max_tokens": 1,
+        "max_tokens": 1024,
         "streaming": False,
     }
     assert gen_props["$ai_input"] == [
@@ -1169,7 +1171,7 @@ def test_anthropic_chain(mock_client):
         <= approximate_latency
     )
     assert gen_props["$ai_input_tokens"] == 17
-    assert gen_props["$ai_output_tokens"] == 1
+    assert gen_props["$ai_output_tokens"] == 4
 
     assert trace_args["event"] == "$ai_trace"
     assert trace_props["$ai_input_state"] == {}
@@ -1186,9 +1188,9 @@ async def test_async_anthropic_streaming(mock_client):
     )
     chain = prompt | ChatAnthropic(
         api_key=ANTHROPIC_API_KEY,
-        model="claude-3-opus-20240229",
+        model="claude-sonnet-4-5-20250929",
         temperature=0,
-        max_tokens=1,
+        max_tokens=1024,
         streaming=True,
         stream_usage=True,
     )
@@ -1268,6 +1270,7 @@ def test_metadata_tools(mock_client):
         name="test",
         tools=tools,
         end_time=None,
+        posthog_properties=None,
     )
     assert callbacks._runs[run_id] == expected
     with patch("time.time", return_value=1234567891):
@@ -1564,9 +1567,9 @@ def test_anthropic_cache_write_and_read_tokens(mock_client):
             AIMessage(
                 content="Using cached analysis to provide quick response.",
                 usage_metadata={
-                    "input_tokens": 200,
+                    "input_tokens": 1200,
                     "output_tokens": 30,
-                    "total_tokens": 1030,
+                    "total_tokens": 1230,
                     "cache_read_input_tokens": 800,  # Anthropic cache read
                 },
             )
@@ -1583,13 +1586,58 @@ def test_anthropic_cache_write_and_read_tokens(mock_client):
     generation_props = generation_args["properties"]
 
     assert generation_args["event"] == "$ai_generation"
-    assert generation_props["$ai_input_tokens"] == 200
+    assert (
+        generation_props["$ai_input_tokens"] == 1200
+    )  # No provider metadata, no subtraction
     assert generation_props["$ai_output_tokens"] == 30
     assert generation_props["$ai_cache_creation_input_tokens"] == 0
     assert generation_props["$ai_cache_read_input_tokens"] == 800
     assert generation_props["$ai_reasoning_tokens"] == 0
 
 
+def test_anthropic_provider_subtracts_cache_tokens(mock_client):
+    """Test that Anthropic provider correctly subtracts cache tokens from input tokens."""
+    from langchain_core.outputs import LLMResult, ChatGeneration
+    from langchain_core.messages import AIMessage
+    from uuid import uuid4
+
+    cb = CallbackHandler(mock_client)
+    run_id = uuid4()
+
+    # Set up with Anthropic provider
+    cb._set_llm_metadata(
+        serialized={},
+        run_id=run_id,
+        messages=[{"role": "user", "content": "test"}],
+        metadata={"ls_provider": "anthropic", "ls_model_name": "claude-3-sonnet"},
+    )
+
+    # Response with cache tokens: 1200 input (includes 800 cached)
+    response = LLMResult(
+        generations=[
+            [
+                ChatGeneration(
+                    message=AIMessage(content="Response"),
+                    generation_info={
+                        "usage_metadata": {
+                            "input_tokens": 1200,
+                            "output_tokens": 50,
+                            "cache_read_input_tokens": 800,
+                        }
+                    },
+                )
+            ]
+        ],
+        llm_output={},
+    )
+
+    cb._pop_run_and_capture_generation(run_id, None, response)
+
+    generation_args = mock_client.capture.call_args_list[0][1]
+    assert generation_args["properties"]["$ai_input_tokens"] == 400  # 1200 - 800
+    assert generation_args["properties"]["$ai_cache_read_input_tokens"] == 800
+
+
 def test_openai_cache_read_tokens(mock_client):
     """Test that OpenAI cache read tokens are captured correctly."""
     prompt = ChatPromptTemplate.from_messages(
@@ -1625,7 +1673,7 @@ def test_openai_cache_read_tokens(mock_client):
     generation_props = generation_args["properties"]
 
     assert generation_args["event"] == "$ai_generation"
-    assert generation_props["$ai_input_tokens"] == 150
+    assert generation_props["$ai_input_tokens"] == 150  # No subtraction for OpenAI
     assert generation_props["$ai_output_tokens"] == 40
     assert generation_props["$ai_cache_read_input_tokens"] == 100
     assert generation_props["$ai_cache_creation_input_tokens"] == 0
@@ -1707,7 +1755,7 @@ def test_combined_reasoning_and_cache_tokens(mock_client):
     generation_props = generation_args["properties"]
 
     assert generation_args["event"] == "$ai_generation"
-    assert generation_props["$ai_input_tokens"] == 500
+    assert generation_props["$ai_input_tokens"] == 500  # No subtraction for OpenAI
     assert generation_props["$ai_output_tokens"] == 100
     assert generation_props["$ai_cache_read_input_tokens"] == 300
     assert generation_props["$ai_cache_creation_input_tokens"] == 0
@@ -1715,7 +1763,7 @@ def test_combined_reasoning_and_cache_tokens(mock_client):
 
 
 @pytest.mark.skipif(not OPENAI_API_KEY, reason="OPENAI_API_KEY is not set")
-def test_openai_reasoning_tokens(mock_client):
+def test_openai_reasoning_tokens_o4_mini(mock_client):
     model = ChatOpenAI(
         api_key=OPENAI_API_KEY, model="o4-mini", max_completion_tokens=10
     )
@@ -1876,3 +1924,429 @@ def test_tool_definition(mock_client):
     assert props["$ai_latency"] == 1.0
     # Verify that tools are captured in the $ai_tools property
     assert props["$ai_tools"] == tools
+
+
+def test_cache_read_tokens_subtraction_from_input_tokens(mock_client):
+    """Test that cache_read_tokens are properly subtracted from input_tokens.
+
+    This tests the logic in callbacks.py lines 757-758:
+        if normalized_usage.input_tokens and normalized_usage.cache_read_tokens:
+            normalized_usage.input_tokens = max(normalized_usage.input_tokens - normalized_usage.cache_read_tokens, 0)
+    """
+    prompt = ChatPromptTemplate.from_messages(
+        [("user", "Use the cached prompt for this request")]
+    )
+
+    # Scenario 1: input_tokens includes cache_read_tokens (typical case)
+    # input_tokens=150 includes 100 cache_read tokens, so actual input is 50
+    model = FakeMessagesListChatModel(
+        responses=[
+            AIMessage(
+                content="Response using cached prompt context.",
+                usage_metadata={
+                    "input_tokens": 150,  # Total includes cache reads
+                    "output_tokens": 40,
+                    "total_tokens": 190,
+                    "cache_read_input_tokens": 100,  # 100 tokens read from cache
+                },
+            )
+        ]
+    )
+
+    callbacks = [CallbackHandler(mock_client)]
+    chain = prompt | model
+    result = chain.invoke({}, config={"callbacks": callbacks})
+
+    assert result.content == "Response using cached prompt context."
+    assert mock_client.capture.call_count == 3
+
+    generation_args = mock_client.capture.call_args_list[1][1]
+    generation_props = generation_args["properties"]
+
+    assert generation_args["event"] == "$ai_generation"
+    # Input tokens not reduced without provider metadata
+    assert generation_props["$ai_input_tokens"] == 150
+    assert generation_props["$ai_output_tokens"] == 40
+    assert generation_props["$ai_cache_read_input_tokens"] == 100
+
+
+def test_cache_read_tokens_subtraction_prevents_negative(mock_client):
+    """Test that cache_read_tokens subtraction doesn't result in negative input_tokens.
+
+    This tests the max(..., 0) part of the logic in callbacks.py lines 757-758.
+    """
+    prompt = ChatPromptTemplate.from_messages(
+        [("user", "Edge case with large cache read")]
+    )
+
+    # Edge case: cache_read_tokens >= input_tokens
+    # This could happen in some API responses where accounting differs
+    model = FakeMessagesListChatModel(
+        responses=[
+            AIMessage(
+                content="Response with edge case token counts.",
+                usage_metadata={
+                    "input_tokens": 80,
+                    "output_tokens": 20,
+                    "total_tokens": 100,
+                    "cache_read_input_tokens": 100,  # More than input_tokens
+                },
+            )
+        ]
+    )
+
+    callbacks = [CallbackHandler(mock_client)]
+    chain = prompt | model
+    result = chain.invoke({}, config={"callbacks": callbacks})
+
+    assert result.content == "Response with edge case token counts."
+    assert mock_client.capture.call_count == 3
+
+    generation_args = mock_client.capture.call_args_list[1][1]
+    generation_props = generation_args["properties"]
+
+    assert generation_args["event"] == "$ai_generation"
+    # Input tokens not reduced without provider metadata
+    assert generation_props["$ai_input_tokens"] == 80
+    assert generation_props["$ai_output_tokens"] == 20
+    assert generation_props["$ai_cache_read_input_tokens"] == 100
+
+
+def test_no_cache_read_tokens_no_subtraction(mock_client):
+    """Test that when there are no cache_read_tokens, input_tokens remain unchanged.
+
+    This tests the conditional check before the subtraction in callbacks.py line 757.
+    """
+    prompt = ChatPromptTemplate.from_messages(
+        [("user", "Normal request without cache")]
+    )
+
+    # No cache usage - input_tokens should remain as-is
+    model = FakeMessagesListChatModel(
+        responses=[
+            AIMessage(
+                content="Response without cache.",
+                usage_metadata={
+                    "input_tokens": 100,
+                    "output_tokens": 30,
+                    "total_tokens": 130,
+                    # No cache_read_input_tokens
+                },
+            )
+        ]
+    )
+
+    callbacks = [CallbackHandler(mock_client)]
+    chain = prompt | model
+    result = chain.invoke({}, config={"callbacks": callbacks})
+
+    assert result.content == "Response without cache."
+    assert mock_client.capture.call_count == 3
+
+    generation_args = mock_client.capture.call_args_list[1][1]
+    generation_props = generation_args["properties"]
+
+    assert generation_args["event"] == "$ai_generation"
+    # Input tokens should remain unchanged at 100
+    assert generation_props["$ai_input_tokens"] == 100
+    assert generation_props["$ai_output_tokens"] == 30
+    assert generation_props["$ai_cache_read_input_tokens"] == 0
+
+
+def test_zero_input_tokens_with_cache_read(mock_client):
+    """Test edge case where input_tokens is 0 but cache_read_tokens exist.
+
+    This tests the falsy check in the conditional (line 757).
+    """
+    prompt = ChatPromptTemplate.from_messages([("user", "Edge case query")])
+
+    # Edge case: input_tokens is 0 (falsy), should skip subtraction
+    model = FakeMessagesListChatModel(
+        responses=[
+            AIMessage(
+                content="Response.",
+                usage_metadata={
+                    "input_tokens": 0,
+                    "output_tokens": 10,
+                    "total_tokens": 10,
+                    "cache_read_input_tokens": 50,
+                },
+            )
+        ]
+    )
+
+    callbacks = [CallbackHandler(mock_client)]
+    chain = prompt | model
+    result = chain.invoke({}, config={"callbacks": callbacks})
+
+    assert result.content == "Response."
+    assert mock_client.capture.call_count == 3
+
+    generation_args = mock_client.capture.call_args_list[1][1]
+    generation_props = generation_args["properties"]
+
+    assert generation_args["event"] == "$ai_generation"
+    # Input tokens should remain 0 (no subtraction because input_tokens is falsy)
+    assert generation_props["$ai_input_tokens"] == 0
+    assert generation_props["$ai_output_tokens"] == 10
+    assert generation_props["$ai_cache_read_input_tokens"] == 50
+
+
+def test_cache_write_tokens_not_subtracted_from_input(mock_client):
+    """Test that cache_creation_input_tokens (cache write) do NOT affect input_tokens.
+
+    Only cache_read_tokens should be subtracted from input_tokens, not cache_write_tokens.
+    """
+    prompt = ChatPromptTemplate.from_messages([("user", "Create cache")])
+
+    # Cache creation without cache read
+    model = FakeMessagesListChatModel(
+        responses=[
+            AIMessage(
+                content="Creating cache.",
+                usage_metadata={
+                    "input_tokens": 1000,
+                    "output_tokens": 20,
+                    "total_tokens": 1020,
+                    "cache_creation_input_tokens": 800,  # Cache write, not read
+                },
+            )
+        ]
+    )
+
+    callbacks = [CallbackHandler(mock_client)]
+    chain = prompt | model
+    result = chain.invoke({}, config={"callbacks": callbacks})
+
+    assert result.content == "Creating cache."
+    assert mock_client.capture.call_count == 3
+
+    generation_args = mock_client.capture.call_args_list[1][1]
+    generation_props = generation_args["properties"]
+
+    assert generation_args["event"] == "$ai_generation"
+    # Input tokens should NOT be reduced by cache_creation_input_tokens
+    assert generation_props["$ai_input_tokens"] == 1000
+    assert generation_props["$ai_output_tokens"] == 20
+    assert generation_props["$ai_cache_creation_input_tokens"] == 800
+    assert generation_props["$ai_cache_read_input_tokens"] == 0
+
+
+def test_agent_action_and_finish_imports():
+    """
+    Regression test for LangChain 1.0+ compatibility (Issue #362).
+    Verifies that AgentAction and AgentFinish can be imported and used.
+    This test ensures the imports work with both LangChain 0.x and 1.0+.
+    """
+    # Import the types that caused the compatibility issue
+    try:
+        from langchain_core.agents import AgentAction, AgentFinish
+    except (ImportError, ModuleNotFoundError):
+        from langchain.schema.agent import AgentAction, AgentFinish  # type: ignore
+
+    # Verify they're available in the callbacks module
+    from posthog.ai.langchain.callbacks import CallbackHandler
+
+    # Test on_agent_action with mock data
+    mock_client = MagicMock()
+    callbacks = CallbackHandler(mock_client)
+    run_id = uuid.uuid4()
+    parent_run_id = uuid.uuid4()
+
+    # Create mock AgentAction
+    action = AgentAction(tool="test_tool", tool_input="test_input", log="test_log")
+
+    # Should not raise an exception
+    callbacks.on_agent_action(action, run_id=run_id, parent_run_id=parent_run_id)
+
+    # Verify parent was set
+    assert run_id in callbacks._parent_tree
+    assert callbacks._parent_tree[run_id] == parent_run_id
+
+    # Test on_agent_finish with mock data
+    finish = AgentFinish(return_values={"output": "test_output"}, log="finish_log")
+
+    # Should not raise an exception
+    callbacks.on_agent_finish(finish, run_id=run_id, parent_run_id=parent_run_id)
+
+    # Verify capture was called
+    assert mock_client.capture.call_count == 1
+    call_args = mock_client.capture.call_args[1]
+    assert call_args["event"] == "$ai_span"
+
+
+def test_posthog_properties_field_in_generation_metadata(mock_client):
+    """Test that posthog_properties is properly stored in GenerationMetadata."""
+    callbacks = CallbackHandler(mock_client)
+    run_id = uuid.uuid4()
+
+    # Test with billable=True
+    with patch("time.time", return_value=1234567890):
+        callbacks._set_llm_metadata(
+            {"kwargs": {"openai_api_base": "https://api.openai.com"}},
+            run_id,
+            messages=[{"role": "user", "content": "Test message"}],
+            invocation_params={"temperature": 0.5},
+            metadata={
+                "ls_model_name": "gpt-4o",
+                "ls_provider": "openai",
+                "posthog_properties": {"$ai_billable": True},
+            },
+            name="test",
+        )
+
+    expected = GenerationMetadata(
+        model="gpt-4o",
+        input=[{"role": "user", "content": "Test message"}],
+        start_time=1234567890,
+        model_params={"temperature": 0.5},
+        provider="openai",
+        base_url="https://api.openai.com",
+        name="test",
+        posthog_properties={"$ai_billable": True},
+        end_time=None,
+    )
+    assert callbacks._runs[run_id] == expected
+    assert callbacks._runs[run_id].posthog_properties == {"$ai_billable": True}
+
+    callbacks._pop_run_metadata(run_id)
+
+    # Test with billable=False (explicit)
+    run_id2 = uuid.uuid4()
+    with patch("time.time", return_value=1234567890):
+        callbacks._set_llm_metadata(
+            {"kwargs": {"openai_api_base": "https://api.openai.com"}},
+            run_id2,
+            messages=[{"role": "user", "content": "Test message"}],
+            invocation_params={"temperature": 0.5},
+            metadata={
+                "ls_model_name": "gpt-4o",
+                "ls_provider": "openai",
+                "posthog_properties": {"$ai_billable": False},
+            },
+            name="test",
+        )
+
+    assert callbacks._runs[run_id2].posthog_properties == {"$ai_billable": False}
+    callbacks._pop_run_metadata(run_id2)
+
+    # Test when posthog_properties not provided
+    run_id3 = uuid.uuid4()
+    with patch("time.time", return_value=1234567890):
+        callbacks._set_llm_metadata(
+            {"kwargs": {"openai_api_base": "https://api.openai.com"}},
+            run_id3,
+            messages=[{"role": "user", "content": "Test message"}],
+            invocation_params={"temperature": 0.5},
+            metadata={"ls_model_name": "gpt-4o", "ls_provider": "openai"},
+            name="test",
+        )
+
+    assert callbacks._runs[run_id3].posthog_properties is None
+
+
+def test_billable_property_in_generation_event(mock_client):
+    """Test that the billable property is captured in the $ai_generation event."""
+    callbacks = CallbackHandler(mock_client)
+
+    # We need to test the _set_llm_metadata directly since FakeMessagesListChatModel
+    # doesn't support metadata in the same way as real models
+    run_id = uuid.uuid4()
+    with patch("time.time", return_value=1234567890):
+        callbacks._set_llm_metadata(
+            {},
+            run_id,
+            messages=[{"role": "user", "content": "Test"}],
+            metadata={
+                "posthog_properties": {"$ai_billable": True},
+                "ls_model_name": "test-model",
+            },
+            invocation_params={},
+        )
+
+    mock_response = MagicMock()
+    mock_response.generations = [[MagicMock()]]
+
+    with patch("time.time", return_value=1234567891):
+        run = callbacks._pop_run_metadata(run_id)
+
+    callbacks._capture_generation(
+        trace_id=run_id,
+        run_id=run_id,
+        run=run,
+        output=mock_response,
+        parent_run_id=None,
+    )
+
+    assert mock_client.capture.call_count == 1
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_billable"] is True
+
+
+def test_billable_defaults_to_false_in_event(mock_client):
+    """Test that $ai_billable is not present when not specified."""
+    prompt = ChatPromptTemplate.from_messages([("user", "Test query")])
+    model = FakeMessagesListChatModel(
+        responses=[AIMessage(content="Test response")],
+    )
+
+    callbacks = [CallbackHandler(mock_client)]
+    chain = prompt | model
+    chain.invoke({}, config={"callbacks": callbacks})
+
+    generation_call = None
+    for call in mock_client.capture.call_args_list:
+        if call[1]["event"] == "$ai_generation":
+            generation_call = call
+            break
+
+    assert generation_call is not None
+    props = generation_call[1]["properties"]
+    assert "$ai_billable" not in props
+
+
+def test_billable_with_real_chain(mock_client):
+    """Test billable tracking through a complete chain execution with mocked metadata."""
+    callbacks = CallbackHandler(mock_client)
+    run_id = uuid.uuid4()
+
+    with patch("time.time", return_value=1000.0):
+        callbacks._set_llm_metadata(
+            {},
+            run_id,
+            messages=[{"role": "user", "content": "What's the weather?"}],
+            metadata={
+                "ls_model_name": "fake-model",
+                "ls_provider": "fake",
+                "posthog_properties": {"$ai_billable": True},
+            },
+            invocation_params={"temperature": 0.7},
+        )
+
+    assert callbacks._runs[run_id].posthog_properties == {"$ai_billable": True}
+
+    mock_response = MagicMock()
+    mock_response.generations = [[MagicMock()]]
+
+    with patch("time.time", return_value=1001.0):
+        run = callbacks._pop_run_metadata(run_id)
+
+    callbacks._capture_generation(
+        trace_id=run_id,
+        run_id=run_id,
+        run=run,
+        output=mock_response,
+        parent_run_id=None,
+    )
+
+    assert mock_client.capture.call_count == 1
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_billable"] is True
+    assert props["$ai_model"] == "fake-model"
+    assert props["$ai_provider"] == "fake"

posthog/test/ai/openai/test_openai.py

@@ -1,5 +1,5 @@
 import time
-from unittest.mock import patch
+from unittest.mock import AsyncMock, patch
 
 import pytest
 
@@ -34,6 +34,7 @@ try:
     )
 
     from posthog.ai.openai import OpenAI
+    from posthog.ai.openai.openai_async import AsyncOpenAI
 
     OPENAI_AVAILABLE = True
 except ImportError:
@@ -218,6 +219,106 @@ def mock_openai_response_with_cached_tokens():
     )
 
 
+@pytest.fixture
+def streaming_tool_call_chunks():
+    return [
+        ChatCompletionChunk(
+            id="chunk1",
+            model="gpt-4",
+            object="chat.completion.chunk",
+            created=1234567890,
+            choices=[
+                ChoiceChunk(
+                    index=0,
+                    delta=ChoiceDelta(
+                        role="assistant",
+                        tool_calls=[
+                            ChoiceDeltaToolCall(
+                                index=0,
+                                id="call_abc123",
+                                type="function",
+                                function=ChoiceDeltaToolCallFunction(
+                                    name="get_weather",
+                                    arguments='{"location": "',
+                                ),
+                            )
+                        ],
+                    ),
+                    finish_reason=None,
+                )
+            ],
+        ),
+        ChatCompletionChunk(
+            id="chunk2",
+            model="gpt-4",
+            object="chat.completion.chunk",
+            created=1234567891,
+            choices=[
+                ChoiceChunk(
+                    index=0,
+                    delta=ChoiceDelta(
+                        tool_calls=[
+                            ChoiceDeltaToolCall(
+                                index=0,
+                                id="call_abc123",
+                                type="function",
+                                function=ChoiceDeltaToolCallFunction(
+                                    arguments='San Francisco"',
+                                ),
+                            )
+                        ],
+                    ),
+                    finish_reason=None,
+                )
+            ],
+        ),
+        ChatCompletionChunk(
+            id="chunk3",
+            model="gpt-4",
+            object="chat.completion.chunk",
+            created=1234567892,
+            choices=[
+                ChoiceChunk(
+                    index=0,
+                    delta=ChoiceDelta(
+                        tool_calls=[
+                            ChoiceDeltaToolCall(
+                                index=0,
+                                id="call_abc123",
+                                type="function",
+                                function=ChoiceDeltaToolCallFunction(
+                                    arguments=', "unit": "celsius"}',
+                                ),
+                            )
+                        ],
+                    ),
+                    finish_reason=None,
+                )
+            ],
+        ),
+        ChatCompletionChunk(
+            id="chunk4",
+            model="gpt-4",
+            object="chat.completion.chunk",
+            created=1234567893,
+            choices=[
+                ChoiceChunk(
+                    index=0,
+                    delta=ChoiceDelta(
+                        content="The weather in San Francisco is 15°C.",
+                    ),
+                    finish_reason=None,
+                )
+            ],
+            usage=CompletionUsage(
+                prompt_tokens=20,
+                completion_tokens=15,
+                total_tokens=35,
+            ),
+        ),
+    ]
+
+
 @pytest.fixture
 def mock_openai_response_with_tool_calls():
     return ChatCompletion(
@@ -734,109 +835,11 @@ def test_responses_api_tool_calls(mock_client, mock_responses_api_with_tool_call
         assert props["$ai_http_status"] == 200
 
 
-def test_streaming_with_tool_calls(mock_client):
-    # Create mock tool call chunks that will be returned in sequence
-    tool_call_chunks = [
-        ChatCompletionChunk(
-            id="chunk1",
-            model="gpt-4",
-            object="chat.completion.chunk",
-            created=1234567890,
-            choices=[
-                ChoiceChunk(
-                    index=0,
-                    delta=ChoiceDelta(
-                        role="assistant",
-                        tool_calls=[
-                            ChoiceDeltaToolCall(
-                                index=0,
-                                id="call_abc123",
-                                type="function",
-                                function=ChoiceDeltaToolCallFunction(
-                                    name="get_weather",
-                                    arguments='{"location": "',
-                                ),
-                            )
-                        ],
-                    ),
-                    finish_reason=None,
-                )
-            ],
-        ),
-        ChatCompletionChunk(
-            id="chunk2",
-            model="gpt-4",
-            object="chat.completion.chunk",
-            created=1234567891,
-            choices=[
-                ChoiceChunk(
-                    index=0,
-                    delta=ChoiceDelta(
-                        tool_calls=[
-                            ChoiceDeltaToolCall(
-                                index=0,
-                                id="call_abc123",
-                                type="function",
-                                function=ChoiceDeltaToolCallFunction(
-                                    arguments='San Francisco"',
-                                ),
-                            )
-                        ],
-                    ),
-                    finish_reason=None,
-                )
-            ],
-        ),
-        ChatCompletionChunk(
-            id="chunk3",
-            model="gpt-4",
-            object="chat.completion.chunk",
-            created=1234567892,
-            choices=[
-                ChoiceChunk(
-                    index=0,
-                    delta=ChoiceDelta(
-                        tool_calls=[
-                            ChoiceDeltaToolCall(
-                                index=0,
-                                id="call_abc123",
-                                type="function",
-                                function=ChoiceDeltaToolCallFunction(
-                                    arguments=', "unit": "celsius"}',
-                                ),
-                            )
-                        ],
-                    ),
-                    finish_reason=None,
-                )
-            ],
-        ),
-        ChatCompletionChunk(
-            id="chunk4",
-            model="gpt-4",
-            object="chat.completion.chunk",
-            created=1234567893,
-            choices=[
-                ChoiceChunk(
-                    index=0,
-                    delta=ChoiceDelta(
-                        content="The weather in San Francisco is 15°C.",
-                    ),
-                    finish_reason=None,
-                )
-            ],
-            usage=CompletionUsage(
-                prompt_tokens=20,
-                completion_tokens=15,
-                total_tokens=35,
-            ),
-        ),
-    ]
-
+def test_streaming_with_tool_calls(mock_client, streaming_tool_call_chunks):
     # Mock the create method to return our chunks
     with patch("openai.resources.chat.completions.Completions.create") as mock_create:
         # Set up the mock to return our chunks when iterated
-        mock_create.return_value = tool_call_chunks
+        mock_create.return_value = streaming_tool_call_chunks
 
         client = OpenAI(api_key="test-key", posthog_client=mock_client)
 
@@ -865,7 +868,7 @@ def test_streaming_with_tool_calls(mock_client):
 
         # Verify the chunks were returned correctly
         assert len(chunks) == 4
-        assert chunks == tool_call_chunks
+        assert chunks == streaming_tool_call_chunks
 
         # Verify the capture was called with the right arguments
         assert mock_client.capture.call_count == 1
@@ -890,10 +893,29 @@ def test_streaming_with_tool_calls(mock_client):
         assert defined_tool["function"]["description"] == "Get weather"
         assert defined_tool["function"]["parameters"] == {}
 
-        # Check that the content was also accumulated
+        # Check that both text content and tool calls were accumulated
+        output_content = props["$ai_output_choices"][0]["content"]
+
+        # Find text content and tool call in the output
+        text_content = None
+        tool_call_content = None
+        for item in output_content:
+            if item["type"] == "text":
+                text_content = item
+            elif item["type"] == "function":
+                tool_call_content = item
+
+        # Verify text content
+        assert text_content is not None
+        assert text_content["text"] == "The weather in San Francisco is 15°C."
+
+        # Verify tool call was captured
+        assert tool_call_content is not None
+        assert tool_call_content["id"] == "call_abc123"
+        assert tool_call_content["function"]["name"] == "get_weather"
         assert (
-            props["$ai_output_choices"][0]["content"]
-            == "The weather in San Francisco is 15°C."
+            tool_call_content["function"]["arguments"]
+            == '{"location": "San Francisco", "unit": "celsius"}'
         )
 
         # Check token usage
@@ -1014,6 +1036,248 @@ def test_responses_parse(mock_client, mock_parsed_response):
         assert isinstance(props["$ai_latency"], float)
 
 
+def test_responses_api_streaming_with_tokens(mock_client):
+    """Test that Responses API streaming properly captures token usage from response.usage."""
+    from openai.types.responses import ResponseUsage
+    from unittest.mock import MagicMock
+
+    # Create mock response chunks with usage data in the correct location
+    chunks = []
+
+    # First chunk - just content, no usage
+    chunk1 = MagicMock()
+    chunk1.type = "response.text.delta"
+    chunk1.text = "Test "
+    chunks.append(chunk1)
+
+    # Second chunk - more content
+    chunk2 = MagicMock()
+    chunk2.type = "response.text.delta"
+    chunk2.text = "response"
+    chunks.append(chunk2)
+
+    # Final chunk - completed event with usage in response.usage
+    chunk3 = MagicMock()
+    chunk3.type = "response.completed"
+    chunk3.response = MagicMock()
+    chunk3.response.usage = ResponseUsage(
+        input_tokens=25,
+        output_tokens=30,
+        total_tokens=55,
+        input_tokens_details={"prompt_tokens": 25, "cached_tokens": 0},
+        output_tokens_details={"reasoning_tokens": 0},
+    )
+    chunk3.response.output = ["Test response"]
+    chunks.append(chunk3)
+
+    captured_kwargs = {}
+
+    def mock_streaming_response(**kwargs):
+        # Capture the kwargs to verify stream_options was NOT added
+        captured_kwargs.update(kwargs)
+        return iter(chunks)
+
+    with patch(
+        "openai.resources.responses.Responses.create",
+        side_effect=mock_streaming_response,
+    ):
+        client = OpenAI(api_key="test-key", posthog_client=mock_client)
+
+        # Consume the streaming response
+        response = client.responses.create(
+            model="gpt-4o-mini",
+            input=[{"role": "user", "content": "Test message"}],
+            stream=True,
+            posthog_distinct_id="test-id",
+            posthog_properties={"test": "streaming"},
+        )
+
+        # Consume all chunks
+        list(response)
+
+        # Verify stream_options was NOT added (Responses API doesn't support it)
+        assert "stream_options" not in captured_kwargs
+
+        # Verify capture was called
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        # Verify tokens are captured correctly from response.usage (not 0)
+        assert call_args["distinct_id"] == "test-id"
+        assert call_args["event"] == "$ai_generation"
+        assert props["$ai_provider"] == "openai"
+        assert props["$ai_model"] == "gpt-4o-mini"
+        assert props["$ai_input_tokens"] == 25  # Should not be 0
+        assert props["$ai_output_tokens"] == 30  # Should not be 0
+        assert props["test"] == "streaming"
+        assert isinstance(props["$ai_latency"], float)
+
+
+@pytest.mark.asyncio
+async def test_async_chat_streaming_with_tool_calls(
+    mock_client, streaming_tool_call_chunks
+):
+    captured_kwargs = {}
+
+    async def mock_create(self, **kwargs):
+        captured_kwargs["kwargs"] = kwargs
+
+        async def chunk_iterable():
+            for chunk in streaming_tool_call_chunks:
+                yield chunk
+
+        return chunk_iterable()
+
+    with patch(
+        "openai.resources.chat.completions.AsyncCompletions.create", new=mock_create
+    ):
+        client = AsyncOpenAI(api_key="test-key", posthog_client=mock_client)
+
+        response_stream = await client.chat.completions.create(
+            model="gpt-4",
+            messages=[
+                {"role": "user", "content": "What's the weather in San Francisco?"}
+            ],
+            tools=[
+                {
+                    "type": "function",
+                    "function": {
+                        "name": "get_weather",
+                        "description": "Get weather",
+                        "parameters": {},
+                    },
+                }
+            ],
+            stream=True,
+            posthog_distinct_id="test-id",
+        )
+
+        chunks = []
+        async for chunk in response_stream:
+            chunks.append(chunk)
+
+    kwargs = captured_kwargs["kwargs"]
+    assert kwargs["stream_options"]["include_usage"] is True
+
+    assert len(chunks) == len(streaming_tool_call_chunks)
+    assert chunks == streaming_tool_call_chunks
+
+    assert mock_client.capture.call_count == 1
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "openai"
+    assert props["$ai_model"] == "gpt-4"
+    assert props["$ai_output_tokens"] == 15
+    assert props["$ai_input_tokens"] == 20
+    assert isinstance(props["$ai_latency"], float)
+
+
+@pytest.mark.asyncio
+async def test_async_responses_streaming_with_tokens(mock_client):
+    from openai.types.responses import ResponseUsage
+    from unittest.mock import MagicMock
+
+    chunks = []
+
+    chunk1 = MagicMock()
+    chunk1.type = "response.text.delta"
+    chunk1.text = "Test "
+    chunks.append(chunk1)
+
+    chunk2 = MagicMock()
+    chunk2.type = "response.text.delta"
+    chunk2.text = "response"
+    chunks.append(chunk2)
+
+    chunk3 = MagicMock()
+    chunk3.type = "response.completed"
+    chunk3.response = MagicMock()
+    chunk3.response.usage = ResponseUsage(
+        input_tokens=25,
+        output_tokens=30,
+        total_tokens=55,
+        input_tokens_details={"prompt_tokens": 25, "cached_tokens": 0},
+        output_tokens_details={"reasoning_tokens": 0},
+    )
+    chunk3.response.output = ["Test response"]
+    chunks.append(chunk3)
+
+    captured_kwargs = {}
+
+    async def mock_create(self, **kwargs):
+        captured_kwargs["kwargs"] = kwargs
+
+        async def chunk_iterable():
+            for chunk in chunks:
+                yield chunk
+
+        return chunk_iterable()
+
+    with patch("openai.resources.responses.AsyncResponses.create", new=mock_create):
+        client = AsyncOpenAI(api_key="test-key", posthog_client=mock_client)
+
+        response_stream = await client.responses.create(
+            model="gpt-4o-mini",
+            input=[{"role": "user", "content": "Test message"}],
+            stream=True,
+            posthog_distinct_id="test-id",
+            posthog_properties={"test": "streaming"},
+        )
+
+        async for _ in response_stream:
+            pass
+
+    kwargs = captured_kwargs["kwargs"]
+    assert "stream_options" not in kwargs
+
+    assert mock_client.capture.call_count == 1
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "openai"
+    assert props["$ai_model"] == "gpt-4o-mini"
+    assert props["$ai_input_tokens"] == 25
+    assert props["$ai_output_tokens"] == 30
+    assert props["test"] == "streaming"
+    assert isinstance(props["$ai_latency"], float)
+
+
+@pytest.mark.asyncio
+async def test_async_embeddings_create(mock_client, mock_embedding_response):
+    mock_create = AsyncMock(return_value=mock_embedding_response)
+
+    with patch("openai.resources.embeddings.AsyncEmbeddings.create", new=mock_create):
+        client = AsyncOpenAI(api_key="test-key", posthog_client=mock_client)
+
+        response = await client.embeddings.create(
+            model="text-embedding-3-small",
+            input="Hello world",
+            posthog_distinct_id="test-id",
+            posthog_properties={"foo": "bar"},
+        )
+
+    assert response == mock_embedding_response
+    assert mock_create.await_count == 1
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_embedding"
+    assert props["$ai_provider"] == "openai"
+    assert props["$ai_model"] == "text-embedding-3-small"
+    assert props["foo"] == "bar"
+    assert isinstance(props["$ai_latency"], float)
+
+
 def test_tool_definition(mock_client, mock_openai_response):
     """Test that tools defined in the create function are captured in $ai_tools property"""
     with patch(
@@ -1075,3 +1339,340 @@ def test_tool_definition(mock_client, mock_openai_response):
         assert isinstance(props["$ai_latency"], float)
         # Verify that tools are captured in the $ai_tools property
         assert props["$ai_tools"] == tools
+
+
+def test_web_search_perplexity_style(mock_client):
+    """Test web search detection via annotations (Perplexity-style)."""
+
+    class MockAnnotation:
+        def __init__(self):
+            self.type = "url_citation"
+
+    class MockMessage:
+        def __init__(self):
+            self.role = "assistant"
+            self.content = "Based on recent search results..."
+            self.annotations = [MockAnnotation(), MockAnnotation()]
+
+    class MockChoice:
+        def __init__(self):
+            self.message = MockMessage()
+
+    class MockUsage:
+        def __init__(self):
+            self.prompt_tokens = 50
+            self.completion_tokens = 30
+
+    class MockResponseWithAnnotations:
+        def __init__(self):
+            self.choices = [MockChoice()]
+            self.usage = MockUsage()
+            self.model = "gpt-4-turbo"
+
+    mock_response = MockResponseWithAnnotations()
+
+    with patch("openai.resources.chat.Completions.create", return_value=mock_response):
+        client = OpenAI(api_key="test-key", posthog_client=mock_client)
+        response = client.chat.completions.create(
+            model="gpt-4-turbo",
+            messages=[{"role": "user", "content": "What's happening in tech?"}],
+            posthog_distinct_id="test-id",
+        )
+
+        assert response == mock_response
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        # Verify web search count is detected (binary detection)
+        assert props["$ai_web_search_count"] == 1
+
+
+def test_web_search_responses_api(mock_client):
+    """Test exact web search count from Responses API."""
+
+    class MockWebSearchItem:
+        def __init__(self):
+            self.type = "web_search_call"
+
+    class MockMessageItem:
+        def __init__(self):
+            self.type = "message"
+            self.role = "assistant"
+            self.content = "Here are the results..."
+
+    class MockUsage:
+        def __init__(self):
+            self.input_tokens = 100
+            self.output_tokens = 75
+
+    class MockResponsesAPIResponse:
+        def __init__(self):
+            self.output = [MockWebSearchItem(), MockWebSearchItem(), MockMessageItem()]
+            self.usage = MockUsage()
+            self.model = "gpt-4o"
+
+    mock_response = MockResponsesAPIResponse()
+
+    with patch(
+        "openai.resources.responses.Responses.create", return_value=mock_response
+    ):
+        # Manually call the tracking since we're testing the converter logic
+        from posthog.ai.utils import call_llm_and_track_usage
+
+        def mock_create_call(**kwargs):
+            return mock_response
+
+        result = call_llm_and_track_usage(
+            posthog_distinct_id="test-id",
+            ph_client=mock_client,
+            provider="openai",
+            posthog_trace_id=None,
+            posthog_properties=None,
+            posthog_privacy_mode=False,
+            posthog_groups=None,
+            base_url="https://api.openai.com/v1",
+            call_method=mock_create_call,
+            model="gpt-4o",
+            messages=[{"role": "user", "content": "Search query"}],
+        )
+
+        assert result == mock_response
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        # Verify exact web search count
+        assert props["$ai_web_search_count"] == 2
+
+
+@pytest.fixture
+def streaming_web_search_chunks():
+    """Streaming chunks with web search indicators (Perplexity-style)."""
+    return [
+        ChatCompletionChunk(
+            id="chunk1",
+            model="gpt-4",
+            object="chat.completion.chunk",
+            created=1234567890,
+            choices=[
+                ChoiceChunk(
+                    index=0,
+                    delta=ChoiceDelta(
+                        role="assistant",
+                        content="Based on my search, ",
+                    ),
+                    finish_reason=None,
+                )
+            ],
+        ),
+        ChatCompletionChunk(
+            id="chunk2",
+            model="gpt-4",
+            object="chat.completion.chunk",
+            created=1234567891,
+            choices=[
+                ChoiceChunk(
+                    index=0,
+                    delta=ChoiceDelta(
+                        content="here are the latest news...",
+                    ),
+                    finish_reason=None,
+                )
+            ],
+        ),
+        ChatCompletionChunk(
+            id="chunk3",
+            model="gpt-4",
+            object="chat.completion.chunk",
+            created=1234567892,
+            choices=[
+                ChoiceChunk(
+                    index=0,
+                    delta=ChoiceDelta(),
+                    finish_reason="stop",
+                )
+            ],
+            usage=CompletionUsage(
+                prompt_tokens=20,
+                completion_tokens=15,
+                total_tokens=35,
+            ),
+        ),
+    ]
+
+
+def test_streaming_with_web_search(mock_client, streaming_web_search_chunks):
+    """Test that web search count is properly captured in streaming mode."""
+
+    # Add citations attribute to the last chunk to indicate web search was used
+    streaming_web_search_chunks[-1].citations = ["https://example.com/news"]
+
+    with patch("openai.resources.chat.completions.Completions.create") as mock_create:
+        mock_create.return_value = streaming_web_search_chunks
+
+        client = OpenAI(api_key="test-key", posthog_client=mock_client)
+        response_generator = client.chat.completions.create(
+            model="gpt-4",
+            messages=[{"role": "user", "content": "Search for recent news"}],
+            stream=True,
+            posthog_distinct_id="test-id",
+        )
+
+        # Consume the generator to trigger the event capture
+        chunks = list(response_generator)
+
+        # Verify the chunks were returned correctly
+        assert len(chunks) == 3
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        # Verify web search count is captured (binary detection = 1)
+        assert props["$ai_web_search_count"] == 1
+        assert props["$ai_input_tokens"] == 20
+        assert props["$ai_output_tokens"] == 15
+
+
+def test_streaming_with_web_search_on_non_usage_chunk(
+    mock_client, streaming_web_search_chunks
+):
+    """Test that web search count is captured even when citations appear on chunks without usage data."""
+
+    # Add citations attribute to the FIRST chunk (which has no usage data)
+    # This tests the fix for the bug where web search indicators on non-usage chunks were ignored
+    streaming_web_search_chunks[0].citations = ["https://example.com/news"]
+
+    with patch("openai.resources.chat.completions.Completions.create") as mock_create:
+        mock_create.return_value = streaming_web_search_chunks
+
+        client = OpenAI(api_key="test-key", posthog_client=mock_client)
+        response_generator = client.chat.completions.create(
+            model="gpt-4",
+            messages=[{"role": "user", "content": "Search for recent news"}],
+            stream=True,
+            posthog_distinct_id="test-id",
+        )
+
+        # Consume the generator to trigger the event capture
+        chunks = list(response_generator)
+
+        # Verify the chunks were returned correctly
+        assert len(chunks) == 3
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        # Verify web search count is captured even though citations were on first chunk
+        assert props["$ai_web_search_count"] == 1
+        assert props["$ai_input_tokens"] == 20
+        assert props["$ai_output_tokens"] == 15
+
+
+@pytest.mark.asyncio
+async def test_async_chat_with_web_search(mock_client):
+    """Test that web search count is properly tracked in async non-streaming mode."""
+
+    # Create mock response with citations (Perplexity-style)
+    mock_response = ChatCompletion(
+        id="chatcmpl-test",
+        model="gpt-4",
+        object="chat.completion",
+        created=1234567890,
+        choices=[
+            Choice(
+                index=0,
+                message=ChatCompletionMessage(
+                    role="assistant",
+                    content="Here are the search results...",
+                ),
+                finish_reason="stop",
+            )
+        ],
+        usage=CompletionUsage(
+            prompt_tokens=20,
+            completion_tokens=15,
+            total_tokens=35,
+        ),
+    )
+
+    # Add citations attribute to indicate web search
+    mock_response.citations = ["https://example.com/result1"]
+
+    async def mock_create(self, **kwargs):
+        return mock_response
+
+    with patch(
+        "openai.resources.chat.completions.AsyncCompletions.create", new=mock_create
+    ):
+        client = AsyncOpenAI(api_key="test-key", posthog_client=mock_client)
+
+        response = await client.chat.completions.create(
+            model="gpt-4",
+            messages=[{"role": "user", "content": "Search for recent news"}],
+            posthog_distinct_id="test-id",
+        )
+
+    assert response == mock_response
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Verify web search count is captured (binary detection = 1)
+    assert props["$ai_web_search_count"] == 1
+    assert props["$ai_input_tokens"] == 20
+    assert props["$ai_output_tokens"] == 15
+
+
+@pytest.mark.asyncio
+async def test_async_chat_streaming_with_web_search(
+    mock_client, streaming_web_search_chunks
+):
+    """Test that web search count is properly captured in async streaming mode."""
+
+    # Add citations attribute to the last chunk to indicate web search was used
+    streaming_web_search_chunks[-1].citations = ["https://example.com/news"]
+
+    captured_kwargs = {}
+
+    async def mock_create(self, **kwargs):
+        captured_kwargs["kwargs"] = kwargs
+
+        async def chunk_iterable():
+            for chunk in streaming_web_search_chunks:
+                yield chunk
+
+        return chunk_iterable()
+
+    with patch(
+        "openai.resources.chat.completions.AsyncCompletions.create", new=mock_create
+    ):
+        client = AsyncOpenAI(api_key="test-key", posthog_client=mock_client)
+
+        response_stream = await client.chat.completions.create(
+            model="gpt-4",
+            messages=[{"role": "user", "content": "Search for recent news"}],
+            stream=True,
+            posthog_distinct_id="test-id",
+        )
+
+        chunks = []
+        async for chunk in response_stream:
+            chunks.append(chunk)
+
+    # Verify the chunks were returned correctly
+    assert len(chunks) == 3
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    # Verify web search count is captured (binary detection = 1)
+    assert props["$ai_web_search_count"] == 1
+    assert props["$ai_input_tokens"] == 20
+    assert props["$ai_output_tokens"] == 15

posthog/test/ai/test_sanitization.py

@@ -0,0 +1,335 @@
+import unittest
+
+from posthog.ai.sanitization import (
+    redact_base64_data_url,
+    sanitize_openai,
+    sanitize_openai_response,
+    sanitize_anthropic,
+    sanitize_gemini,
+    sanitize_langchain,
+    is_base64_data_url,
+    is_raw_base64,
+    REDACTED_IMAGE_PLACEHOLDER,
+)
+
+
+class TestSanitization(unittest.TestCase):
+    def setUp(self):
+        self.sample_base64_image = "data:image/jpeg;base64,/9j/4AAQSkZJRgABAQAAAQ..."
+        self.sample_base64_png = "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAUA..."
+        self.regular_url = "https://example.com/image.jpg"
+        self.raw_base64 = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUl=="
+
+    def test_is_base64_data_url(self):
+        self.assertTrue(is_base64_data_url(self.sample_base64_image))
+        self.assertTrue(is_base64_data_url(self.sample_base64_png))
+        self.assertFalse(is_base64_data_url(self.regular_url))
+        self.assertFalse(is_base64_data_url("regular text"))
+
+    def test_is_raw_base64(self):
+        self.assertTrue(is_raw_base64(self.raw_base64))
+        self.assertFalse(is_raw_base64("short"))
+        self.assertFalse(is_raw_base64(self.regular_url))
+        self.assertFalse(is_raw_base64("/path/to/file"))
+
+    def test_redact_base64_data_url(self):
+        self.assertEqual(
+            redact_base64_data_url(self.sample_base64_image), REDACTED_IMAGE_PLACEHOLDER
+        )
+        self.assertEqual(
+            redact_base64_data_url(self.sample_base64_png), REDACTED_IMAGE_PLACEHOLDER
+        )
+        self.assertEqual(redact_base64_data_url(self.regular_url), self.regular_url)
+        self.assertEqual(redact_base64_data_url(None), None)
+        self.assertEqual(redact_base64_data_url(123), 123)
+
+    def test_sanitize_openai(self):
+        input_data = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": "What is in this image?"},
+                    {
+                        "type": "image_url",
+                        "image_url": {
+                            "url": self.sample_base64_image,
+                            "detail": "high",
+                        },
+                    },
+                ],
+            }
+        ]
+
+        result = sanitize_openai(input_data)
+
+        self.assertEqual(result[0]["content"][0]["text"], "What is in this image?")
+        self.assertEqual(
+            result[0]["content"][1]["image_url"]["url"], REDACTED_IMAGE_PLACEHOLDER
+        )
+        self.assertEqual(result[0]["content"][1]["image_url"]["detail"], "high")
+
+    def test_sanitize_openai_preserves_regular_urls(self):
+        input_data = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image_url",
+                        "image_url": {"url": self.regular_url},
+                    }
+                ],
+            }
+        ]
+
+        result = sanitize_openai(input_data)
+        self.assertEqual(result[0]["content"][0]["image_url"]["url"], self.regular_url)
+
+    def test_sanitize_openai_response(self):
+        input_data = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "input_image",
+                        "image_url": self.sample_base64_image,
+                    }
+                ],
+            }
+        ]
+
+        result = sanitize_openai_response(input_data)
+        self.assertEqual(
+            result[0]["content"][0]["image_url"], REDACTED_IMAGE_PLACEHOLDER
+        )
+
+    def test_sanitize_anthropic(self):
+        input_data = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": "What is in this image?"},
+                    {
+                        "type": "image",
+                        "source": {
+                            "type": "base64",
+                            "media_type": "image/jpeg",
+                            "data": "base64data",
+                        },
+                    },
+                ],
+            }
+        ]
+
+        result = sanitize_anthropic(input_data)
+
+        self.assertEqual(result[0]["content"][0]["text"], "What is in this image?")
+        self.assertEqual(
+            result[0]["content"][1]["source"]["data"], REDACTED_IMAGE_PLACEHOLDER
+        )
+        self.assertEqual(result[0]["content"][1]["source"]["type"], "base64")
+        self.assertEqual(result[0]["content"][1]["source"]["media_type"], "image/jpeg")
+
+    def test_sanitize_gemini(self):
+        input_data = [
+            {
+                "parts": [
+                    {"text": "What is in this image?"},
+                    {
+                        "inline_data": {
+                            "mime_type": "image/jpeg",
+                            "data": "base64data",
+                        }
+                    },
+                ]
+            }
+        ]
+
+        result = sanitize_gemini(input_data)
+
+        self.assertEqual(result[0]["parts"][0]["text"], "What is in this image?")
+        self.assertEqual(
+            result[0]["parts"][1]["inline_data"]["data"], REDACTED_IMAGE_PLACEHOLDER
+        )
+        self.assertEqual(
+            result[0]["parts"][1]["inline_data"]["mime_type"], "image/jpeg"
+        )
+
+    def test_sanitize_langchain_openai_style(self):
+        input_data = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image_url",
+                        "image_url": {"url": self.sample_base64_image},
+                    }
+                ],
+            }
+        ]
+
+        result = sanitize_langchain(input_data)
+        self.assertEqual(
+            result[0]["content"][0]["image_url"]["url"], REDACTED_IMAGE_PLACEHOLDER
+        )
+
+    def test_sanitize_langchain_anthropic_style(self):
+        input_data = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image",
+                        "source": {"data": "base64data"},
+                    }
+                ],
+            }
+        ]
+
+        result = sanitize_langchain(input_data)
+        self.assertEqual(
+            result[0]["content"][0]["source"]["data"], REDACTED_IMAGE_PLACEHOLDER
+        )
+
+    def test_sanitize_with_data_url_format(self):
+        # Test that data URLs are properly detected and redacted across providers
+        data_url = "data:image/jpeg;base64,/9j/4AAQSkZJRgABAQAAAQABAAD"
+
+        # OpenAI format
+        openai_data = [
+            {
+                "role": "user",
+                "content": [{"type": "image_url", "image_url": {"url": data_url}}],
+            }
+        ]
+        result = sanitize_openai(openai_data)
+        self.assertEqual(
+            result[0]["content"][0]["image_url"]["url"], REDACTED_IMAGE_PLACEHOLDER
+        )
+
+        # Anthropic format
+        anthropic_data = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image",
+                        "source": {
+                            "type": "base64",
+                            "media_type": "image/jpeg",
+                            "data": data_url,
+                        },
+                    }
+                ],
+            }
+        ]
+        result = sanitize_anthropic(anthropic_data)
+        self.assertEqual(
+            result[0]["content"][0]["source"]["data"], REDACTED_IMAGE_PLACEHOLDER
+        )
+
+        # LangChain format
+        langchain_data = [
+            {"role": "user", "content": [{"type": "image", "data": data_url}]}
+        ]
+        result = sanitize_langchain(langchain_data)
+        self.assertEqual(result[0]["content"][0]["data"], REDACTED_IMAGE_PLACEHOLDER)
+
+    def test_sanitize_with_raw_base64(self):
+        # Test that raw base64 strings (without data URL prefix) are detected
+        raw_base64 = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUl=="
+
+        # Test with Anthropic format
+        anthropic_data = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image",
+                        "source": {
+                            "type": "base64",
+                            "media_type": "image/png",
+                            "data": raw_base64,
+                        },
+                    }
+                ],
+            }
+        ]
+        result = sanitize_anthropic(anthropic_data)
+        self.assertEqual(
+            result[0]["content"][0]["source"]["data"], REDACTED_IMAGE_PLACEHOLDER
+        )
+
+        # Test with Gemini format
+        gemini_data = [
+            {"parts": [{"inline_data": {"mime_type": "image/png", "data": raw_base64}}]}
+        ]
+        result = sanitize_gemini(gemini_data)
+        self.assertEqual(
+            result[0]["parts"][0]["inline_data"]["data"], REDACTED_IMAGE_PLACEHOLDER
+        )
+
+    def test_sanitize_preserves_regular_content(self):
+        # Ensure non-base64 content is preserved across all providers
+        regular_url = "https://example.com/image.jpg"
+        text_content = "What do you see?"
+
+        # OpenAI
+        openai_data = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": text_content},
+                    {"type": "image_url", "image_url": {"url": regular_url}},
+                ],
+            }
+        ]
+        result = sanitize_openai(openai_data)
+        self.assertEqual(result[0]["content"][0]["text"], text_content)
+        self.assertEqual(result[0]["content"][1]["image_url"]["url"], regular_url)
+
+        # Anthropic
+        anthropic_data = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": text_content},
+                    {"type": "image", "source": {"type": "url", "url": regular_url}},
+                ],
+            }
+        ]
+        result = sanitize_anthropic(anthropic_data)
+        self.assertEqual(result[0]["content"][0]["text"], text_content)
+        # URL-based images should remain unchanged
+        self.assertEqual(result[0]["content"][1]["source"]["url"], regular_url)
+
+    def test_sanitize_handles_non_dict_content(self):
+        input_data = [{"role": "user", "content": "Just text"}]
+
+        result = sanitize_openai(input_data)
+        self.assertEqual(result, input_data)
+
+    def test_sanitize_handles_none_input(self):
+        self.assertIsNone(sanitize_openai(None))
+        self.assertIsNone(sanitize_anthropic(None))
+        self.assertIsNone(sanitize_gemini(None))
+        self.assertIsNone(sanitize_langchain(None))
+
+    def test_sanitize_handles_single_message(self):
+        input_data = {
+            "role": "user",
+            "content": [
+                {
+                    "type": "image_url",
+                    "image_url": {"url": self.sample_base64_image},
+                }
+            ],
+        }
+
+        result = sanitize_openai(input_data)
+        self.assertEqual(
+            result["content"][0]["image_url"]["url"], REDACTED_IMAGE_PLACEHOLDER
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

posthog/test/ai/test_system_prompts.py

@@ -0,0 +1,354 @@
+"""
+Tests for system prompt capture across all LLM providers.
+
+This test suite ensures that system prompts are correctly captured in analytics
+regardless of how they're passed to the providers:
+- As first message in messages/contents array (standard format)
+- As separate system parameter (Anthropic, OpenAI)
+- As instructions parameter (OpenAI Responses API)
+- As system_instruction parameter (Gemini)
+"""
+
+import time
+import unittest
+from unittest.mock import patch, MagicMock
+
+
+class TestSystemPromptCapture(unittest.TestCase):
+    """Test system prompt capture for all providers."""
+
+    def setUp(self):
+        super().setUp()
+        self.test_system_prompt = "You are a helpful AI assistant."
+        self.test_user_message = "Hello, how are you?"
+        self.test_response = "I'm doing well, thank you!"
+
+        # Create mock PostHog client
+        self.client = MagicMock()
+        self.client.privacy_mode = False
+
+    def _assert_system_prompt_captured(self, captured_input):
+        """Helper to assert system prompt is correctly captured."""
+        self.assertEqual(
+            len(captured_input), 2, "Should have 2 messages (system + user)"
+        )
+        self.assertEqual(
+            captured_input[0]["role"], "system", "First message should be system"
+        )
+        self.assertEqual(
+            captured_input[0]["content"],
+            self.test_system_prompt,
+            "System content should match",
+        )
+        self.assertEqual(
+            captured_input[1]["role"], "user", "Second message should be user"
+        )
+        self.assertEqual(
+            captured_input[1]["content"],
+            self.test_user_message,
+            "User content should match",
+        )
+
+    # OpenAI Tests
+    def test_openai_messages_array_system_prompt(self):
+        """Test OpenAI with system prompt in messages array."""
+        try:
+            from posthog.ai.openai import OpenAI
+            from openai.types.chat import ChatCompletion, ChatCompletionMessage
+            from openai.types.chat.chat_completion import Choice
+            from openai.types.completion_usage import CompletionUsage
+        except ImportError:
+            self.skipTest("OpenAI package not available")
+
+        mock_response = ChatCompletion(
+            id="test",
+            model="gpt-4",
+            object="chat.completion",
+            created=int(time.time()),
+            choices=[
+                Choice(
+                    finish_reason="stop",
+                    index=0,
+                    message=ChatCompletionMessage(
+                        content=self.test_response, role="assistant"
+                    ),
+                )
+            ],
+            usage=CompletionUsage(
+                completion_tokens=10, prompt_tokens=20, total_tokens=30
+            ),
+        )
+
+        with patch(
+            "openai.resources.chat.completions.Completions.create",
+            return_value=mock_response,
+        ):
+            client = OpenAI(posthog_client=self.client, api_key="test")
+
+            messages = [
+                {"role": "system", "content": self.test_system_prompt},
+                {"role": "user", "content": self.test_user_message},
+            ]
+
+            client.chat.completions.create(
+                model="gpt-4", messages=messages, posthog_distinct_id="test-user"
+            )
+
+            self.assertEqual(len(self.client.capture.call_args_list), 1)
+            properties = self.client.capture.call_args_list[0][1]["properties"]
+            self._assert_system_prompt_captured(properties["$ai_input"])
+
+    def test_openai_separate_system_parameter(self):
+        """Test OpenAI with system prompt as separate parameter."""
+        try:
+            from posthog.ai.openai import OpenAI
+            from openai.types.chat import ChatCompletion, ChatCompletionMessage
+            from openai.types.chat.chat_completion import Choice
+            from openai.types.completion_usage import CompletionUsage
+        except ImportError:
+            self.skipTest("OpenAI package not available")
+
+        mock_response = ChatCompletion(
+            id="test",
+            model="gpt-4",
+            object="chat.completion",
+            created=int(time.time()),
+            choices=[
+                Choice(
+                    finish_reason="stop",
+                    index=0,
+                    message=ChatCompletionMessage(
+                        content=self.test_response, role="assistant"
+                    ),
+                )
+            ],
+            usage=CompletionUsage(
+                completion_tokens=10, prompt_tokens=20, total_tokens=30
+            ),
+        )
+
+        with patch(
+            "openai.resources.chat.completions.Completions.create",
+            return_value=mock_response,
+        ):
+            client = OpenAI(posthog_client=self.client, api_key="test")
+
+            messages = [{"role": "user", "content": self.test_user_message}]
+
+            client.chat.completions.create(
+                model="gpt-4",
+                messages=messages,
+                system=self.test_system_prompt,
+                posthog_distinct_id="test-user",
+            )
+
+            self.assertEqual(len(self.client.capture.call_args_list), 1)
+            properties = self.client.capture.call_args_list[0][1]["properties"]
+            self._assert_system_prompt_captured(properties["$ai_input"])
+
+    def test_openai_streaming_system_parameter(self):
+        """Test OpenAI streaming with system parameter."""
+        try:
+            from posthog.ai.openai import OpenAI
+            from openai.types.chat.chat_completion_chunk import ChatCompletionChunk
+            from openai.types.chat.chat_completion_chunk import Choice as ChoiceChunk
+            from openai.types.chat.chat_completion_chunk import ChoiceDelta
+            from openai.types.completion_usage import CompletionUsage
+        except ImportError:
+            self.skipTest("OpenAI package not available")
+
+        chunk1 = ChatCompletionChunk(
+            id="test",
+            model="gpt-4",
+            object="chat.completion.chunk",
+            created=int(time.time()),
+            choices=[
+                ChoiceChunk(
+                    finish_reason=None,
+                    index=0,
+                    delta=ChoiceDelta(content="Hello", role="assistant"),
+                )
+            ],
+        )
+
+        chunk2 = ChatCompletionChunk(
+            id="test",
+            model="gpt-4",
+            object="chat.completion.chunk",
+            created=int(time.time()),
+            choices=[
+                ChoiceChunk(
+                    finish_reason="stop",
+                    index=0,
+                    delta=ChoiceDelta(content=" there!", role=None),
+                )
+            ],
+            usage=CompletionUsage(
+                completion_tokens=10, prompt_tokens=20, total_tokens=30
+            ),
+        )
+
+        with patch(
+            "openai.resources.chat.completions.Completions.create",
+            return_value=[chunk1, chunk2],
+        ):
+            client = OpenAI(posthog_client=self.client, api_key="test")
+
+            messages = [{"role": "user", "content": self.test_user_message}]
+
+            response_generator = client.chat.completions.create(
+                model="gpt-4",
+                messages=messages,
+                system=self.test_system_prompt,
+                stream=True,
+                posthog_distinct_id="test-user",
+            )
+
+            list(response_generator)  # Consume generator
+
+            self.assertEqual(len(self.client.capture.call_args_list), 1)
+            properties = self.client.capture.call_args_list[0][1]["properties"]
+            self._assert_system_prompt_captured(properties["$ai_input"])
+
+    # Anthropic Tests
+    def test_anthropic_messages_array_system_prompt(self):
+        """Test Anthropic with system prompt in messages array."""
+        try:
+            from posthog.ai.anthropic import Anthropic
+        except ImportError:
+            self.skipTest("Anthropic package not available")
+
+        with patch("anthropic.resources.messages.Messages.create") as mock_create:
+            mock_response = MagicMock()
+            mock_response.usage.input_tokens = 20
+            mock_response.usage.output_tokens = 10
+            mock_response.usage.cache_read_input_tokens = None
+            mock_response.usage.cache_creation_input_tokens = None
+            mock_create.return_value = mock_response
+
+            client = Anthropic(posthog_client=self.client, api_key="test")
+
+            messages = [
+                {"role": "system", "content": self.test_system_prompt},
+                {"role": "user", "content": self.test_user_message},
+            ]
+
+            client.messages.create(
+                model="claude-3-5-sonnet-20241022",
+                messages=messages,
+                posthog_distinct_id="test-user",
+            )
+
+            self.assertEqual(len(self.client.capture.call_args_list), 1)
+            properties = self.client.capture.call_args_list[0][1]["properties"]
+            self._assert_system_prompt_captured(properties["$ai_input"])
+
+    def test_anthropic_separate_system_parameter(self):
+        """Test Anthropic with system prompt as separate parameter."""
+        try:
+            from posthog.ai.anthropic import Anthropic
+        except ImportError:
+            self.skipTest("Anthropic package not available")
+
+        with patch("anthropic.resources.messages.Messages.create") as mock_create:
+            mock_response = MagicMock()
+            mock_response.usage.input_tokens = 20
+            mock_response.usage.output_tokens = 10
+            mock_response.usage.cache_read_input_tokens = None
+            mock_response.usage.cache_creation_input_tokens = None
+            mock_create.return_value = mock_response
+
+            client = Anthropic(posthog_client=self.client, api_key="test")
+
+            messages = [{"role": "user", "content": self.test_user_message}]
+
+            client.messages.create(
+                model="claude-3-5-sonnet-20241022",
+                messages=messages,
+                system=self.test_system_prompt,
+                posthog_distinct_id="test-user",
+            )
+
+            self.assertEqual(len(self.client.capture.call_args_list), 1)
+            properties = self.client.capture.call_args_list[0][1]["properties"]
+            self._assert_system_prompt_captured(properties["$ai_input"])
+
+    # Gemini Tests
+    def test_gemini_contents_array_system_prompt(self):
+        """Test Gemini with system prompt in contents array."""
+        try:
+            from posthog.ai.gemini import Client
+        except ImportError:
+            self.skipTest("Gemini package not available")
+
+        with patch("google.genai.Client") as mock_genai_class:
+            mock_response = MagicMock()
+            mock_response.candidates = [MagicMock()]
+            mock_response.candidates[0].content.parts = [MagicMock()]
+            mock_response.candidates[0].content.parts[0].text = self.test_response
+            mock_response.usage_metadata.prompt_token_count = 20
+            mock_response.usage_metadata.candidates_token_count = 10
+            mock_response.usage_metadata.cached_content_token_count = None
+            mock_response.usage_metadata.thoughts_token_count = None
+
+            mock_client_instance = MagicMock()
+            mock_models_instance = MagicMock()
+            mock_models_instance.generate_content.return_value = mock_response
+            mock_client_instance.models = mock_models_instance
+            mock_genai_class.return_value = mock_client_instance
+
+            client = Client(posthog_client=self.client, api_key="test")
+
+            contents = [
+                {"role": "system", "content": self.test_system_prompt},
+                {"role": "user", "content": self.test_user_message},
+            ]
+
+            client.models.generate_content(
+                model="gemini-2.0-flash",
+                contents=contents,
+                posthog_distinct_id="test-user",
+            )
+
+            self.assertEqual(len(self.client.capture.call_args_list), 1)
+            properties = self.client.capture.call_args_list[0][1]["properties"]
+            self._assert_system_prompt_captured(properties["$ai_input"])
+
+    def test_gemini_system_instruction_parameter(self):
+        """Test Gemini with system_instruction in config parameter."""
+        try:
+            from posthog.ai.gemini import Client
+        except ImportError:
+            self.skipTest("Gemini package not available")
+
+        with patch("google.genai.Client") as mock_genai_class:
+            mock_response = MagicMock()
+            mock_response.candidates = [MagicMock()]
+            mock_response.candidates[0].content.parts = [MagicMock()]
+            mock_response.candidates[0].content.parts[0].text = self.test_response
+            mock_response.usage_metadata.prompt_token_count = 20
+            mock_response.usage_metadata.candidates_token_count = 10
+            mock_response.usage_metadata.cached_content_token_count = None
+            mock_response.usage_metadata.thoughts_token_count = None
+
+            mock_client_instance = MagicMock()
+            mock_models_instance = MagicMock()
+            mock_models_instance.generate_content.return_value = mock_response
+            mock_client_instance.models = mock_models_instance
+            mock_genai_class.return_value = mock_client_instance
+
+            client = Client(posthog_client=self.client, api_key="test")
+
+            contents = [{"role": "user", "content": self.test_user_message}]
+            config = {"system_instruction": self.test_system_prompt}
+
+            client.models.generate_content(
+                model="gemini-2.0-flash",
+                contents=contents,
+                config=config,
+                posthog_distinct_id="test-user",
+            )
+
+            self.assertEqual(len(self.client.capture.call_args_list), 1)
+            properties = self.client.capture.call_args_list[0][1]["properties"]
+            self._assert_system_prompt_captured(properties["$ai_input"])

posthog/test/integrations/test_middleware.py

@@ -4,7 +4,21 @@ from posthog.contexts import (
     get_context_distinct_id,
 )
 import unittest
-from unittest.mock import Mock
+from unittest.mock import Mock, patch
+import asyncio
+
+# Configure Django settings before importing middleware
+import django
+from django.conf import settings
+
+if not settings.configured:
+    settings.configure(
+        DEBUG=True,
+        SECRET_KEY="test-secret-key",
+        INSTALLED_APPS=[],
+        MIDDLEWARE=[],
+    )
+    django.setup()
 
 from posthog.integrations.django import PosthogContextMiddleware
 
@@ -38,14 +52,33 @@ class TestPosthogContextMiddleware(unittest.TestCase):
         request_filter=None,
         tag_map=None,
         capture_exceptions=True,
+        get_response=None,
     ):
-        """Helper to create middleware instance without calling __init__"""
-        middleware = PosthogContextMiddleware.__new__(PosthogContextMiddleware)
-        middleware.get_response = Mock()
-        middleware.extra_tags = extra_tags
-        middleware.request_filter = request_filter
-        middleware.tag_map = tag_map
-        middleware.capture_exceptions = capture_exceptions
+        """Helper to create middleware instance with mock Django settings"""
+        if get_response is None:
+            get_response = Mock()
+
+        with patch("django.conf.settings") as mock_settings:
+            # Configure mock settings
+            mock_settings.POSTHOG_MW_EXTRA_TAGS = extra_tags
+            mock_settings.POSTHOG_MW_REQUEST_FILTER = request_filter
+            mock_settings.POSTHOG_MW_TAG_MAP = tag_map
+            mock_settings.POSTHOG_MW_CAPTURE_EXCEPTIONS = capture_exceptions
+            mock_settings.POSTHOG_MW_CLIENT = None
+
+            # Make hasattr work correctly
+            def mock_hasattr(obj, name):
+                return name in [
+                    "POSTHOG_MW_EXTRA_TAGS",
+                    "POSTHOG_MW_REQUEST_FILTER",
+                    "POSTHOG_MW_TAG_MAP",
+                    "POSTHOG_MW_CAPTURE_EXCEPTIONS",
+                    "POSTHOG_MW_CLIENT",
+                ]
+
+            with patch("builtins.hasattr", side_effect=mock_hasattr):
+                middleware = PosthogContextMiddleware(get_response)
+
         return middleware
 
     def test_extract_tags_basic(self):
@@ -168,6 +201,573 @@ class TestPosthogContextMiddleware(unittest.TestCase):
 
         self.assertEqual(tags["$request_method"], "PATCH")
 
+    def test_process_exception_called_during_view_exception(self):
+        """
+        Unit test verifying process_exception captures exceptions per Django's contract.
+
+        Since this is a library test (no Django runtime), we simulate how Django
+        would invoke our middleware in production:
+        1. Middleware.__call__ creates context with request tags
+        2. View raises exception inside get_response
+        3. Django's BaseHandler catches it, calls process_exception, returns error response
+        4. Exception never propagates to middleware's context manager
+
+        We manually call process_exception to simulate Django's behavior - this is
+        the only way to test the hook without a full Django integration test.
+        """
+        mock_client = Mock()
+        view_exception = ValueError("View raised this error")
+        error_response = Mock(status_code=500)
+
+        def mock_get_response(request):
+            # Simulate Django's exception handling: catches view exception,
+            # calls process_exception hook if it exists, returns error response
+            if hasattr(middleware, "process_exception"):
+                middleware.process_exception(request, view_exception)
+            return error_response
+
+        middleware = self.create_middleware(get_response=mock_get_response)
+        middleware.client = mock_client
+
+        request = MockRequest(
+            headers={"X-POSTHOG-DISTINCT-ID": "test-user"},
+            method="POST",
+            path="/api/endpoint",
+        )
+        response = middleware(request)
+
+        self.assertEqual(response.status_code, 500)
+        mock_client.capture_exception.assert_called_once_with(view_exception)
+
+    def test_process_exception_respects_capture_exceptions_false(self):
+        """Verify process_exception respects capture_exceptions=False setting"""
+        mock_client = Mock()
+        view_exception = ValueError("Should not be captured")
+
+        def mock_get_response(request):
+            if hasattr(middleware, "process_exception"):
+                middleware.process_exception(request, view_exception)
+            return Mock(status_code=500)
+
+        middleware = self.create_middleware(
+            capture_exceptions=False, get_response=mock_get_response
+        )
+        middleware.client = mock_client
+
+        request = MockRequest()
+        middleware(request)
+
+        mock_client.capture_exception.assert_not_called()
+
+    def test_process_exception_respects_request_filter(self):
+        """Verify process_exception respects request_filter setting"""
+        mock_client = Mock()
+        view_exception = ValueError("Should be filtered")
+
+        def mock_get_response(request):
+            if hasattr(middleware, "process_exception"):
+                middleware.process_exception(request, view_exception)
+            return Mock(status_code=500)
+
+        middleware = self.create_middleware(
+            request_filter=lambda req: False,
+            capture_exceptions=True,
+            get_response=mock_get_response,
+        )
+        middleware.client = mock_client
+
+        request = MockRequest()
+        middleware(request)
+
+        mock_client.capture_exception.assert_not_called()
+
+
+class TestPosthogContextMiddlewareSync(unittest.TestCase):
+    """Test synchronous middleware behavior"""
+
+    def test_sync_middleware_call(self):
+        """Test that sync middleware correctly processes requests"""
+        mock_response = Mock()
+        get_response = Mock(return_value=mock_response)
+
+        # Create middleware with sync get_response
+        middleware = PosthogContextMiddleware(get_response)
+
+        # Verify sync mode detected
+        self.assertFalse(middleware._is_coroutine)
+
+        request = MockRequest(
+            headers={"X-POSTHOG-SESSION-ID": "test-session"},
+            method="GET",
+            path="/test",
+        )
+
+        with new_context():
+            response = middleware(request)
+
+            # Verify response returned
+            self.assertEqual(response, mock_response)
+            get_response.assert_called_once_with(request)
+
+    def test_sync_middleware_with_filter(self):
+        """Test sync middleware respects request filter"""
+        mock_response = Mock()
+        get_response = Mock(return_value=mock_response)
+
+        # Create middleware with request filter that filters all requests
+        def request_filter(req):
+            return False
+
+        middleware = PosthogContextMiddleware.__new__(PosthogContextMiddleware)
+        middleware.get_response = get_response
+        middleware._is_coroutine = False
+        middleware.request_filter = request_filter
+        middleware.capture_exceptions = True
+        middleware.client = None
+
+        request = MockRequest()
+
+        # Should skip context creation and return response directly
+        response = middleware(request)
+        self.assertEqual(response, mock_response)
+        get_response.assert_called_once_with(request)
+
+    def test_view_exceptions_only_captured_via_process_exception(self):
+        """
+        Demonstrates that process_exception is required to capture view exceptions.
+
+        In production Django, view exceptions don't propagate to middleware's context
+        manager because Django's BaseHandler catches them first and converts them to
+        error responses. Django provides the exception via process_exception hook instead.
+
+        This unit test proves:
+        1. Context manager in __call__ never sees view exceptions (Django intercepts)
+        2. Only process_exception can capture them
+        3. Without process_exception, exceptions are silently lost (v6.7.5 regression)
+
+        We manually call process_exception to verify the hook works - in production,
+        Django's BaseHandler would call it when a view raises.
+        """
+        mock_client = Mock()
+        get_response = Mock(return_value=Mock(status_code=500))
+
+        middleware = PosthogContextMiddleware(get_response)
+        middleware.client = mock_client
+
+        def get_response_simulating_django(request):
+            # Simulates Django behavior: view exception converted to error response,
+            # never propagates to middleware's context manager
+            return Mock(status_code=500)
+
+        middleware._sync_get_response = get_response_simulating_django
+
+        request = MockRequest()
+
+        response = middleware(request)
+        self.assertEqual(response.status_code, 500)
+
+        # Context manager didn't capture anything - exception was intercepted by Django
+        mock_client.capture_exception.assert_not_called()
+
+        # Verify process_exception hook exists and captures exceptions when called
+        if hasattr(middleware, "process_exception"):
+            exception = ValueError("View error")
+            middleware.process_exception(request, exception)
+            mock_client.capture_exception.assert_called_once_with(exception)
+        else:
+            self.fail(
+                "process_exception missing - view exceptions will not be captured!"
+            )
+
+
+class TestPosthogContextMiddlewareAsync(unittest.TestCase):
+    """Test asynchronous middleware behavior"""
+
+    def test_async_middleware_detection(self):
+        """Test that async get_response is correctly detected"""
+
+        async def async_get_response(request):
+            return Mock()
+
+        middleware = PosthogContextMiddleware(async_get_response)
+
+        # Verify async mode detected
+        self.assertTrue(middleware._is_coroutine)
+
+    def test_async_middleware_call(self):
+        """Test that async middleware correctly processes requests"""
+
+        async def run_test():
+            mock_response = Mock()
+
+            async def async_get_response(request):
+                return mock_response
+
+            middleware = PosthogContextMiddleware(async_get_response)
+
+            request = MockRequest(
+                headers={"X-POSTHOG-SESSION-ID": "async-session"},
+                method="POST",
+                path="/async-test",
+            )
+
+            with new_context():
+                # Call should return the coroutine from __acall__
+                result = middleware(request)
+
+                # Verify it's a coroutine
+                self.assertTrue(asyncio.iscoroutine(result))
+
+                # Await the result
+                response = await result
+                self.assertEqual(response, mock_response)
+
+        asyncio.run(run_test())
+
+    def test_async_middleware_with_filter(self):
+        """Test async middleware respects request filter"""
+
+        async def run_test():
+            mock_response = Mock()
+
+            async def async_get_response(request):
+                return mock_response
+
+            # Properly initialize middleware
+            middleware = PosthogContextMiddleware(async_get_response)
+            # Override request filter after initialization
+            middleware.request_filter = lambda req: False
+
+            request = MockRequest()
+
+            # Should skip context creation and return response directly
+            result = middleware(request)
+            response = await result
+            self.assertEqual(response, mock_response)
+
+        asyncio.run(run_test())
+
+    def test_async_middleware_context_propagation(self):
+        """Test that async middleware properly propagates context"""
+
+        async def run_test():
+            mock_response = Mock()
+
+            async def async_get_response(request):
+                # Verify context is available during async processing
+                session_id = get_context_session_id()
+                self.assertEqual(session_id, "async-session-123")
+                return mock_response
+
+            middleware = PosthogContextMiddleware(async_get_response)
+
+            request = MockRequest(
+                headers={"X-POSTHOG-SESSION-ID": "async-session-123"},
+                method="GET",
+            )
+
+            with new_context():
+                result = middleware(request)
+                await result
+
+        asyncio.run(run_test())
+
+    def test_async_middleware_exception_capture(self):
+        """Test that async middleware captures exceptions during request processing"""
+
+        async def run_test():
+            mock_client = Mock()
+
+            # Make async_get_response raise an exception
+            async def raise_exception(request):
+                raise ValueError("Async test exception")
+
+            # Properly initialize middleware
+            middleware = PosthogContextMiddleware(raise_exception)
+            middleware.client = mock_client  # Override with mock client
+
+            request = MockRequest()
+
+            # Should capture exception and re-raise
+            with self.assertRaises(ValueError):
+                result = middleware(request)
+                await result
+
+            # Verify exception was captured by middleware
+            mock_client.capture_exception.assert_called_once()
+            captured_exception = mock_client.capture_exception.call_args[0][0]
+            self.assertIsInstance(captured_exception, ValueError)
+            self.assertEqual(str(captured_exception), "Async test exception")
+
+        asyncio.run(run_test())
+
+    def test_async_middleware_with_authenticated_user(self):
+        """
+        Test that async middleware correctly extracts user info in async context.
+
+        Django's request.user is a SimpleLazyObject that defers DB access.
+        In async context, accessing it directly raises SynchronousOnlyOperation.
+        The middleware should use request.auser() instead.
+
+        This tests the fix for issue #355.
+        """
+
+        async def run_test():
+            mock_response = Mock()
+            mock_user = Mock()
+            mock_user.is_authenticated = True
+            mock_user.pk = 123
+            mock_user.email = "test@example.com"
+
+            async def async_get_response(request):
+                # Verify user info was extracted and set as distinct_id
+                distinct_id = get_context_distinct_id()
+                self.assertEqual(distinct_id, "123")
+                return mock_response
+
+            middleware = PosthogContextMiddleware(async_get_response)
+            middleware.client = Mock()
+
+            request = MockRequest(
+                headers={"X-POSTHOG-SESSION-ID": "test-session"}, method="GET"
+            )
+
+            # Mock auser() to return authenticated user
+            async def mock_auser():
+                return mock_user
+
+            request.auser = mock_auser
+
+            with new_context():
+                result = middleware(request)
+                response = await result
+                self.assertEqual(response, mock_response)
+
+        asyncio.run(run_test())
+
+    def test_async_middleware_with_unauthenticated_user(self):
+        """
+        Test that async middleware handles unauthenticated users correctly.
+        """
+
+        async def run_test():
+            mock_response = Mock()
+            mock_user = Mock()
+            mock_user.is_authenticated = False  # Not authenticated
+
+            async def async_get_response(request):
+                # Verify no distinct_id was set (no user)
+                distinct_id = get_context_distinct_id()
+                self.assertIsNone(distinct_id)
+                return mock_response
+
+            middleware = PosthogContextMiddleware(async_get_response)
+            middleware.client = Mock()
+
+            request = MockRequest(
+                headers={"X-POSTHOG-SESSION-ID": "test-session"}, method="GET"
+            )
+
+            async def mock_auser():
+                return mock_user
+
+            request.auser = mock_auser
+
+            with new_context():
+                result = middleware(request)
+                response = await result
+                self.assertEqual(response, mock_response)
+
+        asyncio.run(run_test())
+
+    def test_async_middleware_without_user_attribute(self):
+        """
+        Test that async middleware handles requests without user attribute (no auth middleware).
+        """
+
+        async def run_test():
+            mock_response = Mock()
+
+            async def async_get_response(request):
+                return mock_response
+
+            middleware = PosthogContextMiddleware(async_get_response)
+            middleware.client = Mock()
+
+            # Request without auser method (no auth middleware)
+            request = MockRequest(
+                headers={"X-POSTHOG-SESSION-ID": "test-session"}, method="GET"
+            )
+
+            with new_context():
+                result = middleware(request)
+                response = await result
+                self.assertEqual(response, mock_response)
+
+        asyncio.run(run_test())
+
+    def test_async_middleware_with_extra_tags(self):
+        """
+        Test that async middleware works with extra_tags callback.
+        """
+
+        async def run_test():
+            mock_response = Mock()
+
+            def extra_tags_callback(request):
+                # Simple sync callback - should work
+                return {"custom_tag": "custom_value"}
+
+            async def async_get_response(request):
+                return mock_response
+
+            middleware = PosthogContextMiddleware(async_get_response)
+            middleware.extra_tags = extra_tags_callback
+            middleware.client = Mock()
+
+            request = MockRequest(
+                headers={"X-POSTHOG-SESSION-ID": "test-session"}, method="GET"
+            )
+
+            # Mock auser for no user
+            async def mock_auser():
+                return None
+
+            request.auser = mock_auser
+
+            with new_context():
+                result = middleware(request)
+                response = await result
+                self.assertEqual(response, mock_response)
+
+        asyncio.run(run_test())
+
+    def test_async_middleware_with_tag_map(self):
+        """
+        Test that async middleware works with tag_map callback.
+        """
+
+        async def run_test():
+            mock_response = Mock()
+
+            def tag_map_callback(tags):
+                # Simple sync callback - should work
+                tags["mapped"] = "yes"
+                return tags
+
+            async def async_get_response(request):
+                return mock_response
+
+            middleware = PosthogContextMiddleware(async_get_response)
+            middleware.tag_map = tag_map_callback
+            middleware.client = Mock()
+
+            request = MockRequest(
+                headers={"X-POSTHOG-SESSION-ID": "test-session"}, method="GET"
+            )
+
+            # Mock auser for no user
+            async def mock_auser():
+                return None
+
+            request.auser = mock_auser
+
+            with new_context():
+                result = middleware(request)
+                response = await result
+                self.assertEqual(response, mock_response)
+
+        asyncio.run(run_test())
+
+    def test_async_middleware_user_extraction_with_all_headers(self):
+        """
+        Test async middleware extracts all request info correctly.
+        """
+
+        async def run_test():
+            mock_response = Mock()
+            mock_user = Mock()
+            mock_user.is_authenticated = True
+            mock_user.pk = 456
+            mock_user.email = "async@test.com"
+
+            async def async_get_response(request):
+                # Verify all context was set correctly
+                distinct_id = get_context_distinct_id()
+                session_id = get_context_session_id()
+                self.assertEqual(distinct_id, "456")
+                self.assertEqual(session_id, "async-sess-123")
+                return mock_response
+
+            middleware = PosthogContextMiddleware(async_get_response)
+            middleware.client = Mock()
+
+            request = MockRequest(
+                headers={
+                    "X-POSTHOG-SESSION-ID": "async-sess-123",
+                    "X-Forwarded-For": "192.168.1.1",
+                    "User-Agent": "TestAgent/1.0",
+                },
+                method="POST",
+                path="/api/test",
+            )
+
+            async def mock_auser():
+                return mock_user
+
+            request.auser = mock_auser
+
+            with new_context():
+                result = middleware(request)
+                response = await result
+                self.assertEqual(response, mock_response)
+
+        asyncio.run(run_test())
+
+
+class TestPosthogContextMiddlewareHybrid(unittest.TestCase):
+    """Test hybrid middleware behavior with mixed sync/async chains"""
+
+    def test_hybrid_flags_set(self):
+        """Test that both capability flags are set"""
+        self.assertTrue(PosthogContextMiddleware.sync_capable)
+        self.assertTrue(PosthogContextMiddleware.async_capable)
+
+    def test_sync_to_async_routing(self):
+        """Test that __call__ routes to __acall__ when async"""
+
+        async def run_test():
+            async def async_get_response(request):
+                return Mock()
+
+            middleware = PosthogContextMiddleware(async_get_response)
+
+            # Verify routing happens
+            request = MockRequest()
+            result = middleware(request)
+
+            # Should be a coroutine from __acall__
+            self.assertTrue(asyncio.iscoroutine(result))
+            await result  # Clean up
+
+        asyncio.run(run_test())
+
+    def test_sync_path_direct_return(self):
+        """Test that sync path returns directly without coroutine"""
+        mock_response = Mock()
+
+        def sync_get_response(request):
+            return mock_response
+
+        middleware = PosthogContextMiddleware(sync_get_response)
+
+        request = MockRequest()
+        result = middleware(request)
+
+        # Should NOT be a coroutine
+        self.assertFalse(asyncio.iscoroutine(result))
+        self.assertEqual(result, mock_response)
+
 
 if __name__ == "__main__":
     unittest.main()

posthog/test/test_client.py

@@ -2,17 +2,18 @@ import time
 import unittest
 from datetime import datetime
 from uuid import uuid4
-from posthog.contexts import get_context_session_id, set_context_session, new_context
 
 import mock
 import six
 from parameterized import parameterized
 
 from posthog.client import Client
-from posthog.request import APIError
+from posthog.contexts import get_context_session_id, new_context, set_context_session
+from posthog.request import APIError, GetResponse
 from posthog.test.test_utils import FAKE_TEST_API_KEY
 from posthog.types import FeatureFlag, LegacyFlagMetadata
 from posthog.version import VERSION
+from posthog.contexts import tag
 
 
 class TestClient(unittest.TestCase):
@@ -197,12 +198,6 @@ class TestClient(unittest.TestCase):
             print(capture_call)
             self.assertEqual(capture_call[1]["distinct_id"], "distinct_id")
             self.assertEqual(capture_call[0][0], "$exception")
-            self.assertEqual(
-                capture_call[1]["properties"]["$exception_type"], "Exception"
-            )
-            self.assertEqual(
-                capture_call[1]["properties"]["$exception_message"], "test exception"
-            )
             self.assertEqual(
                 capture_call[1]["properties"]["$exception_list"][0]["mechanism"][
                     "type"
@@ -751,6 +746,96 @@ class TestClient(unittest.TestCase):
 
             self.assertEqual(patch_flags.call_count, 0)
 
+    @mock.patch("posthog.client.flags")
+    def test_capture_with_send_feature_flags_true_and_local_evaluation_uses_local_flags(
+        self, patch_flags
+    ):
+        """Test that send_feature_flags=True with local evaluation enabled uses local flags without API call"""
+        patch_flags.return_value = {"featureFlags": {"remote-flag": "remote-variant"}}
+
+        multivariate_flag = {
+            "id": 1,
+            "name": "Beta Feature",
+            "key": "beta-feature-local",
+            "active": True,
+            "rollout_percentage": 100,
+            "filters": {
+                "groups": [
+                    {
+                        "rollout_percentage": 100,
+                    },
+                ],
+                "multivariate": {
+                    "variants": [
+                        {
+                            "key": "first-variant",
+                            "name": "First Variant",
+                            "rollout_percentage": 50,
+                        },
+                        {
+                            "key": "second-variant",
+                            "name": "Second Variant",
+                            "rollout_percentage": 50,
+                        },
+                    ]
+                },
+            },
+        }
+        simple_flag = {
+            "id": 2,
+            "name": "Simple Flag",
+            "key": "simple-flag",
+            "active": True,
+            "filters": {
+                "groups": [
+                    {
+                        "rollout_percentage": 100,
+                    }
+                ],
+            },
+        }
+
+        with mock.patch("posthog.client.batch_post") as mock_post:
+            client = Client(
+                FAKE_TEST_API_KEY,
+                on_error=self.set_fail,
+                personal_api_key=FAKE_TEST_API_KEY,
+                sync_mode=True,
+            )
+            client.feature_flags = [multivariate_flag, simple_flag]
+
+            msg_uuid = client.capture(
+                "python test event",
+                distinct_id="distinct_id",
+                send_feature_flags=True,
+            )
+            self.assertIsNotNone(msg_uuid)
+            self.assertFalse(self.failed)
+
+            # Get the enqueued message from the mock
+            mock_post.assert_called_once()
+            batch_data = mock_post.call_args[1]["batch"]
+            msg = batch_data[0]
+
+            self.assertEqual(msg["event"], "python test event")
+            self.assertEqual(msg["distinct_id"], "distinct_id")
+
+            # Verify local flags are included in the event
+            self.assertIn("$feature/beta-feature-local", msg["properties"])
+            self.assertIn("$feature/simple-flag", msg["properties"])
+            self.assertEqual(msg["properties"]["$feature/simple-flag"], True)
+
+            # Verify active feature flags are set correctly
+            active_flags = msg["properties"]["$active_feature_flags"]
+            self.assertIn("beta-feature-local", active_flags)
+            self.assertIn("simple-flag", active_flags)
+
+            # The remote flag should NOT be included since we used local evaluation
+            self.assertNotIn("$feature/remote-flag", msg["properties"])
+
+            # CRITICAL: Verify the /flags API was NOT called
+            self.assertEqual(patch_flags.call_count, 0)
+
     @mock.patch("posthog.client.flags")
     def test_capture_with_send_feature_flags_options_only_evaluate_locally_true(
         self, patch_flags
@@ -1741,6 +1826,7 @@ class TestClient(unittest.TestCase):
             person_properties={"distinct_id": "some_id"},
             group_properties={},
             geoip_disable=True,
+            flag_keys_to_evaluate=["random_key"],
         )
         patch_flags.reset_mock()
         client.feature_enabled(
@@ -1755,6 +1841,7 @@ class TestClient(unittest.TestCase):
             person_properties={"distinct_id": "feature_enabled_distinct_id"},
             group_properties={},
             geoip_disable=True,
+            flag_keys_to_evaluate=["random_key"],
         )
         patch_flags.reset_mock()
         client.get_all_flags_and_payloads("all_flags_payloads_id")
@@ -1815,6 +1902,7 @@ class TestClient(unittest.TestCase):
                 "instance": {"$group_key": "app.posthog.com"},
             },
             geoip_disable=False,
+            flag_keys_to_evaluate=["random_key"],
         )
 
         patch_flags.reset_mock()
@@ -1841,6 +1929,7 @@ class TestClient(unittest.TestCase):
                 "instance": {"$group_key": "app.posthog.com"},
             },
             geoip_disable=False,
+            flag_keys_to_evaluate=["random_key"],
         )
 
         patch_flags.reset_mock()
@@ -2057,7 +2146,7 @@ class TestClient(unittest.TestCase):
 
     def test_set_context_session_override_in_capture(self):
         """Test that explicit session ID overrides context session ID in capture"""
-        from posthog.contexts import set_context_session, new_context
+        from posthog.contexts import new_context, set_context_session
 
         with mock.patch("posthog.client.batch_post") as mock_post:
             client = Client(FAKE_TEST_API_KEY, on_error=self.set_fail, sync_mode=True)
@@ -2090,13 +2179,21 @@ class TestClient(unittest.TestCase):
         self, patch_get, patch_poller
     ):
         """Test that when enable_local_evaluation=False, the poller is not started"""
-        patch_get.return_value = {
-            "flags": [
-                {"id": 1, "name": "Beta Feature", "key": "beta-feature", "active": True}
-            ],
-            "group_type_mapping": {},
-            "cohorts": {},
-        }
+        patch_get.return_value = GetResponse(
+            data={
+                "flags": [
+                    {
+                        "id": 1,
+                        "name": "Beta Feature",
+                        "key": "beta-feature",
+                        "active": True,
+                    }
+                ],
+                "group_type_mapping": {},
+                "cohorts": {},
+            },
+            etag='"test-etag"',
+        )
 
         client = Client(
             FAKE_TEST_API_KEY,
@@ -2118,13 +2215,21 @@ class TestClient(unittest.TestCase):
     @mock.patch("posthog.client.get")
     def test_enable_local_evaluation_true_starts_poller(self, patch_get, patch_poller):
         """Test that when enable_local_evaluation=True (default), the poller is started"""
-        patch_get.return_value = {
-            "flags": [
-                {"id": 1, "name": "Beta Feature", "key": "beta-feature", "active": True}
-            ],
-            "group_type_mapping": {},
-            "cohorts": {},
-        }
+        patch_get.return_value = GetResponse(
+            data={
+                "flags": [
+                    {
+                        "id": 1,
+                        "name": "Beta Feature",
+                        "key": "beta-feature",
+                        "active": True,
+                    }
+                ],
+                "group_type_mapping": {},
+                "cohorts": {},
+            },
+            etag='"test-etag"',
+        )
 
         client = Client(
             FAKE_TEST_API_KEY,
@@ -2158,6 +2263,7 @@ class TestClient(unittest.TestCase):
         self.assertEqual(result, {"test": "payload"})
         patch_remote_config.assert_called_once_with(
             "test-personal-key",
+            FAKE_TEST_API_KEY,
             client.host,
             "test-flag",
             timeout=client.feature_flags_request_timeout_seconds,
@@ -2185,6 +2291,7 @@ class TestClient(unittest.TestCase):
             "only_evaluate_locally": None,
             "person_properties": None,
             "group_properties": None,
+            "flag_keys_filter": None,
         }
         self.assertEqual(result, expected)
 
@@ -2195,6 +2302,7 @@ class TestClient(unittest.TestCase):
             "only_evaluate_locally": None,
             "person_properties": None,
             "group_properties": None,
+            "flag_keys_filter": None,
         }
         self.assertEqual(result, expected)
 
@@ -2210,6 +2318,7 @@ class TestClient(unittest.TestCase):
             "only_evaluate_locally": True,
             "person_properties": {"plan": "premium"},
             "group_properties": {"company": {"type": "enterprise"}},
+            "flag_keys_filter": None,
         }
         self.assertEqual(result, expected)
 
@@ -2221,6 +2330,7 @@ class TestClient(unittest.TestCase):
             "only_evaluate_locally": None,
             "person_properties": {"user_id": "123"},
             "group_properties": None,
+            "flag_keys_filter": None,
         }
         self.assertEqual(result, expected)
 
@@ -2231,6 +2341,7 @@ class TestClient(unittest.TestCase):
             "only_evaluate_locally": None,
             "person_properties": None,
             "group_properties": None,
+            "flag_keys_filter": None,
         }
         self.assertEqual(result, expected)
 
@@ -2247,6 +2358,53 @@ class TestClient(unittest.TestCase):
             client._parse_send_feature_flags(None)
         self.assertIn("Invalid type for send_feature_flags", str(cm.exception))
 
+    @mock.patch("posthog.client.flags")
+    def test_capture_with_send_feature_flags_flag_keys_filter(self, patch_flags):
+        """Test that SendFeatureFlagsOptions with flag_keys_filter only evaluates specified flags"""
+        # When flag_keys_to_evaluate is provided, the API should only return the requested flags
+        patch_flags.return_value = {
+            "featureFlags": {
+                "flag1": "value1",
+                "flag3": "value3",
+            }
+        }
+
+        with mock.patch("posthog.client.batch_post") as mock_post:
+            client = Client(
+                FAKE_TEST_API_KEY,
+                on_error=self.set_fail,
+                personal_api_key=FAKE_TEST_API_KEY,
+                sync_mode=True,
+            )
+
+            send_options = {
+                "flag_keys_filter": ["flag1", "flag3"],
+                "person_properties": {"subscription": "pro"},
+            }
+
+            msg_uuid = client.capture(
+                "test event", distinct_id="distinct_id", send_feature_flags=send_options
+            )
+
+            self.assertIsNotNone(msg_uuid)
+            self.assertFalse(self.failed)
+
+            # Verify flags() was called with flag_keys_to_evaluate
+            patch_flags.assert_called_once()
+            call_args = patch_flags.call_args[1]
+            self.assertEqual(call_args["flag_keys_to_evaluate"], ["flag1", "flag3"])
+            self.assertEqual(call_args["person_properties"], {"subscription": "pro"})
+
+            # Check the message includes only the filtered flags
+            mock_post.assert_called_once()
+            batch_data = mock_post.call_args[1]["batch"]
+            msg = batch_data[0]
+
+            self.assertEqual(msg["properties"]["$feature/flag1"], "value1")
+            self.assertEqual(msg["properties"]["$feature/flag3"], "value3")
+            # flag2 should not be included since it wasn't requested
+            self.assertNotIn("$feature/flag2", msg["properties"])
+
     @mock.patch("posthog.client.batch_post")
     def test_get_feature_flag_result_with_empty_string_payload(self, patch_batch_post):
         """Test that get_feature_flag_result returns a FeatureFlagResult when payload is empty string"""
@@ -2282,9 +2440,7 @@ class TestClient(unittest.TestCase):
                             }
                         ]
                     },
-                    "payloads": {
-                        "empty-variant": ""  # Empty string payload
-                    },
+                    "payloads": {"empty-variant": ""},  # Empty string payload
                 },
             }
         ]
@@ -2355,3 +2511,58 @@ class TestClient(unittest.TestCase):
         self.assertEqual(
             result["featureFlagPayloads"]["normal-payload-flag"], "normal payload"
         )
+
+    def test_context_tags_added(self):
+        with mock.patch("posthog.client.batch_post") as mock_post:
+            client = Client(FAKE_TEST_API_KEY, on_error=self.set_fail, sync_mode=True)
+
+            with new_context():
+                tag("random_tag", 12345)
+                client.capture("python test event", distinct_id="distinct_id")
+
+            batch_data = mock_post.call_args[1]["batch"]
+            msg = batch_data[0]
+            self.assertEqual(msg["properties"]["$context_tags"], ["random_tag"])
+
+    @mock.patch(
+        "posthog.client.Client._enqueue", side_effect=Exception("Unexpected error")
+    )
+    def test_methods_handle_exceptions(self, mock_enqueue):
+        """Test that all decorated methods handle exceptions gracefully."""
+        client = Client("test-key")
+
+        test_cases = [
+            ("capture", ["test_event"], {}),
+            ("set", [], {"distinct_id": "some-id", "properties": {"a": "b"}}),
+            ("set_once", [], {"distinct_id": "some-id", "properties": {"a": "b"}}),
+            ("group_identify", ["group-type", "group-key"], {}),
+            ("alias", ["some-id", "new-id"], {}),
+        ]
+
+        for method_name, args, kwargs in test_cases:
+            with self.subTest(method=method_name):
+                method = getattr(client, method_name)
+                result = method(*args, **kwargs)
+                self.assertEqual(result, None)
+
+    @mock.patch(
+        "posthog.client.Client._enqueue", side_effect=Exception("Expected error")
+    )
+    def test_debug_flag_re_raises_exceptions(self, mock_enqueue):
+        """Test that methods re-raise exceptions when debug=True."""
+        client = Client("test-key", debug=True)
+
+        test_cases = [
+            ("capture", ["test_event"], {}),
+            ("set", [], {"distinct_id": "some-id", "properties": {"a": "b"}}),
+            ("set_once", [], {"distinct_id": "some-id", "properties": {"a": "b"}}),
+            ("group_identify", ["group-type", "group-key"], {}),
+            ("alias", ["some-id", "new-id"], {}),
+        ]
+
+        for method_name, args, kwargs in test_cases:
+            with self.subTest(method=method_name):
+                method = getattr(client, method_name)
+                with self.assertRaises(Exception) as cm:
+                    method(*args, **kwargs)
+                self.assertEqual(str(cm.exception), "Expected error")

posthog/test/test_exception_capture.py

@@ -32,3 +32,421 @@ def test_excepthook(tmpdir):
         b'"$exception_list": [{"mechanism": {"type": "generic", "handled": true}, "module": null, "type": "ZeroDivisionError", "value": "division by zero", "stacktrace": {"frames": [{"platform": "python", "filename": "app.py", "abs_path"'
         in output
     )
+
+
+def test_code_variables_capture(tmpdir):
+    app = tmpdir.join("app.py")
+    app.write(
+        dedent(
+            """
+    import os
+    from posthog import Posthog
+    
+    class UnserializableObject:
+        pass
+    
+    posthog = Posthog(
+        'phc_x', 
+        host='https://eu.i.posthog.com', 
+        debug=True, 
+        enable_exception_autocapture=True,
+        capture_exception_code_variables=True,
+        project_root=os.path.dirname(os.path.abspath(__file__))
+    )
+    
+    def trigger_error():
+        my_string = "hello world"
+        my_number = 42
+        my_bool = True
+        my_dict = {"name": "test", "value": 123}
+        my_sensitive_dict = {
+            "safe_key": "safe_value",
+            "password": "secret123",  # key matches pattern -> should be masked
+            "other_key": "contains_password_here",  # value matches pattern -> should be masked
+        }
+        my_nested_dict = {
+            "level1": {
+                "level2": {
+                    "api_key": "nested_secret",  # deeply nested key matches
+                    "data": "contains_token_here",  # deeply nested value matches
+                    "safe": "visible",
+                }
+            }
+        }
+        my_list = ["safe_item", "has_password_inside", "another_safe"]
+        my_tuple = ("tuple_safe", "secret_in_value", "tuple_also_safe")
+        my_list_of_dicts = [
+            {"id": 1, "password": "list_dict_secret"},
+            {"id": 2, "value": "safe_value"},
+        ]
+        my_obj = UnserializableObject()
+        my_password = "secret123"  # Should be masked by default (name matches)
+        my_innocent_var = "contains_password_here"  # Should be masked by default (value matches)
+        __should_be_ignored = "hidden"  # Should be ignored by default
+        
+        1/0  # Trigger exception
+    
+    def intermediate_function():
+        request_id = "abc-123"
+        user_count = 100
+        is_active = True
+        
+        trigger_error()
+    
+    def process_data():
+        batch_size = 50
+        retry_count = 3
+        
+        intermediate_function()
+    
+    process_data()
+    """
+        )
+    )
+
+    with pytest.raises(subprocess.CalledProcessError) as excinfo:
+        subprocess.check_output([sys.executable, str(app)], stderr=subprocess.STDOUT)
+
+    output = excinfo.value.output
+
+    assert b"ZeroDivisionError" in output
+    assert b"code_variables" in output
+
+    # Variables from trigger_error frame
+    assert b"'my_string': 'hello world'" in output
+    assert b"'my_number': 42" in output
+    assert b"'my_bool': 'True'" in output
+    assert b'"my_dict": "{\\"name\\": \\"test\\", \\"value\\": 123}"' in output
+    assert (
+        b'{\\"safe_key\\": \\"safe_value\\", \\"password\\": \\"$$_posthog_redacted_based_on_masking_rules_$$\\", \\"other_key\\": \\"$$_posthog_redacted_based_on_masking_rules_$$\\"}'
+        in output
+    )
+    assert (
+        b'{\\"level1\\": {\\"level2\\": {\\"api_key\\": \\"$$_posthog_redacted_based_on_masking_rules_$$\\", \\"data\\": \\"$$_posthog_redacted_based_on_masking_rules_$$\\", \\"safe\\": \\"visible\\"}}}'
+        in output
+    )
+    assert (
+        b'[\\"safe_item\\", \\"$$_posthog_redacted_based_on_masking_rules_$$\\", \\"another_safe\\"]'
+        in output
+    )
+    assert (
+        b'[\\"tuple_safe\\", \\"$$_posthog_redacted_based_on_masking_rules_$$\\", \\"tuple_also_safe\\"]'
+        in output
+    )
+    assert (
+        b'[{\\"id\\": 1, \\"password\\": \\"$$_posthog_redacted_based_on_masking_rules_$$\\"}, {\\"id\\": 2, \\"value\\": \\"safe_value\\"}]'
+        in output
+    )
+    assert b"<__main__.UnserializableObject object at" in output
+    assert b"'my_password': '$$_posthog_redacted_based_on_masking_rules_$$'" in output
+    assert (
+        b"'my_innocent_var': '$$_posthog_redacted_based_on_masking_rules_$$'" in output
+    )
+    assert b"'__should_be_ignored':" not in output
+
+    # Variables from intermediate_function frame
+    assert b"'request_id': 'abc-123'" in output
+    assert b"'user_count': 100" in output
+    assert b"'is_active': 'True'" in output
+
+    # Variables from process_data frame
+    assert b"'batch_size': 50" in output
+    assert b"'retry_count': 3" in output
+
+
+def test_code_variables_context_override(tmpdir):
+    app = tmpdir.join("app.py")
+    app.write(
+        dedent(
+            """
+    import os
+    import posthog
+    from posthog import Posthog
+    
+    posthog_client = Posthog(
+        'phc_x', 
+        host='https://eu.i.posthog.com', 
+        debug=True, 
+        enable_exception_autocapture=True,
+        capture_exception_code_variables=False,
+        project_root=os.path.dirname(os.path.abspath(__file__))
+    )
+    
+    def process_data():
+        bank = "should_be_masked"
+        __dunder_var = "should_be_visible"
+        
+        1/0
+    
+    with posthog.new_context(client=posthog_client):
+        posthog.set_capture_exception_code_variables_context(True)
+        posthog.set_code_variables_mask_patterns_context([r"(?i).*bank.*"])
+        posthog.set_code_variables_ignore_patterns_context([])
+        
+        process_data()
+    """
+        )
+    )
+
+    with pytest.raises(subprocess.CalledProcessError) as excinfo:
+        subprocess.check_output([sys.executable, str(app)], stderr=subprocess.STDOUT)
+
+    output = excinfo.value.output
+
+    assert b"ZeroDivisionError" in output
+    assert b"code_variables" in output
+    assert b"'bank': '$$_posthog_redacted_based_on_masking_rules_$$'" in output
+    assert b"'__dunder_var': 'should_be_visible'" in output
+
+
+def test_code_variables_size_limiter(tmpdir):
+    app = tmpdir.join("app.py")
+    app.write(
+        dedent(
+            """
+    import os
+    from posthog import Posthog
+    
+    posthog = Posthog(
+        'phc_x', 
+        host='https://eu.i.posthog.com', 
+        debug=True, 
+        enable_exception_autocapture=True,
+        capture_exception_code_variables=True,
+        project_root=os.path.dirname(os.path.abspath(__file__))
+    )
+    
+    def trigger_error():
+        var_a = "a" * 2000
+        var_b = "b" * 2000
+        var_c = "c" * 2000
+        var_d = "d" * 2000
+        var_e = "e" * 2000
+        var_f = "f" * 2000
+        var_g = "g" * 2000
+        
+        1/0
+    
+    def intermediate_function():
+        var_h = "h" * 2000
+        var_i = "i" * 2000
+        var_j = "j" * 2000
+        var_k = "k" * 2000
+        var_l = "l" * 2000
+        var_m = "m" * 2000
+        var_n = "n" * 2000
+        
+        trigger_error()
+    
+    def process_data():
+        var_o = "o" * 2000
+        var_p = "p" * 2000
+        var_q = "q" * 2000
+        var_r = "r" * 2000
+        var_s = "s" * 2000
+        var_t = "t" * 2000
+        var_u = "u" * 2000
+        
+        intermediate_function()
+    
+    process_data()
+    """
+        )
+    )
+
+    with pytest.raises(subprocess.CalledProcessError) as excinfo:
+        subprocess.check_output([sys.executable, str(app)], stderr=subprocess.STDOUT)
+
+    output = excinfo.value.output.decode("utf-8")
+
+    assert "ZeroDivisionError" in output
+    assert "code_variables" in output
+
+    captured_vars = []
+    for var_name in [
+        "var_a",
+        "var_b",
+        "var_c",
+        "var_d",
+        "var_e",
+        "var_f",
+        "var_g",
+        "var_h",
+        "var_i",
+        "var_j",
+        "var_k",
+        "var_l",
+        "var_m",
+        "var_n",
+        "var_o",
+        "var_p",
+        "var_q",
+        "var_r",
+        "var_s",
+        "var_t",
+        "var_u",
+    ]:
+        if f"'{var_name}'" in output:
+            captured_vars.append(var_name)
+
+    assert len(captured_vars) > 0
+    assert len(captured_vars) < 21
+
+
+def test_code_variables_disabled_capture(tmpdir):
+    app = tmpdir.join("app.py")
+    app.write(
+        dedent(
+            """
+    import os
+    from posthog import Posthog
+    
+    posthog = Posthog(
+        'phc_x', 
+        host='https://eu.i.posthog.com', 
+        debug=True, 
+        enable_exception_autocapture=True,
+        capture_exception_code_variables=False,
+        project_root=os.path.dirname(os.path.abspath(__file__))
+    )
+    
+    def trigger_error():
+        my_string = "hello world"
+        my_number = 42
+        my_bool = True
+        
+        1/0
+    
+    trigger_error()
+    """
+        )
+    )
+
+    with pytest.raises(subprocess.CalledProcessError) as excinfo:
+        subprocess.check_output([sys.executable, str(app)], stderr=subprocess.STDOUT)
+
+    output = excinfo.value.output.decode("utf-8")
+
+    assert "ZeroDivisionError" in output
+    assert "'code_variables':" not in output
+    assert '"code_variables":' not in output
+    assert "'my_string'" not in output
+    assert "'my_number'" not in output
+
+
+def test_code_variables_enabled_then_disabled_in_context(tmpdir):
+    app = tmpdir.join("app.py")
+    app.write(
+        dedent(
+            """
+    import os
+    import posthog
+    from posthog import Posthog
+    
+    posthog_client = Posthog(
+        'phc_x', 
+        host='https://eu.i.posthog.com', 
+        debug=True, 
+        enable_exception_autocapture=True,
+        capture_exception_code_variables=True,
+        project_root=os.path.dirname(os.path.abspath(__file__))
+    )
+    
+    def process_data():
+        my_var = "should not be captured"
+        important_value = 123
+        
+        1/0
+    
+    with posthog.new_context(client=posthog_client):
+        posthog.set_capture_exception_code_variables_context(False)
+        
+        process_data()
+    """
+        )
+    )
+
+    with pytest.raises(subprocess.CalledProcessError) as excinfo:
+        subprocess.check_output([sys.executable, str(app)], stderr=subprocess.STDOUT)
+
+    output = excinfo.value.output.decode("utf-8")
+
+    assert "ZeroDivisionError" in output
+    assert "'code_variables':" not in output
+    assert '"code_variables":' not in output
+    assert "'my_var'" not in output
+    assert "'important_value'" not in output
+
+
+def test_code_variables_repr_fallback(tmpdir):
+    app = tmpdir.join("app.py")
+    app.write(
+        dedent(
+            """
+    import os
+    import re
+    from datetime import datetime, timedelta
+    from decimal import Decimal
+    from fractions import Fraction
+    from posthog import Posthog
+    
+    class CustomReprClass:
+        def __repr__(self):
+            return '<CustomReprClass: custom representation>'
+    
+    posthog = Posthog(
+        'phc_x', 
+        host='https://eu.i.posthog.com', 
+        debug=True, 
+        enable_exception_autocapture=True,
+        capture_exception_code_variables=True,
+        project_root=os.path.dirname(os.path.abspath(__file__))
+    )
+    
+    def trigger_error():
+        my_regex = re.compile(r'\\d+')
+        my_datetime = datetime(2024, 1, 15, 10, 30, 45)
+        my_timedelta = timedelta(days=5, hours=3)
+        my_decimal = Decimal('123.456')
+        my_fraction = Fraction(3, 4)
+        my_set = {1, 2, 3}
+        my_frozenset = frozenset([4, 5, 6])
+        my_bytes = b'hello bytes'
+        my_bytearray = bytearray(b'mutable bytes')
+        my_memoryview = memoryview(b'memory view')
+        my_complex = complex(3, 4)
+        my_range = range(10)
+        my_custom = CustomReprClass()
+        my_lambda = lambda x: x * 2
+        my_function = trigger_error
+        
+        1/0
+    
+    trigger_error()
+    """
+        )
+    )
+
+    with pytest.raises(subprocess.CalledProcessError) as excinfo:
+        subprocess.check_output([sys.executable, str(app)], stderr=subprocess.STDOUT)
+
+    output = excinfo.value.output.decode("utf-8")
+
+    assert "ZeroDivisionError" in output
+    assert "code_variables" in output
+
+    assert "re.compile(" in output and "\\\\d+" in output
+    assert "datetime.datetime(2024, 1, 15, 10, 30, 45)" in output
+    assert "datetime.timedelta(days=5, seconds=10800)" in output
+    assert "Decimal('123.456')" in output
+    assert "Fraction(3, 4)" in output
+    assert "{1, 2, 3}" in output
+    assert "frozenset({4, 5, 6})" in output
+    assert "b'hello bytes'" in output
+    assert "bytearray(b'mutable bytes')" in output
+    assert "<memory at" in output
+    assert "(3+4j)" in output
+    assert "range(0, 10)" in output
+    assert "<CustomReprClass: custom representation>" in output
+    assert "<lambda>" in output
+    assert "<function trigger_error at" in output

posthog/test/test_module.py

@@ -18,14 +18,6 @@ class TestModule(unittest.TestCase):
             "testsecret", host="http://localhost:8000", on_error=self.failed
         )
 
-    def test_no_api_key(self):
-        self.posthog.api_key = None
-        self.assertRaises(Exception, self.posthog.capture)
-
-    def test_no_host(self):
-        self.posthog.host = None
-        self.assertRaises(Exception, self.posthog.capture)
-
     def test_track(self):
         res = self.posthog.capture("python module event", distinct_id="distinct_id")
         self._assert_enqueue_result(res)

posthog/test/test_request.py

@@ -6,16 +6,59 @@ import mock
 import pytest
 import requests
 
+import posthog.request as request_module
 from posthog.request import (
+    APIError,
     DatetimeSerializer,
+    GetResponse,
+    KEEP_ALIVE_SOCKET_OPTIONS,
     QuotaLimitError,
+    _mask_tokens_in_url,
     batch_post,
     decide,
     determine_server_host,
+    disable_connection_reuse,
+    enable_keep_alive,
+    get,
+    set_socket_options,
 )
 from posthog.test.test_utils import TEST_API_KEY
 
 
+@pytest.mark.parametrize(
+    "url, expected",
+    [
+        # Token with params after - masks keeping first 10 chars
+        (
+            "https://example.com/api/flags?token=phc_abc123xyz789&send_cohorts",
+            "https://example.com/api/flags?token=phc_abc123...&send_cohorts",
+        ),
+        # Token at end of URL
+        (
+            "https://example.com/api/flags?token=phc_abc123xyz789",
+            "https://example.com/api/flags?token=phc_abc123...",
+        ),
+        # No token - unchanged
+        (
+            "https://example.com/api/flags?other=value",
+            "https://example.com/api/flags?other=value",
+        ),
+        # Short token (<10 chars) - unchanged
+        (
+            "https://example.com/api/flags?token=short",
+            "https://example.com/api/flags?token=short",
+        ),
+        # Exactly 10 char token - gets ellipsis
+        (
+            "https://example.com/api/flags?token=1234567890",
+            "https://example.com/api/flags?token=1234567890...",
+        ),
+    ],
+)
+def test_mask_tokens_in_url(url, expected):
+    assert _mask_tokens_in_url(url) == expected
+
+
 class TestRequests(unittest.TestCase):
     def test_valid_request(self):
         res = batch_post(
@@ -107,6 +150,184 @@ class TestRequests(unittest.TestCase):
             self.assertEqual(response["featureFlags"], {"flag1": True})
 
 
+class TestGet(unittest.TestCase):
+    """Unit tests for the get() function HTTP-level behavior."""
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_returns_data_and_etag(self, mock_get):
+        """Test that get() returns GetResponse with data and etag from headers."""
+        mock_response = requests.Response()
+        mock_response.status_code = 200
+        mock_response.headers["ETag"] = '"abc123"'
+        mock_response._content = json.dumps({"flags": [{"key": "test-flag"}]}).encode(
+            "utf-8"
+        )
+        mock_get.return_value = mock_response
+
+        response = get("api_key", "/test-url", host="https://example.com")
+
+        self.assertIsInstance(response, GetResponse)
+        self.assertEqual(response.data, {"flags": [{"key": "test-flag"}]})
+        self.assertEqual(response.etag, '"abc123"')
+        self.assertFalse(response.not_modified)
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_sends_if_none_match_header_when_etag_provided(self, mock_get):
+        """Test that If-None-Match header is sent when etag parameter is provided."""
+        mock_response = requests.Response()
+        mock_response.status_code = 200
+        mock_response.headers["ETag"] = '"new-etag"'
+        mock_response._content = json.dumps({"flags": []}).encode("utf-8")
+        mock_get.return_value = mock_response
+
+        get("api_key", "/test-url", host="https://example.com", etag='"previous-etag"')
+
+        call_kwargs = mock_get.call_args[1]
+        self.assertEqual(call_kwargs["headers"]["If-None-Match"], '"previous-etag"')
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_does_not_send_if_none_match_when_no_etag(self, mock_get):
+        """Test that If-None-Match header is not sent when no etag provided."""
+        mock_response = requests.Response()
+        mock_response.status_code = 200
+        mock_response._content = json.dumps({"flags": []}).encode("utf-8")
+        mock_get.return_value = mock_response
+
+        get("api_key", "/test-url", host="https://example.com")
+
+        call_kwargs = mock_get.call_args[1]
+        self.assertNotIn("If-None-Match", call_kwargs["headers"])
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_handles_304_not_modified(self, mock_get):
+        """Test that 304 Not Modified response returns not_modified=True with no data."""
+        mock_response = requests.Response()
+        mock_response.status_code = 304
+        mock_response.headers["ETag"] = '"unchanged-etag"'
+        mock_get.return_value = mock_response
+
+        response = get(
+            "api_key", "/test-url", host="https://example.com", etag='"unchanged-etag"'
+        )
+
+        self.assertIsInstance(response, GetResponse)
+        self.assertIsNone(response.data)
+        self.assertEqual(response.etag, '"unchanged-etag"')
+        self.assertTrue(response.not_modified)
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_304_without_etag_header_uses_request_etag(self, mock_get):
+        """Test that 304 response without ETag header falls back to request etag."""
+        mock_response = requests.Response()
+        mock_response.status_code = 304
+        # Server doesn't return ETag header on 304
+        mock_get.return_value = mock_response
+
+        response = get(
+            "api_key", "/test-url", host="https://example.com", etag='"original-etag"'
+        )
+
+        self.assertTrue(response.not_modified)
+        self.assertEqual(response.etag, '"original-etag"')
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_200_without_etag_header(self, mock_get):
+        """Test that 200 response without ETag header returns None for etag."""
+        mock_response = requests.Response()
+        mock_response.status_code = 200
+        mock_response._content = json.dumps({"flags": []}).encode("utf-8")
+        # No ETag header
+        mock_get.return_value = mock_response
+
+        response = get("api_key", "/test-url", host="https://example.com")
+
+        self.assertFalse(response.not_modified)
+        self.assertIsNone(response.etag)
+        self.assertEqual(response.data, {"flags": []})
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_error_response_raises_api_error(self, mock_get):
+        """Test that error responses raise APIError."""
+        mock_response = requests.Response()
+        mock_response.status_code = 401
+        mock_response._content = json.dumps({"detail": "Unauthorized"}).encode("utf-8")
+        mock_get.return_value = mock_response
+
+        with self.assertRaises(APIError) as ctx:
+            get("bad_key", "/test-url", host="https://example.com")
+
+        self.assertEqual(ctx.exception.status, 401)
+        self.assertEqual(ctx.exception.message, "Unauthorized")
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_sends_authorization_header(self, mock_get):
+        """Test that Authorization header is sent with Bearer token."""
+        mock_response = requests.Response()
+        mock_response.status_code = 200
+        mock_response._content = json.dumps({}).encode("utf-8")
+        mock_get.return_value = mock_response
+
+        get("my-api-key", "/test-url", host="https://example.com")
+
+        call_kwargs = mock_get.call_args[1]
+        self.assertEqual(call_kwargs["headers"]["Authorization"], "Bearer my-api-key")
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_sends_user_agent_header(self, mock_get):
+        """Test that User-Agent header is sent."""
+        mock_response = requests.Response()
+        mock_response.status_code = 200
+        mock_response._content = json.dumps({}).encode("utf-8")
+        mock_get.return_value = mock_response
+
+        get("api_key", "/test-url", host="https://example.com")
+
+        call_kwargs = mock_get.call_args[1]
+        self.assertIn("User-Agent", call_kwargs["headers"])
+        self.assertTrue(
+            call_kwargs["headers"]["User-Agent"].startswith("posthog-python/")
+        )
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_passes_timeout(self, mock_get):
+        """Test that timeout parameter is passed to the request."""
+        mock_response = requests.Response()
+        mock_response.status_code = 200
+        mock_response._content = json.dumps({}).encode("utf-8")
+        mock_get.return_value = mock_response
+
+        get("api_key", "/test-url", host="https://example.com", timeout=30)
+
+        call_kwargs = mock_get.call_args[1]
+        self.assertEqual(call_kwargs["timeout"], 30)
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_constructs_full_url(self, mock_get):
+        """Test that host and url are combined correctly."""
+        mock_response = requests.Response()
+        mock_response.status_code = 200
+        mock_response._content = json.dumps({}).encode("utf-8")
+        mock_get.return_value = mock_response
+
+        get("api_key", "/api/flags", host="https://example.com")
+
+        call_args = mock_get.call_args[0]
+        self.assertEqual(call_args[0], "https://example.com/api/flags")
+
+    @mock.patch("posthog.request._session.get")
+    def test_get_removes_trailing_slash_from_host(self, mock_get):
+        """Test that trailing slash is removed from host."""
+        mock_response = requests.Response()
+        mock_response.status_code = 200
+        mock_response._content = json.dumps({}).encode("utf-8")
+        mock_get.return_value = mock_response
+
+        get("api_key", "/api/flags", host="https://example.com/")
+
+        call_args = mock_get.call_args[0]
+        self.assertEqual(call_args[0], "https://example.com/api/flags")
+
+
 @pytest.mark.parametrize(
     "host, expected",
     [
@@ -128,3 +349,47 @@ class TestRequests(unittest.TestCase):
 )
 def test_routing_to_custom_host(host, expected):
     assert determine_server_host(host) == expected
+
+
+def test_enable_keep_alive_sets_socket_options():
+    try:
+        enable_keep_alive()
+        from posthog.request import _session
+
+        adapter = _session.get_adapter("https://example.com")
+        assert adapter.socket_options == KEEP_ALIVE_SOCKET_OPTIONS
+    finally:
+        set_socket_options(None)
+
+
+def test_set_socket_options_clears_with_none():
+    try:
+        enable_keep_alive()
+        set_socket_options(None)
+        from posthog.request import _session
+
+        adapter = _session.get_adapter("https://example.com")
+        assert adapter.socket_options is None
+    finally:
+        set_socket_options(None)
+
+
+def test_disable_connection_reuse_creates_fresh_sessions():
+    try:
+        disable_connection_reuse()
+        session1 = request_module._get_session()
+        session2 = request_module._get_session()
+        assert session1 is not session2
+    finally:
+        request_module._pooling_enabled = True
+
+
+def test_set_socket_options_is_idempotent():
+    try:
+        enable_keep_alive()
+        session1 = request_module._session
+        enable_keep_alive()
+        session2 = request_module._session
+        assert session1 is session2
+    finally:
+        set_socket_options(None)

posthog/test/test_utils.py

@@ -1,3 +1,4 @@
+import sys
 import time
 import unittest
 from dataclasses import dataclass
@@ -122,7 +123,9 @@ class TestUtils(unittest.TestCase):
             "bar": 2,
             "baz": None,
         }
-        assert utils.clean(ModelV1(foo=1, bar="2")) == {"foo": 1, "bar": "2"}
+        # Pydantic V1 is not compatible with Python 3.14+
+        if sys.version_info < (3, 14):
+            assert utils.clean(ModelV1(foo=1, bar="2")) == {"foo": 1, "bar": "2"}
         assert utils.clean(NestedModel(foo=ModelV2(foo="1", bar=2, baz="3"))) == {
             "foo": {"foo": "1", "bar": 2, "baz": "3"}
         }

posthog/types.py

@@ -9,6 +9,7 @@ FlagValue = Union[bool, str]
 BeforeSendCallback = Callable[[dict[str, Any]], Optional[dict[str, Any]]]
 
 
+# Type alias for the send_feature_flags parameter
 class SendFeatureFlagsOptions(TypedDict, total=False):
     """Options for sending feature flags with capture events.
 
@@ -22,9 +23,11 @@ class SendFeatureFlagsOptions(TypedDict, total=False):
             Format: { group_type_name: { group_properties } }
     """
 
+    should_send: bool
     only_evaluate_locally: Optional[bool]
     person_properties: Optional[dict[str, Any]]
     group_properties: Optional[dict[str, dict[str, Any]]]
+    flag_keys_filter: Optional[list[str]]
 
 
 @dataclass(frozen=True)
@@ -120,6 +123,7 @@ class FlagsResponse(TypedDict, total=False):
     errorsWhileComputingFlags: bool
     requestId: str
     quotaLimit: Optional[List[str]]
+    evaluatedAt: Optional[int]
 
 
 class FlagsAndPayloads(TypedDict, total=True):

posthog/version.py

@@ -1,4 +1,4 @@
-VERSION = "6.3.4"
+VERSION = "7.3.1"
 
 if __name__ == "__main__":
     print(VERSION, end="")  # noqa: T201

pyproject.toml

@@ -10,18 +10,18 @@ authors = [{ name = "PostHog", email = "hey@posthog.com" }]
 maintainers = [{ name = "PostHog", email = "hey@posthog.com" }]
 license = { text = "MIT" }
 readme = "README.md"
-requires-python = ">=3.9"
+requires-python = ">=3.10"
 classifiers = [
     "Development Status :: 5 - Production/Stable",
     "Intended Audience :: Developers",
     "Operating System :: OS Independent",
     "License :: OSI Approved :: MIT License",
     "Programming Language :: Python",
-    "Programming Language :: Python :: 3.9",
     "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
     "Programming Language :: Python :: 3.13",
+    "Programming Language :: Python :: 3.14",
 ]
 dependencies = [
     "requests>=2.7,<3.0",
@@ -66,13 +66,13 @@ test = [
     "pytest-timeout",
     "pytest-asyncio",
     "django",
-    "openai",
-    "anthropic",
-    "langgraph>=0.4.8",
-    "langchain-core>=0.3.65",
-    "langchain-community>=0.3.25",
-    "langchain-openai>=0.3.22",
-    "langchain-anthropic>=0.3.15",
+    "openai>=2.0",
+    "anthropic>=0.72",
+    "langgraph>=1.0",
+    "langchain-core>=1.0",
+    "langchain-community>=0.4",
+    "langchain-openai>=1.0",
+    "langchain-anthropic>=1.0",
     "google-genai",
     "pydantic",
     "parameterized>=0.8.1",
@@ -96,3 +96,5 @@ version = { attr = "posthog.version.VERSION" }
 [tool.pytest.ini_options]
 asyncio_mode = "auto"
 asyncio_default_fixture_loop_scope = "function"
+testpaths = ["posthog/test"]
+norecursedirs = ["integration_tests"]

remote_config_example.py

@@ -0,0 +1,32 @@
+#!/usr/bin/env python3
+"""
+Simple test script for PostHog remote config endpoint.
+"""
+
+import posthog
+
+# Initialize PostHog client
+posthog.api_key = "phc_..."
+posthog.personal_api_key = "phs_..."  # or "phx_..."
+posthog.host = "http://localhost:8000"  # or "https://us.posthog.com"
+posthog.debug = True
+
+
+def test_remote_config():
+    """Test remote config payload retrieval."""
+    print("Testing remote config endpoint...")
+
+    # Test feature flag key - replace with an actual flag key from your project
+    flag_key = "unencrypted-remote-config-setting"
+
+    try:
+        # Get remote config payload
+        payload = posthog.get_remote_config_payload(flag_key)
+        print(f"✅ Success! Remote config payload for '{flag_key}': {payload}")
+
+    except Exception as e:
+        print(f"❌ Error getting remote config: {e}")
+
+
+if __name__ == "__main__":
+    test_remote_config()

setup.py

@@ -14,7 +14,7 @@ long_description = """
 PostHog is developer-friendly, self-hosted product analytics.
 posthog-python is the python package.
 
-This package requires Python 3.9 or higher.
+This package requires Python 3.10 or higher.
 """
 
 # Minimal setup.py for backward compatibility

setup_analytics.py

@@ -47,7 +47,7 @@ long_description = """
 PostHog is developer-friendly, self-hosted product analytics.
 posthog-python is the python package.
 
-This package requires Python 3.9 or higher.
+This package requires Python 3.10 or higher.
 """
 
 # Minimal setup.py for backward compatibility