feat(llmo): support Vertex AI (#302)

* feat(llmo): support Vertex AI

* chore(llmo): run formatter

* fix(llmo): fix types error

* chore(llmo): run formatter

* chore(llmo): bump version

.gitignore

@@ -18,3 +18,4 @@ posthog-analytics
 pyrightconfig.json
 .env
 .DS_Store
+posthog-python-references.json

CHANGELOG.md

@@ -1,3 +1,82 @@
+# 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
+
+# 6.3.3 - 2025-08-01
+
+- fix: `get_feature_flag_result` now correctly returns FeatureFlagResult when payload is empty string instead of None
+
+# 6.3.2 - 2025-07-31
+
+- fix: Anthropic's tool calls are now handled properly
+
+# 6.3.0 - 2025-07-22
+
+- feat: Enhanced `send_feature_flags` parameter to accept `SendFeatureFlagsOptions` object for declarative control over local/remote evaluation and custom properties
+
+# 6.2.1 - 2025-07-21
+
+- feat: make `posthog_client` an optional argument in PostHog AI providers wrappers (`posthog.ai.*`), intuitively using the default client as the default
+
+# 6.1.1 - 2025-07-16
+
+- fix: correctly capture exceptions processed by Django from views or middleware
+
+# 6.1.0 - 2025-07-10
+
+- feat: decouple feature flag local evaluation from personal API keys; support decrypting remote config payloads without relying on the feature flags poller
+
+# 6.0.4 - 2025-07-09
+
+- fix: add POSTHOG_MW_CLIENT setting to django middleware, to support custom clients for exception capture.
+
+# 6.0.3 - 2025-07-07
+
+- feat: add a feature flag evaluation cache (local storage or redis) to support returning flag evaluations when the service is down
+
+# 6.0.2 - 2025-07-02
+
+- fix: send_feature_flags changed to default to false in `Client::capture_exception`
+
+# 6.0.1
+
+- fix: response `$process_person_profile` property when passed to capture
+
+# 6.0.0
+
+This release contains a number of major breaking changes:
+
+- feat: make distinct_id an optional parameter in posthog.capture and related functions
+- feat: make capture and related functions return `Optional[str]`, which is the UUID of the sent event, if it was sent
+- fix: remove `identify` (prefer `posthog.set()`), and `page` and `screen` (prefer `posthog.capture()`)
+- fix: delete exception-capture specific integrations module. Prefer the general-purpose django middleware as a replacement for the django `Integration`.
+
+To migrate to this version, you'll mostly just need to switch to using named keyword arguments, rather than positional ones. For example:
+
+```python
+# Old calling convention
+posthog.capture("user123", "button_clicked", {"button_id": "123"})
+# New calling convention
+posthog.capture(distinct_id="user123", event="button_clicked", properties={"button_id": "123"})
+
+# Better pattern
+with posthog.new_context():
+    posthog.identify_context("user123")
+
+    # The event name is the first argument, and can be passed positionally, or as a keyword argument in a later position
+    posthog.capture("button_pressed")
+```
+
+Generally, arguments are now appropriately typed, and docstrings have been updated. If something is unclear, please open an issue, or submit a PR!
+
+# 5.4.0 - 2025-06-20
+
+- feat: add support to session_id context on page method
+
 # 5.3.0 - 2025-06-19
 
 - fix: safely handle exception values

README.md

@@ -32,7 +32,7 @@ We recommend using [uv](https://docs.astral.sh/uv/). It's super fast.
 ```bash
 uv python install 3.9.19
 uv python pin 3.9.19
-uv venv env
+uv venv
 source env/bin/activate
 uv sync --extra dev --extra test
 pre-commit install
@@ -45,7 +45,9 @@ Assuming you have a [local version of PostHog](https://posthog.com/docs/developi
 
 ### Releasing Versions
 
-Updates are released using GitHub Actions: after bumping `version.py` in `master` and adding to `CHANGELOG.md`, go to [our release workflow's page](https://github.com/PostHog/posthog-python/actions/workflows/release.yaml) and dispatch it manually, using workflow from `master`.
+Updates are released automatically using GitHub Actions when `version.py` is updated on `master`. After bumping `version.py` in `master` and adding to `CHANGELOG.md`, the [release workflow](https://github.com/PostHog/posthog-python/blob/master/.github/workflows/release.yaml) will automatically trigger and deploy the new version.
+
+If you need to check the latest runs or manually trigger a release, you can go to [our release workflow's page](https://github.com/PostHog/posthog-python/actions/workflows/release.yaml) and dispatch it manually, using workflow from `master`.
 
 
 ### Testing changes locally with the PostHog app

bin/docs

@@ -0,0 +1,8 @@
+#!/usr/bin/env bash
+#/ Usage: bin/docs
+#/ Description: Generate documentation for the PostHog Python SDK
+source bin/helpers/_utils.sh
+set_source_and_root_dir
+ensure_virtual_env
+
+exec python3 "$(dirname "$0")/docs_scripts/generate_json_schemas.py" "$@" 

bin/docs_scripts/doc_constant.py

@@ -0,0 +1,81 @@
+"""
+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",
+]
+
+# Documentation generation metadata
+DOCUMENTATION_METADATA = {
+    "hogRef": "0.1",
+    "slugPrefix": "posthog-python",
+    "specUrl": "https://github.com/PostHog/posthog-python",
+}
+
+# Docstring parsing patterns for new format
+DOCSTRING_PATTERNS = {
+    "examples_section": r"Examples:\s*\n(.*?)(?=\n\s*\n\s*Category:|\Z)",
+    "args_section": r"Args:\s*\n(.*?)(?=\n\s*\n\s*Examples:|\n\s*\n\s*Details:|\n\s*\n\s*Category:|\Z)",
+    "details_section": r"Details:\s*\n(.*?)(?=\n\s*\n\s*Examples:|\n\s*\n\s*Category:|\Z)",
+    "category_section": r"Category:\s*\n\s*(.+?)\s*(?:\n|$)",
+    "code_block": r"```(?:python)?\n(.*?)```",
+    "param_description": r"^\s*{param_name}:\s*(.+?)(?=\n\s*\w+:|\Z)",
+    "args_marker": r"\n\s*Args:\s*\n",
+    "examples_marker": r"\n\s*Examples:\s*\n",
+    "details_marker": r"\n\s*Details:\s*\n",
+    "category_marker": r"\n\s*Category:\s*\n",
+}
+
+# Output file configuration
+OUTPUT_CONFIG: Dict[str, Union[str, int]] = {
+    "output_dir": ".",
+    "filename": "posthog-python-references.json",
+    "indent": 2,
+}
+
+# Documentation structure defaults
+DOC_DEFAULTS = {
+    "showDocs": True,
+    "releaseTag": "public",
+    "return_type_void": "None",
+    "max_optional_params": 3,
+}

bin/docs_scripts/generate_json_schemas.py

@@ -0,0 +1,472 @@
+#!/usr/bin/env python3
+"""
+Generate comprehensive SDK documentation JSON from PostHog Python SDK.
+This script inspects the code and docstrings to create documentation in the specified format.
+"""
+
+import json
+import inspect
+import re
+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,
+    DOC_DEFAULTS,
+)
+import os
+
+
+def extract_examples_from_docstring(docstring: str) -> list:
+    """Extract code examples from docstring."""
+    if not docstring:
+        return []
+
+    examples = []
+
+    # Look for Examples section in the new format
+    examples_section_match = re.search(
+        DOCSTRING_PATTERNS["examples_section"], docstring, re.DOTALL
+    )
+    if examples_section_match:
+        examples_content = examples_section_match.group(1).strip()
+        # Extract code blocks from the Examples section
+        code_blocks = re.findall(
+            DOCSTRING_PATTERNS["code_block"], examples_content, re.DOTALL
+        )
+        for i, code_block in enumerate(code_blocks):
+            # Remove common leading whitespace while preserving relative indentation
+            code = dedent(code_block).strip()
+
+            # Extract name from first comment line if present
+            lines = code.split("\n")
+            name = f"Example {i + 1}"  # Default fallback
+
+            if lines and lines[0].strip().startswith("#"):
+                # Extract name from first comment, keep the comment in the code
+                comment_text = lines[0].strip()[1:].strip()
+                if comment_text:
+                    name = comment_text
+
+            examples.append({"id": f"example_{i + 1}", "name": name, "code": code})
+
+    return examples
+
+
+def extract_details_from_docstring(docstring: str) -> str:
+    """Extract details section from docstring."""
+    if not docstring:
+        return ""
+
+    # Look for Details section
+    details_match = re.search(
+        DOCSTRING_PATTERNS["details_section"], docstring, re.DOTALL
+    )
+    if details_match:
+        details_content = details_match.group(1).strip()
+        # Clean up formatting
+        return details_content.replace("\n", " ")
+
+    return ""
+
+
+def parse_docstring_tags(docstring: str) -> dict:
+    """Parse tags from docstring Category section."""
+    if not docstring:
+        return {}
+
+    tags = {}
+
+    # Extract Category section
+    category_match = re.search(DOCSTRING_PATTERNS["category_section"], docstring)
+    if category_match:
+        category_value = category_match.group(1).strip()
+        tags["category"] = category_value
+
+    return tags
+
+
+def extract_description_from_docstring(docstring: str) -> str:
+    """Extract main description from docstring."""
+    if not docstring:
+        return ""
+
+    # Clean up the docstring
+    cleaned = dedent(docstring).strip()
+
+    # Find the end of the description by looking for first section marker
+    # Check for Args:, Examples:, Details:, or Category: sections
+    section_patterns = [
+        DOCSTRING_PATTERNS["args_marker"],
+        DOCSTRING_PATTERNS["examples_marker"],
+        DOCSTRING_PATTERNS["details_marker"],
+        DOCSTRING_PATTERNS["category_marker"],
+    ]
+
+    end_pos = len(cleaned)
+    for pattern in section_patterns:
+        match = re.search(pattern, cleaned)
+        if match:
+            end_pos = min(end_pos, match.start())
+
+    # Extract description up to the first section marker
+    description = cleaned[:end_pos].strip()
+
+    # Remove one level of \n since it will be rendered as markdown
+    # and \n will be padded in later steps
+    description = description.replace("\n", " ")
+
+    return description
+
+
+def get_type_name(type_annotation) -> str:
+    """Convert type annotation to string name."""
+    if type_annotation is None or type_annotation is type(None):
+        return "any"
+
+    # Handle typing constructs
+    origin = get_origin(type_annotation)
+    if origin is not None:
+        # Handle Union types (including Optional)
+        if origin is Union:
+            args = get_args(type_annotation)
+            if len(args) == 2 and type(None) in args:
+                # This is Optional[Type] - get the non-None type
+                non_none_type = next(arg for arg in args if arg is not type(None))
+                return f"Optional[{get_type_name(non_none_type)}]"
+            else:
+                # Regular Union - list all types
+                type_names = [get_type_name(arg) for arg in args]
+                return f"Union[{', '.join(type_names)}]"
+
+        # Handle other generic types (List, Dict, etc.)
+        origin_name = getattr(origin, "__name__", str(origin))
+        args = get_args(type_annotation)
+        if args:
+            arg_names = [get_type_name(arg) for arg in args]
+            return f"{origin_name}[{', '.join(arg_names)}]"
+        else:
+            return origin_name
+
+    # Handle regular types
+    elif hasattr(type_annotation, "__name__"):
+        return type_annotation.__name__
+    else:
+        return str(type_annotation)
+
+
+def analyze_parameter(param: inspect.Parameter, docstring: str = "") -> dict:
+    """Analyze a function parameter and return its documentation."""
+    # Determine if parameter is optional (has default value)
+    is_optional = param.default == inspect.Parameter.empty
+
+    # Get the type annotation
+    type_annotation = param.annotation
+    param_type = "any"
+
+    if type_annotation != inspect.Parameter.empty:
+        # Handle Union/Optional types first
+        origin = get_origin(type_annotation)
+        if origin is Union:
+            args = get_args(type_annotation)
+            if len(args) == 2 and type(None) in args:
+                # This is Optional[Type]
+                non_none_type = next(arg for arg in args if arg is not type(None))
+                param_type = get_type_name(non_none_type)
+                is_optional = True
+            else:
+                # Other Union types, use first type
+                param_type = get_type_name(args[0]) if args else "any"
+        else:
+            param_type = get_type_name(type_annotation)
+    elif param.default != inspect.Parameter.empty:
+        # No type annotation, but has default value - infer type from default
+        param_type = get_type_name(type(param.default))
+
+    # Extract parameter description from Args section
+    param_description = f"Parameter: {param.name}"
+    if docstring:
+        # Look for Args section and extract description for this parameter
+        args_section_match = re.search(
+            DOCSTRING_PATTERNS["args_section"], docstring, re.DOTALL
+        )
+        if args_section_match:
+            args_content = args_section_match.group(1)
+            # Look for the parameter description
+            param_pattern = DOCSTRING_PATTERNS["param_description"].format(
+                param_name=re.escape(param.name)
+            )
+            param_match = re.search(
+                param_pattern, args_content, re.MULTILINE | re.DOTALL
+            )
+            if param_match:
+                param_description = param_match.group(1).strip().replace("\n", " ")
+
+    param_info = {
+        "name": param.name,
+        "description": param_description,
+        "isOptional": is_optional,
+        "type": param_type,
+    }
+
+    return param_info
+
+
+def analyze_function(func, name: str) -> dict:
+    """Analyze a function and return its documentation."""
+    try:
+        sig = inspect.signature(func)
+        docstring = inspect.getdoc(func) or ""
+
+        # Skip functions with empty docstrings
+        if not docstring.strip():
+            return {}
+
+        # Extract parameters (excluding 'self')
+        params = []
+        for param_name, param in sig.parameters.items():
+            if param_name != "self":
+                params.append(analyze_parameter(param, docstring))
+
+        # Special handling for constructor
+        display_name = name
+        if name == "__init__":
+            display_name = func.__qualname__.split(".")[0]
+
+        # Parse tags from docstring
+        tags = parse_docstring_tags(docstring)
+
+        category = tags.get("category", None)
+
+        # Extract description
+        description = extract_description_from_docstring(docstring)
+
+        # Skip if no meaningful description
+        if not description.strip():
+            return {}
+
+        # Extract details section (only if it exists)
+        details = extract_details_from_docstring(docstring)
+
+        # Get examples from docstring, do not generate fallback examples
+        examples = extract_examples_from_docstring(docstring)
+        # If no examples, do not include the examples key or set to empty list
+
+        result = {
+            "id": name,
+            "title": display_name,
+            "description": description,
+            "details": details,
+            "category": category,
+            "params": params,
+            "showDocs": DOC_DEFAULTS["showDocs"],
+            "releaseTag": DOC_DEFAULTS["releaseTag"],
+            "returnType": {
+                "id": "return_type",
+                "name": get_type_name(sig.return_annotation)
+                if sig.return_annotation != inspect.Signature.empty
+                else DOC_DEFAULTS["return_type_void"],
+            },
+        }
+        if examples:
+            result["examples"] = examples
+        return result
+    except Exception as e:
+        print(f"Error analyzing function {name}: {e}")
+        return {}
+
+
+def analyze_class(cls) -> dict:
+    """Analyze a class and return its documentation."""
+    class_doc = inspect.getdoc(cls) or f"Class: {cls.__name__}"
+
+    # Get all public methods and constructor
+    functions = []
+    for method_name in dir(cls):
+        if method_name.startswith("_") and method_name != "__init__":
+            continue
+
+        method = getattr(cls, method_name)
+        if callable(method):
+            func_info = analyze_function(method, method_name)
+            if func_info:  # Only add if not None (empty docstring check)
+                functions.append(func_info)
+
+    return {
+        "id": cls.__name__,
+        "title": cls.__name__,
+        "description": extract_description_from_docstring(class_doc),
+        "functions": functions,
+    }
+
+
+def analyze_type(cls) -> dict:
+    """Analyze a type/dataclass and return its documentation."""
+    type_info = {
+        "id": cls.__name__,
+        "name": cls.__name__,
+        "path": f"{cls.__module__}.{cls.__name__}",
+        "properties": [],
+        "example": "",
+    }
+
+    if is_dataclass(cls):
+        # Handle dataclass
+        for field in fields(cls):
+            prop = {
+                "name": field.name,
+                "type": get_type_name(field.type),
+                "description": f"Field: {field.name}",
+            }
+            type_info["properties"].append(prop)
+    elif hasattr(cls, "__annotations__"):
+        # Handle TypedDict or annotated class
+        for field_name, field_type in cls.__annotations__.items():
+            prop = {
+                "name": field_name,
+                "type": get_type_name(field_type),
+                "description": f"Field: {field_name}",
+            }
+            type_info["properties"].append(prop)
+
+    return type_info
+
+
+def generate_sdk_documentation():
+    """Generate complete SDK documentation in the requested format."""
+
+    # Import PostHog components
+    import posthog
+    from posthog.client import Client
+    import posthog.types as types_module
+    import posthog.args as args_module
+    from posthog.version import VERSION
+
+    # Main SDK info
+    sdk_info = {
+        "version": VERSION,
+        "id": "posthog-python",
+        "title": "PostHog Python SDK",
+        "description": "Integrate PostHog into any python application.",
+        "slugPrefix": DOCUMENTATION_METADATA["slugPrefix"],
+        "specUrl": DOCUMENTATION_METADATA["specUrl"],
+    }
+
+    # Collect types
+    types_list = []
+
+    # Types from posthog.types
+    for name in dir(types_module):
+        obj = getattr(types_module, name)
+        if inspect.isclass(obj) and not name.startswith("_"):
+            try:
+                type_info = analyze_type(obj)
+                types_list.append(type_info)
+            except Exception as e:
+                print(f"Error analyzing type {name}: {e}")
+
+    # Types from posthog.args
+    for name in dir(args_module):
+        obj = getattr(args_module, name)
+        if inspect.isclass(obj) and not name.startswith("_"):
+            try:
+                type_info = analyze_type(obj)
+                types_list.append(type_info)
+            except Exception as e:
+                print(f"Error analyzing type {name}: {e}")
+
+    # Collect classes
+    classes_list = []
+
+    # Main PostHog class (renamed from Client)
+    client_class = analyze_class(Client)
+    client_class["id"] = "PostHog"
+    client_class["title"] = "PostHog"
+    classes_list.append(client_class)
+
+    # Global module functions (functions callable as posthog.function_name)
+    global_functions = []
+    for func_name in dir(posthog):
+        # Skip private functions and non-callables
+        if func_name.startswith("_") or not callable(getattr(posthog, func_name)):
+            continue
+
+        func = getattr(posthog, func_name)
+        # Only include functions actually defined in the posthog module (not imported)
+        # and exclude class references
+        if (
+            func_name not in ["Client", "Posthog"]
+            and hasattr(func, "__module__")
+            and func.__module__ == "posthog"
+        ):
+            try:
+                func_info = analyze_function(func, func_name)
+                if func_info:  # Only add if not None (has proper docstring)
+                    global_functions.append(func_info)
+            except Exception:
+                continue
+
+    # Add global functions as a "class"
+    if global_functions:
+        classes_list.append(
+            {
+                "id": "PostHogModule",
+                "title": "PostHog Module Functions",
+                "description": "Global functions available in the PostHog module",
+                "functions": global_functions,
+            }
+        )
+
+    # 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,
+    }
+
+    return result
+
+
+if __name__ == "__main__":
+    print("Generating PostHog Python SDK documentation...")
+
+    try:
+        documentation = generate_sdk_documentation()
+
+        # Write to file
+        output_file = os.path.join(
+            str(OUTPUT_CONFIG["output_dir"]), str(OUTPUT_CONFIG["filename"])
+        )
+        with open(output_file, "w") as f:
+            json.dump(documentation, f, indent=int(OUTPUT_CONFIG["indent"]))
+
+        print(f"✓ Generated {output_file}")
+
+        # Print summary
+        types_count = len(documentation["types"])
+        classes_count = len(documentation["classes"])
+
+        total_functions = sum(len(cls["functions"]) for cls in documentation["classes"])
+
+        print("📊 Documentation Summary:")
+        print(f"   • {types_count} types documented")
+        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
+
+        traceback.print_exc()

example.py

@@ -41,9 +41,9 @@ print(
 
 # Capture an event
 posthog.capture(
-    "distinct_id",
     "event",
-    {"property1": "value", "property2": "value"},
+    distinct_id="distinct_id",
+    properties={"property1": "value", "property2": "value"},
     send_feature_flags=True,
 )
 
@@ -59,37 +59,52 @@ 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"))
-exit()
-# # Alias a previous distinct id with a new one
+
+# 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(
-    "new_distinct_id", "event2", {"property1": "value", "property2": "value"}
+    "event2",
+    distinct_id="new_distinct_id",
+    properties={"property1": "value", "property2": "value"},
 )
 posthog.capture(
-    "new_distinct_id",
     "event-with-groups",
-    {"property1": "value", "property2": "value"},
+    distinct_id="new_distinct_id",
+    properties={"property1": "value", "property2": "value"},
     groups={"company": "id:5"},
 )
 
 # # Add properties to the person
-posthog.identify("new_distinct_id", {"email": "something@something.com"})
+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("new_distinct_id", {"self_serve_signup": True})
+posthog.set_once(distinct_id="new_distinct_id", properties={"self_serve_signup": True})
 
 
 posthog.set_once(
-    "new_distinct_id", {"self_serve_signup": False}
+    distinct_id="new_distinct_id", properties={"self_serve_signup": False}
 )  # this will not change the property (because it was already set)
 
-posthog.set("new_distinct_id", {"current_browser": "Chrome"})
-posthog.set("new_distinct_id", {"current_browser": "Firefox"})
+posthog.set(distinct_id="new_distinct_id", properties={"current_browser": "Chrome"})
+posthog.set(distinct_id="new_distinct_id", properties={"current_browser": "Firefox"})
 
 
 # #############################################################################

mypy-baseline.txt

@@ -22,16 +22,25 @@ posthog/client.py:0: error: Library stubs not installed for "six"  [import-untyp
 posthog/client.py:0: note: Hint: "python3 -m pip install types-six"
 posthog/client.py:0: error: Name "queue" already defined (by an import)  [no-redef]
 posthog/client.py:0: error: Need type annotation for "queue"  [var-annotated]
-posthog/client.py:0: error: Item "None" of "Any | None" has no attribute "get"  [union-attr]
-simulator.py:0: error: Unexpected keyword argument "anonymous_id" for "capture"  [call-arg]
-posthog/__init__.py:0: note: "capture" defined here
-simulator.py:0: error: Unexpected keyword argument "anonymous_id" for "identify"  [call-arg]
-posthog/__init__.py:0: note: "identify" defined here
-simulator.py:0: error: Unexpected keyword argument "traits" for "identify"  [call-arg]
-posthog/__init__.py:0: note: "identify" defined here
+posthog/client.py:0: error: Incompatible types in assignment (expression has type "Any | list[Any]", variable has type "None")  [assignment]
+posthog/client.py:0: error: Incompatible types in assignment (expression has type "dict[Any, Any]", variable has type "None")  [assignment]
+posthog/client.py:0: error: "None" has no attribute "__iter__" (not iterable)  [attr-defined]
+posthog/client.py:0: error: Statement is unreachable  [unreachable]
+posthog/client.py:0: error: Incompatible types in assignment (expression has type "Any | dict[Any, Any]", variable has type "None")  [assignment]
+posthog/client.py:0: error: Incompatible types in assignment (expression has type "Any | dict[Any, Any]", variable has type "None")  [assignment]
+posthog/client.py:0: error: Incompatible types in assignment (expression has type "dict[Never, Never]", variable has type "None")  [assignment]
+posthog/client.py:0: error: Incompatible types in assignment (expression has type "dict[Never, Never]", variable has type "None")  [assignment]
+posthog/client.py:0: error: Right operand of "and" is never evaluated  [unreachable]
+posthog/client.py:0: error: Incompatible types in assignment (expression has type "Poller", variable has type "None")  [assignment]
+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/ai/anthropic/anthropic.py

@@ -8,7 +8,7 @@ except ImportError:
 
 import time
 import uuid
-from typing import Any, Dict, Optional
+from typing import Any, Dict, Optional, cast
 
 from posthog.ai.utils import (
     call_llm_and_track_usage,
@@ -17,6 +17,7 @@ from posthog.ai.utils import (
     with_privacy_mode,
 )
 from posthog.client import Client as PostHogClient
+from posthog import setup
 
 
 class Anthropic(anthropic.Anthropic):
@@ -26,14 +27,14 @@ class Anthropic(anthropic.Anthropic):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         """
         Args:
             posthog_client: PostHog client for tracking usage
             **kwargs: Additional arguments passed to the Anthropic client
         """
         super().__init__(**kwargs)
-        self._ph_client = posthog_client
+        self._ph_client = posthog_client or setup()
         self.messages = WrappedMessages(self)
 
 

posthog/ai/anthropic/anthropic_async.py

@@ -10,6 +10,7 @@ import time
 import uuid
 from typing import Any, Dict, Optional
 
+from posthog import setup
 from posthog.ai.utils import (
     call_llm_and_track_usage_async,
     get_model_params,
@@ -26,14 +27,14 @@ class AsyncAnthropic(anthropic.AsyncAnthropic):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         """
         Args:
             posthog_client: PostHog client for tracking usage
             **kwargs: Additional arguments passed to the Anthropic client
         """
         super().__init__(**kwargs)
-        self._ph_client = posthog_client
+        self._ph_client = posthog_client or setup()
         self.messages = AsyncWrappedMessages(self)
 
 

posthog/ai/anthropic/anthropic_providers.py

@@ -5,9 +5,12 @@ except ImportError:
         "Please install the Anthropic SDK to use this feature: 'pip install anthropic'"
     )
 
+from typing import Optional
+
 from posthog.ai.anthropic.anthropic import WrappedMessages
 from posthog.ai.anthropic.anthropic_async import AsyncWrappedMessages
 from posthog.client import Client as PostHogClient
+from posthog import setup
 
 
 class AnthropicBedrock(anthropic.AnthropicBedrock):
@@ -17,9 +20,9 @@ class AnthropicBedrock(anthropic.AnthropicBedrock):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         super().__init__(**kwargs)
-        self._ph_client = posthog_client
+        self._ph_client = posthog_client or setup()
         self.messages = WrappedMessages(self)
 
 
@@ -30,9 +33,9 @@ class AsyncAnthropicBedrock(anthropic.AsyncAnthropicBedrock):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         super().__init__(**kwargs)
-        self._ph_client = posthog_client
+        self._ph_client = posthog_client or setup()
         self.messages = AsyncWrappedMessages(self)
 
 
@@ -43,9 +46,9 @@ class AnthropicVertex(anthropic.AnthropicVertex):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         super().__init__(**kwargs)
-        self._ph_client = posthog_client
+        self._ph_client = posthog_client or setup()
         self.messages = WrappedMessages(self)
 
 
@@ -56,7 +59,7 @@ class AsyncAnthropicVertex(anthropic.AsyncAnthropicVertex):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         super().__init__(**kwargs)
-        self._ph_client = posthog_client
+        self._ph_client = posthog_client or setup()
         self.messages = AsyncWrappedMessages(self)

posthog/ai/gemini/gemini.py

@@ -10,6 +10,7 @@ except ImportError:
         "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,
     get_model_params,
@@ -36,9 +37,17 @@ class Client:
         )
     """
 
+    _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,
@@ -48,7 +57,13 @@ 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)
@@ -56,12 +71,20 @@ class Client:
             posthog_groups: Default groups for all calls (can be overridden per call)
             **kwargs: Additional arguments (for future compatibility)
         """
-        if posthog_client is None:
+        self._ph_client = posthog_client or setup()
+
+        if self._ph_client is None:
             raise ValueError("posthog_client is required for PostHog tracking")
 
         self.models = Models(
             api_key=api_key,
-            posthog_client=posthog_client,
+            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,
@@ -80,6 +103,12 @@ 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,
@@ -89,7 +118,13 @@ 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
@@ -97,10 +132,10 @@ class Models:
             posthog_groups: Default groups for all calls
             **kwargs: Additional arguments (for future compatibility)
         """
-        if posthog_client is None:
-            raise ValueError("posthog_client is required for PostHog tracking")
+        self._ph_client = posthog_client or setup()
 
-        self._ph_client = posthog_client
+        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
@@ -108,16 +143,40 @@ 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
+        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
 
-        self._client = genai.Client(api_key=api_key)
+        # 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(

posthog/ai/langchain/callbacks.py

@@ -5,6 +5,7 @@ except ImportError:
         "Please install LangChain to use this feature: 'pip install langchain'"
     )
 
+import json
 import logging
 import time
 from dataclasses import dataclass
@@ -29,11 +30,12 @@ from langchain_core.messages import (
     HumanMessage,
     SystemMessage,
     ToolMessage,
+    ToolCall,
 )
 from langchain_core.outputs import ChatGeneration, LLMResult
 from pydantic import BaseModel
 
-from posthog import default_client
+from posthog import setup
 from posthog.ai.utils import get_model_params, with_privacy_mode
 from posthog.client import Client
 
@@ -81,10 +83,10 @@ class CallbackHandler(BaseCallbackHandler):
     The PostHog LLM observability callback handler for LangChain.
     """
 
-    _client: Client
+    _ph_client: Client
     """PostHog client instance."""
 
-    _distinct_id: Optional[Union[str, int, float, UUID]]
+    _distinct_id: Optional[Union[str, int, UUID]]
     """Distinct ID of the user to associate the trace with."""
 
     _trace_id: Optional[Union[str, int, float, UUID]]
@@ -112,7 +114,7 @@ class CallbackHandler(BaseCallbackHandler):
         self,
         client: Optional[Client] = None,
         *,
-        distinct_id: Optional[Union[str, int, float, UUID]] = None,
+        distinct_id: Optional[Union[str, int, UUID]] = None,
         trace_id: Optional[Union[str, int, float, UUID]] = None,
         properties: Optional[Dict[str, Any]] = None,
         privacy_mode: bool = False,
@@ -127,10 +129,7 @@ class CallbackHandler(BaseCallbackHandler):
             privacy_mode: Whether to redact the input and output of the trace.
             groups: Optional additional PostHog groups to use for the trace.
         """
-        posthog_client = client or default_client
-        if posthog_client is None:
-            raise ValueError("PostHog client is required")
-        self._client = posthog_client
+        self._ph_client = client or setup()
         self._distinct_id = distinct_id
         self._trace_id = trace_id
         self._properties = properties or {}
@@ -481,7 +480,7 @@ class CallbackHandler(BaseCallbackHandler):
         event_properties = {
             "$ai_trace_id": trace_id,
             "$ai_input_state": with_privacy_mode(
-                self._client, self._privacy_mode, run.input
+                self._ph_client, self._privacy_mode, run.input
             ),
             "$ai_latency": run.latency,
             "$ai_span_name": run.name,
@@ -497,13 +496,13 @@ class CallbackHandler(BaseCallbackHandler):
             event_properties["$ai_is_error"] = True
         elif outputs is not None:
             event_properties["$ai_output_state"] = with_privacy_mode(
-                self._client, self._privacy_mode, outputs
+                self._ph_client, self._privacy_mode, outputs
             )
 
         if self._distinct_id is None:
             event_properties["$process_person_profile"] = False
 
-        self._client.capture(
+        self._ph_client.capture(
             distinct_id=self._distinct_id or run_id,
             event=event_name,
             properties=event_properties,
@@ -550,17 +549,16 @@ class CallbackHandler(BaseCallbackHandler):
             "$ai_provider": run.provider,
             "$ai_model": run.model,
             "$ai_model_parameters": run.model_params,
-            "$ai_input": with_privacy_mode(self._client, self._privacy_mode, run.input),
+            "$ai_input": with_privacy_mode(
+                self._ph_client, self._privacy_mode, run.input
+            ),
             "$ai_http_status": 200,
             "$ai_latency": run.latency,
             "$ai_base_url": run.base_url,
         }
+
         if run.tools:
-            event_properties["$ai_tools"] = with_privacy_mode(
-                self._client,
-                self._privacy_mode,
-                run.tools,
-            )
+            event_properties["$ai_tools"] = run.tools
 
         if isinstance(output, BaseException):
             event_properties["$ai_http_status"] = _get_http_status(output)
@@ -586,10 +584,11 @@ class CallbackHandler(BaseCallbackHandler):
                 ]
             else:
                 completions = [
-                    _extract_raw_esponse(generation) for generation in generation_result
+                    _extract_raw_response(generation)
+                    for generation in generation_result
                 ]
             event_properties["$ai_output_choices"] = with_privacy_mode(
-                self._client, self._privacy_mode, completions
+                self._ph_client, self._privacy_mode, completions
             )
 
         if self._properties:
@@ -598,7 +597,7 @@ class CallbackHandler(BaseCallbackHandler):
         if self._distinct_id is None:
             event_properties["$process_person_profile"] = False
 
-        self._client.capture(
+        self._ph_client.capture(
             distinct_id=self._distinct_id or trace_id,
             event="$ai_generation",
             properties=event_properties,
@@ -617,7 +616,7 @@ class CallbackHandler(BaseCallbackHandler):
         )
 
 
-def _extract_raw_esponse(last_response):
+def _extract_raw_response(last_response):
     """Extract the response from the last response of the LLM call."""
     # We return the text of the response if not empty
     if last_response.text is not None and last_response.text.strip() != "":
@@ -630,12 +629,35 @@ def _extract_raw_esponse(last_response):
         return ""
 
 
-def _convert_message_to_dict(message: BaseMessage) -> Dict[str, Any]:
+def _convert_lc_tool_calls_to_oai(
+    tool_calls: list[ToolCall],
+) -> list[dict[str, Any]]:
+    try:
+        return [
+            {
+                "type": "function",
+                "id": tool_call["id"],
+                "function": {
+                    "name": tool_call["name"],
+                    "arguments": json.dumps(tool_call["args"]),
+                },
+            }
+            for tool_call in tool_calls
+        ]
+    except KeyError:
+        return tool_calls
+
+
+def _convert_message_to_dict(message: BaseMessage) -> dict[str, Any]:
     # assistant message
     if isinstance(message, HumanMessage):
         message_dict = {"role": "user", "content": message.content}
     elif isinstance(message, AIMessage):
         message_dict = {"role": "assistant", "content": message.content}
+        if message.tool_calls:
+            message_dict["tool_calls"] = _convert_lc_tool_calls_to_oai(
+                message.tool_calls
+            )
     elif isinstance(message, SystemMessage):
         message_dict = {"role": "system", "content": message.content}
     elif isinstance(message, ToolMessage):
@@ -648,6 +670,9 @@ def _convert_message_to_dict(message: BaseMessage) -> Dict[str, Any]:
     if message.additional_kwargs:
         message_dict.update(message.additional_kwargs)
 
+    if "content" in message_dict and not message_dict["content"]:
+        message_dict["content"] = ""
+
     return message_dict
 
 

posthog/ai/openai/openai.py

@@ -11,10 +11,12 @@ except ImportError:
 
 from posthog.ai.utils import (
     call_llm_and_track_usage,
+    extract_available_tool_calls,
     get_model_params,
     with_privacy_mode,
 )
 from posthog.client import Client as PostHogClient
+from posthog import setup
 
 
 class OpenAI(openai.OpenAI):
@@ -24,16 +26,15 @@ class OpenAI(openai.OpenAI):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         """
         Args:
             api_key: OpenAI API key.
-            posthog_client: If provided, events will be captured via this client instead
-                            of the global posthog.
+            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
+        self._ph_client = posthog_client or setup()
 
         # Store original objects after parent initialization (only if they exist)
         self._original_chat = getattr(self, "chat", None)
@@ -167,6 +168,7 @@ class WrappedResponses:
                     usage_stats,
                     latency,
                     output,
+                    extract_available_tool_calls("openai", kwargs),
                 )
 
         return generator()
@@ -182,7 +184,7 @@ class WrappedResponses:
         usage_stats: Dict[str, int],
         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())
@@ -212,12 +214,8 @@ class WrappedResponses:
             **(posthog_properties or {}),
         }
 
-        if tool_calls:
-            event_properties["$ai_tools"] = with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                tool_calls,
-            )
+        if available_tool_calls:
+            event_properties["$ai_tools"] = available_tool_calls
 
         if posthog_distinct_id is None:
             event_properties["$process_person_profile"] = False
@@ -341,7 +339,6 @@ class WrappedCompletions:
         start_time = time.time()
         usage_stats: Dict[str, int] = {}
         accumulated_content = []
-        accumulated_tools = {}
         if "stream_options" not in kwargs:
             kwargs["stream_options"] = {}
         kwargs["stream_options"]["include_usage"] = True
@@ -350,7 +347,6 @@ class WrappedCompletions:
         def generator():
             nonlocal usage_stats
             nonlocal accumulated_content  # noqa: F824
-            nonlocal accumulated_tools  # noqa: F824
 
             try:
                 for chunk in response:
@@ -389,31 +385,12 @@ class WrappedCompletions:
                             if content:
                                 accumulated_content.append(content)
 
-                        # Process tool calls
-                        tool_calls = getattr(chunk.choices[0].delta, "tool_calls", None)
-                        if tool_calls:
-                            for tool_call in tool_calls:
-                                index = tool_call.index
-                                if index not in accumulated_tools:
-                                    accumulated_tools[index] = tool_call
-                                else:
-                                    # Append arguments for existing tool calls
-                                    if hasattr(tool_call, "function") and hasattr(
-                                        tool_call.function, "arguments"
-                                    ):
-                                        accumulated_tools[
-                                            index
-                                        ].function.arguments += (
-                                            tool_call.function.arguments
-                                        )
-
                     yield chunk
 
             finally:
                 end_time = time.time()
                 latency = end_time - start_time
                 output = "".join(accumulated_content)
-                tools = list(accumulated_tools.values()) if accumulated_tools else None
                 self._capture_streaming_event(
                     posthog_distinct_id,
                     posthog_trace_id,
@@ -424,7 +401,7 @@ class WrappedCompletions:
                     usage_stats,
                     latency,
                     output,
-                    tools,
+                    extract_available_tool_calls("openai", kwargs),
                 )
 
         return generator()
@@ -440,7 +417,7 @@ class WrappedCompletions:
         usage_stats: Dict[str, int],
         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())
@@ -470,12 +447,8 @@ class WrappedCompletions:
             **(posthog_properties or {}),
         }
 
-        if tool_calls:
-            event_properties["$ai_tools"] = with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                tool_calls,
-            )
+        if available_tool_calls:
+            event_properties["$ai_tools"] = available_tool_calls
 
         if posthog_distinct_id is None:
             event_properties["$process_person_profile"] = False

posthog/ai/openai/openai_async.py

@@ -1,6 +1,6 @@
 import time
 import uuid
-from typing import Any, Dict, List, Optional
+from typing import Any, Dict, List, Optional, cast
 
 try:
     import openai
@@ -9,8 +9,10 @@ except ImportError:
         "Please install the OpenAI SDK to use this feature: 'pip install openai'"
     )
 
+from posthog import setup
 from posthog.ai.utils import (
     call_llm_and_track_usage_async,
+    extract_available_tool_calls,
     get_model_params,
     with_privacy_mode,
 )
@@ -24,7 +26,7 @@ class AsyncOpenAI(openai.AsyncOpenAI):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         """
         Args:
             api_key: OpenAI API key.
@@ -33,7 +35,7 @@ class AsyncOpenAI(openai.AsyncOpenAI):
             **openai_config: Any additional keyword args to set on openai (e.g. organization="xxx").
         """
         super().__init__(**kwargs)
-        self._ph_client = posthog_client
+        self._ph_client = posthog_client or setup()
 
         # Store original objects after parent initialization (only if they exist)
         self._original_chat = getattr(self, "chat", None)
@@ -167,6 +169,7 @@ class WrappedResponses:
                     usage_stats,
                     latency,
                     output,
+                    extract_available_tool_calls("openai", kwargs),
                 )
 
         return async_generator()
@@ -182,7 +185,7 @@ class WrappedResponses:
         usage_stats: Dict[str, int],
         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())
@@ -212,12 +215,8 @@ class WrappedResponses:
             **(posthog_properties or {}),
         }
 
-        if tool_calls:
-            event_properties["$ai_tools"] = with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                tool_calls,
-            )
+        if available_tool_calls:
+            event_properties["$ai_tools"] = available_tool_calls
 
         if posthog_distinct_id is None:
             event_properties["$process_person_profile"] = False
@@ -343,7 +342,6 @@ class WrappedCompletions:
         start_time = time.time()
         usage_stats: Dict[str, int] = {}
         accumulated_content = []
-        accumulated_tools = {}
 
         if "stream_options" not in kwargs:
             kwargs["stream_options"] = {}
@@ -353,7 +351,6 @@ class WrappedCompletions:
         async def async_generator():
             nonlocal usage_stats
             nonlocal accumulated_content  # noqa: F824
-            nonlocal accumulated_tools  # noqa: F824
 
             try:
                 async for chunk in response:
@@ -392,31 +389,12 @@ class WrappedCompletions:
                             if content:
                                 accumulated_content.append(content)
 
-                        # Process tool calls
-                        tool_calls = getattr(chunk.choices[0].delta, "tool_calls", None)
-                        if tool_calls:
-                            for tool_call in tool_calls:
-                                index = tool_call.index
-                                if index not in accumulated_tools:
-                                    accumulated_tools[index] = tool_call
-                                else:
-                                    # Append arguments for existing tool calls
-                                    if hasattr(tool_call, "function") and hasattr(
-                                        tool_call.function, "arguments"
-                                    ):
-                                        accumulated_tools[
-                                            index
-                                        ].function.arguments += (
-                                            tool_call.function.arguments
-                                        )
-
                     yield chunk
 
             finally:
                 end_time = time.time()
                 latency = end_time - start_time
                 output = "".join(accumulated_content)
-                tools = list(accumulated_tools.values()) if accumulated_tools else None
                 await self._capture_streaming_event(
                     posthog_distinct_id,
                     posthog_trace_id,
@@ -427,7 +405,7 @@ class WrappedCompletions:
                     usage_stats,
                     latency,
                     output,
-                    tools,
+                    extract_available_tool_calls("openai", kwargs),
                 )
 
         return async_generator()
@@ -443,7 +421,7 @@ class WrappedCompletions:
         usage_stats: Dict[str, int],
         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())
@@ -473,12 +451,8 @@ class WrappedCompletions:
             **(posthog_properties or {}),
         }
 
-        if tool_calls:
-            event_properties["$ai_tools"] = with_privacy_mode(
-                self._client._ph_client,
-                posthog_privacy_mode,
-                tool_calls,
-            )
+        if available_tool_calls:
+            event_properties["$ai_tools"] = available_tool_calls
 
         if posthog_distinct_id is None:
             event_properties["$process_person_profile"] = False

posthog/ai/openai/openai_providers.py

@@ -15,7 +15,10 @@ from posthog.ai.openai.openai_async import WrappedBeta as AsyncWrappedBeta
 from posthog.ai.openai.openai_async import WrappedChat as AsyncWrappedChat
 from posthog.ai.openai.openai_async import WrappedEmbeddings as AsyncWrappedEmbeddings
 from posthog.ai.openai.openai_async import WrappedResponses as AsyncWrappedResponses
+from typing import Optional
+
 from posthog.client import Client as PostHogClient
+from posthog import setup
 
 
 class AzureOpenAI(openai.AzureOpenAI):
@@ -25,7 +28,7 @@ class AzureOpenAI(openai.AzureOpenAI):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         """
         Args:
             api_key: Azure OpenAI API key.
@@ -34,7 +37,7 @@ class AzureOpenAI(openai.AzureOpenAI):
             **openai_config: Any additional keyword args to set on Azure OpenAI (e.g. azure_endpoint="xxx").
         """
         super().__init__(**kwargs)
-        self._ph_client = posthog_client
+        self._ph_client = posthog_client or setup()
 
         # Store original objects after parent initialization (only if they exist)
         self._original_chat = getattr(self, "chat", None)
@@ -63,7 +66,7 @@ class AsyncAzureOpenAI(openai.AsyncAzureOpenAI):
 
     _ph_client: PostHogClient
 
-    def __init__(self, posthog_client: PostHogClient, **kwargs):
+    def __init__(self, posthog_client: Optional[PostHogClient] = None, **kwargs):
         """
         Args:
             api_key: Azure OpenAI API key.
@@ -72,7 +75,7 @@ class AsyncAzureOpenAI(openai.AsyncAzureOpenAI):
             **openai_config: Any additional keyword args to set on Azure OpenAI (e.g. azure_endpoint="xxx").
         """
         super().__init__(**kwargs)
-        self._ph_client = posthog_client
+        self._ph_client = posthog_client or setup()
 
         # Store original objects after parent initialization (only if they exist)
         self._original_chat = getattr(self, "chat", None)

posthog/ai/utils.py

@@ -117,34 +117,87 @@ def format_response(response, provider: str):
 
 def format_response_anthropic(response):
     output = []
+    content = []
+
     for choice in response.content:
-        if choice.text:
-            output.append(
-                {
-                    "role": "assistant",
-                    "content": choice.text,
-                }
-            )
+        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 and choice.message.content:
-                output.append(
-                    {
-                        "content": choice.message.content,
-                        "role": choice.message.role,
-                    }
-                )
+            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":
-                # Extract text content from the content list
+                role = item.role
+
                 if hasattr(item, "content") and isinstance(item.content, list):
                     for content_item in item.content:
                         if (
@@ -152,99 +205,110 @@ def format_response_openai(response):
                             and content_item.type == "output_text"
                             and hasattr(content_item, "text")
                         ):
-                            output.append(
-                                {
-                                    "content": content_item.text,
-                                    "role": item.role,
-                                }
-                            )
+                            content.append({"type": "text", "text": content_item.text})
                         elif hasattr(content_item, "text"):
-                            output.append(
-                                {
-                                    "content": content_item.text,
-                                    "role": item.role,
-                                }
-                            )
+                            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")
                         ):
-                            output.append(
+                            content.append(
                                 {
-                                    "content": {
-                                        "type": "image",
-                                        "image": content_item.image_url,
-                                    },
-                                    "role": item.role,
+                                    "type": "image",
+                                    "image": content_item.image_url,
                                 }
                             )
-                else:
-                    output.append(
-                        {
-                            "content": item.content,
-                            "role": item.role,
-                        }
-                    )
+                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_text = ""
+                content = []
+
                 if hasattr(candidate.content, "parts") and candidate.content.parts:
                     for part in candidate.content.parts:
                         if hasattr(part, "text") and part.text:
-                            content_text += part.text
-                if content_text:
-                    output.append(
-                        {
-                            "role": "assistant",
-                            "content": content_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": candidate.text,
+                        "content": [{"type": "text", "text": candidate.text}],
                     }
                 )
     elif hasattr(response, "text") and response.text:
         output.append(
             {
                 "role": "assistant",
-                "content": response.text,
+                "content": [{"type": "text", "text": response.text}],
             }
         )
+
     return output
 
 
-def format_tool_calls(response, provider: str):
+def extract_available_tool_calls(provider: str, kwargs: Dict[str, Any]):
     if provider == "anthropic":
-        if hasattr(response, "tools") and response.tools and len(response.tools) > 0:
-            return response.tools
-    elif provider == "openai":
-        # Handle both Chat Completions and Responses API
-        if hasattr(response, "choices") and response.choices:
-            # Check for tool_calls in message (Chat Completions format)
-            if (
-                hasattr(response.choices[0], "message")
-                and hasattr(response.choices[0].message, "tool_calls")
-                and response.choices[0].message.tool_calls
-            ):
-                return response.choices[0].message.tool_calls
+        if "tools" in kwargs:
+            return kwargs["tools"]
 
-            # Check for tool_calls directly in response (Responses API format)
-            if (
-                hasattr(response.choices[0], "tool_calls")
-                and response.choices[0].tool_calls
-            ):
-                return response.choices[0].tool_calls
-    return None
+        return None
+    elif provider == "gemini":
+        if "config" in kwargs and hasattr(kwargs["config"], "tools"):
+            return kwargs["config"].tools
+
+        return None
+    elif provider == "openai":
+        if "tools" in kwargs:
+            return kwargs["tools"]
+
+        return None
 
 
 def merge_system_prompt(kwargs: Dict[str, Any], provider: str):
@@ -377,11 +441,10 @@ def call_llm_and_track_usage(
             **(error_params or {}),
         }
 
-        tool_calls = format_tool_calls(response, provider)
-        if tool_calls:
-            event_properties["$ai_tools"] = with_privacy_mode(
-                ph_client, posthog_privacy_mode, tool_calls
-            )
+        available_tool_calls = extract_available_tool_calls(provider, kwargs)
+
+        if available_tool_calls:
+            event_properties["$ai_tools"] = available_tool_calls
 
         if (
             usage.get("cache_read_input_tokens") is not None
@@ -492,11 +555,10 @@ async def call_llm_and_track_usage_async(
             **(error_params or {}),
         }
 
-        tool_calls = format_tool_calls(response, provider)
-        if tool_calls:
-            event_properties["$ai_tools"] = with_privacy_mode(
-                ph_client, posthog_privacy_mode, tool_calls
-            )
+        available_tool_calls = extract_available_tool_calls(provider, kwargs)
+
+        if available_tool_calls:
+            event_properties["$ai_tools"] = available_tool_calls
 
         if (
             usage.get("cache_read_input_tokens") is not None

posthog/args.py

@@ -0,0 +1,71 @@
+from typing import TypedDict, Optional, Any, Dict, Union, Tuple, Type
+from types import TracebackType
+from typing_extensions import NotRequired  # For Python < 3.11 compatibility
+from datetime import datetime
+import numbers
+from uuid import UUID
+
+from posthog.types import SendFeatureFlagsOptions
+
+ID_TYPES = Union[numbers.Number, str, UUID, int]
+
+
+class OptionalCaptureArgs(TypedDict):
+    """Optional arguments for the capture method.
+
+    Args:
+        distinct_id: Unique identifier for the person associated with this event. If not set, the context
+            distinct_id is used, if available, otherwise a UUID is generated, and the event is marked
+            as personless. Setting context-level distinct_id's is recommended.
+        properties: Dictionary of properties to track with the event
+        timestamp: When the event occurred (defaults to current time)
+        uuid: Unique identifier for this specific event. If not provided, one is generated. The event
+            UUID is returned, so you can correlate it with actions in your app (like showing users an
+            error ID if you capture an exception).
+        groups: Group identifiers to associate with this event (format: {group_type: group_key})
+        send_feature_flags: Whether to include currently active feature flags in the event properties.
+            Can be a boolean (True/False) or a SendFeatureFlagsOptions object for advanced configuration.
+            Defaults to False.
+        disable_geoip: Whether to disable GeoIP lookup for this event. Defaults to False.
+    """
+
+    distinct_id: NotRequired[Optional[ID_TYPES]]
+    properties: NotRequired[Optional[Dict[str, Any]]]
+    timestamp: NotRequired[Optional[Union[datetime, str]]]
+    uuid: NotRequired[Optional[str]]
+    groups: NotRequired[Optional[Dict[str, str]]]
+    send_feature_flags: NotRequired[
+        Optional[Union[bool, SendFeatureFlagsOptions]]
+    ]  # Updated to support both boolean and options object
+    disable_geoip: NotRequired[
+        Optional[bool]
+    ]  # As above, optional so we can tell if the user is intentionally overriding a client setting or not
+
+
+class OptionalSetArgs(TypedDict):
+    """Optional arguments for the set method.
+
+    Args:
+        distinct_id: Unique identifier for the user to set properties on. If not set, the context
+            distinct_id is used, if available, otherwise this function does nothing. Setting
+            context-level distinct_id's is recommended.
+        properties: Dictionary of properties to set on the person
+        timestamp: When the properties were set (defaults to current time)
+        uuid: Unique identifier for this operation. If not provided, one is generated. This
+            UUID is returned, so you can correlate it with actions in your app.
+        disable_geoip: Whether to disable GeoIP lookup for this operation. Defaults to False.
+    """
+
+    distinct_id: NotRequired[Optional[ID_TYPES]]
+    properties: NotRequired[Optional[Dict[str, Any]]]
+    timestamp: NotRequired[Optional[Union[datetime, str]]]
+    uuid: NotRequired[Optional[str]]
+    disable_geoip: NotRequired[Optional[bool]]
+
+
+ExcInfo = Union[
+    Tuple[Type[BaseException], BaseException, Optional[TracebackType]],
+    Tuple[None, None, None],
+]
+
+ExceptionArg = Union[BaseException, ExcInfo]

posthog/contexts.py

@@ -1,6 +1,10 @@
 import contextvars
 from contextlib import contextmanager
-from typing import Optional, Any, Callable, Dict, TypeVar, cast
+from typing import Optional, Any, Callable, Dict, TypeVar, cast, TYPE_CHECKING
+
+if TYPE_CHECKING:
+    # To avoid circular imports
+    from posthog.client import Client
 
 
 class ContextScope:
@@ -9,7 +13,9 @@ class ContextScope:
         parent=None,
         fresh: bool = False,
         capture_exceptions: bool = True,
+        client: Optional["Client"] = None,
     ):
+        self.client: Optional[Client] = client
         self.parent = parent
         self.fresh = fresh
         self.capture_exceptions = capture_exceptions
@@ -64,7 +70,11 @@ def _get_current_context() -> Optional[ContextScope]:
 
 
 @contextmanager
-def new_context(fresh=False, capture_exceptions=True):
+def new_context(
+    fresh: bool = False,
+    capture_exceptions: bool = True,
+    client: Optional["Client"] = None,
+):
     """
     Create a new context scope that will be active for the duration of the with block.
     Any tags set within this scope will be isolated to this context. Any exceptions raised
@@ -77,34 +87,49 @@ def new_context(fresh=False, capture_exceptions=True):
         capture_exceptions: Whether to capture exceptions raised within the context (default: True).
                If True, captures exceptions and tags them with the context tags before propagating them.
                If False, exceptions will propagate without being tagged or captured.
+        client: Optional client instance to use for capturing exceptions (default: None).
+                If provided, the client will be used to capture exceptions within the context.
+                If not provided, the default (global) client will be used. Note that the passed
+                client is only used to capture exceptions within the context - other events captured
+                within the context via `Client.capture` or `posthog.capture` will still carry the context
+                state (tags, identity, session id), but will be captured by the client directly used (or
+                the global one, in the case of `posthog.capture`)
 
     Examples:
+        ```python
         # Inherit parent context tags
         with posthog.new_context():
             posthog.tag("request_id", "123")
             # Both this event and the exception will be tagged with the context tags
             posthog.capture("event_name", {"property": "value"})
             raise ValueError("Something went wrong")
-
+        ```
+        ```python
         # Start with fresh context (no inherited tags)
         with posthog.new_context(fresh=True):
             posthog.tag("request_id", "123")
             # Both this event and the exception will be tagged with the context tags
             posthog.capture("event_name", {"property": "value"})
             raise ValueError("Something went wrong")
+        ```
 
+    Category:
+        Contexts
     """
     from posthog import capture_exception
 
     current_context = _get_current_context()
-    new_context = ContextScope(current_context, fresh, capture_exceptions)
+    new_context = ContextScope(current_context, fresh, capture_exceptions, client)
     _context_stack.set(new_context)
 
     try:
         yield
     except Exception as e:
         if new_context.capture_exceptions:
-            capture_exception(e)
+            if new_context.client:
+                new_context.client.capture_exception(e)
+            else:
+                capture_exception(e)
         raise
     finally:
         _context_stack.set(new_context.get_parent())
@@ -112,22 +137,26 @@ def new_context(fresh=False, capture_exceptions=True):
 
 def tag(key: str, value: Any) -> None:
     """
-    Add a tag to the current context.
+    Add a tag to the current context. All tags are added as properties to any event, including exceptions, captured
+    within the context.
 
     Args:
         key: The tag key
         value: The tag value
 
     Example:
+        ```python
         posthog.tag("user_id", "123")
+        ```
+
+    Category:
+        Contexts
     """
     current_context = _get_current_context()
     if current_context:
         current_context.add_tag(key, value)
 
 
-# NOTE: we should probably also remove this - there's no reason for the user to ever
-# need to manually interact with the current tag set
 def get_tags() -> Dict[str, Any]:
     """
     Get all tags from the current context. Note, modifying
@@ -135,6 +164,9 @@ def get_tags() -> Dict[str, Any]:
 
     Returns:
         Dict of all tags in the current context
+
+    Category:
+        Contexts
     """
     current_context = _get_current_context()
     if current_context:
@@ -142,25 +174,20 @@ def get_tags() -> Dict[str, Any]:
     return {}
 
 
-# NOTE: We should probably remove this function - the way to clear scope context
-# is by entering a new, fresh context, rather than by clearing the tags or other
-# scope data directly.
-def clear_tags() -> None:
-    """Clear all tags in the current context. Does not clear parent tags"""
-    current_context = _get_current_context()
-    if current_context:
-        current_context.tags.clear()
-
-
 def identify_context(distinct_id: str) -> None:
     """
     Identify the current context with a distinct ID, associating all events captured in this or
     child contexts with the given distinct ID (unless identify_context is called again). This is overridden by
     distinct id's passed directly to posthog.capture and related methods (identify, set etc). Entering a
-    fresh context will clear the context-level distinct ID.
+    fresh context will clear the context-level distinct ID. The distinct-id passed should be uniquely associated
+    with one of your users. Events captured outside of a context, or in a context with no associated distinct
+    ID, will be assigned a random UUID, and captured as "personless".
 
     Args:
         distinct_id: The distinct ID to associate with the current context and its children.
+
+    Category:
+        Contexts
     """
     current_context = _get_current_context()
     if current_context:
@@ -175,6 +202,9 @@ def set_context_session(session_id: str) -> None:
 
     Args:
         session_id: The session ID to associate with the current context and its children. See https://posthog.com/docs/data/sessions
+
+    Category:
+        Contexts
     """
     current_context = _get_current_context()
     if current_context:
@@ -187,6 +217,9 @@ def get_context_session_id() -> Optional[str]:
 
     Returns:
         The session ID if set, None otherwise
+
+    Category:
+        Contexts
     """
     current_context = _get_current_context()
     if current_context:
@@ -200,6 +233,9 @@ def get_context_distinct_id() -> Optional[str]:
 
     Returns:
         The distinct ID if set, None otherwise
+
+    Category:
+        Contexts
     """
     current_context = _get_current_context()
     if current_context:
@@ -210,7 +246,7 @@ def get_context_distinct_id() -> Optional[str]:
 F = TypeVar("F", bound=Callable[..., Any])
 
 
-def scoped(fresh=False, capture_exceptions=True):
+def scoped(fresh: bool = False, capture_exceptions: bool = True):
     """
     Decorator that creates a new context for the function. Simply wraps
     the function in a with posthog.new_context(): block.
@@ -230,6 +266,9 @@ def scoped(fresh=False, capture_exceptions=True):
             # If this raises an exception, it will be captured with tags
             # and then re-raised
             some_risky_function()
+
+    Category:
+        Contexts
     """
 
     def decorator(func: F) -> F:

posthog/exception_capture.py

@@ -6,47 +6,25 @@
 import logging
 import sys
 import threading
-from enum import Enum
-from typing import TYPE_CHECKING, List, Optional
+from typing import TYPE_CHECKING
 
 if TYPE_CHECKING:
     from posthog.client import Client
 
 
-class Integrations(str, Enum):
-    Django = "django"
-
-
 class ExceptionCapture:
     # TODO: Add client side rate limiting to prevent spamming the server with exceptions
 
     log = logging.getLogger("posthog")
 
-    def __init__(
-        self, client: "Client", integrations: Optional[List[Integrations]] = None
-    ):
+    def __init__(self, client: "Client"):
         self.client = client
         self.original_excepthook = sys.excepthook
         sys.excepthook = self.exception_handler
         threading.excepthook = self.thread_exception_handler
-        self.enabled_integrations = []
-
-        for integration in integrations or []:
-            # TODO: Maybe find a better way of enabling integrations
-            # This is very annoying currently if we had to add any configuration per integration
-            if integration == Integrations.Django:
-                try:
-                    from posthog.exception_integrations.django import DjangoIntegration
-
-                    enabled_integration = DjangoIntegration(self.exception_receiver)
-                    self.enabled_integrations.append(enabled_integration)
-                except Exception as e:
-                    self.log.exception(f"Failed to enable Django integration: {e}")
 
     def close(self):
         sys.excepthook = self.original_excepthook
-        for integration in self.enabled_integrations:
-            integration.uninstall()
 
     def exception_handler(self, exc_type, exc_value, exc_traceback):
         # don't affect default behaviour.
@@ -66,6 +44,6 @@ class ExceptionCapture:
     def capture_exception(self, exception, metadata=None):
         try:
             distinct_id = metadata.get("distinct_id") if metadata else None
-            self.client.capture_exception(exception, distinct_id)
+            self.client.capture_exception(exception, distinct_id=distinct_id)
         except Exception as e:
             self.log.exception(f"Failed to capture exception: {e}")

posthog/exception_integrations/__init__.py

@@ -1,5 +0,0 @@
-class IntegrationEnablingError(Exception):
-    """
-    The integration could not be enabled due to a user error like
-    `django` not being installed for the `DjangoIntegration`.
-    """

posthog/exception_integrations/django.py

@@ -1,117 +0,0 @@
-# Portions of this file are derived from getsentry/sentry-javascript by Software, Inc. dba Sentry
-# Licensed under the MIT License
-
-# 💖open source (under MIT License)
-
-import re
-import sys
-from typing import TYPE_CHECKING
-
-from posthog.exception_integrations import IntegrationEnablingError
-
-try:
-    from django import VERSION as DJANGO_VERSION
-    from django.core import signals
-
-except ImportError:
-    raise IntegrationEnablingError("Django not installed")
-
-
-if TYPE_CHECKING:
-    from typing import Any, Dict  # noqa: F401
-
-    from django.core.handlers.wsgi import WSGIRequest  # noqa: F401
-
-
-class DjangoIntegration:
-    # TODO: Abstract integrations one we have more and can see patterns
-    """
-    Autocapture errors from a Django application.
-    """
-
-    identifier = "django"
-
-    def __init__(self, capture_exception_fn=None):
-        if DJANGO_VERSION < (4, 2):
-            raise IntegrationEnablingError("Django 4.2 or newer is required.")
-
-        # TODO: Right now this seems too complicated / overkill for us, but seems like we can automatically plug in middlewares
-        # which is great for users (they don't need to do this) and everything should just work.
-        # We should consider this in the future, but for now we can just use the middleware and signals handlers.
-        # See: https://github.com/getsentry/sentry-python/blob/269d96d6e9821122fbff280e6a26956e5ed03c0b/sentry_sdk/integrations/django/__init__.py
-
-        self.capture_exception_fn = capture_exception_fn
-
-        def _got_request_exception(request=None, **kwargs):
-            # type: (WSGIRequest, **Any) -> None
-
-            extra_props = {}
-            if request is not None:
-                # get headers metadata
-                extra_props = DjangoRequestExtractor(request).extract_person_data()
-
-            self.capture_exception_fn(sys.exc_info(), extra_props)
-
-        signals.got_request_exception.connect(_got_request_exception)
-
-    def uninstall(self):
-        pass
-
-
-class DjangoRequestExtractor:
-    def __init__(self, request):
-        # type: (Any) -> None
-        self.request = request
-
-    def extract_person_data(self):
-        headers = self.headers()
-
-        # Extract traceparent and tracestate headers
-        traceparent = headers.get("Traceparent")
-        tracestate = headers.get("Tracestate")
-
-        # Extract the distinct_id from tracestate
-        distinct_id = None
-        if tracestate:
-            # TODO: Align on the format of the distinct_id in tracestate
-            # We can't have comma or equals in header values here, so maybe we should base64 encode it?
-            match = re.search(r"posthog-distinct-id=([^,]+)", tracestate)
-            if match:
-                distinct_id = match.group(1)
-
-        return {
-            **self.user(),
-            "distinct_id": distinct_id,
-            "ip": headers.get("X-Forwarded-For"),
-            "user_agent": headers.get("User-Agent"),
-            "traceparent": traceparent,
-            "$request_path": self.request.path,
-        }
-
-    def user(self):
-        user_data: dict[str, str] = {}
-
-        user = getattr(self.request, "user", None)
-
-        if user is None or not user.is_authenticated:
-            return user_data
-
-        try:
-            user_id = str(user.pk)
-            if user_id:
-                user_data.setdefault("$user_id", user_id)
-        except Exception:
-            pass
-
-        try:
-            email = str(user.email)
-            if email:
-                user_data.setdefault("email", email)
-        except Exception:
-            pass
-
-        return user_data
-
-    def headers(self):
-        # type: () -> Dict[str, str]
-        return dict(self.request.headers)

posthog/exception_utils.py

@@ -11,7 +11,24 @@ import re
 import sys
 import types
 from datetime import datetime
-from typing import TYPE_CHECKING
+from types import FrameType, TracebackType  # noqa: F401
+from typing import (  # noqa: F401
+    Any,
+    Dict,
+    Iterator,
+    List,
+    Literal,
+    Optional,
+    Set,
+    Tuple,
+    TypedDict,
+    TypeVar,
+    Union,
+    cast,
+    TYPE_CHECKING,
+)
+
+from posthog.args import ExcInfo, ExceptionArg  # noqa: F401
 
 try:
     # Python 3.11
@@ -23,85 +40,61 @@ except ImportError:
 
 DEFAULT_MAX_VALUE_LENGTH = 1024
 
+LogLevelStr = Literal["fatal", "critical", "error", "warning", "info", "debug"]
 
-if TYPE_CHECKING:
-    from types import FrameType, TracebackType
-    from typing import (  # noqa: F401
-        Any,
-        Callable,
-        Dict,
-        Iterator,
-        List,
-        Literal,
-        Optional,
-        Set,
-        Tuple,
-        Type,
-        TypedDict,
-        TypeVar,
-        Union,
-        cast,
-    )
-
-    ExcInfo = Union[
-        Tuple[Type[BaseException], BaseException, Optional[TracebackType]],
-        Tuple[None, None, None],
-    ]
-    LogLevelStr = Literal["fatal", "critical", "error", "warning", "info", "debug"]
-
-    Event = TypedDict(
-        "Event",
-        {
-            "breadcrumbs": Dict[
-                Literal["values"], List[Dict[str, Any]]
-            ],  # TODO: We can expand on this type
-            "check_in_id": str,
-            "contexts": Dict[str, Dict[str, object]],
-            "dist": str,
-            "duration": Optional[float],
-            "environment": str,
-            "errors": List[Dict[str, Any]],  # TODO: We can expand on this type
-            "event_id": str,
-            "exception": Dict[
-                Literal["values"], List[Dict[str, Any]]
-            ],  # TODO: We can expand on this type
-            # "extra": MutableMapping[str, object],
-            # "fingerprint": List[str],
-            "level": LogLevelStr,
-            # "logentry": Mapping[str, object],
-            "logger": str,
-            # "measurements": Dict[str, MeasurementValue],
-            "message": str,
-            "modules": Dict[str, str],
-            # "monitor_config": Mapping[str, object],
-            "monitor_slug": Optional[str],
-            "platform": Literal["python"],
-            "profile": object,
-            "release": str,
-            "request": Dict[str, object],
-            # "sdk": Mapping[str, object],
-            "server_name": str,
-            "spans": List[Dict[str, object]],
-            "stacktrace": Dict[
-                str, object
-            ],  # We access this key in the code, but I am unsure whether we ever set it
-            "start_timestamp": datetime,
-            "status": Optional[str],
-            # "tags": MutableMapping[
-            #     str, str
-            # ],  # Tags must be less than 200 characters each
-            "threads": Dict[
-                Literal["values"], List[Dict[str, Any]]
-            ],  # TODO: We can expand on this type
-            "timestamp": Optional[datetime],  # Must be set before sending the event
-            "transaction": str,
-            # "transaction_info": Mapping[str, Any],  # TODO: We can expand on this type
-            "type": Literal["check_in", "transaction"],
-            "user": Dict[str, object],
-            "_metrics_summary": Dict[str, object],
-        },
-        total=False,
-    )
+Event = TypedDict(
+    "Event",
+    {
+        "breadcrumbs": Dict[
+            Literal["values"], List[Dict[str, Any]]
+        ],  # TODO: We can expand on this type
+        "check_in_id": str,
+        "contexts": Dict[str, Dict[str, object]],
+        "dist": str,
+        "duration": Optional[float],
+        "environment": str,
+        "errors": List[Dict[str, Any]],  # TODO: We can expand on this type
+        "event_id": str,
+        "exception": Dict[
+            Literal["values"], List[Dict[str, Any]]
+        ],  # TODO: We can expand on this type
+        # "extra": MutableMapping[str, object],
+        # "fingerprint": List[str],
+        "level": LogLevelStr,
+        # "logentry": Mapping[str, object],
+        "logger": str,
+        # "measurements": Dict[str, MeasurementValue],
+        "message": str,
+        "modules": Dict[str, str],
+        # "monitor_config": Mapping[str, object],
+        "monitor_slug": Optional[str],
+        "platform": Literal["python"],
+        "profile": object,
+        "release": str,
+        "request": Dict[str, object],
+        # "sdk": Mapping[str, object],
+        "server_name": str,
+        "spans": List[Dict[str, object]],
+        "stacktrace": Dict[
+            str, object
+        ],  # We access this key in the code, but I am unsure whether we ever set it
+        "start_timestamp": datetime,
+        "status": Optional[str],
+        # "tags": MutableMapping[
+        #     str, str
+        # ],  # Tags must be less than 200 characters each
+        "threads": Dict[
+            Literal["values"], List[Dict[str, Any]]
+        ],  # TODO: We can expand on this type
+        "timestamp": Optional[datetime],  # Must be set before sending the event
+        "transaction": str,
+        # "transaction_info": Mapping[str, Any],  # TODO: We can expand on this type
+        "type": Literal["check_in", "transaction"],
+        "user": Dict[str, object],
+        "_metrics_summary": Dict[str, object],
+    },
+    total=False,
+)
 
 
 epoch = datetime(1970, 1, 1)
@@ -362,12 +355,9 @@ def filename_for_module(module, abs_path):
 def serialize_frame(
     frame,
     tb_lineno=None,
-    include_local_variables=True,
-    include_source_context=True,
     max_value_length=None,
-    custom_repr=None,
 ):
-    # type: (FrameType, Optional[int], bool, bool, Optional[int], Optional[Callable[..., Optional[str]]]) -> Dict[str, Any]
+    # type: (FrameType, Optional[int], Optional[int]) -> Dict[str, Any]
     f_code = getattr(frame, "f_code", None)
     if not f_code:
         abs_path = None
@@ -392,45 +382,13 @@ def serialize_frame(
         "lineno": tb_lineno,
     }  # type: Dict[str, Any]
 
-    if include_source_context:
-        rv["pre_context"], rv["context_line"], rv["post_context"] = get_source_context(
-            frame, tb_lineno, max_value_length
-        )
-
-    if include_local_variables:
-        # TODO - we don't support local variables, yet
-        pass
+    rv["pre_context"], rv["context_line"], rv["post_context"] = get_source_context(
+        frame, tb_lineno, max_value_length
+    )
 
     return rv
 
 
-def current_stacktrace(
-    include_local_variables=True,  # type: bool
-    include_source_context=True,  # type: bool
-    max_value_length=None,  # type: Optional[int]
-):
-    # type: (...) -> Dict[str, Any]
-    __tracebackhide__ = True
-    frames = []
-
-    f = sys._getframe()  # type: Optional[FrameType]
-    while f is not None:
-        if not should_hide_frame(f):
-            frames.append(
-                serialize_frame(
-                    f,
-                    include_local_variables=include_local_variables,
-                    include_source_context=include_source_context,
-                    max_value_length=max_value_length,
-                )
-            )
-        f = f.f_back
-
-    frames.reverse()
-
-    return {"frames": frames, "type": "raw"}
-
-
 def get_errno(exc_value):
     # type: (BaseException) -> Optional[Any]
     return getattr(exc_value, "errno", None)
@@ -451,7 +409,6 @@ def single_exception_from_error_tuple(
     exc_type,  # type: Optional[type]
     exc_value,  # type: Optional[BaseException]
     tb,  # type: Optional[TracebackType]
-    client_options=None,  # type: Optional[Dict[str, Any]]
     mechanism=None,  # type: Optional[Dict[str, Any]]
     exception_id=None,  # type: Optional[int]
     parent_id=None,  # type: Optional[int]
@@ -499,25 +456,13 @@ def single_exception_from_error_tuple(
     exception_value["type"] = get_type_name(exc_type)
     exception_value["value"] = get_error_message(exc_value)
 
-    if client_options is None:
-        include_local_variables = True
-        include_source_context = True
-        max_value_length = DEFAULT_MAX_VALUE_LENGTH  # fallback
-        custom_repr = None
-    else:
-        include_local_variables = client_options["include_local_variables"]
-        include_source_context = client_options["include_source_context"]
-        max_value_length = client_options["max_value_length"]
-        custom_repr = client_options.get("custom_repr")
+    max_value_length = DEFAULT_MAX_VALUE_LENGTH  # fallback
 
     frames = [
         serialize_frame(
             tb.tb_frame,
             tb_lineno=tb.tb_lineno,
-            include_local_variables=include_local_variables,
-            include_source_context=include_source_context,
             max_value_length=max_value_length,
-            custom_repr=custom_repr,
         )
         for tb in iter_stacks(tb)
     ]
@@ -573,7 +518,6 @@ def exceptions_from_error(
     exc_type,  # type: Optional[type]
     exc_value,  # type: Optional[BaseException]
     tb,  # type: Optional[TracebackType]
-    client_options=None,  # type: Optional[Dict[str, Any]]
     mechanism=None,  # type: Optional[Dict[str, Any]]
     exception_id=0,  # type: int
     parent_id=0,  # type: int
@@ -589,7 +533,6 @@ def exceptions_from_error(
         exc_type=exc_type,
         exc_value=exc_value,
         tb=tb,
-        client_options=client_options,
         mechanism=mechanism,
         exception_id=exception_id,
         parent_id=parent_id,
@@ -617,7 +560,6 @@ def exceptions_from_error(
                 exc_type=type(cause),
                 exc_value=cause,
                 tb=getattr(cause, "__traceback__", None),
-                client_options=client_options,
                 mechanism=mechanism,
                 exception_id=exception_id,
                 source="__cause__",
@@ -638,7 +580,6 @@ def exceptions_from_error(
                 exc_type=type(context),
                 exc_value=context,
                 tb=getattr(context, "__traceback__", None),
-                client_options=client_options,
                 mechanism=mechanism,
                 exception_id=exception_id,
                 source="__context__",
@@ -653,7 +594,6 @@ def exceptions_from_error(
                 exc_type=type(e),
                 exc_value=e,
                 tb=getattr(e, "__traceback__", None),
-                client_options=client_options,
                 mechanism=mechanism,
                 exception_id=exception_id,
                 parent_id=parent_id,
@@ -666,7 +606,6 @@ def exceptions_from_error(
 
 def exceptions_from_error_tuple(
     exc_info,  # type: ExcInfo
-    client_options=None,  # type: Optional[Dict[str, Any]]
     mechanism=None,  # type: Optional[Dict[str, Any]]
 ):
     # type: (...) -> List[Dict[str, Any]]
@@ -681,7 +620,6 @@ def exceptions_from_error_tuple(
             exc_type=exc_type,
             exc_value=exc_value,
             tb=tb,
-            client_options=client_options,
             mechanism=mechanism,
             exception_id=0,
             parent_id=0,
@@ -691,9 +629,7 @@ def exceptions_from_error_tuple(
         exceptions = []
         for exc_type, exc_value, tb in walk_exception_chain(exc_info):
             exceptions.append(
-                single_exception_from_error_tuple(
-                    exc_type, exc_value, tb, client_options, mechanism
-                )
+                single_exception_from_error_tuple(exc_type, exc_value, tb, mechanism)
             )
 
     exceptions.reverse()
@@ -783,7 +719,7 @@ def set_in_app_in_frames(frames, in_app_exclude, in_app_include, project_root=No
 
 
 def exception_is_already_captured(error):
-    # type: (Union[BaseException, ExcInfo]) -> bool
+    # type: (ExceptionArg) -> bool
     if isinstance(error, BaseException):
         return hasattr(error, "__posthog_exception_captured")
     # Autocaptured exceptions are passed as a tuple from our system hooks,
@@ -796,19 +732,21 @@ def exception_is_already_captured(error):
         return False  # type: ignore[unreachable]
 
 
-def mark_exception_as_captured(error):
-    # type: (Union[BaseException, ExcInfo]) -> None
+def mark_exception_as_captured(error, uuid):
+    # type: (ExceptionArg, str) -> None
     if isinstance(error, BaseException):
         setattr(error, "__posthog_exception_captured", True)
+        setattr(error, "__posthog_exception_uuid", uuid)
     # Autocaptured exceptions are passed as a tuple from our system hooks,
     # the second item is the exception value (the first is the exception type)
     elif isinstance(error, tuple) and len(error) > 1:
         if error[1] is not None:
             setattr(error[1], "__posthog_exception_captured", True)
+            setattr(error[1], "__posthog_exception_uuid", uuid)
 
 
 def exc_info_from_error(error):
-    # type: (Union[BaseException, ExcInfo]) -> ExcInfo
+    # type: (ExceptionArg) -> ExcInfo
     if isinstance(error, tuple) and len(error) == 3:
         exc_type, exc_value, tb = error
     elif isinstance(error, BaseException):
@@ -865,27 +803,6 @@ def construct_artificial_traceback(e):
     setattr(e, "__traceback__", tb)
 
 
-def event_from_exception(
-    exc_info,  # type: Union[BaseException, ExcInfo]
-    client_options=None,  # type: Optional[Dict[str, Any]]
-    mechanism=None,  # type: Optional[Dict[str, Any]]
-):
-    # type: (...) -> Tuple[Event, Dict[str, Any]]
-    exc_info = exc_info_from_error(exc_info)
-    hint = event_hint_with_exc_info(exc_info)
-    return (
-        {
-            "level": "error",
-            "exception": {
-                "values": exceptions_from_error_tuple(
-                    exc_info, client_options, mechanism
-                )
-            },
-        },
-        hint,
-    )
-
-
 def _module_in_list(name, items):
     # type: (str | None, Optional[List[str]]) -> bool
     if name is None:

posthog/feature_flags.py

@@ -109,6 +109,14 @@ def is_condition_match(
             property_type = prop.get("type")
             if property_type == "cohort":
                 matches = match_cohort(prop, properties, cohort_properties)
+            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"),
+                )
+                continue
             else:
                 matches = match_property(prop, properties)
             if not matches:
@@ -317,6 +325,13 @@ def match_property_group(property_group, property_values, cohort_properties) ->
             try:
                 if prop.get("type") == "cohort":
                     matches = match_cohort(prop, property_values, cohort_properties)
+                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"),
+                    )
+                    continue
                 else:
                     matches = match_property(prop, property_values)
 

posthog/integrations/django.py

@@ -1,5 +1,6 @@
 from typing import TYPE_CHECKING, cast
-from posthog import scopes
+from posthog import contexts, capture_exception
+from posthog.client import Client
 
 if TYPE_CHECKING:
     from django.http import HttpRequest, HttpResponse  # noqa: F401
@@ -16,7 +17,8 @@ class PosthogContextMiddleware:
     - Request Method as $request_method
 
     The context will also auto-capture exceptions and send them to PostHog, unless you disable it by setting
-    `POSTHOG_MW_CAPTURE_EXCEPTIONS` to `False` in your Django settings.
+    `POSTHOG_MW_CAPTURE_EXCEPTIONS` to `False` in your Django settings. The exceptions are captured using the
+    global client, unless the setting `POSTHOG_MW_CLIENT` is set to a custom client instance
 
     The middleware behaviour is customisable through 3 additional functions:
     - `POSTHOG_MW_EXTRA_TAGS`, which is a Callable[[HttpRequest], Dict[str, Any]] expected to return a dictionary of additional tags to be added to the context.
@@ -74,19 +76,32 @@ class PosthogContextMiddleware:
         else:
             self.capture_exceptions = True
 
+        if hasattr(settings, "POSTHOG_MW_CLIENT") and isinstance(
+            settings.POSTHOG_MW_CLIENT, Client
+        ):
+            self.client = cast("Optional[Client]", settings.POSTHOG_MW_CLIENT)
+        else:
+            self.client = None
+
     def extract_tags(self, request):
         # type: (HttpRequest) -> Dict[str, Any]
         tags = {}
 
+        (user_id, user_email) = self.extract_request_user(request)
+
         # Extract session ID from X-POSTHOG-SESSION-ID header
         session_id = request.headers.get("X-POSTHOG-SESSION-ID")
         if session_id:
-            scopes.set_context_session(session_id)
+            contexts.set_context_session(session_id)
 
-        # Extract distinct ID from X-POSTHOG-DISTINCT-ID header
-        distinct_id = request.headers.get("X-POSTHOG-DISTINCT-ID")
+        # Extract distinct ID from X-POSTHOG-DISTINCT-ID header or request user id
+        distinct_id = request.headers.get("X-POSTHOG-DISTINCT-ID") or user_id
         if distinct_id:
-            scopes.identify_context(distinct_id)
+            contexts.identify_context(distinct_id)
+
+        # Extract user email
+        if user_email:
+            tags["email"] = user_email
 
         # Extract current URL
         absolute_url = request.build_absolute_uri()
@@ -97,6 +112,20 @@ class PosthogContextMiddleware:
         if request.method:
             tags["$request_method"] = request.method
 
+        # Extract request path
+        if request.path:
+            tags["$request_path"] = request.path
+
+        # Extract IP address
+        ip_address = request.headers.get("X-Forwarded-For")
+        if ip_address:
+            tags["$ip_address"] = ip_address
+
+        # Extract user agent
+        user_agent = request.headers.get("User-Agent")
+        if user_agent:
+            tags["$user_agent"] = user_agent
+
         # Apply extra tags if configured
         if self.extra_tags:
             extra = self.extra_tags(request)
@@ -109,13 +138,44 @@ class PosthogContextMiddleware:
 
         return tags
 
+    def extract_request_user(self, request):
+        user_id = None
+        email = None
+
+        user = getattr(request, "user", None)
+
+        if user and getattr(user, "is_authenticated", False):
+            try:
+                user_id = str(user.pk)
+            except Exception:
+                pass
+
+            try:
+                email = str(user.email)
+            except Exception:
+                pass
+
+        return user_id, email
+
     def __call__(self, request):
         # type: (HttpRequest) -> HttpResponse
         if self.request_filter and not self.request_filter(request):
             return self.get_response(request)
 
-        with scopes.new_context(self.capture_exceptions):
+        with contexts.new_context(self.capture_exceptions, client=self.client):
             for k, v in self.extract_tags(request).items():
-                scopes.tag(k, v)
+                contexts.tag(k, v)
 
             return self.get_response(request)
+
+    def process_exception(self, request, exception):
+        if self.request_filter and not self.request_filter(request):
+            return
+
+        if not self.capture_exceptions:
+            return
+
+        if self.client:
+            self.client.capture_exception(exception)
+        else:
+            capture_exception(exception)

posthog/test/ai/anthropic/test_anthropic.py

@@ -88,6 +88,56 @@ def mock_anthropic_response_with_cached_tokens():
     )
 
 
+@pytest.fixture
+def mock_anthropic_response_with_tool_calls():
+    return Message(
+        id="msg_456",
+        type="message",
+        role="assistant",
+        content=[
+            {"type": "text", "text": "I'll help you check the weather."},
+            {"type": "text", "text": " Let me look that up."},
+            {
+                "type": "tool_use",
+                "id": "toolu_abc123",
+                "name": "get_weather",
+                "input": {"location": "San Francisco"},
+            },
+        ],
+        model="claude-3-5-sonnet-20241022",
+        usage=Usage(
+            input_tokens=25,
+            output_tokens=15,
+        ),
+        stop_reason="tool_use",
+        stop_sequence=None,
+    )
+
+
+@pytest.fixture
+def mock_anthropic_response_tool_calls_only():
+    return Message(
+        id="msg_789",
+        type="message",
+        role="assistant",
+        content=[
+            {
+                "type": "tool_use",
+                "id": "toolu_def456",
+                "name": "get_weather",
+                "input": {"location": "New York", "unit": "fahrenheit"},
+            }
+        ],
+        model="claude-3-5-sonnet-20241022",
+        usage=Usage(
+            input_tokens=30,
+            output_tokens=12,
+        ),
+        stop_reason="tool_use",
+        stop_sequence=None,
+    )
+
+
 def test_basic_completion(mock_client, mock_anthropic_response):
     with patch(
         "anthropic.resources.Messages.create", return_value=mock_anthropic_response
@@ -112,7 +162,10 @@ def test_basic_completion(mock_client, mock_anthropic_response):
         assert props["$ai_model"] == "claude-3-opus-20240229"
         assert props["$ai_input"] == [{"role": "user", "content": "Hello"}]
         assert props["$ai_output_choices"] == [
-            {"role": "assistant", "content": "Test response"}
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": "Test response"}],
+            }
         ]
         assert props["$ai_input_tokens"] == 20
         assert props["$ai_output_tokens"] == 10
@@ -286,7 +339,9 @@ def test_basic_integration(mock_client):
         {"role": "user", "content": "Foo"},
     ]
     assert props["$ai_output_choices"][0]["role"] == "assistant"
-    assert props["$ai_output_choices"][0]["content"] == "Bar"
+    assert props["$ai_output_choices"][0]["content"] == [
+        {"type": "text", "text": "Bar"}
+    ]
     assert props["$ai_input_tokens"] == 18
     assert props["$ai_output_tokens"] == 1
     assert props["$ai_http_status"] == 200
@@ -425,7 +480,10 @@ def test_cached_tokens(mock_client, mock_anthropic_response_with_cached_tokens):
         assert props["$ai_model"] == "claude-3-opus-20240229"
         assert props["$ai_input"] == [{"role": "user", "content": "Hello"}]
         assert props["$ai_output_choices"] == [
-            {"role": "assistant", "content": "Test response"}
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": "Test response"}],
+            }
         ]
         assert props["$ai_input_tokens"] == 20
         assert props["$ai_output_tokens"] == 10
@@ -434,3 +492,257 @@ def test_cached_tokens(mock_client, mock_anthropic_response_with_cached_tokens):
         assert props["$ai_http_status"] == 200
         assert props["foo"] == "bar"
         assert isinstance(props["$ai_latency"], float)
+
+
+def test_tool_definition(mock_client, mock_anthropic_response):
+    with patch(
+        "anthropic.resources.Messages.create",
+        return_value=mock_anthropic_response,
+    ):
+        client = Anthropic(api_key="test-key", posthog_client=mock_client)
+
+        tools = [
+            {
+                "name": "get_weather",
+                "description": "Get the current weather for a specific location",
+                "input_schema": {
+                    "type": "object",
+                    "properties": {
+                        "location": {
+                            "type": "string",
+                            "description": "The city or location name to get weather for",
+                        }
+                    },
+                    "required": ["location"],
+                },
+            }
+        ]
+
+        response = client.messages.create(
+            model="claude-3-5-sonnet-20241022",
+            max_tokens=200,
+            temperature=0.7,
+            tools=tools,
+            messages=[{"role": "user", "content": "hey"}],
+            posthog_distinct_id="test-id",
+            posthog_properties={"foo": "bar"},
+        )
+
+        assert response == mock_anthropic_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"] == "anthropic"
+        assert props["$ai_model"] == "claude-3-5-sonnet-20241022"
+        assert props["$ai_input"] == [{"role": "user", "content": "hey"}]
+        assert props["$ai_output_choices"] == [
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": "Test response"}],
+            }
+        ]
+        assert props["$ai_input_tokens"] == 20
+        assert props["$ai_output_tokens"] == 10
+        assert props["$ai_http_status"] == 200
+        assert props["foo"] == "bar"
+        assert isinstance(props["$ai_latency"], float)
+        # Verify that tools are captured in the $ai_tools property
+        assert props["$ai_tools"] == tools
+
+
+def test_tool_calls_in_output_choices(
+    mock_client, mock_anthropic_response_with_tool_calls
+):
+    with patch(
+        "anthropic.resources.Messages.create",
+        return_value=mock_anthropic_response_with_tool_calls,
+    ):
+        client = Anthropic(api_key="test-key", posthog_client=mock_client)
+        response = client.messages.create(
+            model="claude-3-5-sonnet-20241022",
+            max_tokens=200,
+            messages=[
+                {"role": "user", "content": "What's the weather in San Francisco?"}
+            ],
+            tools=[
+                {
+                    "name": "get_weather",
+                    "description": "Get weather",
+                    "input_schema": {
+                        "type": "object",
+                        "properties": {"location": {"type": "string"}},
+                        "required": ["location"],
+                    },
+                }
+            ],
+            posthog_distinct_id="test-id",
+        )
+
+        assert response == mock_anthropic_response_with_tool_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"] == "anthropic"
+        assert props["$ai_model"] == "claude-3-5-sonnet-20241022"
+        assert props["$ai_output_choices"] == [
+            {
+                "role": "assistant",
+                "content": [
+                    {"type": "text", "text": "I'll help you check the weather."},
+                    {"type": "text", "text": " Let me look that up."},
+                    {
+                        "type": "function",
+                        "id": "toolu_abc123",
+                        "function": {
+                            "name": "get_weather",
+                            "arguments": {"location": "San Francisco"},
+                        },
+                    },
+                ],
+            }
+        ]
+
+        # Check token usage
+        assert props["$ai_input_tokens"] == 25
+        assert props["$ai_output_tokens"] == 15
+        assert props["$ai_http_status"] == 200
+
+
+def test_tool_calls_only_no_content(
+    mock_client, mock_anthropic_response_tool_calls_only
+):
+    with patch(
+        "anthropic.resources.Messages.create",
+        return_value=mock_anthropic_response_tool_calls_only,
+    ):
+        client = Anthropic(api_key="test-key", posthog_client=mock_client)
+        response = client.messages.create(
+            model="claude-3-5-sonnet-20241022",
+            max_tokens=200,
+            messages=[{"role": "user", "content": "Get weather for New York"}],
+            tools=[
+                {
+                    "name": "get_weather",
+                    "description": "Get weather",
+                    "input_schema": {
+                        "type": "object",
+                        "properties": {
+                            "location": {"type": "string"},
+                            "unit": {"type": "string"},
+                        },
+                        "required": ["location"],
+                    },
+                }
+            ],
+            posthog_distinct_id="test-id",
+        )
+
+        assert response == mock_anthropic_response_tool_calls_only
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        assert call_args["distinct_id"] == "test-id"
+        assert call_args["event"] == "$ai_generation"
+        assert props["$ai_provider"] == "anthropic"
+        assert props["$ai_model"] == "claude-3-5-sonnet-20241022"
+        assert props["$ai_output_choices"] == [
+            {
+                "role": "assistant",
+                "content": [
+                    {
+                        "type": "function",
+                        "id": "toolu_def456",
+                        "function": {
+                            "name": "get_weather",
+                            "arguments": {"location": "New York", "unit": "fahrenheit"},
+                        },
+                    }
+                ],
+            }
+        ]
+
+        # Check token usage
+        assert props["$ai_input_tokens"] == 30
+        assert props["$ai_output_tokens"] == 12
+        assert props["$ai_http_status"] == 200
+
+
+def test_async_tool_calls_in_output_choices(
+    mock_client, mock_anthropic_response_with_tool_calls
+):
+    import asyncio
+
+    async def mock_async_create(**kwargs):
+        return mock_anthropic_response_with_tool_calls
+
+    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():
+            return await async_client.messages.create(
+                model="claude-3-5-sonnet-20241022",
+                max_tokens=200,
+                messages=[
+                    {"role": "user", "content": "What's the weather in San Francisco?"}
+                ],
+                tools=[
+                    {
+                        "name": "get_weather",
+                        "description": "Get weather",
+                        "input_schema": {
+                            "type": "object",
+                            "properties": {"location": {"type": "string"}},
+                            "required": ["location"],
+                        },
+                    }
+                ],
+                posthog_distinct_id="test-id",
+            )
+
+        response = asyncio.run(run_test())
+
+        assert response == mock_anthropic_response_with_tool_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"] == "anthropic"
+        assert props["$ai_model"] == "claude-3-5-sonnet-20241022"
+        assert props["$ai_output_choices"] == [
+            {
+                "role": "assistant",
+                "content": [
+                    {"type": "text", "text": "I'll help you check the weather."},
+                    {"type": "text", "text": " Let me look that up."},
+                    {
+                        "type": "function",
+                        "id": "toolu_abc123",
+                        "function": {
+                            "name": "get_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

posthog/test/ai/gemini/test_gemini.py

@@ -56,6 +56,87 @@ def mock_google_genai_client():
         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_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."
+    # Make hasattr(part, "text") return True
+    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 - need to ensure hasattr() works correctly
+    mock_function_part = MagicMock()
+    mock_function_part.function_call = mock_function_call
+    # Make hasattr(part, "function_call") return True
+    type(mock_function_part).function_call = mock_function_part.function_call
+    # Ensure hasattr(part, "text") returns False for the function part
+    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
+
+
+@pytest.fixture
+def mock_gemini_response_function_calls_only():
+    mock_response = MagicMock()
+
+    # Mock usage metadata
+    mock_usage = MagicMock()
+    mock_usage.prompt_token_count = 30
+    mock_usage.candidates_token_count = 12
+    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": "New York", "unit": "fahrenheit"}
+
+    # Mock function call part (no text part) - need to ensure hasattr() works correctly
+    mock_function_part = MagicMock()
+    mock_function_part.function_call = mock_function_call
+    # Make hasattr(part, "function_call") return True
+    type(mock_function_part).function_call = mock_function_part.function_call
+    # Ensure hasattr(part, "text") returns False for the function part
+    del mock_function_part.text
+
+    # Mock content with only function call part
+    mock_content = MagicMock()
+    mock_content.parts = [mock_function_part]
+
+    # Mock candidate
+    mock_candidate = MagicMock()
+    mock_candidate.content = mock_content
+    mock_response.candidates = [mock_candidate]
+
+    return mock_response
+
+
 def test_new_client_basic_generation(
     mock_client, mock_google_genai_client, mock_gemini_response
 ):
@@ -318,3 +399,233 @@ def test_new_client_override_defaults(
     assert props["team"] == "ai"  # from defaults
     assert props["feature"] == "chat"  # from call
     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 $ai_tools property"""
+    mock_google_genai_client.models.generate_content.return_value = mock_gemini_response
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+
+    # Create mock tools configuration
+    mock_tool = MagicMock()
+    mock_tool.function_declarations = [
+        MagicMock(
+            name="get_current_weather",
+            description="Gets the current weather for a given location.",
+            parameters=MagicMock(
+                type="OBJECT",
+                properties={
+                    "location": MagicMock(
+                        type="STRING",
+                        description="The city and state, e.g. San Francisco, CA",
+                    )
+                },
+                required=["location"],
+            ),
+        )
+    ]
+
+    mock_config = MagicMock()
+    mock_config.tools = [mock_tool]
+
+    response = client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents=["hey"],
+        config=mock_config,
+        posthog_distinct_id="test-id",
+        posthog_properties={"foo": "bar"},
+    )
+
+    assert response == mock_gemini_response
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "gemini"
+    assert props["$ai_model"] == "gemini-2.5-flash"
+    assert props["$ai_input"] == [{"role": "user", "content": "hey"}]
+    assert props["$ai_output_choices"] == [
+        {
+            "role": "assistant",
+            "content": [{"type": "text", "text": "Test response from Gemini"}],
+        }
+    ]
+    assert props["$ai_input_tokens"] == 20
+    assert props["$ai_output_tokens"] == 10
+    assert props["$ai_http_status"] == 200
+    assert props["foo"] == "bar"
+    assert isinstance(props["$ai_latency"], float)
+    # Verify that tools are captured in the $ai_tools property
+    assert props["$ai_tools"] == [mock_tool]
+
+
+def test_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 $ai_output_choices"""
+    mock_google_genai_client.models.generate_content.return_value = (
+        mock_gemini_response_with_function_calls
+    )
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+
+    response = client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents=["What's the weather in San Francisco?"],
+        posthog_distinct_id="test-id",
+    )
+
+    assert response == mock_gemini_response_with_function_calls
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$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
+
+
+def test_function_calls_only_no_content(
+    mock_client, mock_google_genai_client, mock_gemini_response_function_calls_only
+):
+    """Test function calls without text content in $ai_output_choices"""
+    mock_google_genai_client.models.generate_content.return_value = (
+        mock_gemini_response_function_calls_only
+    )
+
+    client = Client(api_key="test-key", posthog_client=mock_client)
+
+    response = client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents=["Get weather for New York"],
+        posthog_distinct_id="test-id",
+    )
+
+    assert response == mock_gemini_response_function_calls_only
+    assert mock_client.capture.call_count == 1
+
+    call_args = mock_client.capture.call_args[1]
+    props = call_args["properties"]
+
+    assert call_args["distinct_id"] == "test-id"
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "gemini"
+    assert props["$ai_model"] == "gemini-2.5-flash"
+    assert props["$ai_output_choices"] == [
+        {
+            "role": "assistant",
+            "content": [
+                {
+                    "type": "function",
+                    "function": {
+                        "name": "get_current_weather",
+                        "arguments": {"location": "New York", "unit": "fahrenheit"},
+                    },
+                }
+            ],
+        }
+    ]
+
+    # Check token usage
+    assert props["$ai_input_tokens"] == 30
+    assert props["$ai_output_tokens"] == 12
+    assert props["$ai_http_status"] == 200

posthog/test/ai/langchain/test_callbacks.py

@@ -5,7 +5,7 @@ import os
 import time
 import uuid
 from typing import List, Literal, Optional, TypedDict, Union
-from unittest.mock import patch
+from unittest.mock import patch, MagicMock
 
 import pytest
 
@@ -1727,3 +1727,152 @@ def test_openai_reasoning_tokens(mock_client):
     assert call["properties"]["$ai_reasoning_tokens"] is not None
     assert call["properties"]["$ai_input_tokens"] is not None
     assert call["properties"]["$ai_output_tokens"] is not None
+
+
+def test_callback_handler_without_client():
+    """Test that CallbackHandler works properly when no PostHog client is passed."""
+    with patch("posthog.ai.langchain.callbacks.setup") as mock_setup:
+        mock_client = mock_setup.return_value
+
+        callbacks = CallbackHandler()
+
+        # Verify that setup() was called
+        mock_setup.assert_called_once()
+
+        # Verify that the client was set to the result of setup()
+        assert callbacks._ph_client == mock_client
+
+        # Test that the callback handler works with a simple chain
+        prompt = ChatPromptTemplate.from_messages([("user", "Foo")])
+        model = FakeMessagesListChatModel(responses=[AIMessage(content="Bar")])
+        chain = prompt | model
+
+        # This should work and call the mock client
+        result = chain.invoke({}, config={"callbacks": [callbacks]})
+        assert result.content == "Bar"
+
+        # Verify that the mock client was used for capturing events
+        assert mock_client.capture.call_count == 3
+
+
+def test_convert_message_to_dict_tool_calls():
+    """Test that _convert_message_to_dict properly converts tool calls in AIMessage."""
+    from posthog.ai.langchain.callbacks import _convert_message_to_dict
+    from langchain_core.messages import AIMessage
+    from langchain_core.messages.tool import ToolCall
+
+    # Create an AIMessage with tool calls
+    tool_calls = [
+        ToolCall(
+            id="call_123",
+            name="get_weather",
+            args={"city": "San Francisco", "units": "celsius"},
+        )
+    ]
+
+    ai_message = AIMessage(
+        content="I'll check the weather for you.", tool_calls=tool_calls
+    )
+
+    # Convert to dict
+    result = _convert_message_to_dict(ai_message)
+
+    # Verify the conversion
+    assert result["role"] == "assistant"
+    assert result["content"] == "I'll check the weather for you."
+    assert result["tool_calls"] == [
+        {
+            "type": "function",
+            "id": "call_123",
+            "function": {
+                "name": "get_weather",
+                "arguments": '{"city": "San Francisco", "units": "celsius"}',
+            },
+        }
+    ]
+
+
+def test_tool_definition(mock_client):
+    """Test that tools defined in invocation parameters are captured in $ai_tools property"""
+    callbacks = CallbackHandler(mock_client)
+    run_id = uuid.uuid4()
+
+    # Define tools to be passed to the invocation parameters
+    tools = [
+        {
+            "type": "function",
+            "function": {
+                "name": "get_weather",
+                "description": "Get the current weather for a specific location",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "location": {
+                            "type": "string",
+                            "description": "The city or location name to get weather for",
+                        }
+                    },
+                    "required": ["location"],
+                },
+            },
+        }
+    ]
+
+    with patch("time.time", return_value=1234567890):
+        callbacks._set_llm_metadata(
+            {"kwargs": {"openai_api_base": "https://api.openai.com/v1"}},
+            run_id,
+            messages=[{"role": "user", "content": "hey"}],
+            invocation_params={"temperature": 0.7, "tools": tools},
+            metadata={"ls_model_name": "gpt-4o-mini", "ls_provider": "openai"},
+            name="test",
+        )
+
+    expected = GenerationMetadata(
+        model="gpt-4o-mini",
+        input=[{"role": "user", "content": "hey"}],
+        start_time=1234567890,
+        model_params={"temperature": 0.7},
+        provider="openai",
+        base_url="https://api.openai.com/v1",
+        name="test",
+        tools=tools,
+        end_time=None,
+    )
+    assert callbacks._runs[run_id] == expected
+
+    with patch("time.time", return_value=1234567891):
+        run = callbacks._pop_run_metadata(run_id)
+    expected.end_time = 1234567891
+    assert run == expected
+    assert callbacks._runs == {}
+
+    # Now test that the tools are properly captured in the PostHog event
+    mock_response = MagicMock()
+    mock_response.generations = [[MagicMock()]]
+
+    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["distinct_id"] == run_id
+    assert call_args["event"] == "$ai_generation"
+    assert props["$ai_provider"] == "openai"
+    assert props["$ai_model"] == "gpt-4o-mini"
+    assert props["$ai_input"] == [{"role": "user", "content": "hey"}]
+    assert props["$ai_model_parameters"] == {"temperature": 0.7}
+    assert props["$ai_base_url"] == "https://api.openai.com/v1"
+    assert props["$ai_span_name"] == "test"
+    assert props["$ai_span_id"] == run_id
+    assert props["$ai_trace_id"] == run_id
+    assert props["$ai_latency"] == 1.0
+    # Verify that tools are captured in the $ai_tools property
+    assert props["$ai_tools"] == tools

posthog/test/ai/openai/test_openai.py

@@ -1,4 +1,3 @@
-import json
 import time
 from unittest.mock import patch
 
@@ -26,6 +25,7 @@ try:
         ResponseOutputMessage,
         ResponseOutputText,
         ResponseUsage,
+        ResponseFunctionToolCall,
         ParsedResponse,
     )
     from openai.types.responses.parsed_response import (
@@ -227,11 +227,27 @@ def mock_openai_response_with_tool_calls():
         created=int(time.time()),
         choices=[
             Choice(
-                finish_reason="tool_calls",
+                finish_reason="stop",
                 index=0,
                 message=ChatCompletionMessage(
                     content="I'll check the weather for you.",
                     role="assistant",
+                ),
+            ),
+            Choice(
+                finish_reason="stop",
+                index=1,
+                message=ChatCompletionMessage(
+                    content=" Let me look that up.",
+                    role="assistant",
+                ),
+            ),
+            Choice(
+                finish_reason="tool_calls",
+                index=2,
+                message=ChatCompletionMessage(
+                    content=None,
+                    role="assistant",
                     tool_calls=[
                         ChatCompletionMessageToolCall(
                             id="call_abc123",
@@ -243,7 +259,7 @@ def mock_openai_response_with_tool_calls():
                         )
                     ],
                 ),
-            )
+            ),
         ],
         usage=CompletionUsage(
             completion_tokens=15,
@@ -253,6 +269,97 @@ def mock_openai_response_with_tool_calls():
     )
 
 
+@pytest.fixture
+def mock_openai_response_tool_calls_only():
+    return ChatCompletion(
+        id="test",
+        model="gpt-4",
+        object="chat.completion",
+        created=int(time.time()),
+        choices=[
+            Choice(
+                finish_reason="tool_calls",
+                index=0,
+                message=ChatCompletionMessage(
+                    content=None,
+                    role="assistant",
+                    tool_calls=[
+                        ChatCompletionMessageToolCall(
+                            id="call_def456",
+                            type="function",
+                            function=Function(
+                                name="get_weather",
+                                arguments='{"location": "New York"}',
+                            ),
+                        )
+                    ],
+                ),
+            )
+        ],
+        usage=CompletionUsage(
+            completion_tokens=10,
+            prompt_tokens=25,
+            total_tokens=35,
+        ),
+    )
+
+
+@pytest.fixture
+def mock_responses_api_with_tool_calls():
+    return Response(
+        id="resp_123",
+        object="response",
+        created_at=int(time.time()),
+        model="gpt-4o-mini",
+        status="completed",
+        error=None,
+        incomplete_details=None,
+        instructions=None,
+        max_output_tokens=None,
+        tools=[],
+        tool_choice="auto",
+        parallel_tool_calls=True,
+        output=[
+            ResponseOutputMessage(
+                id="msg_456",
+                type="message",
+                role="assistant",
+                status="completed",
+                content=[
+                    ResponseOutputText(
+                        type="output_text",
+                        text="I'll help you with the weather.",
+                        annotations=[],
+                    ),
+                    ResponseOutputText(
+                        type="output_text",
+                        text=" Let me check that for you.",
+                        annotations=[],
+                    ),
+                ],
+            ),
+            ResponseFunctionToolCall(
+                id="fc_789",
+                type="function_call",
+                name="get_weather",
+                call_id="call_xyz789",
+                arguments='{"location": "Chicago"}',
+                status="completed",
+            ),
+        ],
+        usage=ResponseUsage(
+            input_tokens=30,
+            output_tokens=20,
+            input_tokens_details={"prompt_tokens": 30, "cached_tokens": 0},
+            output_tokens_details={"reasoning_tokens": 0},
+            total_tokens=50,
+        ),
+        previous_response_id=None,
+        user=None,
+        metadata={},
+    )
+
+
 def test_basic_completion(mock_client, mock_openai_response):
     with patch(
         "openai.resources.chat.completions.Completions.create",
@@ -278,7 +385,10 @@ def test_basic_completion(mock_client, mock_openai_response):
         assert props["$ai_model"] == "gpt-4"
         assert props["$ai_input"] == [{"role": "user", "content": "Hello"}]
         assert props["$ai_output_choices"] == [
-            {"role": "assistant", "content": "Test response"}
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": "Test response"}],
+            }
         ]
         assert props["$ai_input_tokens"] == 20
         assert props["$ai_output_tokens"] == 10
@@ -427,7 +537,10 @@ def test_cached_tokens(mock_client, mock_openai_response_with_cached_tokens):
         assert props["$ai_model"] == "gpt-4"
         assert props["$ai_input"] == [{"role": "user", "content": "Hello"}]
         assert props["$ai_output_choices"] == [
-            {"role": "assistant", "content": "Test response"}
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": "Test response"}],
+            }
         ]
         assert props["$ai_input_tokens"] == 20
         assert props["$ai_output_tokens"] == 10
@@ -475,24 +588,34 @@ def test_tool_calls(mock_client, mock_openai_response_with_tool_calls):
             {"role": "user", "content": "What's the weather in San Francisco?"}
         ]
         assert props["$ai_output_choices"] == [
-            {"role": "assistant", "content": "I'll check the weather for you."}
+            {
+                "role": "assistant",
+                "content": [
+                    {"type": "text", "text": "I'll check the weather for you."},
+                    {"type": "text", "text": " Let me look that up."},
+                    {
+                        "type": "function",
+                        "id": "call_abc123",
+                        "function": {
+                            "name": "get_weather",
+                            "arguments": '{"location": "San Francisco", "unit": "celsius"}',
+                        },
+                    },
+                ],
+            }
         ]
 
-        # Check that tool calls are properly captured
+        # Check that defined tools are properly captured in $ai_tools
         assert "$ai_tools" in props
-        tool_calls = props["$ai_tools"]
-        assert len(tool_calls) == 1
+        defined_tools = props["$ai_tools"]
+        assert len(defined_tools) == 1
 
-        # Verify the tool call details
-        tool_call = tool_calls[0]
-        assert tool_call.id == "call_abc123"
-        assert tool_call.type == "function"
-        assert tool_call.function.name == "get_weather"
-
-        # Verify the arguments
-        arguments = tool_call.function.arguments
-        parsed_args = json.loads(arguments)
-        assert parsed_args == {"location": "San Francisco", "unit": "celsius"}
+        # Verify the defined tool details
+        defined_tool = defined_tools[0]
+        assert defined_tool["type"] == "function"
+        assert defined_tool["function"]["name"] == "get_weather"
+        assert defined_tool["function"]["description"] == "Get weather"
+        assert defined_tool["function"]["parameters"] == {}
 
         # Check token usage
         assert props["$ai_input_tokens"] == 20
@@ -500,6 +623,117 @@ def test_tool_calls(mock_client, mock_openai_response_with_tool_calls):
         assert props["$ai_http_status"] == 200
 
 
+def test_tool_calls_only_no_content(mock_client, mock_openai_response_tool_calls_only):
+    with patch(
+        "openai.resources.chat.completions.Completions.create",
+        return_value=mock_openai_response_tool_calls_only,
+    ):
+        client = OpenAI(api_key="test-key", posthog_client=mock_client)
+        response = client.chat.completions.create(
+            model="gpt-4",
+            messages=[{"role": "user", "content": "Get weather for New York"}],
+            tools=[
+                {
+                    "type": "function",
+                    "function": {
+                        "name": "get_weather",
+                        "description": "Get weather",
+                        "parameters": {},
+                    },
+                }
+            ],
+            posthog_distinct_id="test-id",
+        )
+
+        assert response == mock_openai_response_tool_calls_only
+        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_choices"] == [
+            {
+                "role": "assistant",
+                "content": [
+                    {
+                        "type": "function",
+                        "id": "call_def456",
+                        "function": {
+                            "name": "get_weather",
+                            "arguments": '{"location": "New York"}',
+                        },
+                    }
+                ],
+            }
+        ]
+
+        # Check token usage
+        assert props["$ai_input_tokens"] == 25
+        assert props["$ai_output_tokens"] == 10
+        assert props["$ai_http_status"] == 200
+
+
+def test_responses_api_tool_calls(mock_client, mock_responses_api_with_tool_calls):
+    with patch(
+        "openai.resources.responses.Responses.create",
+        return_value=mock_responses_api_with_tool_calls,
+    ):
+        client = OpenAI(api_key="test-key", posthog_client=mock_client)
+        response = client.responses.create(
+            model="gpt-4o-mini",
+            input=[{"role": "user", "content": "What's the weather in Chicago?"}],
+            tools=[
+                {
+                    "name": "get_weather",
+                    "type": "function",
+                    "function": {
+                        "name": "get_weather",
+                        "description": "Get weather",
+                        "parameters": {},
+                    },
+                }
+            ],
+            posthog_distinct_id="test-id",
+        )
+
+        assert response == mock_responses_api_with_tool_calls
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        assert call_args["distinct_id"] == "test-id"
+        assert call_args["event"] == "$ai_generation"
+        assert props["$ai_provider"] == "openai"
+        assert props["$ai_model"] == "gpt-4o-mini"
+        assert props["$ai_output_choices"] == [
+            {
+                "role": "assistant",
+                "content": [
+                    {"type": "text", "text": "I'll help you with the weather."},
+                    {"type": "text", "text": " Let me check that for you."},
+                    {
+                        "type": "function",
+                        "id": "call_xyz789",
+                        "function": {
+                            "name": "get_weather",
+                            "arguments": '{"location": "Chicago"}',
+                        },
+                    },
+                ],
+            }
+        ]
+
+        # Check token usage
+        assert props["$ai_input_tokens"] == 30
+        assert props["$ai_output_tokens"] == 20
+        assert props["$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 = [
@@ -644,21 +878,17 @@ def test_streaming_with_tool_calls(mock_client):
         assert props["$ai_provider"] == "openai"
         assert props["$ai_model"] == "gpt-4"
 
-        # Check that the tool calls were properly accumulated
+        # Check that defined tools are properly captured in $ai_tools
         assert "$ai_tools" in props
-        tool_calls = props["$ai_tools"]
-        assert len(tool_calls) == 1
+        defined_tools = props["$ai_tools"]
+        assert len(defined_tools) == 1
 
-        # Verify the complete tool call was properly assembled
-        tool_call = tool_calls[0]
-        assert tool_call.id == "call_abc123"
-        assert tool_call.type == "function"
-        assert tool_call.function.name == "get_weather"
-
-        # Verify the arguments were concatenated correctly
-        arguments = tool_call.function.arguments
-        parsed_args = json.loads(arguments)
-        assert parsed_args == {"location": "San Francisco", "unit": "celsius"}
+        # Verify the defined tool details
+        defined_tool = defined_tools[0]
+        assert defined_tool["type"] == "function"
+        assert defined_tool["function"]["name"] == "get_weather"
+        assert defined_tool["function"]["description"] == "Get weather"
+        assert defined_tool["function"]["parameters"] == {}
 
         # Check that the content was also accumulated
         assert (
@@ -696,7 +926,10 @@ def test_responses_api(mock_client, mock_openai_response_with_responses_api):
         assert props["$ai_model"] == "gpt-4o-mini"
         assert props["$ai_input"] == [{"role": "user", "content": "Hello"}]
         assert props["$ai_output_choices"] == [
-            {"role": "assistant", "content": "Test response"}
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": "Test response"}],
+            }
         ]
         assert props["$ai_input_tokens"] == 10
         assert props["$ai_output_tokens"] == 10
@@ -765,7 +998,12 @@ def test_responses_parse(mock_client, mock_parsed_response):
         assert props["$ai_output_choices"] == [
             {
                 "role": "assistant",
-                "content": '{"name": "Science Fair", "date": "Friday", "participants": ["Alice", "Bob"]}',
+                "content": [
+                    {
+                        "type": "text",
+                        "text": '{"name": "Science Fair", "date": "Friday", "participants": ["Alice", "Bob"]}',
+                    }
+                ],
             }
         ]
         assert props["$ai_input_tokens"] == 15
@@ -774,3 +1012,66 @@ def test_responses_parse(mock_client, mock_parsed_response):
         assert props["$ai_http_status"] == 200
         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 $ai_tools property"""
+    with patch(
+        "openai.resources.chat.completions.Completions.create",
+        return_value=mock_openai_response,
+    ):
+        client = OpenAI(api_key="test-key", posthog_client=mock_client)
+
+        # Define tools to be passed to the create function
+        tools = [
+            {
+                "type": "function",
+                "function": {
+                    "name": "get_weather",
+                    "description": "Get the current weather for a specific location",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {
+                            "location": {
+                                "type": "string",
+                                "description": "The city or location name to get weather for",
+                            }
+                        },
+                        "required": ["location"],
+                    },
+                },
+            }
+        ]
+
+        response = client.chat.completions.create(
+            model="gpt-4o-mini",
+            messages=[{"role": "user", "content": "hey"}],
+            tools=tools,
+            posthog_distinct_id="test-id",
+            posthog_properties={"foo": "bar"},
+        )
+
+        assert response == mock_openai_response
+        assert mock_client.capture.call_count == 1
+
+        call_args = mock_client.capture.call_args[1]
+        props = call_args["properties"]
+
+        assert call_args["distinct_id"] == "test-id"
+        assert call_args["event"] == "$ai_generation"
+        assert props["$ai_provider"] == "openai"
+        assert props["$ai_model"] == "gpt-4o-mini"
+        assert props["$ai_input"] == [{"role": "user", "content": "hey"}]
+        assert props["$ai_output_choices"] == [
+            {
+                "role": "assistant",
+                "content": [{"type": "text", "text": "Test response"}],
+            }
+        ]
+        assert props["$ai_input_tokens"] == 20
+        assert props["$ai_output_tokens"] == 10
+        assert props["$ai_http_status"] == 200
+        assert props["foo"] == "bar"
+        assert isinstance(props["$ai_latency"], float)
+        # Verify that tools are captured in the $ai_tools property
+        assert props["$ai_tools"] == tools

posthog/test/exception_integrations/__init__.py

@@ -1,3 +0,0 @@
-import pytest
-
-pytest.importorskip("django")

posthog/test/exception_integrations/test_django.py

@@ -1,121 +0,0 @@
-from posthog.exception_integrations.django import DjangoRequestExtractor
-from django.test import RequestFactory
-from django.conf import settings
-from django.core.management import call_command
-import django
-
-DEFAULT_USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/537.3"
-
-# setup a test app
-if not settings.configured:
-    settings.configure(
-        SECRET_KEY="test",
-        DEFAULT_CHARSET="utf-8",
-        INSTALLED_APPS=[
-            "django.contrib.auth",
-            "django.contrib.contenttypes",
-        ],
-        DATABASES={
-            "default": {
-                "ENGINE": "django.db.backends.sqlite3",
-                "NAME": ":memory:",
-            }
-        },
-    )
-    django.setup()
-
-    call_command("migrate", verbosity=0, interactive=False)
-
-
-def mock_request_factory(override_headers):
-    factory = RequestFactory(
-        headers={
-            "User-Agent": DEFAULT_USER_AGENT,
-            "Referrer": "http://example.com",
-            "X-Forwarded-For": "193.4.5.12",
-            **(override_headers or {}),
-        }
-    )
-
-    request = factory.get("/api/endpoint")
-    return request
-
-
-def test_request_extractor_with_no_trace():
-    request = mock_request_factory(None)
-    extractor = DjangoRequestExtractor(request)
-    assert extractor.extract_person_data() == {
-        "ip": "193.4.5.12",
-        "user_agent": DEFAULT_USER_AGENT,
-        "traceparent": None,
-        "distinct_id": None,
-        "$request_path": "/api/endpoint",
-    }
-
-
-def test_request_extractor_with_trace():
-    request = mock_request_factory(
-        {"traceparent": "00-0af7651916cd43dd8448eb211c80319c-b7ad6b7169203331-01"}
-    )
-
-    extractor = DjangoRequestExtractor(request)
-    assert extractor.extract_person_data() == {
-        "ip": "193.4.5.12",
-        "user_agent": DEFAULT_USER_AGENT,
-        "traceparent": "00-0af7651916cd43dd8448eb211c80319c-b7ad6b7169203331-01",
-        "distinct_id": None,
-        "$request_path": "/api/endpoint",
-    }
-
-
-def test_request_extractor_with_tracestate():
-    request = mock_request_factory(
-        {
-            "traceparent": "00-0af7651916cd43dd8448eb211c80319c-b7ad6b7169203331-01",
-            "tracestate": "posthog-distinct-id=1234",
-        }
-    )
-    extractor = DjangoRequestExtractor(request)
-    assert extractor.extract_person_data() == {
-        "ip": "193.4.5.12",
-        "user_agent": DEFAULT_USER_AGENT,
-        "traceparent": "00-0af7651916cd43dd8448eb211c80319c-b7ad6b7169203331-01",
-        "distinct_id": "1234",
-        "$request_path": "/api/endpoint",
-    }
-
-
-def test_request_extractor_with_complicated_tracestate():
-    request = mock_request_factory(
-        {"tracestate": "posthog-distinct-id=alohaMountainsXUYZ,rojo=00f067aa0ba902b7"}
-    )
-    extractor = DjangoRequestExtractor(request)
-    assert extractor.extract_person_data() == {
-        "ip": "193.4.5.12",
-        "user_agent": DEFAULT_USER_AGENT,
-        "traceparent": None,
-        "distinct_id": "alohaMountainsXUYZ",
-        "$request_path": "/api/endpoint",
-    }
-
-
-def test_request_extractor_with_request_user():
-    from django.contrib.auth.models import User
-
-    user = User.objects.create_user(
-        username="test", email="test@posthog.com", password="top_secret"
-    )
-
-    request = mock_request_factory(None)
-    request.user = user
-
-    extractor = DjangoRequestExtractor(request)
-    assert extractor.extract_person_data() == {
-        "ip": "193.4.5.12",
-        "user_agent": DEFAULT_USER_AGENT,
-        "traceparent": None,
-        "distinct_id": None,
-        "$request_path": "/api/endpoint",
-        "email": "test@posthog.com",
-        "$user_id": "1",
-    }

posthog/test/integrations/test_middleware.py

@@ -1,4 +1,8 @@
-from posthog.scopes import new_context, get_context_session_id, get_context_distinct_id
+from posthog.contexts import (
+    new_context,
+    get_context_session_id,
+    get_context_distinct_id,
+)
 import unittest
 from unittest.mock import Mock
 

posthog/test/test_before_send.py

@@ -40,16 +40,30 @@ class TestClient(unittest.TestCase):
             event["properties"]["processed_by_before_send"] = True
             return event
 
-        client = Client(
-            FAKE_TEST_API_KEY, on_error=self.set_fail, before_send=my_before_send
-        )
-        success, msg = client.capture("user1", "test_event", {"original": "value"})
+        with mock.patch("posthog.client.batch_post") as mock_post:
+            client = Client(
+                FAKE_TEST_API_KEY,
+                on_error=self.set_fail,
+                before_send=my_before_send,
+                sync_mode=True,
+            )
+            msg_uuid = client.capture(
+                "test_event", distinct_id="user1", properties={"original": "value"}
+            )
 
-        self.assertTrue(success)
-        self.assertEqual(msg["properties"]["processed_by_before_send"], True)
-        self.assertEqual(msg["properties"]["original"], "value")
-        self.assertEqual(len(processed_events), 1)
-        self.assertEqual(processed_events[0]["event"], "test_event")
+            self.assertIsNotNone(msg_uuid)
+
+            # Get the enqueued message from the mock
+            mock_post.assert_called_once()
+            batch_data = mock_post.call_args[1]["batch"]
+            enqueued_msg = batch_data[0]
+
+            self.assertEqual(
+                enqueued_msg["properties"]["processed_by_before_send"], True
+            )
+            self.assertEqual(enqueued_msg["properties"]["original"], "value")
+            self.assertEqual(len(processed_events), 1)
+            self.assertEqual(processed_events[0]["event"], "test_event")
 
     def test_before_send_callback_drops_event(self):
         """Test that before_send callback can drop events by returning None."""
@@ -59,20 +73,27 @@ class TestClient(unittest.TestCase):
                 return None
             return event
 
-        client = Client(
-            FAKE_TEST_API_KEY, on_error=self.set_fail, before_send=drop_test_events
-        )
+        with mock.patch("posthog.client.batch_post") as mock_post:
+            client = Client(
+                FAKE_TEST_API_KEY,
+                on_error=self.set_fail,
+                before_send=drop_test_events,
+                sync_mode=True,
+            )
 
-        # Event should be dropped
-        success, msg = client.capture("user1", "test_drop_me")
-        self.assertTrue(success)
-        self.assertIsNone(msg)
+            # Event should be dropped
+            msg_uuid = client.capture("test_drop_me", distinct_id="user1")
+            self.assertIsNone(msg_uuid)
 
-        # Event should go through
-        success, msg = client.capture("user1", "keep_me")
-        self.assertTrue(success)
-        self.assertIsNotNone(msg)
-        self.assertEqual(msg["event"], "keep_me")
+            # Event should go through
+            msg_uuid = client.capture("keep_me", distinct_id="user1")
+            self.assertIsNotNone(msg_uuid)
+
+            # Check the enqueued message
+            mock_post.assert_called_once()
+            batch_data = mock_post.call_args[1]["batch"]
+            enqueued_msg = batch_data[0]
+            self.assertEqual(enqueued_msg["event"], "keep_me")
 
     def test_before_send_callback_handles_exceptions(self):
         """Test that exceptions in before_send don't crash the client."""
@@ -80,18 +101,26 @@ class TestClient(unittest.TestCase):
         def buggy_before_send(event):
             raise ValueError("Oops!")
 
-        client = Client(
-            FAKE_TEST_API_KEY, on_error=self.set_fail, before_send=buggy_before_send
-        )
-        success, msg = client.capture("user1", "robust_event")
+        with mock.patch("posthog.client.batch_post") as mock_post:
+            client = Client(
+                FAKE_TEST_API_KEY,
+                on_error=self.set_fail,
+                before_send=buggy_before_send,
+                sync_mode=True,
+            )
+            msg_uuid = client.capture("robust_event", distinct_id="user1")
 
-        # Event should still be sent despite the exception
-        self.assertTrue(success)
-        self.assertIsNotNone(msg)
-        self.assertEqual(msg["event"], "robust_event")
+            # Event should still be sent despite the exception
+            self.assertIsNotNone(msg_uuid)
+
+            # Check the enqueued message
+            mock_post.assert_called_once()
+            batch_data = mock_post.call_args[1]["batch"]
+            enqueued_msg = batch_data[0]
+            self.assertEqual(enqueued_msg["event"], "robust_event")
 
     def test_before_send_callback_works_with_all_event_types(self):
-        """Test that before_send works with capture, identify, set, etc."""
+        """Test that before_send works with capture, set, etc."""
 
         def add_marker(event):
             if "properties" not in event:
@@ -99,38 +128,46 @@ class TestClient(unittest.TestCase):
             event["properties"]["marked"] = True
             return event
 
-        client = Client(
-            FAKE_TEST_API_KEY, on_error=self.set_fail, before_send=add_marker
-        )
+        with mock.patch("posthog.client.batch_post") as mock_post:
+            client = Client(
+                FAKE_TEST_API_KEY,
+                on_error=self.set_fail,
+                before_send=add_marker,
+                sync_mode=True,
+            )
 
-        # Test capture
-        success, msg = client.capture("user1", "event")
-        self.assertTrue(success)
-        self.assertTrue(msg["properties"]["marked"])
+            # Test capture
+            msg_uuid = client.capture("event", distinct_id="user1")
+            self.assertIsNotNone(msg_uuid)
 
-        # Test identify
-        success, msg = client.identify("user1", {"trait": "value"})
-        self.assertTrue(success)
-        self.assertTrue(msg["properties"]["marked"])
+            # Test set
+            msg_uuid = client.set(distinct_id="user1", properties={"prop": "value"})
+            self.assertIsNotNone(msg_uuid)
 
-        # Test set
-        success, msg = client.set("user1", {"prop": "value"})
-        self.assertTrue(success)
-        self.assertTrue(msg["properties"]["marked"])
-
-        # Test page
-        success, msg = client.page("user1", "https://example.com")
-        self.assertTrue(success)
-        self.assertTrue(msg["properties"]["marked"])
+            # Check all events were marked
+            self.assertEqual(mock_post.call_count, 2)
+            for call in mock_post.call_args_list:
+                batch_data = call[1]["batch"]
+                enqueued_msg = batch_data[0]
+                self.assertTrue(enqueued_msg["properties"]["marked"])
 
     def test_before_send_callback_disabled_when_none(self):
         """Test that client works normally when before_send is None."""
-        client = Client(FAKE_TEST_API_KEY, on_error=self.set_fail, before_send=None)
-        success, msg = client.capture("user1", "normal_event")
+        with mock.patch("posthog.client.batch_post") as mock_post:
+            client = Client(
+                FAKE_TEST_API_KEY,
+                on_error=self.set_fail,
+                before_send=None,
+                sync_mode=True,
+            )
+            msg_uuid = client.capture("normal_event", distinct_id="user1")
+            self.assertIsNotNone(msg_uuid)
 
-        self.assertTrue(success)
-        self.assertIsNotNone(msg)
-        self.assertEqual(msg["event"], "normal_event")
+            # Check the event was sent normally
+            mock_post.assert_called_once()
+            batch_data = mock_post.call_args[1]["batch"]
+            enqueued_msg = batch_data[0]
+            self.assertEqual(enqueued_msg["event"], "normal_event")
 
     def test_before_send_callback_pii_scrubbing_example(self):
         """Test a realistic PII scrubbing use case."""
@@ -152,20 +189,30 @@ class TestClient(unittest.TestCase):
 
             return event
 
-        client = Client(
-            FAKE_TEST_API_KEY, on_error=self.set_fail, before_send=scrub_pii
-        )
-        success, msg = client.capture(
-            "user1",
-            "form_submit",
-            {
-                "email": "user@example.com",
-                "credit_card": "1234-5678-9012-3456",
-                "form_name": "contact",
-            },
-        )
+        with mock.patch("posthog.client.batch_post") as mock_post:
+            client = Client(
+                FAKE_TEST_API_KEY,
+                on_error=self.set_fail,
+                before_send=scrub_pii,
+                sync_mode=True,
+            )
+            msg_uuid = client.capture(
+                "form_submit",
+                distinct_id="user1",
+                properties={
+                    "email": "user@example.com",
+                    "credit_card": "1234-5678-9012-3456",
+                    "form_name": "contact",
+                },
+            )
 
-        self.assertTrue(success)
-        self.assertEqual(msg["properties"]["email"], "***@example.com")
-        self.assertNotIn("credit_card", msg["properties"])
-        self.assertEqual(msg["properties"]["form_name"], "contact")
+            self.assertIsNotNone(msg_uuid)
+
+            # Check the enqueued message was scrubbed
+            mock_post.assert_called_once()
+            batch_data = mock_post.call_args[1]["batch"]
+            enqueued_msg = batch_data[0]
+
+            self.assertEqual(enqueued_msg["properties"]["email"], "***@example.com")
+            self.assertNotIn("credit_card", enqueued_msg["properties"])
+            self.assertEqual(enqueued_msg["properties"]["form_name"], "contact")

posthog/test/test_contexts.py

@@ -1,8 +1,7 @@
 import unittest
 from unittest.mock import patch
 
-from posthog.scopes import (
-    clear_tags,
+from posthog.contexts import (
     get_tags,
     new_context,
     scoped,
@@ -14,11 +13,7 @@ from posthog.scopes import (
 )
 
 
-class TestScopes(unittest.TestCase):
-    def setUp(self):
-        # Reset any context between tests
-        clear_tags()
-
+class TestContexts(unittest.TestCase):
     def test_tag_and_get_tags(self):
         with new_context(fresh=True):
             tag("key1", "value1")
@@ -28,14 +23,6 @@ class TestScopes(unittest.TestCase):
             assert tags["key1"] == "value1"
             assert tags["key2"] == 2
 
-    def test_clear_tags(self):
-        with new_context(fresh=True):
-            tag("key1", "value1")
-            assert get_tags()["key1"] == "value1"
-
-            clear_tags()
-            assert get_tags() == {}
-
     def test_new_context_isolation(self):
         with new_context(fresh=True):
             # Set tag in outer context

posthog/test/test_exception_capture.py

@@ -32,32 +32,3 @@ def test_excepthook(tmpdir):
         b'"$exception_list": [{"mechanism": {"type": "generic", "handled": true}, "module": null, "type": "ZeroDivisionError", "value": "division by zero", "stacktrace": {"frames": [{"platform": "python", "filename": "app.py", "abs_path"'
         in output
     )
-
-
-def test_trying_to_use_django_integration(tmpdir):
-    app = tmpdir.join("app.py")
-    app.write(
-        dedent(
-            """
-    from posthog import Posthog, Integrations
-    posthog = Posthog('phc_x', host='https://eu.i.posthog.com', enable_exception_autocapture=True, exception_autocapture_integrations=[Integrations.Django], debug=True, on_error=lambda e, batch: print('error handling batch: ', e, batch))
-
-    # frame_value = "LOL"
-
-    1/0
-    """
-        )
-    )
-
-    with pytest.raises(subprocess.CalledProcessError) as excinfo:
-        subprocess.check_output([sys.executable, str(app)], stderr=subprocess.STDOUT)
-
-    output = excinfo.value.output
-
-    assert b"ZeroDivisionError" in output
-    assert b"LOL" in output
-    assert b"DEBUG:posthog:data uploaded successfully" in output
-    assert (
-        b'"$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
-    )

posthog/test/test_feature_flag_result.py

@@ -229,9 +229,9 @@ class TestGetFeatureFlagResult(unittest.TestCase):
         self.assertEqual(flag_result.variant, None)
         self.assertEqual(flag_result.payload, 300)
         patch_capture.assert_called_with(
-            "some-distinct-id",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id",
+            properties={
                 "$feature_flag": "person-flag",
                 "$feature_flag_response": True,
                 "locally_evaluated": True,
@@ -283,9 +283,9 @@ class TestGetFeatureFlagResult(unittest.TestCase):
         self.assertEqual(flag_result.payload, {"some": "value"})
 
         patch_capture.assert_called_with(
-            "distinct_id",
             "$feature_flag_called",
-            {
+            distinct_id="distinct_id",
+            properties={
                 "$feature_flag": "person-flag",
                 "$feature_flag_response": "variant-1",
                 "locally_evaluated": True,
@@ -305,9 +305,9 @@ class TestGetFeatureFlagResult(unittest.TestCase):
         self.assertIsNone(another_flag_result.payload)
 
         patch_capture.assert_called_with(
-            "another-distinct-id",
             "$feature_flag_called",
-            {
+            distinct_id="another-distinct-id",
+            properties={
                 "$feature_flag": "person-flag",
                 "$feature_flag_response": "variant-2",
                 "locally_evaluated": True,
@@ -345,9 +345,9 @@ class TestGetFeatureFlagResult(unittest.TestCase):
         self.assertEqual(flag_result.variant, None)
         self.assertEqual(flag_result.payload, 300)
         patch_capture.assert_called_with(
-            "some-distinct-id",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id",
+            properties={
                 "$feature_flag": "person-flag",
                 "$feature_flag_response": True,
                 "locally_evaluated": False,
@@ -388,9 +388,9 @@ class TestGetFeatureFlagResult(unittest.TestCase):
         self.assertEqual(flag_result.get_value(), "variant-1")
         self.assertEqual(flag_result.payload, [1, 2, 3])
         patch_capture.assert_called_with(
-            "distinct_id",
             "$feature_flag_called",
-            {
+            distinct_id="distinct_id",
+            properties={
                 "$feature_flag": "person-flag",
                 "$feature_flag_response": "variant-1",
                 "locally_evaluated": False,
@@ -431,9 +431,9 @@ class TestGetFeatureFlagResult(unittest.TestCase):
 
         self.assertIsNone(flag_result)
         patch_capture.assert_called_with(
-            "some-distinct-id",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id",
+            properties={
                 "$feature_flag": "no-person-flag",
                 "$feature_flag_response": None,
                 "locally_evaluated": False,

posthog/test/test_feature_flags.py

@@ -1355,6 +1355,77 @@ class TestLocalEvaluation(unittest.TestCase):
         self.assertEqual(patch_flags.call_count, 0)
         self.assertEqual(patch_get.call_count, 0)
 
+    @mock.patch("posthog.feature_flags.log")
+    @mock.patch("posthog.client.flags")
+    @mock.patch("posthog.client.get")
+    def test_feature_flags_with_flag_dependencies(
+        self, patch_get, patch_flags, mock_log
+    ):
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+        client.feature_flags = [
+            {
+                "id": 1,
+                "name": "Flag with Dependencies",
+                "key": "flag-with-dependencies",
+                "active": True,
+                "filters": {
+                    "groups": [
+                        {
+                            "properties": [
+                                {
+                                    "key": "beta-feature",
+                                    "operator": "exact",
+                                    "value": True,
+                                    "type": "flag",
+                                },
+                                {
+                                    "key": "email",
+                                    "operator": "icontains",
+                                    "value": "@example.com",
+                                    "type": "person",
+                                },
+                            ],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            }
+        ]
+
+        # Test that flag evaluation doesn't fail when encountering a flag dependency
+        # The flag should evaluate based on other conditions (email contains @example.com)
+        # Since flag dependencies aren't implemented, it should skip the flag condition
+        # and evaluate based on the email condition only
+        feature_flag_match = client.get_feature_flag(
+            "flag-with-dependencies",
+            "test-user",
+            person_properties={"email": "test@example.com"},
+        )
+        self.assertEqual(feature_flag_match, True)
+        self.assertEqual(patch_flags.call_count, 0)
+        self.assertEqual(patch_get.call_count, 0)
+
+        # Verify warning was logged for flag dependency
+        mock_log.warning.assert_called_with(
+            "Flag dependency filters are not supported in local evaluation. "
+            "Skipping condition for flag '%s' with dependency on flag '%s'",
+            "flag-with-dependencies",
+            "beta-feature",
+        )
+
+        # Test with email that doesn't match
+        feature_flag_match = client.get_feature_flag(
+            "flag-with-dependencies",
+            "test-user-2",
+            person_properties={"email": "test@other.com"},
+        )
+        self.assertEqual(feature_flag_match, False)
+        self.assertEqual(patch_flags.call_count, 0)
+        self.assertEqual(patch_get.call_count, 0)
+
+        # Verify warning was logged again for the second evaluation
+        self.assertEqual(mock_log.warning.call_count, 2)
+
     @mock.patch("posthog.client.Poller")
     @mock.patch("posthog.client.get")
     def test_load_feature_flags(self, patch_get, patch_poll):
@@ -2695,9 +2766,9 @@ class TestCaptureCalls(unittest.TestCase):
         )
         self.assertEqual(patch_capture.call_count, 1)
         patch_capture.assert_called_with(
-            "some-distinct-id",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id",
+            properties={
                 "$feature_flag": "complex-flag",
                 "$feature_flag_response": True,
                 "locally_evaluated": True,
@@ -2729,9 +2800,9 @@ class TestCaptureCalls(unittest.TestCase):
         )
         self.assertEqual(patch_capture.call_count, 1)
         patch_capture.assert_called_with(
-            "some-distinct-id2",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id2",
+            properties={
                 "$feature_flag": "complex-flag",
                 "$feature_flag_response": True,
                 "locally_evaluated": True,
@@ -2767,9 +2838,9 @@ class TestCaptureCalls(unittest.TestCase):
         self.assertEqual(patch_flags.call_count, 1)
         self.assertEqual(patch_capture.call_count, 1)
         patch_capture.assert_called_with(
-            "some-distinct-id2",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id2",
+            properties={
                 "$feature_flag": "decide-flag",
                 "$feature_flag_response": "decide-value",
                 "locally_evaluated": False,
@@ -2820,9 +2891,9 @@ class TestCaptureCalls(unittest.TestCase):
         )
         self.assertEqual(patch_capture.call_count, 1)
         patch_capture.assert_called_with(
-            "some-distinct-id",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id",
+            properties={
                 "$feature_flag": "decide-flag",
                 "$feature_flag_response": "decide-variant",
                 "locally_evaluated": False,
@@ -2871,9 +2942,9 @@ class TestCaptureCalls(unittest.TestCase):
         )
         self.assertEqual(patch_capture.call_count, 1)
         patch_capture.assert_called_with(
-            "some-distinct-id",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id",
+            properties={
                 "$feature_flag": "decide-flag-with-payload",
                 "$feature_flag_response": True,
                 "locally_evaluated": False,
@@ -2948,7 +3019,9 @@ class TestCaptureCalls(unittest.TestCase):
             "featureFlags": {"person-flag": True},
             "featureFlagPayloads": {"person-flag": 300},
         }
-        client = Client(api_key=FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+        client = Client(
+            project_api_key=FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY
+        )
 
         client.feature_flags = [
             {
@@ -2977,9 +3050,9 @@ class TestCaptureCalls(unittest.TestCase):
         # Assert that capture was called once, with the correct parameters
         self.assertEqual(patch_capture.call_count, 1)
         patch_capture.assert_called_with(
-            "some-distinct-id",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id",
+            properties={
                 "$feature_flag": "person-flag",
                 "$feature_flag_response": True,
                 "locally_evaluated": True,
@@ -3012,9 +3085,9 @@ class TestCaptureCalls(unittest.TestCase):
 
         self.assertEqual(patch_capture.call_count, 1)
         patch_capture.assert_called_with(
-            "some-distinct-id2",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id2",
+            properties={
                 "$feature_flag": "person-flag",
                 "$feature_flag_response": True,
                 "locally_evaluated": True,
@@ -3058,9 +3131,9 @@ class TestCaptureCalls(unittest.TestCase):
         )
 
         patch_capture.assert_called_with(
-            "some-distinct-id",
             "$feature_flag_called",
-            {
+            distinct_id="some-distinct-id",
+            properties={
                 "$feature_flag": "complex-flag",
                 "$feature_flag_response": True,
                 "locally_evaluated": True,
@@ -3102,9 +3175,9 @@ class TestCaptureCalls(unittest.TestCase):
                 person_properties={"region": "USA", "name": "Aloha"},
             )
             patch_capture.assert_called_with(
-                distinct_id,
                 "$feature_flag_called",
-                {
+                distinct_id=distinct_id,
+                properties={
                     "$feature_flag": "complex-flag",
                     "$feature_flag_response": True,
                     "locally_evaluated": True,
@@ -5229,7 +5302,9 @@ class TestConsistency(unittest.TestCase):
             "featureFlags": {}
         }  # Ensure decide returns empty flags
 
-        client = Client(api_key=FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+        client = Client(
+            project_api_key=FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY
+        )
         client.feature_flags = [
             {
                 "id": 1,
@@ -5253,7 +5328,9 @@ class TestConsistency(unittest.TestCase):
             "featureFlagPayloads": {"Beta-Feature": {"some": "value"}},
         }
 
-        client = Client(api_key=FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+        client = Client(
+            project_api_key=FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY
+        )
         client.feature_flags = [
             {
                 "id": 1,
@@ -5282,7 +5359,9 @@ class TestConsistency(unittest.TestCase):
             "featureFlagPayloads": {"Beta-Feature": {"some": "value"}},
         }
 
-        client = Client(api_key=FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+        client = Client(
+            project_api_key=FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY
+        )
         client.feature_flags = [
             {
                 "id": 1,

posthog/test/test_module.py

@@ -7,8 +7,7 @@ class TestModule(unittest.TestCase):
     posthog = None
 
     def _assert_enqueue_result(self, result):
-        self.assertEqual(type(result[0]), bool)
-        self.assertEqual(type(result[1]), dict)
+        self.assertEqual(type(result[0]), str)
 
     def failed(self):
         self.failed = True
@@ -28,12 +27,7 @@ class TestModule(unittest.TestCase):
         self.assertRaises(Exception, self.posthog.capture)
 
     def test_track(self):
-        res = self.posthog.capture("distinct_id", "python module event")
-        self._assert_enqueue_result(res)
-        self.posthog.flush()
-
-    def test_identify(self):
-        res = self.posthog.identify("distinct_id", {"email": "user@email.com"})
+        res = self.posthog.capture("python module event", distinct_id="distinct_id")
         self._assert_enqueue_result(res)
         self.posthog.flush()
 
@@ -42,9 +36,5 @@ class TestModule(unittest.TestCase):
         self._assert_enqueue_result(res)
         self.posthog.flush()
 
-    def test_page(self):
-        self.posthog.page("distinct_id", "https://posthog.com/contact")
-        self.posthog.flush()
-
     def test_flush(self):
         self.posthog.flush()

posthog/test/test_utils.py

@@ -1,3 +1,4 @@
+import time
 import unittest
 from dataclasses import dataclass
 from datetime import date, datetime, timedelta
@@ -12,6 +13,7 @@ from pydantic import BaseModel
 from pydantic.v1 import BaseModel as BaseModelV1
 
 from posthog import utils
+from posthog.types import FeatureFlagResult
 
 TEST_API_KEY = "kOOlRy2QlMY9jHZQv0bKz0FZyazBUoY8Arj0lFVNjs4"
 FAKE_TEST_API_KEY = "random_key"
@@ -173,3 +175,124 @@ class TestUtils(unittest.TestCase):
                 "inner_optional": None,
             },
         }
+
+
+class TestFlagCache(unittest.TestCase):
+    def setUp(self):
+        self.cache = utils.FlagCache(max_size=3, default_ttl=1)
+        self.flag_result = FeatureFlagResult.from_value_and_payload(
+            "test-flag", True, None
+        )
+
+    def test_cache_basic_operations(self):
+        distinct_id = "user123"
+        flag_key = "test-flag"
+        flag_version = 1
+
+        # Test cache miss
+        result = self.cache.get_cached_flag(distinct_id, flag_key, flag_version)
+        assert result is None
+
+        # Test cache set and hit
+        self.cache.set_cached_flag(
+            distinct_id, flag_key, self.flag_result, flag_version
+        )
+        result = self.cache.get_cached_flag(distinct_id, flag_key, flag_version)
+        assert result is not None
+        assert result.get_value()
+
+    def test_cache_ttl_expiration(self):
+        distinct_id = "user123"
+        flag_key = "test-flag"
+        flag_version = 1
+
+        # Set flag in cache
+        self.cache.set_cached_flag(
+            distinct_id, flag_key, self.flag_result, flag_version
+        )
+
+        # Should be available immediately
+        result = self.cache.get_cached_flag(distinct_id, flag_key, flag_version)
+        assert result is not None
+
+        # Wait for TTL to expire (1 second + buffer)
+        time.sleep(1.1)
+
+        # Should be expired
+        result = self.cache.get_cached_flag(distinct_id, flag_key, flag_version)
+        assert result is None
+
+    def test_cache_version_invalidation(self):
+        distinct_id = "user123"
+        flag_key = "test-flag"
+        old_version = 1
+        new_version = 2
+
+        # Set flag with old version
+        self.cache.set_cached_flag(distinct_id, flag_key, self.flag_result, old_version)
+
+        # Should hit with old version
+        result = self.cache.get_cached_flag(distinct_id, flag_key, old_version)
+        assert result is not None
+
+        # Should miss with new version
+        result = self.cache.get_cached_flag(distinct_id, flag_key, new_version)
+        assert result is None
+
+        # Invalidate old version
+        self.cache.invalidate_version(old_version)
+
+        # Should miss even with old version after invalidation
+        result = self.cache.get_cached_flag(distinct_id, flag_key, old_version)
+        assert result is None
+
+    def test_stale_cache_functionality(self):
+        distinct_id = "user123"
+        flag_key = "test-flag"
+        flag_version = 1
+
+        # Set flag in cache
+        self.cache.set_cached_flag(
+            distinct_id, flag_key, self.flag_result, flag_version
+        )
+
+        # Wait for TTL to expire
+        time.sleep(1.1)
+
+        # Should not get fresh cache
+        result = self.cache.get_cached_flag(distinct_id, flag_key, flag_version)
+        assert result is None
+
+        # Should get stale cache (within 1 hour default)
+        stale_result = self.cache.get_stale_cached_flag(distinct_id, flag_key)
+        assert stale_result is not None
+        assert stale_result.get_value()
+
+    def test_lru_eviction(self):
+        # Cache has max_size=3, so adding 4 users should evict the LRU one
+        flag_version = 1
+
+        # Add 3 users
+        for i in range(3):
+            user_id = f"user{i}"
+            self.cache.set_cached_flag(
+                user_id, "test-flag", self.flag_result, flag_version
+            )
+
+        # Access user0 to make it recently used
+        self.cache.get_cached_flag("user0", "test-flag", flag_version)
+
+        # Add 4th user, should evict user1 (least recently used)
+        self.cache.set_cached_flag("user3", "test-flag", self.flag_result, flag_version)
+
+        # user0 should still be there (was recently accessed)
+        result = self.cache.get_cached_flag("user0", "test-flag", flag_version)
+        assert result is not None
+
+        # user2 should still be there (was recently added)
+        result = self.cache.get_cached_flag("user2", "test-flag", flag_version)
+        assert result is not None
+
+        # user3 should be there (just added)
+        result = self.cache.get_cached_flag("user3", "test-flag", flag_version)
+        assert result is not None

posthog/types.py

@@ -9,6 +9,24 @@ FlagValue = Union[bool, str]
 BeforeSendCallback = Callable[[dict[str, Any]], Optional[dict[str, Any]]]
 
 
+class SendFeatureFlagsOptions(TypedDict, total=False):
+    """Options for sending feature flags with capture events.
+
+    Args:
+        only_evaluate_locally: Whether to only use local evaluation for feature flags.
+            If True, only flags that can be evaluated locally will be included.
+            If False, remote evaluation via /flags API will be used when needed.
+        person_properties: Properties to use for feature flag evaluation specific to this event.
+            These properties will be merged with any existing person properties.
+        group_properties: Group properties to use for feature flag evaluation specific to this event.
+            Format: { group_type_name: { group_properties } }
+    """
+
+    only_evaluate_locally: Optional[bool]
+    person_properties: Optional[dict[str, Any]]
+    group_properties: Optional[dict[str, dict[str, Any]]]
+
+
 @dataclass(frozen=True)
 class FlagReason:
     code: str
@@ -92,7 +110,7 @@ class FeatureFlag:
             variant=variant,
             reason=None,
             metadata=LegacyFlagMetadata(
-                payload=payload if payload else None,
+                payload=payload,
             ),
         )
 
@@ -160,7 +178,9 @@ class FeatureFlagResult:
             key=key,
             enabled=enabled,
             variant=variant,
-            payload=json.loads(payload) if isinstance(payload, str) else payload,
+            payload=json.loads(payload)
+            if isinstance(payload, str) and payload
+            else payload,
             reason=None,
         )
 
@@ -201,6 +221,7 @@ class FeatureFlagResult:
             payload=(
                 json.loads(details.metadata.payload)
                 if isinstance(details.metadata.payload, str)
+                and details.metadata.payload
                 else details.metadata.payload
             ),
             reason=details.reason.description if details.reason else None,
@@ -278,5 +299,7 @@ def to_payloads(response: FlagsResponse) -> Optional[dict[str, str]]:
     return {
         key: value.metadata.payload
         for key, value in response.get("flags", {}).items()
-        if isinstance(value, FeatureFlag) and value.enabled and value.metadata.payload
+        if isinstance(value, FeatureFlag)
+        and value.enabled
+        and value.metadata.payload is not None
     }

posthog/utils.py

@@ -1,12 +1,17 @@
+import json
 import logging
 import numbers
 import re
+import time
 from collections import defaultdict
 from dataclasses import asdict, is_dataclass
 from datetime import date, datetime, timezone
 from decimal import Decimal
 from typing import Any, Optional
 from uuid import UUID
+import sys
+import platform
+import distro  # For Linux OS detection
 
 import six
 from dateutil.tz import tzlocal, tzutc
@@ -154,6 +159,266 @@ class SizeLimitedDict(defaultdict):
         super().__setitem__(key, value)
 
 
+class FlagCacheEntry:
+    def __init__(self, flag_result, flag_definition_version, timestamp=None):
+        self.flag_result = flag_result
+        self.flag_definition_version = flag_definition_version
+        self.timestamp = timestamp or time.time()
+
+    def is_valid(self, current_time, ttl, current_flag_version):
+        time_valid = (current_time - self.timestamp) < ttl
+        version_valid = self.flag_definition_version == current_flag_version
+        return time_valid and version_valid
+
+    def is_stale_but_usable(self, current_time, max_stale_age=3600):
+        return (current_time - self.timestamp) < max_stale_age
+
+
+class FlagCache:
+    def __init__(self, max_size=10000, default_ttl=300):
+        self.cache = {}  # distinct_id -> {flag_key: FlagCacheEntry}
+        self.access_times = {}  # distinct_id -> last_access_time
+        self.max_size = max_size
+        self.default_ttl = default_ttl
+
+    def get_cached_flag(self, distinct_id, flag_key, current_flag_version):
+        current_time = time.time()
+
+        if distinct_id not in self.cache:
+            return None
+
+        user_flags = self.cache[distinct_id]
+        if flag_key not in user_flags:
+            return None
+
+        entry = user_flags[flag_key]
+        if entry.is_valid(current_time, self.default_ttl, current_flag_version):
+            self.access_times[distinct_id] = current_time
+            return entry.flag_result
+
+        return None
+
+    def get_stale_cached_flag(self, distinct_id, flag_key, max_stale_age=3600):
+        current_time = time.time()
+
+        if distinct_id not in self.cache:
+            return None
+
+        user_flags = self.cache[distinct_id]
+        if flag_key not in user_flags:
+            return None
+
+        entry = user_flags[flag_key]
+        if entry.is_stale_but_usable(current_time, max_stale_age):
+            return entry.flag_result
+
+        return None
+
+    def set_cached_flag(
+        self, distinct_id, flag_key, flag_result, flag_definition_version
+    ):
+        current_time = time.time()
+
+        # Evict LRU users if we're at capacity
+        if distinct_id not in self.cache and len(self.cache) >= self.max_size:
+            self._evict_lru()
+
+        # Initialize user cache if needed
+        if distinct_id not in self.cache:
+            self.cache[distinct_id] = {}
+
+        # Store the flag result
+        self.cache[distinct_id][flag_key] = FlagCacheEntry(
+            flag_result, flag_definition_version, current_time
+        )
+        self.access_times[distinct_id] = current_time
+
+    def invalidate_version(self, old_version):
+        users_to_remove = []
+
+        for distinct_id, user_flags in self.cache.items():
+            flags_to_remove = []
+            for flag_key, entry in user_flags.items():
+                if entry.flag_definition_version == old_version:
+                    flags_to_remove.append(flag_key)
+
+            # Remove invalidated flags
+            for flag_key in flags_to_remove:
+                del user_flags[flag_key]
+
+            # Remove user entirely if no flags remain
+            if not user_flags:
+                users_to_remove.append(distinct_id)
+
+        # Clean up empty users
+        for distinct_id in users_to_remove:
+            del self.cache[distinct_id]
+            if distinct_id in self.access_times:
+                del self.access_times[distinct_id]
+
+    def _evict_lru(self):
+        if not self.access_times:
+            return
+
+        # Remove 20% of least recently used entries
+        sorted_users = sorted(self.access_times.items(), key=lambda x: x[1])
+        to_remove = max(1, len(sorted_users) // 5)
+
+        for distinct_id, _ in sorted_users[:to_remove]:
+            if distinct_id in self.cache:
+                del self.cache[distinct_id]
+            if distinct_id in self.access_times:
+                del self.access_times[distinct_id]
+
+    def clear(self):
+        self.cache.clear()
+        self.access_times.clear()
+
+
+class RedisFlagCache:
+    def __init__(
+        self, redis_client, default_ttl=300, stale_ttl=3600, key_prefix="posthog:flags:"
+    ):
+        self.redis = redis_client
+        self.default_ttl = default_ttl
+        self.stale_ttl = stale_ttl
+        self.key_prefix = key_prefix
+        self.version_key = f"{key_prefix}version"
+
+    def _get_cache_key(self, distinct_id, flag_key):
+        return f"{self.key_prefix}{distinct_id}:{flag_key}"
+
+    def _serialize_entry(self, flag_result, flag_definition_version, timestamp=None):
+        if timestamp is None:
+            timestamp = time.time()
+
+        # Use clean to make flag_result JSON-serializable for cross-platform compatibility
+        serialized_result = clean(flag_result)
+
+        entry = {
+            "flag_result": serialized_result,
+            "flag_version": flag_definition_version,
+            "timestamp": timestamp,
+        }
+        return json.dumps(entry)
+
+    def _deserialize_entry(self, data):
+        try:
+            entry = json.loads(data)
+            flag_result = entry["flag_result"]
+            return FlagCacheEntry(
+                flag_result=flag_result,
+                flag_definition_version=entry["flag_version"],
+                timestamp=entry["timestamp"],
+            )
+        except (json.JSONDecodeError, KeyError, ValueError):
+            # If deserialization fails, treat as cache miss
+            return None
+
+    def get_cached_flag(self, distinct_id, flag_key, current_flag_version):
+        try:
+            cache_key = self._get_cache_key(distinct_id, flag_key)
+            data = self.redis.get(cache_key)
+
+            if data:
+                entry = self._deserialize_entry(data)
+                if entry and entry.is_valid(
+                    time.time(), self.default_ttl, current_flag_version
+                ):
+                    return entry.flag_result
+
+            return None
+        except Exception:
+            # Redis error - return None to fall back to normal evaluation
+            return None
+
+    def get_stale_cached_flag(self, distinct_id, flag_key, max_stale_age=None):
+        try:
+            if max_stale_age is None:
+                max_stale_age = self.stale_ttl
+
+            cache_key = self._get_cache_key(distinct_id, flag_key)
+            data = self.redis.get(cache_key)
+
+            if data:
+                entry = self._deserialize_entry(data)
+                if entry and entry.is_stale_but_usable(time.time(), max_stale_age):
+                    return entry.flag_result
+
+            return None
+        except Exception:
+            # Redis error - return None
+            return None
+
+    def set_cached_flag(
+        self, distinct_id, flag_key, flag_result, flag_definition_version
+    ):
+        try:
+            cache_key = self._get_cache_key(distinct_id, flag_key)
+            serialized_entry = self._serialize_entry(
+                flag_result, flag_definition_version
+            )
+
+            # Set with TTL for automatic cleanup (use stale_ttl for total lifetime)
+            self.redis.setex(cache_key, self.stale_ttl, serialized_entry)
+
+            # Update the current version
+            self.redis.set(self.version_key, flag_definition_version)
+
+        except Exception:
+            # Redis error - silently fail, don't break flag evaluation
+            pass
+
+    def invalidate_version(self, old_version):
+        try:
+            # For Redis, we use a simple approach: scan for keys with old version
+            # and delete them. This could be expensive with many keys, but it's
+            # necessary for correctness.
+
+            cursor = 0
+            pattern = f"{self.key_prefix}*"
+
+            while True:
+                cursor, keys = self.redis.scan(cursor, match=pattern, count=100)
+
+                for key in keys:
+                    if key.decode() == self.version_key:
+                        continue
+
+                    try:
+                        data = self.redis.get(key)
+                        if data:
+                            entry_dict = json.loads(data)
+                            if entry_dict.get("flag_version") == old_version:
+                                self.redis.delete(key)
+                    except (json.JSONDecodeError, KeyError):
+                        # If we can't parse the entry, delete it to be safe
+                        self.redis.delete(key)
+
+                if cursor == 0:
+                    break
+
+        except Exception:
+            # Redis error - silently fail
+            pass
+
+    def clear(self):
+        try:
+            # Delete all keys matching our pattern
+            cursor = 0
+            pattern = f"{self.key_prefix}*"
+
+            while True:
+                cursor, keys = self.redis.scan(cursor, match=pattern, count=100)
+                if keys:
+                    self.redis.delete(*keys)
+                if cursor == 0:
+                    break
+        except Exception:
+            # Redis error - silently fail
+            pass
+
+
 def convert_to_datetime_aware(date_obj):
     if date_obj.tzinfo is None:
         date_obj = date_obj.replace(tzinfo=timezone.utc)
@@ -198,3 +463,57 @@ def str_iequals(value, comparand):
         False
     """
     return str(value).casefold() == str(comparand).casefold()
+
+
+def get_os_info():
+    """
+    Returns standardized OS name and version information.
+    Similar to how user agent parsing works in JS.
+    """
+    os_name = ""
+    os_version = ""
+
+    platform_name = sys.platform
+
+    if platform_name.startswith("win"):
+        os_name = "Windows"
+        if hasattr(platform, "win32_ver"):
+            win_version = platform.win32_ver()[0]
+            if win_version:
+                os_version = win_version
+
+    elif platform_name == "darwin":
+        os_name = "Mac OS X"
+        if hasattr(platform, "mac_ver"):
+            mac_version = platform.mac_ver()[0]
+            if mac_version:
+                os_version = mac_version
+
+    elif platform_name.startswith("linux"):
+        os_name = "Linux"
+        linux_info = distro.info()
+        if linux_info["version"]:
+            os_version = linux_info["version"]
+
+    elif platform_name.startswith("freebsd"):
+        os_name = "FreeBSD"
+        if hasattr(platform, "release"):
+            os_version = platform.release()
+
+    else:
+        os_name = platform_name
+        if hasattr(platform, "release"):
+            os_version = platform.release()
+
+    return os_name, os_version
+
+
+def system_context() -> dict[str, Any]:
+    os_name, os_version = get_os_info()
+
+    return {
+        "$python_runtime": platform.python_implementation(),
+        "$python_version": "%s.%s.%s" % (sys.version_info[:3]),
+        "$os": os_name,
+        "$os_version": os_version,
+    }

posthog/version.py

@@ -1,4 +1,4 @@
-VERSION = "5.3.0"
+VERSION = "6.4.0"
 
 if __name__ == "__main__":
     print(VERSION, end="")  # noqa: T201

pyproject.toml

@@ -29,6 +29,7 @@ dependencies = [
     "python-dateutil>=2.2",
     "backoff>=1.10.0",
     "distro>=1.5.0",
+    "typing-extensions>=4.2.0",
 ]
 
 [project.urls]
@@ -86,7 +87,6 @@ packages = [
     "posthog.ai.anthropic",
     "posthog.ai.gemini",
     "posthog.test",
-    "posthog.exception_integrations",
     "posthog.integrations",
 ]
 

simulator.py

@@ -1,112 +0,0 @@
-import argparse
-import json
-import logging
-
-import posthog
-
-__name__ = "simulator.py"
-__version__ = "0.0.1"
-__description__ = "scripting simulator"
-
-
-def json_hash(str):
-    if str:
-        return json.loads(str)
-
-
-# posthog -method=<method> -posthog-write-key=<posthogWriteKey> [options]
-
-
-parser = argparse.ArgumentParser(description="send a posthog message")
-
-parser.add_argument("--writeKey", help="the posthog writeKey")
-parser.add_argument("--type", help="The posthog message type")
-
-parser.add_argument("--distinct_id", help="the user id to send the event as")
-parser.add_argument("--anonymousId", help="the anonymous user id to send the event as")
-
-parser.add_argument("--event", help="the event name to send with the event")
-parser.add_argument("--properties", help="the event properties to send (JSON-encoded)")
-
-parser.add_argument(
-    "--name", help="name of the screen or page to send with the message"
-)
-
-parser.add_argument("--traits", help="the identify/group traits to send (JSON-encoded)")
-
-parser.add_argument("--groupId", help="the group id")
-
-options = parser.parse_args()
-
-
-def failed(status, msg):
-    raise Exception(msg)
-
-
-def capture():
-    posthog.capture(
-        options.distinct_id,
-        options.event,
-        anonymous_id=options.anonymousId,
-        properties=json_hash(options.properties),
-    )
-
-
-def page():
-    posthog.page(
-        options.distinct_id,
-        name=options.name,
-        anonymous_id=options.anonymousId,
-        properties=json_hash(options.properties),
-    )
-
-
-def identify():
-    posthog.identify(
-        options.distinct_id,
-        anonymous_id=options.anonymousId,
-        traits=json_hash(options.traits),
-    )
-
-
-def set_once():
-    posthog.set_once(
-        options.distinct_id,
-        properties=json_hash(options.traits),
-    )
-
-
-def set():
-    posthog.set(
-        options.distinct_id,
-        properties=json_hash(options.traits),
-    )
-
-
-def unknown():
-    print()
-
-
-posthog.api_key = options.writeKey
-posthog.on_error = failed
-posthog.debug = True
-
-log = logging.getLogger("posthog")
-ch = logging.StreamHandler()
-ch.setLevel(logging.DEBUG)
-log.addHandler(ch)
-
-switcher = {
-    "capture": capture,
-    "page": page,
-    "identify": identify,
-    "set_once": set_once,
-    "set": set,
-}
-
-func = switcher.get(options.type)
-if func:
-    func()
-    posthog.shutdown()
-else:
-    print("Invalid Message Type " + options.type)