chore: Release v7.9.2

chore: Migrate releases to sampo (#398 )
* chore: Migrate releases to `sampo` This is much closer to what we have in `posthog-js`, let's see if it's a good thing! There's still a lot to do before deploying this: - updating CI - updating README with instructions * Add sampo changeset * chore: Update to relase Python via Slack + sampo * Update release.yml * fix: Use pyproject.toml version as source of truth
2026-02-18 22:05:00 +00:00 · 2026-02-18 19:02:19 -03:00 · 2026-02-17 16:58:59 +00:00 · 2026-02-17 10:32:05 +01:00 · 2026-02-16 11:05:09 +01:00 · 2026-02-13 15:53:22 +01:00
37 changed files with 3192 additions and 212 deletions
@@ -1,58 +1,249 @@
 name: "Release"

 on:
-  push:
-    branches:
-      - master
-    paths:
-      - "posthog/version.py"
+  pull_request:
+    types: [closed]
+    branches: [master]
  workflow_dispatch:

+permissions:
+  contents: read
+
+# Concurrency control: only one release process can run at a time
+# This prevents race conditions if multiple PRs with 'release' label merge simultaneously
+concurrency:
+  group: release
+  cancel-in-progress: false
+
 jobs:
-  release:
-    name: Publish release
+  check-release-label:
+    name: Check for release label
    runs-on: ubuntu-latest
-    permissions:
-      contents: write
-      id-token: write
+    # Run when PR with 'release' label is merged to master
+    if: |
+      github.event_name == 'workflow_dispatch' ||
+       (github.event_name == 'pull_request' &&
+        github.event.pull_request.merged == true &&
+        contains(github.event.pull_request.labels.*.name, 'release'))
+    outputs:
+      should-release: ${{ steps.check.outputs.should-release }}
    steps:
-      - name: Checkout the repository
-        uses: actions/checkout@85e6279cec87321a52edac9c87bce653a07cf6c2
+      - name: Checkout repository
+        uses: actions/checkout@v4
        with:
+          ref: master
          fetch-depth: 0

+      - name: Check release conditions
+        id: check
+        run: |
+          changeset_count=$(find .sampo/changesets -name '*.md' 2>/dev/null | wc -l)
+          if [ "$changeset_count" -gt 0 ]; then
+            echo "should-release=true" >> "$GITHUB_OUTPUT"
+            echo "Found $changeset_count changeset(s), ready to release"
+          else
+            echo "should-release=false" >> "$GITHUB_OUTPUT"
+            echo "No changesets to release"
+          fi
+
+  notify-approval-needed:
+    name: Notify Slack - Approval Needed
+    needs: check-release-label
+    if: needs.check-release-label.outputs.should-release == 'true'
+    uses: posthog/.github/.github/workflows/notify-approval-needed.yml@main
+    with:
+      slack_channel_id: ${{ vars.SLACK_APPROVALS_CLIENT_LIBRARIES_CHANNEL_ID }}
+      slack_user_group_id: ${{ vars.GROUP_CLIENT_LIBRARIES_SLACK_GROUP_ID }}
+    secrets:
+      slack_bot_token: ${{ secrets.SLACK_CLIENT_LIBRARIES_BOT_TOKEN }}
+      posthog_project_api_key: ${{ secrets.POSTHOG_PROJECT_API_KEY }}
+
+  release:
+    name: Release and publish
+    needs: [check-release-label, notify-approval-needed]
+    runs-on: ubuntu-latest
+    # Use `always()` to ensure the job runs even if notify-approval-needed is skipped,
+    # but still depend on it to access `needs.notify-approval-needed.outputs.slack_ts`
+    if: always() && needs.check-release-label.outputs.should-release == 'true'
+    environment: "Release" # This will require an approval from a maintainer, they are notified in Slack above
+    permissions:
+      contents: write
+      actions: write
+      id-token: write
+    steps:
+      - name: Notify Slack - Approved
+        if: needs.notify-approval-needed.outputs.slack_ts != ''
+        uses: posthog/.github/.github/actions/slack-thread-reply@main
+        with:
+          slack_bot_token: ${{ secrets.SLACK_CLIENT_LIBRARIES_BOT_TOKEN }}
+          slack_channel_id: ${{ vars.SLACK_APPROVALS_CLIENT_LIBRARIES_CHANNEL_ID }}
+          thread_ts: ${{ needs.notify-approval-needed.outputs.slack_ts }}
+          message: "✅ Release approved! Version bump in progress..."
+          emoji_reaction: "white_check_mark"
+
+      - name: Get GitHub App token
+        id: releaser
+        uses: actions/create-github-app-token@v2
+        with:
+          app-id: ${{ secrets.GH_APP_POSTHOG_PYTHON_RELEASER_APP_ID }}
+          private-key: ${{ secrets.GH_APP_POSTHOG_PYTHON_RELEASER_PRIVATE_KEY }}
+
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          ref: master
+          fetch-depth: 0
+          token: ${{ steps.releaser.outputs.token }}
+
      - name: Set up Python
-        uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55
+        uses: actions/setup-python@v5
        with:
          python-version: 3.11.11

      - name: Install uv
-        uses: astral-sh/setup-uv@0c5e2b8115b80b4c7c5ddf6ffdd634974642d182 # v5.4.1
+        uses: astral-sh/setup-uv@v5
        with:
-            enable-cache: true
-            pyproject-file: 'pyproject.toml'
-            
-      - name: Detect version
-        run: echo "REPO_VERSION=$(python3 posthog/version.py)" >> $GITHUB_ENV
+          enable-cache: true
+          pyproject-file: "pyproject.toml"

-      - name: Prepare for building release
+      - name: Install Rust
+        uses: dtolnay/rust-toolchain@0b1efabc08b657293548b77fb76cc02d26091c7e
+        with:
+          toolchain: 1.91.1
+          components: cargo
+
+      - name: Cache Sampo CLI
+        id: cache-sampo
+        uses: actions/cache@v3
+        with:
+          path: ~/.cargo/bin/sampo
+          key: sampo-${{ runner.os }}-${{ runner.arch }}
+
+      - name: Install Sampo CLI
+        if: steps.cache-sampo.outputs.cache-hit != 'true'
+        run: cargo install sampo
+
+      - name: Install dependencies
        run: uv sync --extra dev

-      - name: Push releases to PyPI
-        env:
-          TWINE_USERNAME: __token__
-        run: uv run make release && uv run make release_analytics
+      - name: Configure Git
+        run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"

-      - name: Create GitHub release
+      - name: Prepare release with Sampo
+        id: sampo-release
+        env:
+          GITHUB_TOKEN: ${{ steps.releaser.outputs.token }}
+        run: |
+          sampo release
+          new_version=$(python3 -c "import tomllib; print(tomllib.load(open('pyproject.toml', 'rb'))['project']['version'])")
+          echo "new_version=$new_version" >> "$GITHUB_OUTPUT"
+
+      - name: Sync version to posthog/version.py
+        run: |
+          echo 'VERSION = "${{ steps.sampo-release.outputs.new_version }}"' > posthog/version.py
+
+      - name: Commit release changes
+        id: commit-release
+        env:
+          GITHUB_TOKEN: ${{ steps.releaser.outputs.token }}
+        run: |
+          git add -A
+          if git diff --staged --quiet; then
+            echo "No changes to commit"
+            echo "committed=false" >> "$GITHUB_OUTPUT"
+          else
+            git commit -m "chore: Release v${{ steps.sampo-release.outputs.new_version }}"
+            git push origin master
+            echo "committed=true" >> "$GITHUB_OUTPUT"
+          fi
+
+      # Publishing is done manually (not via `sampo publish`) because we need to
+      # publish both `posthog` and `posthoganalytics` packages to PyPI.
+      # Sampo only knows about the `posthog` package, so we handle both here.
+      # Both packages use PyPI OIDC trusted publishing (no API tokens needed).
+      - name: Build posthog
+        if: steps.commit-release.outputs.committed == 'true'
+        run: uv run make build_release
+
+      - name: Publish posthog to PyPI
+        if: steps.commit-release.outputs.committed == 'true'
+        uses: pypa/gh-action-pypi-publish@release/v1
+
+      # The `posthoganalytics` package is a mirror of `posthog` published under
+      # a different name for backwards compatibility. The make target handles
+      # copying, renaming imports, and building the dist automatically.
+      - name: Build posthoganalytics
+        if: steps.commit-release.outputs.committed == 'true'
+        run: uv run make build_release_analytics
+
+      - name: Publish posthoganalytics to PyPI
+        if: steps.commit-release.outputs.committed == 'true'
+        uses: pypa/gh-action-pypi-publish@release/v1
+
+      # We skip `sampo publish` (which normally creates the tag) because we
+      # need to publish both posthog and posthoganalytics manually, so we
+      # create the tag ourselves.
+      - name: Tag release
+        if: steps.commit-release.outputs.committed == 'true'
+        run: git tag "v${{ steps.sampo-release.outputs.new_version }}"
+
+      - name: Push tags
+        if: steps.commit-release.outputs.committed == 'true'
+        run: git push origin --tags
+
+      - name: Create GitHub Release
+        if: steps.commit-release.outputs.committed == 'true'
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: gh release create "v${{ steps.sampo-release.outputs.new_version }}" --generate-notes
+
+      - name: Dispatch generate-references
+        if: steps.commit-release.outputs.committed == 'true'
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          gh release create "v${{ env.REPO_VERSION }}" \
-            --title "${{ env.REPO_VERSION }}" \
-            --generate-notes
-      
-      - name: Dispatch generate-references for posthog-python
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          gh workflow run generate-references.yml --ref master
+        run: gh workflow run generate-references.yml --ref master
+
+      # Notify in case of a failure
+      - name: Send failure event to PostHog
+        if: ${{ failure() }}
+        uses: PostHog/posthog-github-action@v0.1
+        with:
+          posthog-token: "${{ secrets.POSTHOG_PROJECT_API_KEY }}"
+          event: "posthog-python-github-release-workflow-failure"
+          properties: >-
+            {
+              "commitSha": "${{ github.sha }}",
+              "jobStatus": "${{ job.status }}",
+              "ref": "${{ github.ref }}",
+              "version": "v${{ steps.sampo-release.outputs.new_version }}"
+            }
+
+      - name: Notify Slack - Failed
+        if: ${{ failure() && needs.notify-approval-needed.outputs.slack_ts != '' }}
+        uses: posthog/.github/.github/actions/slack-thread-reply@main
+        with:
+          slack_bot_token: ${{ secrets.SLACK_CLIENT_LIBRARIES_BOT_TOKEN }}
+          slack_channel_id: ${{ vars.SLACK_APPROVALS_CLIENT_LIBRARIES_CHANNEL_ID }}
+          thread_ts: ${{ needs.notify-approval-needed.outputs.slack_ts }}
+          message: "❌ Failed to release `posthog-python@v${{ steps.sampo-release.outputs.new_version }}`! <https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }}|View logs>"
+          emoji_reaction: "x"
+
+  notify-released:
+    name: Notify Slack - Released
+    needs: [check-release-label, notify-approval-needed, release]
+    runs-on: ubuntu-latest
+    if: always() && needs.release.result == 'success' && needs.notify-approval-needed.outputs.slack_ts != ''
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Notify Slack - Released
+        uses: posthog/.github/.github/actions/slack-thread-reply@main
+        with:
+          slack_bot_token: ${{ secrets.SLACK_CLIENT_LIBRARIES_BOT_TOKEN }}
+          slack_channel_id: ${{ vars.SLACK_APPROVALS_CLIENT_LIBRARIES_CHANNEL_ID }}
+          thread_ts: ${{ needs.notify-approval-needed.outputs.slack_ts }}
+          message: "🚀 posthog-python released successfully!"
+          emoji_reaction: "rocket"
@@ -0,0 +1,21 @@
+name: SDK Compliance Tests
+
+permissions:
+  contents: read
+  packages: read
+  pull-requests: write
+
+on:
+  pull_request:
+  push:
+    branches:
+      - master
+
+jobs:
+  compliance:
+    name: PostHog SDK compliance tests
+    uses: PostHog/posthog-sdk-test-harness/.github/workflows/test-sdk-action.yml@main
+    with:
+      adapter-dockerfile: "sdk_compliance_adapter/Dockerfile"
+      adapter-context: "."
+      test-harness-version: "latest"
@@ -0,0 +1,19 @@
+# Sampo configuration
+version = 1
+
+[git]
+default_branch = "master"
+short_tags = "posthog"    # Tag with v1.2.3 rather than posthog-v1.2.3
+
+[github]
+repository = "posthog/posthog-python"
+
+[changelog]
+# Options for release notes generation.
+# show_commit_hash = true (default)
+# show_acknowledgments = true (default)
+
+[packages]
+# Options for package discovery and filtering.
+# ignore_unpublished = false (default)
+# ignore = ["internal-*", "examples/*"]
@@ -1,39 +1,85 @@
-# 7.7.0 - 2026-01-15
+# posthog
+
+## 7.9.2 — 2026-02-18
+
+### Patch changes
+
+- [f1dc4d7](https://github.com/posthog/posthog-python/commit/f1dc4d73914712983a7f715ee4fe1b70e66e770a) Add sampo to the project — Thanks @rafaeelaudibert!
+
+## 7.9.1 - 2026-02-17
+
+fix(llma): make prompt fetches deterministic by requiring project_api_key and sending it as token query param
+
+## 7.9.0 - 2026-02-17
+
+feat: Support device_id as bucketing identifier for local evaluation
+
+## 7.8.6 - 2026-02-09
+
+fix: limit collections scanning in code variables
+
+## 7.8.5 - 2026-02-09
+
+fix: further optimize code variables pattern matching
+
+## 7.8.4 - 2026-02-09
+
+fix: do not pattern match long values in code variables
+
+## 7.8.3 - 2026-02-06
+
+fix: openAI input image sanitization
+
+## 7.8.2 - 2026-02-04
+
+fix(llma): fix prompts default url
+
+## 7.8.1 - 2026-02-03
+
+fix(llma): small fixes for prompt management
+
+## 7.8.0 - 2026-01-28
+
+feat(llma): add prompt management
+
+Adds the Prompt Management feature. At the time of release, this feature is in a closed alpha.
+
+## 7.7.0 - 2026-01-15

 feat(ai): Add OpenAI Agents SDK integration

 Automatic tracing for agent workflows, handoffs, tool calls, guardrails, and custom spans. Includes `$ai_total_tokens`, `$ai_error_type` categorization, and `$ai_framework` property.

-# 7.6.0 - 2026-01-12
+## 7.6.0 - 2026-01-12

 feat: add device_id to flags request payload

 Add device_id parameter to all feature flag methods, allowing the server to track device identifiers for flag evaluation. The device_id can be passed explicitly or set via context using `set_context_device_id()`.

-# 7.5.1 - 2026-01-07
+## 7.5.1 - 2026-01-07

 fix: avoid return from finally block to fix Python 3.14 SyntaxWarning (#361) - thanks @jodal

-# 7.5.0 - 2026-01-06
+## 7.5.0 - 2026-01-06

 feat: Capture Langchain, OpenAI and Anthropic errors as exceptions (if exception autocapture is enabled)
 feat: Add reference to exception in LLMA trace and span events

-# 7.4.3 - 2026-01-02
+## 7.4.3 - 2026-01-02

 Fixes cache creation cost for Langchain with Anthropic

-# 7.4.2 - 2025-12-22
+## 7.4.2 - 2025-12-22

 feat: add `in_app_modules` option to control code variables capturing

-# 7.4.1 - 2025-12-19
+## 7.4.1 - 2025-12-19

 fix: extract model from response for OpenAI stored prompts

 When using OpenAI stored prompts, the model is defined in the OpenAI dashboard rather than passed in the API request. This fix adds a fallback to extract the model from the response object when not provided in kwargs, ensuring generations show up with the correct model and enabling cost calculations.

-# 7.4.0 - 2025-12-16
+## 7.4.0 - 2025-12-16

 feat: Add automatic retries for feature flag requests

@@ -45,32 +91,32 @@ Feature flag API requests now automatically retry on transient failures:

 Rate limit (429) and quota (402) errors are not retried.

-# 7.3.1 - 2025-12-06
+## 7.3.1 - 2025-12-06

 fix: remove unused $exception_message and $exception_type

-# 7.3.0 - 2025-12-05
+## 7.3.0 - 2025-12-05

 feat: improve code variables capture masking

-# 7.2.0 - 2025-12-01
+## 7.2.0 - 2025-12-01

 feat: add $feature_flag_evaluated_at properties to $feature_flag_called events

-# 7.1.0 - 2025-11-26
+## 7.1.0 - 2025-11-26

 Add support for the async version of Gemini.

-# 7.0.2 - 2025-11-18
+## 7.0.2 - 2025-11-18

 Add support for Python 3.14.
 Projects upgrading to Python 3.14 should ensure any Pydantic models passed into the SDK use Pydantic v2, as Pydantic v1 is not compatible with Python 3.14.

-# 7.0.1 - 2025-11-15
+## 7.0.1 - 2025-11-15

 Try to use repr() when formatting code variables

-# 7.0.0 - 2025-11-11
+## 7.0.0 - 2025-11-11

 NB Python 3.9 is no longer supported

@@ -84,155 +130,155 @@ NB Python 3.9 is no longer supported
  - langchain-community: 0.3.29 → 0.4.1
  - langgraph: 0.6.6 → 1.0.2

-# 6.9.3 - 2025-11-10
+## 6.9.3 - 2025-11-10

 - feat(ph-ai): PostHog properties dict in GenerationMetadata

-# 6.9.2 - 2025-11-10
+## 6.9.2 - 2025-11-10

 - fix(llma): fix cache token double subtraction in Langchain for non-Anthropic providers causing negative costs

-# 6.9.1 - 2025-11-07
+## 6.9.1 - 2025-11-07

 - fix(error-tracking): pass code variables config from init to client

-# 6.9.0 - 2025-11-06
+## 6.9.0 - 2025-11-06

 - feat(error-tracking): add local variables capture

-# 6.8.0 - 2025-11-03
+## 6.8.0 - 2025-11-03

 - feat(llma): send web search calls to be used for LLM cost calculations

-# 6.7.14 - 2025-11-03
+## 6.7.14 - 2025-11-03

 - fix(django): Handle request.user access in async middleware context to prevent SynchronousOnlyOperation errors in Django 5+ (fixes #355)
 - test(django): Add Django 5 integration test suite with real ASGI application testing async middleware behavior

-# 6.7.13 - 2025-11-02
+## 6.7.13 - 2025-11-02

 - fix(llma): cache cost calculation in the LangChain callback

-# 6.7.12 - 2025-11-02
+## 6.7.12 - 2025-11-02

 - fix(django): Restore process_exception method to capture view and downstream middleware exceptions (fixes #329)
 - fix(ai/langchain): Add LangChain 1.0+ compatibility for CallbackHandler imports (fixes #362)

-# 6.7.11 - 2025-10-28
+## 6.7.11 - 2025-10-28

 - feat(ai): Add `$ai_framework` property for framework integrations (e.g. LangChain)

-# 6.7.10 - 2025-10-24
+## 6.7.10 - 2025-10-24

 - fix(django): Make middleware truly hybrid - compatible with both sync (WSGI) and async (ASGI) Django stacks without breaking sync-only deployments

-# 6.7.9 - 2025-10-22
+## 6.7.9 - 2025-10-22

 - fix(flags): multi-condition flags with static cohorts returning wrong variants

-# 6.7.8 - 2025-10-16
+## 6.7.8 - 2025-10-16

 - fix(llma): missing async for OpenAI's streaming implementation

-# 6.7.7 - 2025-10-14
+## 6.7.7 - 2025-10-14

 - fix: remove deprecated attribute $exception_personURL from exception events

-# 6.7.6 - 2025-09-16
+## 6.7.6 - 2025-09-16

 - fix: don't sort condition sets with variant overrides to the top
 - fix: Prevent core Client methods from raising exceptions

-# 6.7.5 - 2025-09-16
+## 6.7.5 - 2025-09-16

 - feat: Django middleware now supports async request handling.

-# 6.7.4 - 2025-09-05
+## 6.7.4 - 2025-09-05

 - fix: Missing system prompts for some providers

-# 6.7.3 - 2025-09-04
+## 6.7.3 - 2025-09-04

 - fix: missing usage tokens in Gemini

-# 6.7.2 - 2025-09-03
+## 6.7.2 - 2025-09-03

 - fix: tool call results in streaming providers

-# 6.7.1 - 2025-09-01
+## 6.7.1 - 2025-09-01

 - fix: Add base64 inline image sanitization

-# 6.7.0 - 2025-08-26
+## 6.7.0 - 2025-08-26

 - feat: Add support for feature flag dependencies

-# 6.6.1 - 2025-08-21
+## 6.6.1 - 2025-08-21

 - fix: Prevent `NoneType` error when `group_properties` is `None`

-# 6.6.0 - 2025-08-15
+## 6.6.0 - 2025-08-15

 - feat: Add `flag_keys_to_evaluate` parameter to optimize feature flag evaluation performance by only evaluating specified flags
 - feat: Add `flag_keys_filter` option to `send_feature_flags` for selective flag evaluation in capture events

-# 6.5.0 - 2025-08-08
+## 6.5.0 - 2025-08-08

 - feat: Add `$context_tags` to an event to know which properties were included as tags

-# 6.4.1 - 2025-08-06
+## 6.4.1 - 2025-08-06

 - fix: Always pass project API key in `remote_config` requests for deterministic project routing

-# 6.4.0 - 2025-08-05
+## 6.4.0 - 2025-08-05

 - feat: support Vertex AI for Gemini

-# 6.3.4 - 2025-08-04
+## 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
+## 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
+## 6.3.2 - 2025-07-31

 - fix: Anthropic's tool calls are now handled properly

-# 6.3.0 - 2025-07-22
+## 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
+## 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
+## 6.1.1 - 2025-07-16

 - fix: correctly capture exceptions processed by Django from views or middleware

-# 6.1.0 - 2025-07-10
+## 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
+## 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
+## 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
+## 6.0.2 - 2025-07-02

 - fix: send_feature_flags changed to default to false in `Client::capture_exception`

-# 6.0.1
+## 6.0.1

 - fix: response `$process_person_profile` property when passed to capture

-# 6.0.0
+## 6.0.0

 This release contains a number of major breaking changes:

@@ -259,15 +305,15 @@ with posthog.new_context():

 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
+## 5.4.0 - 2025-06-20

 - feat: add support to session_id context on page method

-# 5.3.0 - 2025-06-19
+## 5.3.0 - 2025-06-19

 - fix: safely handle exception values

-# 5.2.0 - 2025-06-19
+## 5.2.0 - 2025-06-19

 - feat: construct artificial stack traces if no traceback is available on a captured exception

@@ -5,12 +5,26 @@ test:
 	coverage run -m pytest
 	coverage report

-release:
+build_release:
 	rm -rf dist/*
 	python setup.py sdist bdist_wheel
-	twine upload dist/*

-release_analytics:
+# Builds the `posthoganalytics` PyPI package, which is a mirror of `posthog`
+# published under a different name for internal use by posthog/posthog.
+#
+# The process works in three phases:
+#   1. posthog -> posthoganalytics: Copy the source, rewrite all imports,
+#      remove the original posthog/ dir, and build the dist.
+#   2. posthoganalytics -> posthog: Reverse the import rewrites, copy
+#      everything back into posthog/, and clean up.
+#   3. Restore pyproject.toml from backup (setup_analytics.py modifies it).
+#
+# This ensures the working tree is left in the same state it started in.
+#
+# NOTE: This target clears dist/ before building. In the release workflow,
+# `build_release` (posthog) must be published BEFORE running this target,
+# otherwise the posthog dist artifacts will be lost.
+build_release_analytics:
 	rm -rf dist
 	rm -rf build
 	rm -rf posthoganalytics
@@ -21,7 +35,6 @@ release_analytics:
 	find ./posthoganalytics -name "*.bak" -delete
 	rm -rf posthog
 	python setup_analytics.py sdist bdist_wheel
-	twine upload dist/*
 	mkdir posthog
 	find ./posthoganalytics -type f -name "*.py" -exec sed -i.bak -e 's/from posthoganalytics /from posthog /g' {} \;
 	find ./posthoganalytics -type f -name "*.py" -exec sed -i.bak -e 's/from posthoganalytics\./from posthog\./g' {} \;
@@ -54,4 +67,4 @@ prep_local:
 	@echo "Local copy created at ../posthog-python-local"
 	@echo "Install with: pip install -e ../posthog-python-local"

-.PHONY: test lint release e2e_test prep_local
+.PHONY: test lint build_release build_release_analytics e2e_test prep_local
@@ -0,0 +1,11 @@
+# posthoganalytics
+
+> **Do not use this package.** Use [`posthog`](https://pypi.org/project/posthog/) instead.
+
+```bash
+pip install posthog
+```
+
+This package exists solely for internal use by [posthog/posthog](https://github.com/posthog/posthog) to avoid import conflicts with the local `posthog` package in that repository. It is an automatically generated mirror of `posthog` — same code, same versions, just published under a different name.
+
+If you are not working on the PostHog main repository, you should never need this package. All documentation, issues, and development happen in [`posthog-python`](https://github.com/posthog/posthog-python).
@@ -0,0 +1,3 @@
+from posthog.ai.prompts import Prompts
+
+__all__ = ["Prompts"]
@@ -17,6 +17,7 @@ from posthog.ai.types import (
    TokenUsage,
    ToolInProgress,
 )
+from posthog.ai.utils import serialize_raw_usage


 def format_anthropic_response(response: Any) -> List[FormattedMessage]:
@@ -221,6 +222,12 @@ def extract_anthropic_usage_from_response(response: Any) -> TokenUsage:
    if web_search_count > 0:
        result["web_search_count"] = web_search_count

+    # Capture raw usage metadata for backend processing
+    # Serialize to dict here in the converter (not in utils)
+    serialized = serialize_raw_usage(response.usage)
+    if serialized:
+        result["raw_usage"] = serialized
+
    return result


@@ -247,6 +254,11 @@ def extract_anthropic_usage_from_event(event: Any) -> TokenUsage:
            usage["cache_read_input_tokens"] = getattr(
                event.message.usage, "cache_read_input_tokens", 0
            )
+            # Capture raw usage metadata for backend processing
+            # Serialize to dict here in the converter (not in utils)
+            serialized = serialize_raw_usage(event.message.usage)
+            if serialized:
+                usage["raw_usage"] = serialized

    # Handle usage stats from message_delta event
    if hasattr(event, "usage") and event.usage:
@@ -262,6 +274,12 @@ def extract_anthropic_usage_from_event(event: Any) -> TokenUsage:
                if web_search_count > 0:
                    usage["web_search_count"] = web_search_count

+        # Capture raw usage metadata for backend processing
+        # Serialize to dict here in the converter (not in utils)
+        serialized = serialize_raw_usage(event.usage)
+        if serialized:
+            usage["raw_usage"] = serialized
+
    return usage


@@ -12,6 +12,7 @@ from posthog.ai.types import (
    FormattedMessage,
    TokenUsage,
 )
+from posthog.ai.utils import serialize_raw_usage


 class GeminiPart(TypedDict, total=False):
@@ -487,6 +488,12 @@ def _extract_usage_from_metadata(metadata: Any) -> TokenUsage:
        if reasoning_tokens and reasoning_tokens > 0:
            usage["reasoning_tokens"] = reasoning_tokens

+    # Capture raw usage metadata for backend processing
+    # Serialize to dict here in the converter (not in utils)
+    serialized = serialize_raw_usage(metadata)
+    if serialized:
+        usage["raw_usage"] = serialized
+
    return usage


@@ -16,6 +16,7 @@ from posthog.ai.types import (
    FormattedTextContent,
    TokenUsage,
 )
+from posthog.ai.utils import serialize_raw_usage


 def format_openai_response(response: Any) -> List[FormattedMessage]:
@@ -429,6 +430,12 @@ def extract_openai_usage_from_response(response: Any) -> TokenUsage:
    if web_search_count > 0:
        result["web_search_count"] = web_search_count

+    # Capture raw usage metadata for backend processing
+    # Serialize to dict here in the converter (not in utils)
+    serialized = serialize_raw_usage(response.usage)
+    if serialized:
+        result["raw_usage"] = serialized
+
    return result


@@ -482,6 +489,12 @@ def extract_openai_usage_from_chunk(
                chunk.usage.completion_tokens_details.reasoning_tokens
            )

+        # Capture raw usage metadata for backend processing
+        # Serialize to dict here in the converter (not in utils)
+        serialized = serialize_raw_usage(chunk.usage)
+        if serialized:
+            usage["raw_usage"] = serialized
+
    elif provider_type == "responses":
        # For Responses API, usage is only in chunk.response.usage for completed events
        if hasattr(chunk, "type") and chunk.type == "response.completed":
@@ -516,6 +529,12 @@ def extract_openai_usage_from_chunk(
                    if web_search_count > 0:
                        usage["web_search_count"] = web_search_count

+                # Capture raw usage metadata for backend processing
+                # Serialize to dict here in the converter (not in utils)
+                serialized = serialize_raw_usage(response_usage)
+                if serialized:
+                    usage["raw_usage"] = serialized
+
    return usage


@@ -0,0 +1,286 @@
+"""
+Prompt management for PostHog AI SDK.
+
+Fetch and compile LLM prompts from PostHog with caching and fallback support.
+"""
+
+import logging
+import re
+import time
+import urllib.parse
+from typing import Any, Dict, Optional, Union
+
+from posthog.request import USER_AGENT, _get_session
+from posthog.utils import remove_trailing_slash
+
+log = logging.getLogger("posthog")
+
+APP_ENDPOINT = "https://us.posthog.com"
+DEFAULT_CACHE_TTL_SECONDS = 300  # 5 minutes
+
+PromptVariables = Dict[str, Union[str, int, float, bool]]
+
+
+class CachedPrompt:
+    """Cached prompt with metadata."""
+
+    def __init__(self, prompt: str, fetched_at: float):
+        self.prompt = prompt
+        self.fetched_at = fetched_at
+
+
+def _is_prompt_api_response(data: Any) -> bool:
+    """Check if the response is a valid prompt API response."""
+    return (
+        isinstance(data, dict)
+        and "prompt" in data
+        and isinstance(data.get("prompt"), str)
+    )
+
+
+class Prompts:
+    """
+    Fetch and compile LLM prompts from PostHog.
+
+    Can be initialized with a PostHog client or with direct options.
+
+    Examples:
+        ```python
+        from posthog import Posthog
+        from posthog.ai.prompts import Prompts
+
+        # With PostHog client
+        posthog = Posthog('phc_xxx', host='https://us.posthog.com', personal_api_key='phx_xxx')
+        prompts = Prompts(posthog)
+
+        # Or with direct options (no PostHog client needed)
+        prompts = Prompts(
+            personal_api_key='phx_xxx',
+            project_api_key='phc_xxx',
+            host='https://us.posthog.com',
+        )
+
+        # Fetch with caching and fallback
+        template = prompts.get('support-system-prompt', fallback='You are a helpful assistant.')
+
+        # Compile with variables
+        system_prompt = prompts.compile(template, {
+            'company': 'Acme Corp',
+            'tier': 'premium',
+        })
+        ```
+    """
+
+    def __init__(
+        self,
+        posthog: Optional[Any] = None,
+        *,
+        personal_api_key: Optional[str] = None,
+        project_api_key: Optional[str] = None,
+        host: Optional[str] = None,
+        default_cache_ttl_seconds: Optional[int] = None,
+    ):
+        """
+        Initialize Prompts.
+
+        Args:
+            posthog: PostHog client instance (optional if personal_api_key provided)
+            personal_api_key: Direct personal API key (optional if posthog provided)
+            project_api_key: Direct project API key (optional if posthog provided)
+            host: PostHog host (defaults to app endpoint)
+            default_cache_ttl_seconds: Default cache TTL (defaults to 300)
+        """
+        self._default_cache_ttl_seconds = (
+            default_cache_ttl_seconds or DEFAULT_CACHE_TTL_SECONDS
+        )
+        self._cache: Dict[str, CachedPrompt] = {}
+
+        if posthog is not None:
+            self._personal_api_key = getattr(posthog, "personal_api_key", None) or ""
+            self._project_api_key = getattr(posthog, "api_key", None) or ""
+            self._host = remove_trailing_slash(
+                getattr(posthog, "raw_host", None) or APP_ENDPOINT
+            )
+        else:
+            self._personal_api_key = personal_api_key or ""
+            self._project_api_key = project_api_key or ""
+            self._host = remove_trailing_slash(host or APP_ENDPOINT)
+
+    def get(
+        self,
+        name: str,
+        *,
+        cache_ttl_seconds: Optional[int] = None,
+        fallback: Optional[str] = None,
+    ) -> str:
+        """
+        Fetch a prompt by name from the PostHog API.
+
+        Caching behavior:
+        1. If cache is fresh, return cached value
+        2. If fetch fails and cache exists (stale), return stale cache with warning
+        3. If fetch fails and fallback provided, return fallback with warning
+        4. If fetch fails with no cache/fallback, raise exception
+
+        Args:
+            name: The name of the prompt to fetch
+            cache_ttl_seconds: Cache TTL in seconds (defaults to instance default)
+            fallback: Fallback prompt to use if fetch fails and no cache available
+
+        Returns:
+            The prompt string
+
+        Raises:
+            Exception: If the prompt cannot be fetched and no fallback is available
+        """
+        ttl = (
+            cache_ttl_seconds
+            if cache_ttl_seconds is not None
+            else self._default_cache_ttl_seconds
+        )
+
+        # Check cache first
+        cached = self._cache.get(name)
+        now = time.time()
+
+        if cached is not None:
+            is_fresh = (now - cached.fetched_at) < ttl
+
+            if is_fresh:
+                return cached.prompt
+
+        # Try to fetch from API
+        try:
+            prompt = self._fetch_prompt_from_api(name)
+            fetched_at = time.time()
+
+            # Update cache
+            self._cache[name] = CachedPrompt(prompt=prompt, fetched_at=fetched_at)
+
+            return prompt
+
+        except Exception as error:
+            # Fallback order:
+            # 1. Return stale cache (with warning)
+            if cached is not None:
+                log.warning(
+                    '[PostHog Prompts] Failed to fetch prompt "%s", using stale cache: %s',
+                    name,
+                    error,
+                )
+                return cached.prompt
+
+            # 2. Return fallback (with warning)
+            if fallback is not None:
+                log.warning(
+                    '[PostHog Prompts] Failed to fetch prompt "%s", using fallback: %s',
+                    name,
+                    error,
+                )
+                return fallback
+
+            # 3. Raise error
+            raise
+
+    def compile(self, prompt: str, variables: PromptVariables) -> str:
+        """
+        Replace {{variableName}} placeholders with values.
+
+        Unmatched variables are left unchanged.
+        Supports variable names with hyphens and dots (e.g., user-id, company.name).
+
+        Args:
+            prompt: The prompt template string
+            variables: Object containing variable values
+
+        Returns:
+            The compiled prompt string
+        """
+
+        def replace_variable(match: re.Match) -> str:
+            variable_name = match.group(1)
+
+            if variable_name in variables:
+                return str(variables[variable_name])
+
+            return match.group(0)
+
+        return re.sub(r"\{\{([\w.-]+)\}\}", replace_variable, prompt)
+
+    def clear_cache(self, name: Optional[str] = None) -> None:
+        """
+        Clear cached prompts.
+
+        Args:
+            name: Specific prompt to clear. If None, clears all cached prompts.
+        """
+        if name is not None:
+            self._cache.pop(name, None)
+        else:
+            self._cache.clear()
+
+    def _fetch_prompt_from_api(self, name: str) -> str:
+        """
+        Fetch prompt from PostHog API.
+
+        Endpoint: {host}/api/environments/@current/llm_prompts/name/{encoded_name}/?token={encoded_project_api_key}
+        Auth: Bearer {personal_api_key}
+
+        Args:
+            name: The name of the prompt to fetch
+
+        Returns:
+            The prompt string
+
+        Raises:
+            Exception: If the prompt cannot be fetched
+        """
+        if not self._personal_api_key:
+            raise Exception(
+                "[PostHog Prompts] personal_api_key is required to fetch prompts. "
+                "Please provide it when initializing the Prompts instance."
+            )
+        if not self._project_api_key:
+            raise Exception(
+                "[PostHog Prompts] project_api_key is required to fetch prompts. "
+                "Please provide it when initializing the Prompts instance."
+            )
+
+        encoded_name = urllib.parse.quote(name, safe="")
+        encoded_project_api_key = urllib.parse.quote(self._project_api_key, safe="")
+        url = f"{self._host}/api/environments/@current/llm_prompts/name/{encoded_name}/?token={encoded_project_api_key}"
+
+        headers = {
+            "Authorization": f"Bearer {self._personal_api_key}",
+            "User-Agent": USER_AGENT,
+        }
+
+        response = _get_session().get(url, headers=headers, timeout=10)
+
+        if not response.ok:
+            if response.status_code == 404:
+                raise Exception(f'[PostHog Prompts] Prompt "{name}" not found')
+
+            if response.status_code == 403:
+                raise Exception(
+                    f'[PostHog Prompts] Access denied for prompt "{name}". '
+                    "Check that your personal_api_key has the correct permissions and the LLM prompts feature is enabled."
+                )
+
+            raise Exception(
+                f'[PostHog Prompts] Failed to fetch prompt "{name}": HTTP {response.status_code}'
+            )
+
+        try:
+            data = response.json()
+        except Exception:
+            raise Exception(
+                f'[PostHog Prompts] Invalid response format for prompt "{name}"'
+            )
+
+        if not _is_prompt_api_response(data):
+            raise Exception(
+                f'[PostHog Prompts] Invalid response format for prompt "{name}"'
+            )
+
+        return data["prompt"]
@@ -83,6 +83,12 @@ def sanitize_openai_image(item: Any) -> Any:
    if not isinstance(item, dict):
        return item

+    if item.get("type") == "input_image" and isinstance(item.get("image_url"), str):
+        return {
+            **item,
+            "image_url": redact_base64_data_url(item["image_url"]),
+        }
+
    if (
        item.get("type") == "image_url"
        and isinstance(item.get("image_url"), dict)
@@ -64,6 +64,7 @@ class TokenUsage(TypedDict, total=False):
    cache_creation_input_tokens: Optional[int]
    reasoning_tokens: Optional[int]
    web_search_count: Optional[int]
+    raw_usage: Optional[Any]  # Raw provider usage metadata for backend processing


 class ProviderResponse(TypedDict, total=False):
@@ -13,6 +13,54 @@ from posthog.ai.types import FormattedMessage, StreamingEventData, TokenUsage
 from posthog.client import Client as PostHogClient


+def serialize_raw_usage(raw_usage: Any) -> Optional[Dict[str, Any]]:
+    """
+    Convert raw provider usage objects to JSON-serializable dicts.
+
+    Handles Pydantic models (OpenAI/Anthropic) and protobuf-like objects (Gemini)
+    with a fallback chain to ensure we never pass unserializable objects to PostHog.
+
+    Args:
+        raw_usage: Raw usage object from provider SDK
+
+    Returns:
+        Plain dict or None if conversion fails
+    """
+    if raw_usage is None:
+        return None
+
+    # Already a dict
+    if isinstance(raw_usage, dict):
+        return raw_usage
+
+    # Try Pydantic model_dump() (OpenAI/Anthropic)
+    if hasattr(raw_usage, "model_dump") and callable(raw_usage.model_dump):
+        try:
+            return raw_usage.model_dump()
+        except Exception:
+            pass
+
+    # Try to_dict() (some protobuf objects)
+    if hasattr(raw_usage, "to_dict") and callable(raw_usage.to_dict):
+        try:
+            return raw_usage.to_dict()
+        except Exception:
+            pass
+
+    # Try __dict__ / vars() for simple objects
+    try:
+        return vars(raw_usage)
+    except Exception:
+        pass
+
+    # Last resort: convert to string representation
+    # This ensures we always return something rather than failing
+    try:
+        return {"_raw": str(raw_usage)}
+    except Exception:
+        return None
+
+
 def merge_usage_stats(
    target: TokenUsage, source: TokenUsage, mode: str = "incremental"
 ) -> None:
@@ -60,6 +108,17 @@ def merge_usage_stats(
            current = target.get("web_search_count") or 0
            target["web_search_count"] = max(current, source_web_search)

+        # Merge raw_usage to avoid losing data from earlier events
+        # For Anthropic streaming: message_start has input tokens, message_delta has output
+        # Note: raw_usage is already serialized by converters, so it's a dict
+        source_raw_usage = source.get("raw_usage")
+        if source_raw_usage is not None and isinstance(source_raw_usage, dict):
+            current_raw_value = target.get("raw_usage")
+            current_raw: Dict[str, Any] = (
+                current_raw_value if isinstance(current_raw_value, dict) else {}
+            )
+            target["raw_usage"] = {**current_raw, **source_raw_usage}
+
    elif mode == "cumulative":
        # Replace with latest values (already cumulative)
        if source.get("input_tokens") is not None:
@@ -76,6 +135,9 @@ def merge_usage_stats(
            target["reasoning_tokens"] = source["reasoning_tokens"]
        if source.get("web_search_count") is not None:
            target["web_search_count"] = source["web_search_count"]
+        # Note: raw_usage is already serialized by converters, so it's a dict
+        if source.get("raw_usage") is not None:
+            target["raw_usage"] = source["raw_usage"]

    else:
        raise ValueError(f"Invalid mode: {mode}. Must be 'incremental' or 'cumulative'")
@@ -332,6 +394,11 @@ def call_llm_and_track_usage(
            if web_search_count is not None and web_search_count > 0:
                tag("$ai_web_search_count", web_search_count)

+            raw_usage = usage.get("raw_usage")
+            if raw_usage is not None:
+                # Already serialized by converters
+                tag("$ai_usage", raw_usage)
+
            if posthog_distinct_id is None:
                tag("$process_person_profile", False)

@@ -457,6 +524,11 @@ async def call_llm_and_track_usage_async(
            if web_search_count is not None and web_search_count > 0:
                tag("$ai_web_search_count", web_search_count)

+            raw_usage = usage.get("raw_usage")
+            if raw_usage is not None:
+                # Already serialized by converters
+                tag("$ai_usage", raw_usage)
+
            if posthog_distinct_id is None:
                tag("$process_person_profile", False)

@@ -594,6 +666,12 @@ def capture_streaming_event(
    ):
        event_properties["$ai_web_search_count"] = web_search_count

+    # Add raw usage metadata if present (all providers)
+    raw_usage = event_data["usage_stats"].get("raw_usage")
+    if raw_usage is not None:
+        # Already serialized by converters
+        event_properties["$ai_usage"] = raw_usage
+
    # Handle provider-specific fields
    if (
        event_data["provider"] == "openai"
@@ -38,6 +38,7 @@ from posthog.feature_flags import (
    InconclusiveMatchError,
    RequiresServerEvaluation,
    match_feature_flag_properties,
+    resolve_bucketing_value,
 )
 from posthog.flag_definition_cache import (
    FlagDefinitionCacheData,
@@ -1335,6 +1336,13 @@ class Client(object):
                self.log.error(
                    "[FEATURE FLAGS] Error loading feature flags: To use feature flags, please set a valid personal_api_key. More information: https://posthog.com/docs/api/overview"
                )
+                self.feature_flags = []
+                self.group_type_mapping = {}
+                self.cohorts = {}
+
+                if self.flag_cache:
+                    self.flag_cache.clear()
+
                if self.debug:
                    raise APIError(
                        status=401,
@@ -1411,6 +1419,7 @@ class Client(object):
        person_properties=None,
        group_properties=None,
        warn_on_unknown_groups=True,
+        device_id=None,
    ) -> FlagValue:
        groups = groups or {}
        person_properties = person_properties or {}
@@ -1451,22 +1460,35 @@ class Client(object):
                    )
                return False

+            if group_name not in group_properties:
+                raise InconclusiveMatchError(
+                    f"Flag has no group properties for group '{group_name}'"
+                )
            focused_group_properties = group_properties[group_name]
+            group_key = groups[group_name]
            return match_feature_flag_properties(
                feature_flag,
-                groups[group_name],
+                group_key,
                focused_group_properties,
-                self.feature_flags_by_key,
-                evaluation_cache,
+                cohort_properties=self.cohorts,
+                flags_by_key=self.feature_flags_by_key,
+                evaluation_cache=evaluation_cache,
+                device_id=device_id,
+                bucketing_value=group_key,
            )
        else:
+            bucketing_value = resolve_bucketing_value(
+                feature_flag, distinct_id, device_id
+            )
            return match_feature_flag_properties(
                feature_flag,
                distinct_id,
                person_properties,
-                self.cohorts,
-                self.feature_flags_by_key,
-                evaluation_cache,
+                cohort_properties=self.cohorts,
+                flags_by_key=self.feature_flags_by_key,
+                evaluation_cache=evaluation_cache,
+                device_id=device_id,
+                bucketing_value=bucketing_value,
            )

    def feature_enabled(
@@ -1573,8 +1595,12 @@ class Client(object):
        evaluated_at = None
        feature_flag_error: Optional[str] = None

+        # Resolve device_id from context if not provided
+        if device_id is None:
+            device_id = get_context_device_id()
+
        flag_value = self._locally_evaluate_flag(
-            key, distinct_id, groups, person_properties, group_properties
+            key, distinct_id, groups, person_properties, group_properties, device_id
        )
        flag_was_locally_evaluated = flag_value is not None

@@ -1778,6 +1804,7 @@ class Client(object):
        groups: dict[str, str],
        person_properties: dict[str, str],
        group_properties: dict[str, str],
+        device_id: Optional[str] = None,
    ) -> Optional[FlagValue]:
        if self.feature_flags is None and self.personal_api_key:
            self.load_feature_flags()
@@ -1797,6 +1824,7 @@ class Client(object):
                        groups=groups,
                        person_properties=person_properties,
                        group_properties=group_properties,
+                        device_id=device_id,
                    )
                    self.log.debug(
                        f"Successfully computed flag locally: {key} -> {response}"
@@ -1921,10 +1949,12 @@ class Client(object):
            f"{key}_{'::null::' if response is None else str(response)}"
        )

-        if (
-            feature_flag_reported_key
-            not in self.distinct_ids_feature_flags_reported[distinct_id]
-        ):
+        reported_flags = self.distinct_ids_feature_flags_reported.get(distinct_id)
+        if reported_flags is None:
+            reported_flags = set()
+            self.distinct_ids_feature_flags_reported[distinct_id] = reported_flags
+
+        if feature_flag_reported_key not in reported_flags:
            properties: dict[str, Any] = {
                "$feature_flag": key,
                "$feature_flag_response": response,
@@ -1960,9 +1990,7 @@ class Client(object):
                groups=groups,
                disable_geoip=disable_geoip,
            )
-            self.distinct_ids_feature_flags_reported[distinct_id].add(
-                feature_flag_reported_key
-            )
+            reported_flags.add(feature_flag_reported_key)

    def get_remote_config_payload(self, key: str):
        if self.disabled:
@@ -2099,12 +2127,17 @@ class Client(object):
            )
        )

+        # Resolve device_id from context if not provided
+        if device_id is None:
+            device_id = get_context_device_id()
+
        response, fallback_to_flags = self._get_all_flags_and_payloads_locally(
            distinct_id,
            groups=groups,
            person_properties=person_properties,
            group_properties=group_properties,
            flag_keys_to_evaluate=flag_keys_to_evaluate,
+            device_id=device_id,
        )

        if fallback_to_flags and not only_evaluate_locally:
@@ -2135,6 +2168,7 @@ class Client(object):
        group_properties=None,
        warn_on_unknown_groups=False,
        flag_keys_to_evaluate: Optional[list[str]] = None,
+        device_id: Optional[str] = None,
    ) -> tuple[FlagsAndPayloads, bool]:
        person_properties = person_properties or {}
        group_properties = group_properties or {}
@@ -2164,6 +2198,7 @@ class Client(object):
                        person_properties=person_properties,
                        group_properties=group_properties,
                        warn_on_unknown_groups=warn_on_unknown_groups,
+                        device_id=device_id,
                    )
                    matched_payload = self._compute_payload_locally(
                        flag["key"], flags[flag["key"]]
@@ -3,8 +3,6 @@ import logging
 import time
 from threading import Thread

-import backoff
-
 from posthog.request import APIError, DatetimeSerializer, batch_post

 try:
@@ -128,29 +126,41 @@ class Consumer(Thread):
    def request(self, batch):
        """Attempt to upload the batch and retry before raising an error"""

-        def fatal_exception(exc):
+        def is_retryable(exc):
            if isinstance(exc, APIError):
                # retry on server errors and client errors
-                # with 429 status code (rate limited),
+                # with 408 (request timeout) or 429 (rate limited),
                # don't retry on other client errors
                if exc.status == "N/A":
                    return False
-                return (400 <= exc.status < 500) and exc.status != 429
+                return not ((400 <= exc.status < 500) and exc.status not in (408, 429))
            else:
                # retry on all other errors (eg. network)
-                return False
+                return True

-        @backoff.on_exception(
-            backoff.expo, Exception, max_tries=self.retries + 1, giveup=fatal_exception
-        )
-        def send_request():
-            batch_post(
-                self.api_key,
-                self.host,
-                gzip=self.gzip,
-                timeout=self.timeout,
-                batch=batch,
-                historical_migration=self.historical_migration,
-            )
+        last_exc = None
+        for attempt in range(self.retries + 1):
+            try:
+                batch_post(
+                    self.api_key,
+                    self.host,
+                    gzip=self.gzip,
+                    timeout=self.timeout,
+                    batch=batch,
+                    historical_migration=self.historical_migration,
+                )
+                return
+            except Exception as e:
+                last_exc = e
+                if not is_retryable(e):
+                    raise
+                if attempt < self.retries:
+                    # Respect Retry-After header if present, otherwise use exponential backoff
+                    retry_after = getattr(e, "retry_after", None)
+                    if retry_after and retry_after > 0:
+                        time.sleep(retry_after)
+                    else:
+                        time.sleep(min(2**attempt, 30))

-        send_request()
+        if last_exc:
+            raise last_exc
@@ -43,26 +43,31 @@ except ImportError:
 DEFAULT_MAX_VALUE_LENGTH = 1024

 DEFAULT_CODE_VARIABLES_MASK_PATTERNS = [
-    r"(?i).*password.*",
-    r"(?i).*secret.*",
-    r"(?i).*passwd.*",
-    r"(?i).*pwd.*",
-    r"(?i).*api_key.*",
-    r"(?i).*apikey.*",
-    r"(?i).*auth.*",
-    r"(?i).*credentials.*",
-    r"(?i).*privatekey.*",
-    r"(?i).*private_key.*",
-    r"(?i).*token.*",
-    r"(?i).*aws_access_key_id.*",
-    r"(?i).*_pass",
-    r"(?i)sk_.*",
-    r"(?i).*jwt.*",
+    r"(?i)password",
+    r"(?i)secret",
+    r"(?i)passwd",
+    r"(?i)pwd",
+    r"(?i)api_key",
+    r"(?i)apikey",
+    r"(?i)auth",
+    r"(?i)credentials",
+    r"(?i)privatekey",
+    r"(?i)private_key",
+    r"(?i)token",
+    r"(?i)aws_access_key_id",
+    r"(?i)_pass",
+    r"(?i)sk_",
+    r"(?i)jwt",
 ]

 DEFAULT_CODE_VARIABLES_IGNORE_PATTERNS = [r"^__.*"]

 CODE_VARIABLES_REDACTED_VALUE = "$$_posthog_redacted_based_on_masking_rules_$$"
+CODE_VARIABLES_TOO_LONG_VALUE = "$$_posthog_value_too_long_$$"
+
+_MAX_VALUE_LENGTH_FOR_PATTERN_MATCH = 5_000
+_MAX_COLLECTION_ITEMS_TO_SCAN = 100
+_REGEX_METACHARACTERS = frozenset(r"\.^$*+?{}[]|()")

 DEFAULT_TOTAL_VARIABLES_SIZE_LIMIT = 20 * 1024

@@ -928,40 +933,87 @@ def strip_string(value, max_length=None):
    )


+def _extract_plain_substring(pattern):
+    # Matches inline flag groups like (?i), (?ai), (?ims), etc. that include the 'i' flag.
+    # Python regex flags: a=ASCII, i=IGNORECASE, L=LOCALE, m=MULTILINE, s=DOTALL, u=UNICODE, x=VERBOSE
+    inline_flags = re.match(r"^\(\?[aiLmsux]*i[aiLmsux]*\)", pattern)
+    if not inline_flags:
+        return None
+    remainder = pattern[inline_flags.end() :]
+    if not remainder or any(c in _REGEX_METACHARACTERS for c in remainder):
+        return None
+    return remainder.lower()
+
+
 def _compile_patterns(patterns):
-    compiled = []
+    if not patterns:
+        return None
+    substrings = []
+    regexes = []
    for pattern in patterns:
-        try:
-            compiled.append(re.compile(pattern))
-        except Exception:
-            pass
-    return compiled
+        simple = _extract_plain_substring(pattern)
+        if simple is not None:
+            substrings.append(simple)
+        else:
+            try:
+                regexes.append(re.compile(pattern))
+            except Exception:
+                pass
+    if not substrings and not regexes:
+        return None
+    return (substrings, regexes)


 def _pattern_matches(name, patterns):
-    for pattern in patterns:
+    if patterns is None:
+        return False
+    substrings, regexes = patterns
+    if substrings:
+        name_lower = name.lower()
+        for s in substrings:
+            if s in name_lower:
+                return True
+    for pattern in regexes:
        if pattern.search(name):
            return True
    return False


-def _mask_sensitive_data(value, compiled_mask):
+def _mask_sensitive_data(value, compiled_mask, _seen=None):
    if not compiled_mask:
        return value

+    if isinstance(value, (dict, list, tuple)):
+        if _seen is None:
+            _seen = set()
+        obj_id = id(value)
+        if obj_id in _seen:
+            return "<circular ref>"
+        _seen.add(obj_id)
+
    if isinstance(value, dict):
+        if len(value) > _MAX_COLLECTION_ITEMS_TO_SCAN:
+            return CODE_VARIABLES_TOO_LONG_VALUE
        result = {}
        for k, v in value.items():
            key_str = str(k) if not isinstance(k, str) else k
-            if _pattern_matches(key_str, compiled_mask):
+            if len(key_str) > _MAX_VALUE_LENGTH_FOR_PATTERN_MATCH:
+                result[k] = CODE_VARIABLES_TOO_LONG_VALUE
+            elif _pattern_matches(key_str, compiled_mask):
                result[k] = CODE_VARIABLES_REDACTED_VALUE
            else:
-                result[k] = _mask_sensitive_data(v, compiled_mask)
+                result[k] = _mask_sensitive_data(v, compiled_mask, _seen)
        return result
    elif isinstance(value, (list, tuple)):
-        masked_items = [_mask_sensitive_data(item, compiled_mask) for item in value]
+        if len(value) > _MAX_COLLECTION_ITEMS_TO_SCAN:
+            return CODE_VARIABLES_TOO_LONG_VALUE
+        masked_items = [
+            _mask_sensitive_data(item, compiled_mask, _seen) for item in value
+        ]
        return type(value)(masked_items)
    elif isinstance(value, str):
+        if len(value) > _MAX_VALUE_LENGTH_FOR_PATTERN_MATCH:
+            return CODE_VARIABLES_TOO_LONG_VALUE
        if _pattern_matches(value, compiled_mask):
            return CODE_VARIABLES_REDACTED_VALUE
        return value
@@ -982,7 +1034,9 @@ def _serialize_variable_value(value, limiter, max_length=1024, compiled_mask=Non
            limiter.add(result_size)
            return value
        elif isinstance(value, str):
-            if compiled_mask and _pattern_matches(value, compiled_mask):
+            if len(value) > _MAX_VALUE_LENGTH_FOR_PATTERN_MATCH:
+                result = CODE_VARIABLES_TOO_LONG_VALUE
+            elif compiled_mask and _pattern_matches(value, compiled_mask):
                result = CODE_VARIABLES_REDACTED_VALUE
            else:
                result = value
@@ -2,6 +2,7 @@ import datetime
 import hashlib
 import logging
 import re
+import warnings
 from typing import Optional

 from dateutil import parser
@@ -34,18 +35,18 @@ class RequiresServerEvaluation(Exception):
    pass


-# This function takes a distinct_id and a feature flag key and returns a float between 0 and 1.
-# Given the same distinct_id and key, it'll always return the same float. These floats are
+# This function takes a bucketing value and a feature flag key and returns a float between 0 and 1.
+# Given the same bucketing value and key, it'll always return the same float. These floats are
 # uniformly distributed between 0 and 1, so if we want to show this feature to 20% of traffic
-# we can do _hash(key, distinct_id) < 0.2
-def _hash(key: str, distinct_id: str, salt: str = "") -> float:
-    hash_key = f"{key}.{distinct_id}{salt}"
+# we can do _hash(key, bucketing_value) < 0.2
+def _hash(key: str, bucketing_value: str, salt: str = "") -> float:
+    hash_key = f"{key}.{bucketing_value}{salt}"
    hash_val = int(hashlib.sha1(hash_key.encode("utf-8")).hexdigest()[:15], 16)
    return hash_val / __LONG_SCALE__


-def get_matching_variant(flag, distinct_id):
-    hash_value = _hash(flag["key"], distinct_id, salt="variant")
+def get_matching_variant(flag, bucketing_value):
+    hash_value = _hash(flag["key"], bucketing_value, salt="variant")
    for variant in variant_lookup_table(flag):
        if hash_value >= variant["value_min"] and hash_value < variant["value_max"]:
            return variant["key"]
@@ -68,7 +69,13 @@ def variant_lookup_table(feature_flag):


 def evaluate_flag_dependency(
-    property, flags_by_key, evaluation_cache, distinct_id, properties, cohort_properties
+    property,
+    flags_by_key,
+    evaluation_cache,
+    distinct_id,
+    properties,
+    cohort_properties,
+    device_id=None,
 ):
    """
    Evaluate a flag dependency property according to the dependency chain algorithm.
@@ -80,6 +87,7 @@ def evaluate_flag_dependency(
        distinct_id: The distinct ID being evaluated
        properties: Person properties for evaluation
        cohort_properties: Cohort properties for evaluation
+        device_id: The device ID for bucketing (optional)

    Returns:
        bool: True if all dependencies in the chain evaluate to True, False otherwise
@@ -124,13 +132,27 @@ def evaluate_flag_dependency(
                else:
                    # Recursively evaluate the dependency
                    try:
+                        dep_flag_filters = dep_flag.get("filters") or {}
+                        dep_aggregation_group_type_index = dep_flag_filters.get(
+                            "aggregation_group_type_index"
+                        )
+                        if dep_aggregation_group_type_index is not None:
+                            # Group flags should continue bucketing by the group key
+                            # from the current evaluation context.
+                            dep_bucketing_value = distinct_id
+                        else:
+                            dep_bucketing_value = resolve_bucketing_value(
+                                dep_flag, distinct_id, device_id
+                            )
                        dep_result = match_feature_flag_properties(
                            dep_flag,
                            distinct_id,
                            properties,
-                            cohort_properties,
-                            flags_by_key,
-                            evaluation_cache,
+                            cohort_properties=cohort_properties,
+                            flags_by_key=flags_by_key,
+                            evaluation_cache=evaluation_cache,
+                            device_id=device_id,
+                            bucketing_value=dep_bucketing_value,
                        )
                        evaluation_cache[dep_flag_key] = dep_result
                    except InconclusiveMatchError as e:
@@ -215,21 +237,54 @@ def matches_dependency_value(expected_value, actual_value):
    return False


+def resolve_bucketing_value(flag, distinct_id, device_id=None):
+    """Resolve the bucketing value for a flag based on its bucketing_identifier setting.
+
+    Returns:
+        The appropriate identifier string to use for hashing/bucketing.
+
+    Raises:
+        InconclusiveMatchError: If the flag requires device_id but none was provided.
+    """
+    flag_filters = flag.get("filters") or {}
+    bucketing_identifier = flag.get("bucketing_identifier") or flag_filters.get(
+        "bucketing_identifier"
+    )
+    if bucketing_identifier == "device_id":
+        if not device_id:
+            raise InconclusiveMatchError(
+                "Flag requires device_id for bucketing but none was provided"
+            )
+        return device_id
+    return distinct_id
+
+
 def match_feature_flag_properties(
    flag,
    distinct_id,
    properties,
+    *,
    cohort_properties=None,
    flags_by_key=None,
    evaluation_cache=None,
+    device_id=None,
+    bucketing_value=None,
 ) -> FlagValue:
-    flag_conditions = (flag.get("filters") or {}).get("groups") or []
+    if bucketing_value is None:
+        warnings.warn(
+            "Calling match_feature_flag_properties() without bucketing_value is deprecated. "
+            "Pass bucketing_value explicitly. This fallback will be removed in a future major release.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
+        bucketing_value = resolve_bucketing_value(flag, distinct_id, device_id)
+
+    flag_filters = flag.get("filters") or {}
+    flag_conditions = flag_filters.get("groups") or []
    is_inconclusive = False
    cohort_properties = cohort_properties or {}
    # Some filters can be explicitly set to null, which require accessing variants like so
-    flag_variants = ((flag.get("filters") or {}).get("multivariate") or {}).get(
-        "variants"
-    ) or []
+    flag_variants = (flag_filters.get("multivariate") or {}).get("variants") or []
    valid_variant_keys = [variant["key"] for variant in flag_variants]

    for condition in flag_conditions:
@@ -244,12 +299,14 @@ def match_feature_flag_properties(
                cohort_properties,
                flags_by_key,
                evaluation_cache,
+                bucketing_value=bucketing_value,
+                device_id=device_id,
            ):
                variant_override = condition.get("variant")
                if variant_override and variant_override in valid_variant_keys:
                    variant = variant_override
                else:
-                    variant = get_matching_variant(flag, distinct_id)
+                    variant = get_matching_variant(flag, bucketing_value)
                return variant or True
        except RequiresServerEvaluation:
            # Static cohort or other missing server-side data - must fallback to API
@@ -277,6 +334,9 @@ def is_condition_match(
    cohort_properties,
    flags_by_key=None,
    evaluation_cache=None,
+    *,
+    bucketing_value,
+    device_id=None,
 ) -> bool:
    rollout_percentage = condition.get("rollout_percentage")
    if len(condition.get("properties") or []) > 0:
@@ -290,6 +350,7 @@ def is_condition_match(
                    flags_by_key,
                    evaluation_cache,
                    distinct_id,
+                    device_id=device_id,
                )
            elif property_type == "flag":
                matches = evaluate_flag_dependency(
@@ -299,6 +360,7 @@ def is_condition_match(
                    distinct_id,
                    properties,
                    cohort_properties,
+                    device_id=device_id,
                )
            else:
                matches = match_property(prop, properties)
@@ -308,9 +370,9 @@ def is_condition_match(
        if rollout_percentage is None:
            return True

-    if rollout_percentage is not None and _hash(feature_flag["key"], distinct_id) > (
-        rollout_percentage / 100
-    ):
+    if rollout_percentage is not None and _hash(
+        feature_flag["key"], bucketing_value
+    ) > (rollout_percentage / 100):
        return False

    return True
@@ -454,6 +516,7 @@ def match_cohort(
    flags_by_key=None,
    evaluation_cache=None,
    distinct_id=None,
+    device_id=None,
 ) -> bool:
    # Cohort properties are in the form of property groups like this:
    # {
@@ -478,6 +541,7 @@ def match_cohort(
        flags_by_key,
        evaluation_cache,
        distinct_id,
+        device_id=device_id,
    )


@@ -488,6 +552,7 @@ def match_property_group(
    flags_by_key=None,
    evaluation_cache=None,
    distinct_id=None,
+    device_id=None,
 ) -> bool:
    if not property_group:
        return True
@@ -512,6 +577,7 @@ def match_property_group(
                    flags_by_key,
                    evaluation_cache,
                    distinct_id,
+                    device_id=device_id,
                )
                if property_group_type == "AND":
                    if not matches:
@@ -545,6 +611,7 @@ def match_property_group(
                        flags_by_key,
                        evaluation_cache,
                        distinct_id,
+                        device_id=device_id,
                    )
                elif prop.get("type") == "flag":
                    matches = evaluate_flag_dependency(
@@ -554,6 +621,7 @@ def match_property_group(
                        distinct_id,
                        property_values,
                        cohort_properties,
+                        device_id=device_id,
                    )
                else:
                    matches = match_property(prop, property_values)
@@ -3,7 +3,7 @@ import logging
 import re
 import socket
 from dataclasses import dataclass
-from datetime import date, datetime
+from datetime import date, datetime, timezone
 from gzip import GzipFile
 from io import BytesIO
 from typing import Any, List, Optional, Tuple, Union
@@ -235,12 +235,31 @@ def _process_response(
            )
            raise QuotaLimitError(res.status_code, "Feature flags quota limited")
        return response
+    retry_after = None
+    retry_after_header = res.headers.get("Retry-After")
+    if retry_after_header:
+        try:
+            retry_after = float(retry_after_header)
+        except (ValueError, TypeError):
+            try:
+                from email.utils import parsedate_to_datetime
+
+                retry_after = max(
+                    0.0,
+                    (
+                        parsedate_to_datetime(retry_after_header)
+                        - datetime.now(timezone.utc)
+                    ).total_seconds(),
+                )
+            except (ValueError, TypeError):
+                pass
+
    try:
        payload = res.json()
        log.debug("received response: %s", payload)
-        raise APIError(res.status_code, payload["detail"])
+        raise APIError(res.status_code, payload["detail"], retry_after=retry_after)
    except (KeyError, ValueError):
-        raise APIError(res.status_code, res.text)
+        raise APIError(res.status_code, res.text, retry_after=retry_after)


 def decide(
@@ -348,9 +367,12 @@ def get(


 class APIError(Exception):
-    def __init__(self, status: Union[int, str], message: str):
+    def __init__(
+        self, status: Union[int, str], message: str, retry_after: Optional[float] = None
+    ):
        self.message = message
        self.status = status
+        self.retry_after = retry_after

    def __str__(self):
        msg = "[PostHog] {0} ({1})"
@@ -1,3 +1,4 @@
+import json
 from unittest.mock import patch

 import pytest
@@ -306,6 +307,15 @@ def test_basic_completion(mock_client, mock_anthropic_response):
        assert props["$ai_http_status"] == 200
        assert props["foo"] == "bar"
        assert isinstance(props["$ai_latency"], float)
+        # Verify raw usage metadata is passed for backend processing
+        assert "$ai_usage" in props
+        assert props["$ai_usage"] is not None
+        # Verify it's JSON-serializable
+        json.dumps(props["$ai_usage"])
+        # Verify it has expected structure
+        assert isinstance(props["$ai_usage"], dict)
+        assert "input_tokens" in props["$ai_usage"]
+        assert "output_tokens" in props["$ai_usage"]


 def test_groups(mock_client, mock_anthropic_response):
@@ -918,6 +928,16 @@ def test_streaming_with_tool_calls(mock_client, mock_anthropic_stream_with_tools
        assert props["$ai_cache_read_input_tokens"] == 5
        assert props["$ai_cache_creation_input_tokens"] == 0

+        # Verify raw usage is captured in streaming mode (merged from events)
+        assert "$ai_usage" in props
+        assert props["$ai_usage"] is not None
+        # Verify it's JSON-serializable
+        json.dumps(props["$ai_usage"])
+        # Verify it has expected structure (merged from message_start and message_delta)
+        assert isinstance(props["$ai_usage"], dict)
+        assert "input_tokens" in props["$ai_usage"]
+        assert "output_tokens" in props["$ai_usage"]
+

 def test_async_streaming_with_tool_calls(mock_client, mock_anthropic_stream_with_tools):
    """Test that tool calls are properly captured in async streaming mode."""
@@ -1,3 +1,4 @@
+import json
 from unittest.mock import MagicMock, patch

 import pytest
@@ -34,6 +35,13 @@ def mock_gemini_response():
    # Ensure cache and reasoning tokens are not present (not MagicMock)
    mock_usage.cached_content_token_count = 0
    mock_usage.thoughts_token_count = 0
+    # Make model_dump() return a proper dict for serialization
+    mock_usage.model_dump.return_value = {
+        "prompt_token_count": 20,
+        "candidates_token_count": 10,
+        "cached_content_token_count": 0,
+        "thoughts_token_count": 0,
+    }
    mock_response.usage_metadata = mock_usage

    mock_candidate = MagicMock()
@@ -69,6 +77,13 @@ def mock_gemini_response_with_function_calls():
    mock_usage.candidates_token_count = 15
    mock_usage.cached_content_token_count = 0
    mock_usage.thoughts_token_count = 0
+    # Make model_dump() return a proper dict for serialization
+    mock_usage.model_dump.return_value = {
+        "prompt_token_count": 25,
+        "candidates_token_count": 15,
+        "cached_content_token_count": 0,
+        "thoughts_token_count": 0,
+    }
    mock_response.usage_metadata = mock_usage

    # Mock function call
@@ -117,6 +132,13 @@ def mock_gemini_response_function_calls_only():
    mock_usage.candidates_token_count = 12
    mock_usage.cached_content_token_count = 0
    mock_usage.thoughts_token_count = 0
+    # Make model_dump() return a proper dict for serialization
+    mock_usage.model_dump.return_value = {
+        "prompt_token_count": 30,
+        "candidates_token_count": 12,
+        "cached_content_token_count": 0,
+        "thoughts_token_count": 0,
+    }
    mock_response.usage_metadata = mock_usage

    # Mock function call
@@ -174,6 +196,15 @@ def test_new_client_basic_generation(
    assert props["foo"] == "bar"
    assert "$ai_trace_id" in props
    assert props["$ai_latency"] > 0
+    # Verify raw usage metadata is passed for backend processing
+    assert "$ai_usage" in props
+    assert props["$ai_usage"] is not None
+    # Verify it's JSON-serializable
+    json.dumps(props["$ai_usage"])
+    # Verify it has expected structure
+    assert isinstance(props["$ai_usage"], dict)
+    assert "prompt_token_count" in props["$ai_usage"]
+    assert "candidates_token_count" in props["$ai_usage"]


 def test_new_client_streaming_with_generate_content_stream(
@@ -810,6 +841,13 @@ def test_streaming_cache_and_reasoning_tokens(mock_client, mock_google_genai_cli
    chunk1_usage.candidates_token_count = 5
    chunk1_usage.cached_content_token_count = 30  # Cache tokens
    chunk1_usage.thoughts_token_count = 0
+    # Make model_dump() return a proper dict for serialization
+    chunk1_usage.model_dump.return_value = {
+        "prompt_token_count": 100,
+        "candidates_token_count": 5,
+        "cached_content_token_count": 30,
+        "thoughts_token_count": 0,
+    }
    chunk1.usage_metadata = chunk1_usage

    chunk2 = MagicMock()
@@ -819,6 +857,13 @@ def test_streaming_cache_and_reasoning_tokens(mock_client, mock_google_genai_cli
    chunk2_usage.candidates_token_count = 10
    chunk2_usage.cached_content_token_count = 30  # Same cache tokens
    chunk2_usage.thoughts_token_count = 5  # Reasoning tokens
+    # Make model_dump() return a proper dict for serialization
+    chunk2_usage.model_dump.return_value = {
+        "prompt_token_count": 100,
+        "candidates_token_count": 10,
+        "cached_content_token_count": 30,
+        "thoughts_token_count": 5,
+    }
    chunk2.usage_metadata = chunk2_usage

    mock_stream = iter([chunk1, chunk2])
@@ -848,6 +893,16 @@ def test_streaming_cache_and_reasoning_tokens(mock_client, mock_google_genai_cli
    assert props["$ai_cache_read_input_tokens"] == 30
    assert props["$ai_reasoning_tokens"] == 5

+    # Verify raw usage is captured in streaming mode (merged from chunks)
+    assert "$ai_usage" in props
+    assert props["$ai_usage"] is not None
+    # Verify it's JSON-serializable
+    json.dumps(props["$ai_usage"])
+    # Verify it has expected structure
+    assert isinstance(props["$ai_usage"], dict)
+    assert "prompt_token_count" in props["$ai_usage"]
+    assert "candidates_token_count" in props["$ai_usage"]
+

 def test_web_search_grounding(mock_client, mock_google_genai_client):
    """Test web search detection via grounding_metadata."""
@@ -1,3 +1,4 @@
+import json
 import time
 from unittest.mock import AsyncMock, patch

@@ -496,6 +497,15 @@ def test_basic_completion(mock_client, mock_openai_response):
        assert props["$ai_http_status"] == 200
        assert props["foo"] == "bar"
        assert isinstance(props["$ai_latency"], float)
+        # Verify raw usage metadata is passed for backend processing
+        assert "$ai_usage" in props
+        assert props["$ai_usage"] is not None
+        # Verify it's JSON-serializable
+        json.dumps(props["$ai_usage"])
+        # Verify it has expected structure
+        assert isinstance(props["$ai_usage"], dict)
+        assert "prompt_tokens" in props["$ai_usage"]
+        assert "completion_tokens" in props["$ai_usage"]


 def test_embeddings(mock_client, mock_embedding_response):
@@ -922,6 +932,16 @@ def test_streaming_with_tool_calls(mock_client, streaming_tool_call_chunks):
        assert props["$ai_input_tokens"] == 20
        assert props["$ai_output_tokens"] == 15

+        # Verify raw usage is captured in streaming mode
+        assert "$ai_usage" in props
+        assert props["$ai_usage"] is not None
+        # Verify it's JSON-serializable
+        json.dumps(props["$ai_usage"])
+        # Verify it has expected structure (merged from chunks)
+        assert isinstance(props["$ai_usage"], dict)
+        assert "prompt_tokens" in props["$ai_usage"]
+        assert "completion_tokens" in props["$ai_usage"]
+

 # test responses api
 def test_responses_api(mock_client, mock_openai_response_with_responses_api):
@@ -0,0 +1,607 @@
+import unittest
+from unittest.mock import MagicMock, patch
+
+from posthog.ai.prompts import Prompts
+
+
+class MockResponse:
+    """Mock HTTP response for testing."""
+
+    def __init__(self, json_data=None, status_code=200, ok=True):
+        self._json_data = json_data
+        self.status_code = status_code
+        self.ok = ok
+
+    def json(self):
+        if self._json_data is None:
+            raise ValueError("No JSON data")
+        return self._json_data
+
+
+class TestPrompts(unittest.TestCase):
+    """Tests for the Prompts class."""
+
+    mock_prompt_response = {
+        "id": 1,
+        "name": "test-prompt",
+        "prompt": "Hello, {{name}}! You are a helpful assistant for {{company}}.",
+        "version": 1,
+        "created_by": "user@example.com",
+        "created_at": "2024-01-01T00:00:00Z",
+        "updated_at": "2024-01-01T00:00:00Z",
+        "deleted": False,
+    }
+
+    def create_mock_posthog(
+        self,
+        personal_api_key="phx_test_key",
+        project_api_key="phc_test_key",
+        host="https://us.posthog.com",
+    ):
+        """Create a mock PostHog client."""
+        mock = MagicMock()
+        mock.personal_api_key = personal_api_key
+        mock.api_key = project_api_key
+        mock.raw_host = host
+        return mock
+
+
+class TestPromptsGet(TestPrompts):
+    """Tests for the Prompts.get() method."""
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_successfully_fetch_a_prompt(self, mock_get_session):
+        """Should successfully fetch a prompt."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data=self.mock_prompt_response)
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        result = prompts.get("test-prompt")
+
+        self.assertEqual(result, self.mock_prompt_response["prompt"])
+        mock_get.assert_called_once()
+        call_args = mock_get.call_args
+        self.assertEqual(
+            call_args[0][0],
+            "https://us.posthog.com/api/environments/@current/llm_prompts/name/test-prompt/?token=phc_test_key",
+        )
+        self.assertIn("Authorization", call_args[1]["headers"])
+        self.assertEqual(
+            call_args[1]["headers"]["Authorization"], "Bearer phx_test_key"
+        )
+
+    @patch("posthog.ai.prompts._get_session")
+    @patch("posthog.ai.prompts.time.time")
+    def test_return_cached_prompt_when_fresh(self, mock_time, mock_get_session):
+        """Should return cached prompt when fresh (no API call)."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data=self.mock_prompt_response)
+        mock_time.return_value = 1000.0
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        # First call - fetches from API
+        result1 = prompts.get("test-prompt", cache_ttl_seconds=300)
+        self.assertEqual(result1, self.mock_prompt_response["prompt"])
+        self.assertEqual(mock_get.call_count, 1)
+
+        # Advance time by 60 seconds (still within TTL)
+        mock_time.return_value = 1060.0
+
+        # Second call - should use cache
+        result2 = prompts.get("test-prompt", cache_ttl_seconds=300)
+        self.assertEqual(result2, self.mock_prompt_response["prompt"])
+        self.assertEqual(mock_get.call_count, 1)  # No additional fetch
+
+    @patch("posthog.ai.prompts._get_session")
+    @patch("posthog.ai.prompts.time.time")
+    def test_refetch_when_cache_is_stale(self, mock_time, mock_get_session):
+        """Should refetch when cache is stale."""
+        mock_get = mock_get_session.return_value.get
+        updated_prompt_response = {
+            **self.mock_prompt_response,
+            "prompt": "Updated prompt: Hello, {{name}}!",
+        }
+
+        mock_get.side_effect = [
+            MockResponse(json_data=self.mock_prompt_response),
+            MockResponse(json_data=updated_prompt_response),
+        ]
+        mock_time.return_value = 1000.0
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        # First call - fetches from API
+        result1 = prompts.get("test-prompt", cache_ttl_seconds=60)
+        self.assertEqual(result1, self.mock_prompt_response["prompt"])
+        self.assertEqual(mock_get.call_count, 1)
+
+        # Advance time past TTL
+        mock_time.return_value = 1061.0
+
+        # Second call - should refetch
+        result2 = prompts.get("test-prompt", cache_ttl_seconds=60)
+        self.assertEqual(result2, updated_prompt_response["prompt"])
+        self.assertEqual(mock_get.call_count, 2)
+
+    @patch("posthog.ai.prompts._get_session")
+    @patch("posthog.ai.prompts.time.time")
+    @patch("posthog.ai.prompts.log")
+    def test_use_stale_cache_on_fetch_failure_with_warning(
+        self, mock_log, mock_time, mock_get_session
+    ):
+        """Should use stale cache on fetch failure with warning."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.side_effect = [
+            MockResponse(json_data=self.mock_prompt_response),
+            Exception("Network error"),
+        ]
+        mock_time.return_value = 1000.0
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        # First call - populates cache
+        result1 = prompts.get("test-prompt", cache_ttl_seconds=60)
+        self.assertEqual(result1, self.mock_prompt_response["prompt"])
+
+        # Advance time past TTL
+        mock_time.return_value = 1061.0
+
+        # Second call - should use stale cache
+        result2 = prompts.get("test-prompt", cache_ttl_seconds=60)
+        self.assertEqual(result2, self.mock_prompt_response["prompt"])
+
+        # Check warning was logged
+        mock_log.warning.assert_called()
+        warning_call = mock_log.warning.call_args
+        self.assertIn("using stale cache", warning_call[0][0])
+
+    @patch("posthog.ai.prompts._get_session")
+    @patch("posthog.ai.prompts.log")
+    def test_use_fallback_when_no_cache_and_fetch_fails_with_warning(
+        self, mock_log, mock_get_session
+    ):
+        """Should use fallback when no cache and fetch fails with warning."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.side_effect = Exception("Network error")
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        fallback = "Default system prompt."
+        result = prompts.get("test-prompt", fallback=fallback)
+
+        self.assertEqual(result, fallback)
+
+        # Check warning was logged
+        mock_log.warning.assert_called()
+        warning_call = mock_log.warning.call_args
+        self.assertIn("using fallback", warning_call[0][0])
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_throw_when_no_cache_no_fallback_and_fetch_fails(self, mock_get_session):
+        """Should throw when no cache, no fallback, and fetch fails."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.side_effect = Exception("Network error")
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        with self.assertRaises(Exception) as context:
+            prompts.get("test-prompt")
+
+        self.assertIn("Network error", str(context.exception))
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_handle_404_response(self, mock_get_session):
+        """Should handle 404 response."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(status_code=404, ok=False)
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        with self.assertRaises(Exception) as context:
+            prompts.get("nonexistent-prompt")
+
+        self.assertIn('Prompt "nonexistent-prompt" not found', str(context.exception))
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_handle_403_response(self, mock_get_session):
+        """Should handle 403 response."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(status_code=403, ok=False)
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        with self.assertRaises(Exception) as context:
+            prompts.get("restricted-prompt")
+
+        self.assertIn(
+            'Access denied for prompt "restricted-prompt"', str(context.exception)
+        )
+
+    def test_throw_when_no_personal_api_key_configured(self):
+        """Should throw when no personal_api_key is configured."""
+        posthog = self.create_mock_posthog(personal_api_key=None)
+        prompts = Prompts(posthog)
+
+        with self.assertRaises(Exception) as context:
+            prompts.get("test-prompt")
+
+        self.assertIn(
+            "personal_api_key is required to fetch prompts", str(context.exception)
+        )
+
+    def test_throw_when_no_project_api_key_configured(self):
+        """Should throw when no project_api_key is configured."""
+        posthog = self.create_mock_posthog(project_api_key=None)
+        prompts = Prompts(posthog)
+
+        with self.assertRaises(Exception) as context:
+            prompts.get("test-prompt")
+
+        self.assertIn(
+            "project_api_key is required to fetch prompts", str(context.exception)
+        )
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_throw_when_api_returns_invalid_response_format(self, mock_get_session):
+        """Should throw when API returns invalid response format."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data={"invalid": "response"})
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        with self.assertRaises(Exception) as context:
+            prompts.get("test-prompt")
+
+        self.assertIn("Invalid response format", str(context.exception))
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_use_custom_host_from_posthog_options(self, mock_get_session):
+        """Should use custom host from PostHog options."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data=self.mock_prompt_response)
+
+        posthog = self.create_mock_posthog(host="https://eu.posthog.com")
+        prompts = Prompts(posthog)
+
+        prompts.get("test-prompt")
+
+        call_args = mock_get.call_args
+        self.assertTrue(
+            call_args[0][0].startswith(
+                "https://eu.posthog.com/api/environments/@current/llm_prompts/name/test-prompt/?token=phc_test_key"
+            ),
+            f"Expected URL to start with 'https://eu.posthog.com/api/environments/@current/llm_prompts/name/test-prompt/?token=phc_test_key', got {call_args[0][0]}",
+        )
+
+    @patch("posthog.ai.prompts._get_session")
+    @patch("posthog.ai.prompts.time.time")
+    def test_use_default_cache_ttl_5_minutes(self, mock_time, mock_get_session):
+        """Should use default cache TTL (5 minutes) when not specified."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data=self.mock_prompt_response)
+        mock_time.return_value = 1000.0
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        # First call
+        prompts.get("test-prompt")
+        self.assertEqual(mock_get.call_count, 1)
+
+        # Advance time by 4 minutes (within default 5-minute TTL)
+        mock_time.return_value = 1000.0 + (4 * 60)
+
+        # Second call - should use cache
+        prompts.get("test-prompt")
+        self.assertEqual(mock_get.call_count, 1)
+
+        # Advance time past 5-minute TTL
+        mock_time.return_value = 1000.0 + (6 * 60)
+
+        # Third call - should refetch
+        prompts.get("test-prompt")
+        self.assertEqual(mock_get.call_count, 2)
+
+    @patch("posthog.ai.prompts._get_session")
+    @patch("posthog.ai.prompts.time.time")
+    def test_use_custom_default_cache_ttl_from_constructor(
+        self, mock_time, mock_get_session
+    ):
+        """Should use custom default cache TTL from constructor."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data=self.mock_prompt_response)
+        mock_time.return_value = 1000.0
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog, default_cache_ttl_seconds=60)
+
+        # First call
+        prompts.get("test-prompt")
+        self.assertEqual(mock_get.call_count, 1)
+
+        # Advance time past custom TTL
+        mock_time.return_value = 1061.0
+
+        # Second call - should refetch
+        prompts.get("test-prompt")
+        self.assertEqual(mock_get.call_count, 2)
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_url_encode_prompt_names_with_special_characters(self, mock_get_session):
+        """Should URL-encode prompt names with special characters."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data=self.mock_prompt_response)
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        prompts.get("prompt with spaces/and/slashes")
+
+        call_args = mock_get.call_args
+        self.assertEqual(
+            call_args[0][0],
+            "https://us.posthog.com/api/environments/@current/llm_prompts/name/prompt%20with%20spaces%2Fand%2Fslashes/?token=phc_test_key",
+        )
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_work_with_direct_options_no_posthog_client(self, mock_get_session):
+        """Should work with direct options (no PostHog client)."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data=self.mock_prompt_response)
+
+        prompts = Prompts(
+            personal_api_key="phx_direct_key", project_api_key="phc_direct_key"
+        )
+
+        result = prompts.get("test-prompt")
+
+        self.assertEqual(result, self.mock_prompt_response["prompt"])
+        call_args = mock_get.call_args
+        self.assertEqual(
+            call_args[0][0],
+            "https://us.posthog.com/api/environments/@current/llm_prompts/name/test-prompt/?token=phc_direct_key",
+        )
+        self.assertEqual(
+            call_args[1]["headers"]["Authorization"], "Bearer phx_direct_key"
+        )
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_use_custom_host_from_direct_options(self, mock_get_session):
+        """Should use custom host from direct options."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data=self.mock_prompt_response)
+
+        prompts = Prompts(
+            personal_api_key="phx_direct_key",
+            project_api_key="phc_direct_key",
+            host="https://eu.posthog.com",
+        )
+
+        prompts.get("test-prompt")
+
+        call_args = mock_get.call_args
+        self.assertEqual(
+            call_args[0][0],
+            "https://eu.posthog.com/api/environments/@current/llm_prompts/name/test-prompt/?token=phc_direct_key",
+        )
+
+    @patch("posthog.ai.prompts._get_session")
+    @patch("posthog.ai.prompts.time.time")
+    def test_use_custom_default_cache_ttl_from_direct_options(
+        self, mock_time, mock_get_session
+    ):
+        """Should use custom default cache TTL from direct options."""
+        mock_get = mock_get_session.return_value.get
+        mock_get.return_value = MockResponse(json_data=self.mock_prompt_response)
+        mock_time.return_value = 1000.0
+
+        prompts = Prompts(
+            personal_api_key="phx_direct_key",
+            project_api_key="phc_direct_key",
+            default_cache_ttl_seconds=60,
+        )
+
+        # First call
+        prompts.get("test-prompt")
+        self.assertEqual(mock_get.call_count, 1)
+
+        # Advance time past custom TTL
+        mock_time.return_value = 1061.0
+
+        # Second call - should refetch
+        prompts.get("test-prompt")
+        self.assertEqual(mock_get.call_count, 2)
+
+
+class TestPromptsCompile(TestPrompts):
+    """Tests for the Prompts.compile() method."""
+
+    def test_replace_a_single_variable(self):
+        """Should replace a single variable."""
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        result = prompts.compile("Hello, {{name}}!", {"name": "World"})
+
+        self.assertEqual(result, "Hello, World!")
+
+    def test_replace_multiple_variables(self):
+        """Should replace multiple variables."""
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        result = prompts.compile(
+            "Hello, {{name}}! Welcome to {{company}}. Your tier is {{tier}}.",
+            {"name": "John", "company": "Acme Corp", "tier": "premium"},
+        )
+
+        self.assertEqual(
+            result, "Hello, John! Welcome to Acme Corp. Your tier is premium."
+        )
+
+    def test_handle_numbers(self):
+        """Should handle numbers."""
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        result = prompts.compile("You have {{count}} items.", {"count": 42})
+
+        self.assertEqual(result, "You have 42 items.")
+
+    def test_handle_booleans(self):
+        """Should handle booleans."""
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        result = prompts.compile("Feature enabled: {{enabled}}", {"enabled": True})
+
+        self.assertEqual(result, "Feature enabled: True")
+
+    def test_leave_unmatched_variables_unchanged(self):
+        """Should leave unmatched variables unchanged."""
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        result = prompts.compile(
+            "Hello, {{name}}! Your {{unknown}} is ready.", {"name": "World"}
+        )
+
+        self.assertEqual(result, "Hello, World! Your {{unknown}} is ready.")
+
+    def test_handle_prompts_with_no_variables(self):
+        """Should handle prompts with no variables."""
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        result = prompts.compile("You are a helpful assistant.", {})
+
+        self.assertEqual(result, "You are a helpful assistant.")
+
+    def test_handle_empty_variables_dict(self):
+        """Should handle empty variables dict."""
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        result = prompts.compile("Hello, {{name}}!", {})
+
+        self.assertEqual(result, "Hello, {{name}}!")
+
+    def test_handle_multiple_occurrences_of_same_variable(self):
+        """Should handle multiple occurrences of the same variable."""
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        result = prompts.compile(
+            "Hello, {{name}}! Goodbye, {{name}}!", {"name": "World"}
+        )
+
+        self.assertEqual(result, "Hello, World! Goodbye, World!")
+
+    def test_work_with_direct_options_initialization(self):
+        """Should work with direct options initialization."""
+        prompts = Prompts(
+            personal_api_key="phx_test_key", project_api_key="phc_test_key"
+        )
+
+        result = prompts.compile("Hello, {{name}}!", {"name": "World"})
+
+        self.assertEqual(result, "Hello, World!")
+
+    def test_handle_variables_with_hyphens(self):
+        """Should handle variables with hyphens."""
+        prompts = Prompts(
+            personal_api_key="phx_test_key", project_api_key="phc_test_key"
+        )
+
+        result = prompts.compile("User ID: {{user-id}}", {"user-id": "12345"})
+
+        self.assertEqual(result, "User ID: 12345")
+
+    def test_handle_variables_with_dots(self):
+        """Should handle variables with dots."""
+        prompts = Prompts(
+            personal_api_key="phx_test_key", project_api_key="phc_test_key"
+        )
+
+        result = prompts.compile("Company: {{company.name}}", {"company.name": "Acme"})
+
+        self.assertEqual(result, "Company: Acme")
+
+
+class TestPromptsClearCache(TestPrompts):
+    """Tests for the Prompts.clear_cache() method."""
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_clear_a_specific_prompt_from_cache(self, mock_get_session):
+        """Should clear a specific prompt from cache."""
+        mock_get = mock_get_session.return_value.get
+        other_prompt_response = {**self.mock_prompt_response, "name": "other-prompt"}
+
+        mock_get.side_effect = [
+            MockResponse(json_data=self.mock_prompt_response),
+            MockResponse(json_data=other_prompt_response),
+            MockResponse(json_data=self.mock_prompt_response),
+        ]
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        # Populate cache with two prompts
+        prompts.get("test-prompt")
+        prompts.get("other-prompt")
+        self.assertEqual(mock_get.call_count, 2)
+
+        # Clear only test-prompt
+        prompts.clear_cache("test-prompt")
+
+        # test-prompt should be refetched
+        prompts.get("test-prompt")
+        self.assertEqual(mock_get.call_count, 3)
+
+        # other-prompt should still be cached
+        prompts.get("other-prompt")
+        self.assertEqual(mock_get.call_count, 3)
+
+    @patch("posthog.ai.prompts._get_session")
+    def test_clear_all_prompts_from_cache(self, mock_get_session):
+        """Should clear all prompts from cache when no name is provided."""
+        mock_get = mock_get_session.return_value.get
+        other_prompt_response = {**self.mock_prompt_response, "name": "other-prompt"}
+
+        mock_get.side_effect = [
+            MockResponse(json_data=self.mock_prompt_response),
+            MockResponse(json_data=other_prompt_response),
+            MockResponse(json_data=self.mock_prompt_response),
+            MockResponse(json_data=other_prompt_response),
+        ]
+
+        posthog = self.create_mock_posthog()
+        prompts = Prompts(posthog)
+
+        # Populate cache with two prompts
+        prompts.get("test-prompt")
+        prompts.get("other-prompt")
+        self.assertEqual(mock_get.call_count, 2)
+
+        # Clear all cache
+        prompts.clear_cache()
+
+        # Both prompts should be refetched
+        prompts.get("test-prompt")
+        prompts.get("other-prompt")
+        self.assertEqual(mock_get.call_count, 4)
+
+
+if __name__ == "__main__":
+    unittest.main()
@@ -69,6 +69,25 @@ class TestSanitization(unittest.TestCase):
        )
        self.assertEqual(result[0]["content"][1]["image_url"]["detail"], "high")

+    def test_sanitize_openai_input_image(self):
+        input_data = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "input_image",
+                        "image_url": self.sample_base64_image,
+                    }
+                ],
+            }
+        ]
+
+        result = sanitize_openai(input_data)
+
+        self.assertEqual(
+            result[0]["content"][0]["image_url"], REDACTED_IMAGE_PLACEHOLDER
+        )
+
    def test_sanitize_openai_preserves_regular_urls(self):
        input_data = [
            {
@@ -479,6 +479,20 @@ class TestClient(unittest.TestCase):
            self.assertEqual(client.cohorts, {})
            self.assertIn("PostHog feature flags quota limited", logs.output[0])

+    @mock.patch("posthog.client.get")
+    def test_load_feature_flags_unauthorized(self, patch_get):
+        patch_get.side_effect = APIError(401, "Unauthorized")
+
+        client = Client(FAKE_TEST_API_KEY, personal_api_key="test")
+        with self.assertLogs("posthog", level="ERROR") as logs:
+            client._load_feature_flags()
+
+            self.assertEqual(client.feature_flags, [])
+            self.assertEqual(client.feature_flags_by_key, {})
+            self.assertEqual(client.group_type_mapping, {})
+            self.assertEqual(client.cohorts, {})
+            self.assertIn("please set a valid personal_api_key", logs.output[0])
+
    @mock.patch("posthog.client.flags")
    def test_dont_override_capture_with_local_flags(self, patch_flags):
        patch_flags.return_value = {"featureFlags": {"beta-feature": "random-variant"}}
@@ -167,6 +167,63 @@ class TestConsumer(unittest.TestCase):
            q.join()
            self.assertEqual(mock_post.call_count, 2)

+    def test_request_sleeps_with_retry_after(self) -> None:
+        error = APIError(429, "Too Many Requests", retry_after=5.0)
+        call_count = [0]
+
+        def mock_post(*args: Any, **kwargs: Any) -> None:
+            call_count[0] += 1
+            if call_count[0] <= 1:
+                raise error
+
+        consumer = Consumer(None, TEST_API_KEY, retries=3)
+        with (
+            mock.patch("posthog.consumer.batch_post", side_effect=mock_post),
+            mock.patch("posthog.consumer.time.sleep") as mock_sleep,
+        ):
+            consumer.request([_track_event()])
+            mock_sleep.assert_called_once_with(5.0)
+
+    def test_request_uses_exponential_backoff_without_retry_after(self) -> None:
+        error = APIError(503, "Service Unavailable")
+        call_count = [0]
+
+        def mock_post(*args: Any, **kwargs: Any) -> None:
+            call_count[0] += 1
+            if call_count[0] <= 3:
+                raise error
+
+        consumer = Consumer(None, TEST_API_KEY, retries=3)
+        with (
+            mock.patch("posthog.consumer.batch_post", side_effect=mock_post),
+            mock.patch("posthog.consumer.time.sleep") as mock_sleep,
+        ):
+            consumer.request([_track_event()])
+            self.assertEqual(
+                mock_sleep.call_args_list,
+                [
+                    mock.call(1),  # 2^0
+                    mock.call(2),  # 2^1
+                    mock.call(4),  # 2^2
+                ],
+            )
+
+    def test_request_retries_on_408(self) -> None:
+        call_count = [0]
+
+        def mock_post(*args: Any, **kwargs: Any) -> None:
+            call_count[0] += 1
+            if call_count[0] <= 1:
+                raise APIError(408, "Request Timeout")
+
+        consumer = Consumer(None, TEST_API_KEY, retries=3)
+        with (
+            mock.patch("posthog.consumer.batch_post", side_effect=mock_post),
+            mock.patch("posthog.consumer.time.sleep"),
+        ):
+            consumer.request([_track_event()])
+            self.assertEqual(call_count[0], 2)
+
    @parameterized.expand(
        [
            ("on_error_succeeds", False),
@@ -450,3 +450,240 @@ def test_code_variables_repr_fallback(tmpdir):
    assert "<CustomReprClass: custom representation>" in output
    assert "<lambda>" in output
    assert "<function trigger_error at" in output
+
+
+def test_code_variables_too_long_string_value_replaced(tmpdir):
+    app = tmpdir.join("app.py")
+    app.write(
+        dedent(
+            """
+    import os
+    from posthog import Posthog
+
+    posthog = Posthog(
+        'phc_x',
+        host='https://eu.i.posthog.com',
+        debug=True,
+        enable_exception_autocapture=True,
+        capture_exception_code_variables=True,
+        project_root=os.path.dirname(os.path.abspath(__file__))
+    )
+
+    def trigger_error():
+        short_value = "I am short"
+        long_value = "x" * 20000
+        long_blob = "password_" + "a" * 20000
+
+        1/0
+
+    trigger_error()
+    """
+        )
+    )
+
+    with pytest.raises(subprocess.CalledProcessError) as excinfo:
+        subprocess.check_output([sys.executable, str(app)], stderr=subprocess.STDOUT)
+
+    output = excinfo.value.output.decode("utf-8")
+
+    assert "ZeroDivisionError" in output
+    assert "code_variables" in output
+
+    assert "'short_value': 'I am short'" in output
+
+    assert "$$_posthog_value_too_long_$$" in output
+
+    assert "'long_blob': '$$_posthog_value_too_long_$$'" in output
+
+
+def test_code_variables_too_long_string_in_nested_dict(tmpdir):
+    app = tmpdir.join("app.py")
+    app.write(
+        dedent(
+            """
+    import os
+    from posthog import Posthog
+
+    posthog = Posthog(
+        'phc_x',
+        host='https://eu.i.posthog.com',
+        debug=True,
+        enable_exception_autocapture=True,
+        capture_exception_code_variables=True,
+        project_root=os.path.dirname(os.path.abspath(__file__))
+    )
+
+    def trigger_error():
+        my_data = {
+            "short_key": "short_val",
+            "long_key": "y" * 20000,
+            "nested": {
+                "deep_long": "z" * 20000,
+                "deep_short": "ok",
+            },
+        }
+
+        1/0
+
+    trigger_error()
+    """
+        )
+    )
+
+    with pytest.raises(subprocess.CalledProcessError) as excinfo:
+        subprocess.check_output([sys.executable, str(app)], stderr=subprocess.STDOUT)
+
+    output = excinfo.value.output.decode("utf-8")
+
+    assert "ZeroDivisionError" in output
+    assert "code_variables" in output
+
+    assert "short_val" in output
+    assert "ok" in output
+
+    assert "$$_posthog_value_too_long_$$" in output
+    assert "y" * 1000 not in output
+    assert "z" * 1000 not in output
+
+
+def test_mask_sensitive_data_too_long_dict_key():
+    from posthog.exception_utils import (
+        CODE_VARIABLES_TOO_LONG_VALUE,
+        _compile_patterns,
+        _mask_sensitive_data,
+    )
+
+    compiled_mask = _compile_patterns([r"(?i)password"])
+
+    result = _mask_sensitive_data(
+        {
+            "short": "visible",
+            "k" * 20000: "hidden_val",
+            "password": "secret",
+        },
+        compiled_mask,
+    )
+
+    assert result["short"] == "visible"
+    # This then gets shortened by the JSON truncation at 1024 chars anyways so no worries
+    assert result["k" * 20000] == CODE_VARIABLES_TOO_LONG_VALUE
+    assert result["password"] == "$$_posthog_redacted_based_on_masking_rules_$$"
+
+
+def test_mask_sensitive_data_circular_ref():
+    from posthog.exception_utils import _compile_patterns, _mask_sensitive_data
+
+    compiled_mask = _compile_patterns([r"(?i)password"])
+
+    # Circular dict
+    circular_dict = {"key": "value"}
+    circular_dict["self"] = circular_dict
+
+    result = _mask_sensitive_data(circular_dict, compiled_mask)
+    assert result["key"] == "value"
+    assert result["self"] == "<circular ref>"
+
+    # Circular list
+    circular_list = ["item"]
+    circular_list.append(circular_list)
+
+    result = _mask_sensitive_data(circular_list, compiled_mask)
+    assert result[0] == "item"
+    assert result[1] == "<circular ref>"
+
+
+def test_compile_patterns_fast_path_and_regex_fallback():
+    from posthog.exception_utils import _compile_patterns, _pattern_matches
+
+    # Simple case-insensitive patterns should become substrings
+    simple_only = _compile_patterns([r"(?i)password", r"(?i)token", r"(?i)jwt"])
+    substrings, regexes = simple_only
+    assert substrings == ["password", "token", "jwt"]
+    assert regexes == []
+
+    assert _pattern_matches("my_password_var", simple_only) is True
+    assert _pattern_matches("MY_TOKEN", simple_only) is True
+    assert _pattern_matches("safe_variable", simple_only) is False
+
+    # Complex regex patterns should stay as compiled regexes
+    complex_only = _compile_patterns([r"^__.*", r"\d{3,}", r"^sk_live_"])
+    substrings, regexes = complex_only
+    assert substrings == []
+    assert len(regexes) == 3
+
+    assert _pattern_matches("__dunder", complex_only) is True
+    assert _pattern_matches("has_999_numbers", complex_only) is True
+    assert _pattern_matches("sk_live_abc123", complex_only) is True
+    assert _pattern_matches("normal_var", complex_only) is False
+
+    # Mixed: simple substrings + complex regexes together
+    mixed = _compile_patterns(
+        [
+            r"(?i)secret",  # simple
+            r"(?i)api_key",  # simple
+            r"^__.*",  # regex
+            r"\btoken_\w+",  # regex
+        ]
+    )
+    substrings, regexes = mixed
+    assert substrings == ["secret", "api_key"]
+    assert len(regexes) == 2
+
+    # Substring matches
+    assert _pattern_matches("my_secret", mixed) is True
+    assert _pattern_matches("API_KEY_VALUE", mixed) is True
+
+    # Regex matches
+    assert _pattern_matches("__private", mixed) is True
+    assert _pattern_matches("token_abc", mixed) is True
+
+    # No match
+    assert _pattern_matches("safe_var", mixed) is False
+
+
+def test_mask_sensitive_data_large_dict_replaced():
+    from posthog.exception_utils import (
+        CODE_VARIABLES_TOO_LONG_VALUE,
+        _compile_patterns,
+        _mask_sensitive_data,
+    )
+
+    compiled_mask = _compile_patterns([r"(?i)password"])
+
+    large_dict = {f"key_{i}": f"value_{i}" for i in range(300)}
+
+    result = _mask_sensitive_data(large_dict, compiled_mask)
+
+    assert result == CODE_VARIABLES_TOO_LONG_VALUE
+
+
+def test_mask_sensitive_data_large_list_replaced():
+    from posthog.exception_utils import (
+        CODE_VARIABLES_TOO_LONG_VALUE,
+        _compile_patterns,
+        _mask_sensitive_data,
+    )
+
+    compiled_mask = _compile_patterns([r"(?i)password"])
+
+    large_list = [f"item_{i}" for i in range(300)]
+
+    result = _mask_sensitive_data(large_list, compiled_mask)
+
+    assert result == CODE_VARIABLES_TOO_LONG_VALUE
+
+
+def test_mask_sensitive_data_large_tuple_replaced():
+    from posthog.exception_utils import (
+        CODE_VARIABLES_TOO_LONG_VALUE,
+        _compile_patterns,
+        _mask_sensitive_data,
+    )
+
+    compiled_mask = _compile_patterns([r"(?i)password"])
+
+    large_tuple = tuple(f"item_{i}" for i in range(300))
+
+    result = _mask_sensitive_data(large_tuple, compiled_mask)
+
+    assert result == CODE_VARIABLES_TOO_LONG_VALUE
@@ -233,6 +233,27 @@ class TestLocalEvaluation(unittest.TestCase):

        self.assertEqual(patch_flags.call_count, 1)

+    def test_group_flag_is_inconclusive_when_group_properties_missing(self):
+        feature_flag = {
+            "id": 1,
+            "name": "Group Flag Without Property Filters",
+            "key": "group-flag-no-props",
+            "active": True,
+            "filters": {
+                "aggregation_group_type_index": 0,
+                "groups": [{"properties": [], "rollout_percentage": 100}],
+            },
+        }
+        self.client.group_type_mapping = {"0": "company"}
+
+        with self.assertRaises(InconclusiveMatchError):
+            self.client._compute_flag_locally(
+                feature_flag,
+                "some-distinct-id",
+                groups={"company": "acme"},
+                group_properties={},
+            )
+
    @mock.patch("posthog.client.flags")
    @mock.patch("posthog.client.get")
    def test_flag_with_complex_definition(self, patch_get, patch_flags):
@@ -2513,7 +2534,10 @@ class TestLocalEvaluation(unittest.TestCase):
        self.assertIsNone(client._flags_etag)
        self.assertEqual(client.feature_flags[0]["key"], "flag-v2")

-    def test_load_feature_flags_wrong_key(self):
+    @mock.patch("posthog.client.Poller")
+    @mock.patch("posthog.client.get")
+    def test_load_feature_flags_wrong_key(self, patch_get, _patch_poll):
+        patch_get.side_effect = APIError(401, "Unauthorized")
        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)

        with self.assertLogs("posthog", level="ERROR") as logs:
@@ -3220,6 +3244,541 @@ class TestLocalEvaluation(unittest.TestCase):
        # Verify API was called (fallback occurred)
        self.assertEqual(patch_flags.call_count, 1)

+    @mock.patch("posthog.client.flags")
+    def test_device_id_bucketing_uses_device_id_for_hash(self, patch_flags):
+        """
+        When a flag has bucketing_identifier: "device_id", the device_id should be
+        used for hashing instead of distinct_id.
+        """
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        # This flag uses device_id for bucketing
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "device-bucketed-flag",
+                "active": True,
+                "filters": {
+                    "bucketing_identifier": "device_id",
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            }
+        ]
+
+        # Same distinct_id with different device_ids should produce different results
+        # (based on rollout percentage, we check consistency)
+        result1 = client.get_feature_flag(
+            "device-bucketed-flag", "user-123", device_id="device-A"
+        )
+        result2 = client.get_feature_flag(
+            "device-bucketed-flag", "user-123", device_id="device-A"
+        )
+
+        # Same device_id should give consistent results
+        self.assertEqual(result1, result2)
+
+        # No API fallback should occur
+        self.assertEqual(patch_flags.call_count, 0)
+
+    def test_match_feature_flag_properties_without_bucketing_value_is_deprecated(
+        self,
+    ):
+        """
+        match_feature_flag_properties should preserve backward compatibility when
+        bucketing_value is omitted, while warning about deprecation.
+        """
+        from posthog.feature_flags import match_feature_flag_properties
+
+        flag = {
+            "id": 1,
+            "key": "device-bucketed-flag",
+            "active": True,
+            "filters": {
+                "bucketing_identifier": "device_id",
+                "groups": [
+                    {
+                        "properties": [],
+                        "rollout_percentage": 100,
+                    }
+                ],
+            },
+        }
+
+        with self.assertWarnsRegex(
+            DeprecationWarning, "without bucketing_value is deprecated"
+        ):
+            result = match_feature_flag_properties(
+                flag,
+                "user-123",
+                {},
+                device_id="device-123",
+            )
+
+        self.assertTrue(result)
+
+    @mock.patch("posthog.client.flags")
+    def test_device_id_bucketing_same_device_different_users_same_result(
+        self, patch_flags
+    ):
+        """
+        When a flag uses device_id bucketing, different distinct_ids with the same
+        device_id should get the same result.
+        """
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "device-bucketed-flag",
+                "active": True,
+                "filters": {
+                    "bucketing_identifier": "device_id",
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 50,
+                        }
+                    ],
+                },
+            }
+        ]
+
+        # Different distinct_ids with the same device_id should get the same result
+        result1 = client.get_feature_flag(
+            "device-bucketed-flag", "user-A", device_id="shared-device"
+        )
+        result2 = client.get_feature_flag(
+            "device-bucketed-flag", "user-B", device_id="shared-device"
+        )
+        result3 = client.get_feature_flag(
+            "device-bucketed-flag", "user-C", device_id="shared-device"
+        )
+
+        # All should be the same since device_id is the same
+        self.assertEqual(result1, result2)
+        self.assertEqual(result2, result3)
+
+        self.assertEqual(patch_flags.call_count, 0)
+
+    @mock.patch("posthog.client.flags")
+    def test_device_id_bucketing_fallback_when_device_id_missing(self, patch_flags):
+        """
+        When a flag requires device_id for bucketing but none is provided,
+        it should fallback to server evaluation.
+        """
+        patch_flags.return_value = {"featureFlags": {"device-bucketed-flag": True}}
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "device-bucketed-flag",
+                "active": True,
+                "filters": {
+                    "bucketing_identifier": "device_id",
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            }
+        ]
+
+        # No device_id provided - should fallback to API
+        result = client.get_feature_flag("device-bucketed-flag", "user-123")
+
+        self.assertTrue(result)
+        # API should have been called
+        self.assertEqual(patch_flags.call_count, 1)
+
+    @mock.patch("posthog.client.flags")
+    def test_device_id_bucketing_returns_none_when_only_evaluate_locally_and_no_device_id(
+        self, patch_flags
+    ):
+        """
+        When only_evaluate_locally=True and device_id is required but missing,
+        should return None instead of falling back to API.
+        """
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "device-bucketed-flag",
+                "active": True,
+                "filters": {
+                    "bucketing_identifier": "device_id",
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            }
+        ]
+
+        # No device_id + only_evaluate_locally should return None
+        result = client.get_feature_flag(
+            "device-bucketed-flag", "user-123", only_evaluate_locally=True
+        )
+
+        self.assertIsNone(result)
+        # API should NOT have been called
+        self.assertEqual(patch_flags.call_count, 0)
+
+    @mock.patch("posthog.client.flags")
+    def test_default_bucketing_identifier_uses_distinct_id(self, patch_flags):
+        """
+        When bucketing_identifier is not set or is 'distinct_id', should use
+        distinct_id for hashing (default behavior).
+        """
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        # Flag without bucketing_identifier (defaults to distinct_id)
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "normal-flag",
+                "active": True,
+                "filters": {
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 50,
+                        }
+                    ],
+                },
+            }
+        ]
+
+        # Different distinct_ids should potentially produce different results
+        # but same distinct_id should produce same result
+        result1 = client.get_feature_flag("normal-flag", "user-A")
+        result2 = client.get_feature_flag("normal-flag", "user-A")
+
+        self.assertEqual(result1, result2)
+        self.assertEqual(patch_flags.call_count, 0)
+
+    @mock.patch("posthog.client.flags")
+    def test_device_id_bucketing_with_multivariate_flag(self, patch_flags):
+        """
+        Multivariate flag variant selection should use device_id when
+        bucketing_identifier is set to device_id.
+        """
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "multivariate-device-flag",
+                "active": True,
+                "filters": {
+                    "bucketing_identifier": "device_id",
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                    "multivariate": {
+                        "variants": [
+                            {"key": "control", "rollout_percentage": 50},
+                            {"key": "test", "rollout_percentage": 50},
+                        ]
+                    },
+                },
+            }
+        ]
+
+        # Same device_id should give same variant
+        result1 = client.get_feature_flag(
+            "multivariate-device-flag", "user-A", device_id="device-1"
+        )
+        result2 = client.get_feature_flag(
+            "multivariate-device-flag", "user-B", device_id="device-1"
+        )
+
+        # Both should get the same variant because device_id is the same
+        self.assertEqual(result1, result2)
+        self.assertIn(result1, ["control", "test"])
+
+        self.assertEqual(patch_flags.call_count, 0)
+
+    @mock.patch("posthog.client.flags")
+    def test_device_id_bucketing_from_context(self, patch_flags):
+        """
+        When device_id is not passed as a parameter but is set in the context,
+        it should be resolved from context.
+        """
+        from posthog.contexts import new_context, set_context_device_id
+
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "device-bucketed-flag",
+                "active": True,
+                "filters": {
+                    "bucketing_identifier": "device_id",
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            }
+        ]
+
+        # Set device_id in context
+        with new_context():
+            set_context_device_id("context-device-id")
+            result = client.get_feature_flag("device-bucketed-flag", "user-123")
+
+        # Should evaluate locally using the context device_id
+        self.assertTrue(result)
+        self.assertEqual(patch_flags.call_count, 0)
+
+    @mock.patch("posthog.client.flags")
+    def test_group_flags_ignore_bucketing_identifier(self, patch_flags):
+        """
+        Group flags should continue to use the group identifier for hashing,
+        regardless of the bucketing_identifier setting.
+        """
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "group-flag",
+                "active": True,
+                "filters": {
+                    "aggregation_group_type_index": 0,
+                    "bucketing_identifier": "device_id",  # Should be ignored for group flags
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            }
+        ]
+        client.group_type_mapping = {"0": "company"}
+
+        # Even with bucketing_identifier set to device_id, group flag should use group identifier
+        result = client.get_feature_flag(
+            "group-flag",
+            "user-123",
+            groups={"company": "acme-inc"},
+            device_id="some-device",
+        )
+
+        self.assertTrue(result)
+        self.assertEqual(patch_flags.call_count, 0)
+
+    @mock.patch("posthog.client.flags")
+    def test_group_flag_dependency_receives_device_id(self, patch_flags):
+        """
+        Group flag dependency evaluation should receive device_id so dependent
+        device_id-bucketed flags can be evaluated locally.
+        """
+        patch_flags.return_value = {"featureFlags": {"group-parent-flag": "from-api"}}
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "device-dependent-flag",
+                "active": True,
+                "filters": {
+                    "bucketing_identifier": "device_id",
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            },
+            {
+                "id": 2,
+                "key": "group-parent-flag",
+                "active": True,
+                "filters": {
+                    "aggregation_group_type_index": 0,
+                    "groups": [
+                        {
+                            "properties": [
+                                {
+                                    "key": "device-dependent-flag",
+                                    "operator": "flag_evaluates_to",
+                                    "value": True,
+                                    "type": "flag",
+                                    "dependency_chain": ["device-dependent-flag"],
+                                }
+                            ],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            },
+        ]
+        client.group_type_mapping = {"0": "company"}
+
+        result = client.get_feature_flag(
+            "group-parent-flag",
+            "user-123",
+            groups={"company": "acme-inc"},
+            device_id="device-123",
+        )
+
+        self.assertTrue(result)
+        self.assertEqual(patch_flags.call_count, 0)
+
+    @mock.patch("posthog.client.flags")
+    def test_group_flag_dependency_ignores_device_id_bucketing_identifier(
+        self, patch_flags
+    ):
+        """
+        Group flag dependencies should keep bucketing by group key, even when
+        the dependent group flag has bucketing_identifier set to device_id.
+        """
+        patch_flags.return_value = {"featureFlags": {"parent-group-flag": "from-api"}}
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "child-group-flag",
+                "active": True,
+                "filters": {
+                    "aggregation_group_type_index": 0,
+                    "bucketing_identifier": "device_id",
+                    "groups": [
+                        {
+                            "properties": [],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            },
+            {
+                "id": 2,
+                "key": "parent-group-flag",
+                "active": True,
+                "filters": {
+                    "aggregation_group_type_index": 0,
+                    "groups": [
+                        {
+                            "properties": [
+                                {
+                                    "key": "child-group-flag",
+                                    "operator": "flag_evaluates_to",
+                                    "value": True,
+                                    "type": "flag",
+                                    "dependency_chain": ["child-group-flag"],
+                                }
+                            ],
+                            "rollout_percentage": 100,
+                        }
+                    ],
+                },
+            },
+        ]
+        client.group_type_mapping = {"0": "company"}
+
+        result = client.get_feature_flag(
+            "parent-group-flag",
+            "user-123",
+            groups={"company": "acme-inc"},
+        )
+
+        self.assertTrue(result)
+        self.assertEqual(patch_flags.call_count, 0)
+
+    @mock.patch("posthog.client.flags")
+    def test_get_all_flags_with_device_id_bucketing(self, patch_flags):
+        """
+        get_all_flags_and_payloads should properly handle flags with device_id bucketing.
+        """
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "normal-flag",
+                "active": True,
+                "filters": {
+                    "groups": [{"properties": [], "rollout_percentage": 100}],
+                },
+            },
+            {
+                "id": 2,
+                "key": "device-flag",
+                "active": True,
+                "filters": {
+                    "bucketing_identifier": "device_id",
+                    "groups": [{"properties": [], "rollout_percentage": 100}],
+                },
+            },
+        ]
+
+        # With device_id provided, both flags should be evaluated locally
+        result = client.get_all_flags("user-123", device_id="my-device")
+
+        self.assertEqual(result["normal-flag"], True)
+        self.assertEqual(result["device-flag"], True)
+        self.assertEqual(patch_flags.call_count, 0)
+
+    @mock.patch("posthog.client.flags")
+    def test_get_all_flags_fallback_when_device_id_missing_for_some_flags(
+        self, patch_flags
+    ):
+        """
+        When some flags require device_id but it's not provided, those flags
+        should trigger fallback while others can be evaluated locally.
+        """
+        patch_flags.return_value = {
+            "featureFlags": {"normal-flag": True, "device-flag": "from-api"}
+        }
+        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+
+        client.feature_flags = [
+            {
+                "id": 1,
+                "key": "normal-flag",
+                "active": True,
+                "filters": {
+                    "groups": [{"properties": [], "rollout_percentage": 100}],
+                },
+            },
+            {
+                "id": 2,
+                "key": "device-flag",
+                "active": True,
+                "filters": {
+                    "bucketing_identifier": "device_id",
+                    "groups": [{"properties": [], "rollout_percentage": 100}],
+                },
+            },
+        ]
+
+        # Without device_id, device-flag can't be evaluated locally
+        client.get_all_flags("user-123")
+
+        # Should fallback to API for all flags when any can't be evaluated locally
+        self.assertEqual(patch_flags.call_count, 1)
+

 class TestMatchProperties(unittest.TestCase):
    def property(self, key, value, operator=None):
@@ -4266,13 +4825,15 @@ class TestCaptureCalls(unittest.TestCase):
            disable_geoip=False,
        )

-    @mock.patch("posthog.client.MAX_DICT_SIZE", 100)
    @mock.patch.object(Client, "capture")
    @mock.patch("posthog.client.flags")
    def test_capture_multiple_users_doesnt_out_of_memory(
        self, patch_flags, patch_capture
    ):
        client = Client(FAKE_TEST_API_KEY, personal_api_key=FAKE_TEST_API_KEY)
+        # Set on the instance to avoid relying on module-constant patching behavior
+        # across Python/runtime implementations.
+        client.distinct_ids_feature_flags_reported.max_size = 100
        client.feature_flags = [
            {
                "id": 1,
@@ -1,4 +1 @@
-VERSION = "7.7.0"
-
-if __name__ == "__main__":
-    print(VERSION, end="")  # noqa: T201
+VERSION = "7.9.2"
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"

 [project]
 name = "posthog"
-dynamic = ["version"]
+version = "7.9.2"
 description = "Integrate PostHog into any python application."
 authors = [{ name = "PostHog", email = "hey@posthog.com" }]
 maintainers = [{ name = "PostHog", email = "hey@posthog.com" }]
@@ -93,9 +93,6 @@ packages = [
    "posthog.integrations",
 ]

-[tool.setuptools.dynamic]
-version = { attr = "posthog.version.VERSION" }
-
 [tool.pytest.ini_options]
 asyncio_mode = "auto"
 asyncio_default_fixture_loop_scope = "function"
@@ -0,0 +1,22 @@
+FROM python:3.12-slim
+
+WORKDIR /app
+
+# Copy the SDK source code
+COPY posthog/ /app/sdk/posthog/
+COPY setup.py pyproject.toml README.md LICENSE /app/sdk/
+
+# Install the SDK from source
+RUN cd /app/sdk && pip install --no-cache-dir -e .
+
+# Install adapter dependencies
+RUN pip install --no-cache-dir flask python-dateutil
+
+# Copy adapter code
+COPY sdk_compliance_adapter/adapter.py /app/adapter.py
+
+# Expose port 8080
+EXPOSE 8080
+
+# Run the adapter
+CMD ["python", "/app/adapter.py"]
@@ -0,0 +1,76 @@
+# PostHog Python SDK Test Adapter
+
+This adapter wraps the posthog-python SDK for compliance testing with the [PostHog SDK Test Harness](https://github.com/PostHog/posthog-sdk-test-harness).
+
+## What is This?
+
+This is a simple Flask app that:
+1. Wraps the posthog-python SDK
+2. Exposes a REST API for the test harness to control
+3. Tracks internal SDK state for test assertions
+
+## Running Tests
+
+Tests run automatically in CI via GitHub Actions. See the test harness repo for details.
+
+### Locally with Docker Compose
+
+```bash
+# From the posthog-python/sdk_compliance_adapter directory
+docker-compose up --build --abort-on-container-exit
+```
+
+This will:
+1. Build the Python SDK adapter
+2. Pull the test harness image
+3. Run all compliance tests
+4. Show results
+
+### Manually with Docker
+
+```bash
+# Create network
+docker network create test-network
+
+# Build and run adapter
+docker build -f sdk_compliance_adapter/Dockerfile -t posthog-python-adapter .
+docker run -d --name sdk-adapter --network test-network -p 8080:8080 posthog-python-adapter
+
+# Run test harness
+docker run --rm \
+  --name test-harness \
+  --network test-network \
+  ghcr.io/posthog/sdk-test-harness:latest \
+  run --adapter-url http://sdk-adapter:8080 --mock-url http://test-harness:8081
+
+# Cleanup
+docker stop sdk-adapter && docker rm sdk-adapter
+docker network rm test-network
+```
+
+## Adapter Implementation
+
+See [adapter.py](adapter.py) for the implementation.
+
+The adapter implements the standard SDK adapter interface defined in the [test harness CONTRACT](https://github.com/PostHog/posthog-sdk-test-harness/blob/main/CONTRACT.yaml):
+
+- `GET /health` - Return SDK information
+- `POST /init` - Initialize SDK with config
+- `POST /capture` - Capture an event
+- `POST /flush` - Flush pending events
+- `GET /state` - Return internal state
+- `POST /reset` - Reset SDK state
+
+### Key Implementation Details
+
+**Request Tracking**: The adapter monkey-patches `batch_post` to track all HTTP requests made by the SDK, including retries.
+
+**State Management**: Thread-safe state tracking for events captured vs sent, retry attempts, and errors.
+
+**UUID Tracking**: Extracts and tracks UUIDs from batches to verify deduplication.
+
+## Documentation
+
+For complete documentation on the test harness and how to implement adapters, see:
+- [PostHog SDK Test Harness](https://github.com/PostHog/posthog-sdk-test-harness)
+- [Adapter Implementation Guide](https://github.com/PostHog/posthog-sdk-test-harness/blob/main/ADAPTER_GUIDE.md)
@@ -0,0 +1,383 @@
+"""
+PostHog Python SDK Test Adapter
+
+This adapter implements the SDK Test Adapter Interface defined in the PostHog Capture API Contract.
+It wraps the posthog-python SDK and exposes a REST API for the test harness to exercise.
+"""
+
+import logging
+import os
+import threading
+import time
+from typing import Any, Dict, List, Optional
+
+from flask import Flask, jsonify, request
+
+from posthog import Client
+from posthog.request import batch_post as original_batch_post
+from posthog.version import VERSION
+
+# Configure logging
+logging.basicConfig(
+    level=logging.DEBUG, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
+)
+logger = logging.getLogger(__name__)
+
+app = Flask(__name__)
+
+
+class RequestInfo:
+    """Information about an HTTP request made by the SDK"""
+
+    def __init__(
+        self,
+        timestamp_ms: int,
+        status_code: int,
+        retry_attempt: int,
+        event_count: int,
+        uuid_list: List[str],
+    ):
+        self.timestamp_ms = timestamp_ms
+        self.status_code = status_code
+        self.retry_attempt = retry_attempt
+        self.event_count = event_count
+        self.uuid_list = uuid_list
+
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "timestamp_ms": self.timestamp_ms,
+            "status_code": self.status_code,
+            "retry_attempt": self.retry_attempt,
+            "event_count": self.event_count,
+            "uuid_list": self.uuid_list,
+        }
+
+
+class SDKState:
+    """Tracks SDK internal state for test assertions"""
+
+    def __init__(self):
+        self.lock = threading.Lock()
+        self.pending_events = 0
+        self.total_events_captured = 0
+        self.total_events_sent = 0
+        self.total_retries = 0
+        self.last_error: Optional[str] = None
+        self.requests_made: List[RequestInfo] = []
+        self.client: Optional[Client] = None
+        self.retry_attempts: Dict[str, int] = {}  # Track retry attempts by batch ID
+
+    def reset(self):
+        """Reset all state"""
+        with self.lock:
+            self.pending_events = 0
+            self.total_events_captured = 0
+            self.total_events_sent = 0
+            self.total_retries = 0
+            self.last_error = None
+            self.requests_made = []
+            self.retry_attempts = {}
+            if self.client:
+                # Flush and shutdown existing client
+                try:
+                    self.client.shutdown()
+                except Exception as e:
+                    logger.warning(f"Error shutting down client: {e}")
+                self.client = None
+
+    def increment_captured(self):
+        """Increment total events captured"""
+        with self.lock:
+            self.total_events_captured += 1
+            self.pending_events += 1
+
+    def record_request(self, status_code: int, batch: List[Dict], batch_id: str):
+        """Record an HTTP request made by the SDK"""
+        with self.lock:
+            # Determine retry attempt for this batch
+            retry_attempt = self.retry_attempts.get(batch_id, 0)
+
+            # Extract UUIDs from batch
+            uuid_list = [event.get("uuid", "") for event in batch]
+
+            request_info = RequestInfo(
+                timestamp_ms=int(time.time() * 1000),
+                status_code=status_code,
+                retry_attempt=retry_attempt,
+                event_count=len(batch),
+                uuid_list=uuid_list,
+            )
+            self.requests_made.append(request_info)
+
+            # Update counters
+            if status_code == 200:
+                # Success - clear pending events
+                self.total_events_sent += len(batch)
+                self.pending_events = max(0, self.pending_events - len(batch))
+                # Remove batch from retry tracking
+                self.retry_attempts.pop(batch_id, None)
+            else:
+                # Failure - increment retry count
+                self.retry_attempts[batch_id] = retry_attempt + 1
+                if retry_attempt > 0:
+                    self.total_retries += 1
+
+    def record_error(self, error: str):
+        """Record an error"""
+        with self.lock:
+            self.last_error = error
+
+    def get_state(self) -> Dict[str, Any]:
+        """Get current state as dict"""
+        with self.lock:
+            return {
+                "pending_events": self.pending_events,
+                "total_events_captured": self.total_events_captured,
+                "total_events_sent": self.total_events_sent,
+                "total_retries": self.total_retries,
+                "last_error": self.last_error,
+                "requests_made": [r.to_dict() for r in self.requests_made],
+            }
+
+
+# Global state
+state = SDKState()
+
+
+def create_batch_id(batch: List[Dict]) -> str:
+    """Create a unique ID for a batch based on UUIDs"""
+    uuids = sorted([event.get("uuid", "") for event in batch])
+    return "-".join(uuids[:3])  # Use first 3 UUIDs as batch ID
+
+
+def patched_batch_post(
+    api_key: str,
+    host: Optional[str] = None,
+    gzip: bool = False,
+    timeout: int = 15,
+    **kwargs,
+):
+    """Patched version of batch_post that tracks requests"""
+    batch = kwargs.get("batch", [])
+    batch_id = create_batch_id(batch)
+
+    try:
+        # Call original batch_post
+        response = original_batch_post(api_key, host, gzip, timeout, **kwargs)
+        # Record successful request
+        state.record_request(200, batch, batch_id)
+        return response
+    except Exception as e:
+        # Record failed request
+        status_code = (
+            getattr(e, "status_code", 500) if hasattr(e, "status_code") else 500
+        )
+        state.record_request(status_code, batch, batch_id)
+        state.record_error(str(e))
+        raise
+
+
+# Monkey-patch the batch_post function
+import posthog.request  # noqa: E402
+
+posthog.request.batch_post = patched_batch_post
+
+# Also patch in consumer module
+import posthog.consumer  # noqa: E402
+
+posthog.consumer.batch_post = patched_batch_post
+
+
+@app.route("/health", methods=["GET"])
+def health():
+    """Health check endpoint"""
+    return jsonify(
+        {
+            "sdk_name": "posthog-python",
+            "sdk_version": VERSION,
+            "adapter_version": "1.0.0",
+        }
+    )
+
+
+@app.route("/init", methods=["POST"])
+def init():
+    """Initialize the SDK client"""
+    try:
+        data = request.json or {}
+
+        # Reset state
+        state.reset()
+
+        # Extract config
+        api_key = data.get("api_key")
+        host = data.get("host")
+        flush_at = data.get("flush_at", 100)
+        flush_interval_ms = data.get("flush_interval_ms", 500)
+        max_retries = data.get("max_retries", 3)
+        enable_compression = data.get("enable_compression", False)
+
+        if not api_key:
+            return jsonify({"error": "api_key is required"}), 400
+        if not host:
+            return jsonify({"error": "host is required"}), 400
+
+        # Convert flush_interval from ms to seconds
+        flush_interval = flush_interval_ms / 1000.0
+
+        # Create client
+        client = Client(
+            project_api_key=api_key,
+            host=host,
+            flush_at=flush_at,
+            flush_interval=flush_interval,
+            gzip=enable_compression,
+            max_retries=max_retries,
+            debug=True,
+        )
+
+        state.client = client
+
+        logger.info(
+            f"Initialized SDK with api_key={api_key[:10]}..., host={host}, "
+            f"flush_at={flush_at}, flush_interval={flush_interval}, "
+            f"max_retries={max_retries}, gzip={enable_compression}"
+        )
+
+        return jsonify({"success": True})
+    except Exception as e:
+        logger.exception("Error initializing SDK")
+        return jsonify({"error": str(e)}), 500
+
+
+@app.route("/capture", methods=["POST"])
+def capture():
+    """Capture a single event"""
+    try:
+        if not state.client:
+            return jsonify({"error": "SDK not initialized"}), 400
+
+        data = request.json or {}
+
+        distinct_id = data.get("distinct_id")
+        event = data.get("event")
+        properties = data.get("properties")
+        timestamp = data.get("timestamp")
+
+        if not distinct_id:
+            return jsonify({"error": "distinct_id is required"}), 400
+        if not event:
+            return jsonify({"error": "event is required"}), 400
+
+        # Capture event
+        kwargs = {"distinct_id": distinct_id, "properties": properties}
+        if timestamp:
+            # Parse ISO8601 timestamp
+            from dateutil.parser import parse  # type: ignore[import-untyped]
+
+            kwargs["timestamp"] = parse(timestamp)
+
+        uuid = state.client.capture(event, **kwargs)
+
+        # Track that we captured an event
+        state.increment_captured()
+
+        logger.info(f"Captured event: {event} for {distinct_id}, uuid={uuid}")
+
+        return jsonify({"success": True, "uuid": uuid})
+    except Exception as e:
+        logger.exception("Error capturing event")
+        state.record_error(str(e))
+        return jsonify({"error": str(e)}), 500
+
+
+@app.route("/identify", methods=["POST"])
+def identify():
+    """Identify a user"""
+    try:
+        if not state.client:
+            return jsonify({"error": "SDK not initialized"}), 400
+
+        data = request.json or {}
+
+        distinct_id = data.get("distinct_id")
+        properties = data.get("properties")
+        properties_set_once = data.get("properties_set_once")
+
+        if not distinct_id:
+            return jsonify({"error": "distinct_id is required"}), 400
+
+        # Use the identify pattern - set + set_once
+        if properties:
+            state.client.set(distinct_id=distinct_id, properties=properties)
+            state.increment_captured()
+
+        if properties_set_once:
+            state.client.set_once(
+                distinct_id=distinct_id, properties=properties_set_once
+            )
+            state.increment_captured()
+
+        logger.info(f"Identified user: {distinct_id}")
+
+        return jsonify({"success": True})
+    except Exception as e:
+        logger.exception("Error identifying user")
+        state.record_error(str(e))
+        return jsonify({"error": str(e)}), 500
+
+
+@app.route("/flush", methods=["POST"])
+def flush():
+    """Force flush all pending events"""
+    try:
+        if not state.client:
+            return jsonify({"error": "SDK not initialized"}), 400
+
+        # Flush and wait
+        state.client.flush()
+
+        # Wait a bit for flush to complete
+        # The flush() method triggers queue.join() which blocks until all items are processed
+        time.sleep(0.5)
+
+        logger.info("Flushed pending events")
+
+        return jsonify({"success": True, "events_flushed": state.total_events_sent})
+    except Exception as e:
+        logger.exception("Error flushing events")
+        state.record_error(str(e))
+        return jsonify({"error": str(e), "errors": [str(e)]}, 500)
+
+
+@app.route("/state", methods=["GET"])
+def get_state():
+    """Get internal SDK state"""
+    try:
+        return jsonify(state.get_state())
+    except Exception as e:
+        logger.exception("Error getting state")
+        return jsonify({"error": str(e)}), 500
+
+
+@app.route("/reset", methods=["POST"])
+def reset():
+    """Reset SDK state"""
+    try:
+        state.reset()
+        logger.info("Reset SDK state")
+        return jsonify({"success": True})
+    except Exception as e:
+        logger.exception("Error resetting state")
+        return jsonify({"error": str(e)}), 500
+
+
+def main():
+    """Main entry point"""
+    port = int(os.environ.get("PORT", 8080))
+    logger.info(f"Starting SDK Test Adapter on port {port}")
+    app.run(host="0.0.0.0", port=port, debug=False)
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,25 @@
+version: "3.8"
+
+services:
+  # PostHog Python SDK adapter
+  sdk-adapter:
+    build:
+      context: ..
+      dockerfile: sdk_compliance_adapter/Dockerfile
+    ports:
+      - "8080:8080"
+    networks:
+      - test-network
+
+  # Test harness
+  test-harness:
+    image: ghcr.io/posthog/sdk-test-harness:latest
+    command: ["run", "--adapter-url", "http://sdk-adapter:8080", "--mock-url", "http://test-harness:8081"]
+    networks:
+      - test-network
+    depends_on:
+      - sdk-adapter
+
+networks:
+  test-network:
+    driver: bridge
@@ -0,0 +1,3 @@
+# SDK Test Adapter dependencies
+flask>=3.0.0
+python-dateutil>=2.8.0
@@ -23,6 +23,7 @@ with open("pyproject.toml", "rb") as f:

 # Override specific values
 config["project"]["name"] = "posthoganalytics"
+config["project"]["readme"] = "README_ANALYTICS.md"
 config["tool"]["setuptools"]["dynamic"]["version"] = {
    "attr": "posthoganalytics.version.VERSION"
 }
@@ -278,43 +278,31 @@ sdist = { url = "https://files.pythonhosted.org/packages/eb/56/b1ba7935a17738ae8
 wheels = [
    { url = "https://files.pythonhosted.org/packages/50/bd/b1a6362b80628111e6653c961f987faa55262b4002fcec42308cad1db680/cffi-2.0.0-cp310-cp310-manylinux1_i686.manylinux2014_i686.manylinux_2_17_i686.manylinux_2_5_i686.whl", hash = "sha256:53f77cbe57044e88bbd5ed26ac1d0514d2acf0591dd6bb02a3ae37f76811b80c", size = 208811, upload-time = "2025-09-08T23:22:12.267Z" },
    { url = "https://files.pythonhosted.org/packages/4f/27/6933a8b2562d7bd1fb595074cf99cc81fc3789f6a6c05cdabb46284a3188/cffi-2.0.0-cp310-cp310-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:3e837e369566884707ddaf85fc1744b47575005c0a229de3327f8f9a20f4efeb", size = 216402, upload-time = "2025-09-08T23:22:13.455Z" },
-    { url = "https://files.pythonhosted.org/packages/05/eb/b86f2a2645b62adcfff53b0dd97e8dfafb5c8aa864bd0d9a2c2049a0d551/cffi-2.0.0-cp310-cp310-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:5eda85d6d1879e692d546a078b44251cdd08dd1cfb98dfb77b670c97cee49ea0", size = 203217, upload-time = "2025-09-08T23:22:14.596Z" },
-    { url = "https://files.pythonhosted.org/packages/9f/e0/6cbe77a53acf5acc7c08cc186c9928864bd7c005f9efd0d126884858a5fe/cffi-2.0.0-cp310-cp310-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:9332088d75dc3241c702d852d4671613136d90fa6881da7d770a483fd05248b4", size = 203079, upload-time = "2025-09-08T23:22:15.769Z" },
    { url = "https://files.pythonhosted.org/packages/98/29/9b366e70e243eb3d14a5cb488dfd3a0b6b2f1fb001a203f653b93ccfac88/cffi-2.0.0-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:fc7de24befaeae77ba923797c7c87834c73648a05a4bde34b3b7e5588973a453", size = 216475, upload-time = "2025-09-08T23:22:17.427Z" },
    { url = "https://files.pythonhosted.org/packages/21/7a/13b24e70d2f90a322f2900c5d8e1f14fa7e2a6b3332b7309ba7b2ba51a5a/cffi-2.0.0-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:cf364028c016c03078a23b503f02058f1814320a56ad535686f90565636a9495", size = 218829, upload-time = "2025-09-08T23:22:19.069Z" },
    { url = "https://files.pythonhosted.org/packages/60/99/c9dc110974c59cc981b1f5b66e1d8af8af764e00f0293266824d9c4254bc/cffi-2.0.0-cp310-cp310-musllinux_1_2_i686.whl", hash = "sha256:e11e82b744887154b182fd3e7e8512418446501191994dbf9c9fc1f32cc8efd5", size = 211211, upload-time = "2025-09-08T23:22:20.588Z" },
    { url = "https://files.pythonhosted.org/packages/49/72/ff2d12dbf21aca1b32a40ed792ee6b40f6dc3a9cf1644bd7ef6e95e0ac5e/cffi-2.0.0-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:8ea985900c5c95ce9db1745f7933eeef5d314f0565b27625d9a10ec9881e1bfb", size = 218036, upload-time = "2025-09-08T23:22:22.143Z" },
    { url = "https://files.pythonhosted.org/packages/b1/b7/1200d354378ef52ec227395d95c2576330fd22a869f7a70e88e1447eb234/cffi-2.0.0-cp311-cp311-manylinux1_i686.manylinux2014_i686.manylinux_2_17_i686.manylinux_2_5_i686.whl", hash = "sha256:baf5215e0ab74c16e2dd324e8ec067ef59e41125d3eade2b863d294fd5035c92", size = 209613, upload-time = "2025-09-08T23:22:29.475Z" },
    { url = "https://files.pythonhosted.org/packages/b8/56/6033f5e86e8cc9bb629f0077ba71679508bdf54a9a5e112a3c0b91870332/cffi-2.0.0-cp311-cp311-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:730cacb21e1bdff3ce90babf007d0a0917cc3e6492f336c2f0134101e0944f93", size = 216476, upload-time = "2025-09-08T23:22:31.063Z" },
-    { url = "https://files.pythonhosted.org/packages/dc/7f/55fecd70f7ece178db2f26128ec41430d8720f2d12ca97bf8f0a628207d5/cffi-2.0.0-cp311-cp311-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:6824f87845e3396029f3820c206e459ccc91760e8fa24422f8b0c3d1731cbec5", size = 203374, upload-time = "2025-09-08T23:22:32.507Z" },
-    { url = "https://files.pythonhosted.org/packages/84/ef/a7b77c8bdc0f77adc3b46888f1ad54be8f3b7821697a7b89126e829e676a/cffi-2.0.0-cp311-cp311-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:9de40a7b0323d889cf8d23d1ef214f565ab154443c42737dfe52ff82cf857664", size = 202597, upload-time = "2025-09-08T23:22:34.132Z" },
    { url = "https://files.pythonhosted.org/packages/d7/91/500d892b2bf36529a75b77958edfcd5ad8e2ce4064ce2ecfeab2125d72d1/cffi-2.0.0-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:8941aaadaf67246224cee8c3803777eed332a19d909b47e29c9842ef1e79ac26", size = 215574, upload-time = "2025-09-08T23:22:35.443Z" },
    { url = "https://files.pythonhosted.org/packages/44/64/58f6255b62b101093d5df22dcb752596066c7e89dd725e0afaed242a61be/cffi-2.0.0-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:a05d0c237b3349096d3981b727493e22147f934b20f6f125a3eba8f994bec4a9", size = 218971, upload-time = "2025-09-08T23:22:36.805Z" },
    { url = "https://files.pythonhosted.org/packages/ab/49/fa72cebe2fd8a55fbe14956f9970fe8eb1ac59e5df042f603ef7c8ba0adc/cffi-2.0.0-cp311-cp311-musllinux_1_2_i686.whl", hash = "sha256:94698a9c5f91f9d138526b48fe26a199609544591f859c870d477351dc7b2414", size = 211972, upload-time = "2025-09-08T23:22:38.436Z" },
    { url = "https://files.pythonhosted.org/packages/0b/28/dd0967a76aab36731b6ebfe64dec4e981aff7e0608f60c2d46b46982607d/cffi-2.0.0-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:5fed36fccc0612a53f1d4d9a816b50a36702c28a2aa880cb8a122b3466638743", size = 217078, upload-time = "2025-09-08T23:22:39.776Z" },
    { url = "https://files.pythonhosted.org/packages/ff/df/a4f0fbd47331ceeba3d37c2e51e9dfc9722498becbeec2bd8bc856c9538a/cffi-2.0.0-cp312-cp312-manylinux1_i686.manylinux2014_i686.manylinux_2_17_i686.manylinux_2_5_i686.whl", hash = "sha256:21d1152871b019407d8ac3985f6775c079416c282e431a4da6afe7aefd2bccbe", size = 212529, upload-time = "2025-09-08T23:22:47.349Z" },
    { url = "https://files.pythonhosted.org/packages/d5/72/12b5f8d3865bf0f87cf1404d8c374e7487dcf097a1c91c436e72e6badd83/cffi-2.0.0-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:b21e08af67b8a103c71a250401c78d5e0893beff75e28c53c98f4de42f774062", size = 220097, upload-time = "2025-09-08T23:22:48.677Z" },
-    { url = "https://files.pythonhosted.org/packages/c2/95/7a135d52a50dfa7c882ab0ac17e8dc11cec9d55d2c18dda414c051c5e69e/cffi-2.0.0-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:1e3a615586f05fc4065a8b22b8152f0c1b00cdbc60596d187c2a74f9e3036e4e", size = 207983, upload-time = "2025-09-08T23:22:50.06Z" },
-    { url = "https://files.pythonhosted.org/packages/3a/c8/15cb9ada8895957ea171c62dc78ff3e99159ee7adb13c0123c001a2546c1/cffi-2.0.0-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:81afed14892743bbe14dacb9e36d9e0e504cd204e0b165062c488942b9718037", size = 206519, upload-time = "2025-09-08T23:22:51.364Z" },
    { url = "https://files.pythonhosted.org/packages/78/2d/7fa73dfa841b5ac06c7b8855cfc18622132e365f5b81d02230333ff26e9e/cffi-2.0.0-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:3e17ed538242334bf70832644a32a7aae3d83b57567f9fd60a26257e992b79ba", size = 219572, upload-time = "2025-09-08T23:22:52.902Z" },
    { url = "https://files.pythonhosted.org/packages/07/e0/267e57e387b4ca276b90f0434ff88b2c2241ad72b16d31836adddfd6031b/cffi-2.0.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:3925dd22fa2b7699ed2617149842d2e6adde22b262fcbfada50e3d195e4b3a94", size = 222963, upload-time = "2025-09-08T23:22:54.518Z" },
    { url = "https://files.pythonhosted.org/packages/b6/75/1f2747525e06f53efbd878f4d03bac5b859cbc11c633d0fb81432d98a795/cffi-2.0.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:2c8f814d84194c9ea681642fd164267891702542f028a15fc97d4674b6206187", size = 221361, upload-time = "2025-09-08T23:22:55.867Z" },
    { url = "https://files.pythonhosted.org/packages/b0/1e/d22cc63332bd59b06481ceaac49d6c507598642e2230f201649058a7e704/cffi-2.0.0-cp313-cp313-manylinux1_i686.manylinux2014_i686.manylinux_2_17_i686.manylinux_2_5_i686.whl", hash = "sha256:07b271772c100085dd28b74fa0cd81c8fb1a3ba18b21e03d7c27f3436a10606b", size = 212446, upload-time = "2025-09-08T23:23:03.472Z" },
    { url = "https://files.pythonhosted.org/packages/a9/f5/a2c23eb03b61a0b8747f211eb716446c826ad66818ddc7810cc2cc19b3f2/cffi-2.0.0-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:d48a880098c96020b02d5a1f7d9251308510ce8858940e6fa99ece33f610838b", size = 220101, upload-time = "2025-09-08T23:23:04.792Z" },
-    { url = "https://files.pythonhosted.org/packages/f2/7f/e6647792fc5850d634695bc0e6ab4111ae88e89981d35ac269956605feba/cffi-2.0.0-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:f93fd8e5c8c0a4aa1f424d6173f14a892044054871c771f8566e4008eaa359d2", size = 207948, upload-time = "2025-09-08T23:23:06.127Z" },
-    { url = "https://files.pythonhosted.org/packages/cb/1e/a5a1bd6f1fb30f22573f76533de12a00bf274abcdc55c8edab639078abb6/cffi-2.0.0-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:dd4f05f54a52fb558f1ba9f528228066954fee3ebe629fc1660d874d040ae5a3", size = 206422, upload-time = "2025-09-08T23:23:07.753Z" },
    { url = "https://files.pythonhosted.org/packages/98/df/0a1755e750013a2081e863e7cd37e0cdd02664372c754e5560099eb7aa44/cffi-2.0.0-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:c8d3b5532fc71b7a77c09192b4a5a200ea992702734a2e9279a37f2478236f26", size = 219499, upload-time = "2025-09-08T23:23:09.648Z" },
    { url = "https://files.pythonhosted.org/packages/50/e1/a969e687fcf9ea58e6e2a928ad5e2dd88cc12f6f0ab477e9971f2309b57c/cffi-2.0.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:d9b29c1f0ae438d5ee9acb31cadee00a58c46cc9c0b2f9038c6b0b3470877a8c", size = 222928, upload-time = "2025-09-08T23:23:10.928Z" },
    { url = "https://files.pythonhosted.org/packages/36/54/0362578dd2c9e557a28ac77698ed67323ed5b9775ca9d3fe73fe191bb5d8/cffi-2.0.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:6d50360be4546678fc1b79ffe7a66265e28667840010348dd69a314145807a1b", size = 221302, upload-time = "2025-09-08T23:23:12.42Z" },
    { url = "https://files.pythonhosted.org/packages/d6/43/0e822876f87ea8a4ef95442c3d766a06a51fc5298823f884ef87aaad168c/cffi-2.0.0-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:24b6f81f1983e6df8db3adc38562c83f7d4a0c36162885ec7f7b77c7dcbec97b", size = 220049, upload-time = "2025-09-08T23:23:20.853Z" },
-    { url = "https://files.pythonhosted.org/packages/b4/89/76799151d9c2d2d1ead63c2429da9ea9d7aac304603de0c6e8764e6e8e70/cffi-2.0.0-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:12873ca6cb9b0f0d3a0da705d6086fe911591737a59f28b7936bdfed27c0d47c", size = 207793, upload-time = "2025-09-08T23:23:22.08Z" },
-    { url = "https://files.pythonhosted.org/packages/bb/dd/3465b14bb9e24ee24cb88c9e3730f6de63111fffe513492bf8c808a3547e/cffi-2.0.0-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:d9b97165e8aed9272a6bb17c01e3cc5871a594a446ebedc996e2397a1c1ea8ef", size = 206300, upload-time = "2025-09-08T23:23:23.314Z" },
    { url = "https://files.pythonhosted.org/packages/47/d9/d83e293854571c877a92da46fdec39158f8d7e68da75bf73581225d28e90/cffi-2.0.0-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:afb8db5439b81cf9c9d0c80404b60c3cc9c3add93e114dcae767f1477cb53775", size = 219244, upload-time = "2025-09-08T23:23:24.541Z" },
    { url = "https://files.pythonhosted.org/packages/2b/0f/1f177e3683aead2bb00f7679a16451d302c436b5cbf2505f0ea8146ef59e/cffi-2.0.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:737fe7d37e1a1bffe70bd5754ea763a62a066dc5913ca57e957824b72a85e205", size = 222828, upload-time = "2025-09-08T23:23:26.143Z" },
    { url = "https://files.pythonhosted.org/packages/c6/0f/cafacebd4b040e3119dcb32fed8bdef8dfe94da653155f9d0b9dc660166e/cffi-2.0.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:38100abb9d1b1435bc4cc340bb4489635dc2f0da7456590877030c9b3d40b0c1", size = 220926, upload-time = "2025-09-08T23:23:27.873Z" },
    { url = "https://files.pythonhosted.org/packages/be/b4/c56878d0d1755cf9caa54ba71e5d049479c52f9e4afc230f06822162ab2f/cffi-2.0.0-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:7cc09976e8b56f8cebd752f7113ad07752461f48a58cbba644139015ac24954c", size = 221593, upload-time = "2025-09-08T23:23:31.91Z" },
-    { url = "https://files.pythonhosted.org/packages/e0/0d/eb704606dfe8033e7128df5e90fee946bbcb64a04fcdaa97321309004000/cffi-2.0.0-cp314-cp314t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:92b68146a71df78564e4ef48af17551a5ddd142e5190cdf2c5624d0c3ff5b2e8", size = 209354, upload-time = "2025-09-08T23:23:33.214Z" },
-    { url = "https://files.pythonhosted.org/packages/d8/19/3c435d727b368ca475fb8742ab97c9cb13a0de600ce86f62eab7fa3eea60/cffi-2.0.0-cp314-cp314t-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:b1e74d11748e7e98e2f426ab176d4ed720a64412b6a15054378afdb71e0f37dc", size = 208480, upload-time = "2025-09-08T23:23:34.495Z" },
    { url = "https://files.pythonhosted.org/packages/d0/44/681604464ed9541673e486521497406fadcc15b5217c3e326b061696899a/cffi-2.0.0-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:28a3a209b96630bca57cce802da70c266eb08c6e97e5afd61a75611ee6c64592", size = 221584, upload-time = "2025-09-08T23:23:36.096Z" },
    { url = "https://files.pythonhosted.org/packages/25/8e/342a504ff018a2825d395d44d63a767dd8ebc927ebda557fecdaca3ac33a/cffi-2.0.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:7553fb2090d71822f02c629afe6042c299edf91ba1bf94951165613553984512", size = 224443, upload-time = "2025-09-08T23:23:37.328Z" },
    { url = "https://files.pythonhosted.org/packages/e1/5e/b666bacbbc60fbf415ba9988324a132c9a7a0448a9a8f125074671c0f2c3/cffi-2.0.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:6c6c373cfc5c83a975506110d17457138c8c63016b563cc9ed6e056a82f13ce4", size = 223437, upload-time = "2025-09-08T23:23:38.945Z" },
@@ -540,10 +528,8 @@ wheels = [
    { url = "https://files.pythonhosted.org/packages/5c/49/498c86566a1d80e978b42f0d702795f69887005548c041636df6ae1ca64c/cryptography-46.0.3-cp311-abi3-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:01ca9ff2885f3acc98c29f1860552e37f6d7c7d013d7334ff2a9de43a449315d", size = 4450807, upload-time = "2025-10-15T23:16:56.414Z" },
    { url = "https://files.pythonhosted.org/packages/4b/0a/863a3604112174c8624a2ac3c038662d9e59970c7f926acdcfaed8d61142/cryptography-46.0.3-cp311-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:6eae65d4c3d33da080cff9c4ab1f711b15c1d9760809dad6ea763f3812d254cb", size = 4299615, upload-time = "2025-10-15T23:16:58.442Z" },
    { url = "https://files.pythonhosted.org/packages/64/02/b73a533f6b64a69f3cd3872acb6ebc12aef924d8d103133bb3ea750dc703/cryptography-46.0.3-cp311-abi3-manylinux_2_28_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:e5bf0ed4490068a2e72ac03d786693adeb909981cc596425d09032d372bcc849", size = 4016800, upload-time = "2025-10-15T23:17:00.378Z" },
-    { url = "https://files.pythonhosted.org/packages/25/d5/16e41afbfa450cde85a3b7ec599bebefaef16b5c6ba4ec49a3532336ed72/cryptography-46.0.3-cp311-abi3-manylinux_2_28_ppc64le.whl", hash = "sha256:5ecfccd2329e37e9b7112a888e76d9feca2347f12f37918facbb893d7bb88ee8", size = 4984707, upload-time = "2025-10-15T23:17:01.98Z" },
    { url = "https://files.pythonhosted.org/packages/c9/56/e7e69b427c3878352c2fb9b450bd0e19ed552753491d39d7d0a2f5226d41/cryptography-46.0.3-cp311-abi3-manylinux_2_28_x86_64.whl", hash = "sha256:a2c0cd47381a3229c403062f764160d57d4d175e022c1df84e168c6251a22eec", size = 4482541, upload-time = "2025-10-15T23:17:04.078Z" },
    { url = "https://files.pythonhosted.org/packages/78/f6/50736d40d97e8483172f1bb6e698895b92a223dba513b0ca6f06b2365339/cryptography-46.0.3-cp311-abi3-manylinux_2_34_aarch64.whl", hash = "sha256:549e234ff32571b1f4076ac269fcce7a808d3bf98b76c8dd560e42dbc66d7d91", size = 4299464, upload-time = "2025-10-15T23:17:05.483Z" },
-    { url = "https://files.pythonhosted.org/packages/00/de/d8e26b1a855f19d9994a19c702fa2e93b0456beccbcfe437eda00e0701f2/cryptography-46.0.3-cp311-abi3-manylinux_2_34_ppc64le.whl", hash = "sha256:c0a7bb1a68a5d3471880e264621346c48665b3bf1c3759d682fc0864c540bd9e", size = 4950838, upload-time = "2025-10-15T23:17:07.425Z" },
    { url = "https://files.pythonhosted.org/packages/8f/29/798fc4ec461a1c9e9f735f2fc58741b0daae30688f41b2497dcbc9ed1355/cryptography-46.0.3-cp311-abi3-manylinux_2_34_x86_64.whl", hash = "sha256:10b01676fc208c3e6feeb25a8b83d81767e8059e1fe86e1dc62d10a3018fa926", size = 4481596, upload-time = "2025-10-15T23:17:09.343Z" },
    { url = "https://files.pythonhosted.org/packages/15/8d/03cd48b20a573adfff7652b76271078e3045b9f49387920e7f1f631d125e/cryptography-46.0.3-cp311-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:0abf1ffd6e57c67e92af68330d05760b7b7efb243aab8377e583284dbab72c71", size = 4426782, upload-time = "2025-10-15T23:17:11.22Z" },
    { url = "https://files.pythonhosted.org/packages/fa/b1/ebacbfe53317d55cf33165bda24c86523497a6881f339f9aae5c2e13e57b/cryptography-46.0.3-cp311-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:a04bee9ab6a4da801eb9b51f1b708a1b5b5c9eb48c03f74198464c66f0d344ac", size = 4698381, upload-time = "2025-10-15T23:17:12.829Z" },
@@ -551,10 +537,8 @@ wheels = [
    { url = "https://files.pythonhosted.org/packages/c5/fd/bc1daf8230eaa075184cbbf5f8cd00ba9db4fd32d63fb83da4671b72ed8a/cryptography-46.0.3-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:39b6755623145ad5eff1dab323f4eae2a32a77a7abef2c5089a04a3d04366715", size = 4435078, upload-time = "2025-10-15T23:17:23.042Z" },
    { url = "https://files.pythonhosted.org/packages/82/98/d3bd5407ce4c60017f8ff9e63ffee4200ab3e23fe05b765cab805a7db008/cryptography-46.0.3-cp314-cp314t-manylinux_2_28_aarch64.whl", hash = "sha256:db391fa7c66df6762ee3f00c95a89e6d428f4d60e7abc8328f4fe155b5ac6e54", size = 4293460, upload-time = "2025-10-15T23:17:24.885Z" },
    { url = "https://files.pythonhosted.org/packages/26/e9/e23e7900983c2b8af7a08098db406cf989d7f09caea7897e347598d4cd5b/cryptography-46.0.3-cp314-cp314t-manylinux_2_28_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:78a97cf6a8839a48c49271cdcbd5cf37ca2c1d6b7fdd86cc864f302b5e9bf459", size = 3995237, upload-time = "2025-10-15T23:17:26.449Z" },
-    { url = "https://files.pythonhosted.org/packages/91/15/af68c509d4a138cfe299d0d7ddb14afba15233223ebd933b4bbdbc7155d3/cryptography-46.0.3-cp314-cp314t-manylinux_2_28_ppc64le.whl", hash = "sha256:dfb781ff7eaa91a6f7fd41776ec37c5853c795d3b358d4896fdbb5df168af422", size = 4967344, upload-time = "2025-10-15T23:17:28.06Z" },
    { url = "https://files.pythonhosted.org/packages/ca/e3/8643d077c53868b681af077edf6b3cb58288b5423610f21c62aadcbe99f4/cryptography-46.0.3-cp314-cp314t-manylinux_2_28_x86_64.whl", hash = "sha256:6f61efb26e76c45c4a227835ddeae96d83624fb0d29eb5df5b96e14ed1a0afb7", size = 4466564, upload-time = "2025-10-15T23:17:29.665Z" },
    { url = "https://files.pythonhosted.org/packages/0e/43/c1e8726fa59c236ff477ff2b5dc071e54b21e5a1e51aa2cee1676f1c986f/cryptography-46.0.3-cp314-cp314t-manylinux_2_34_aarch64.whl", hash = "sha256:23b1a8f26e43f47ceb6d6a43115f33a5a37d57df4ea0ca295b780ae8546e8044", size = 4292415, upload-time = "2025-10-15T23:17:31.686Z" },
-    { url = "https://files.pythonhosted.org/packages/42/f9/2f8fefdb1aee8a8e3256a0568cffc4e6d517b256a2fe97a029b3f1b9fe7e/cryptography-46.0.3-cp314-cp314t-manylinux_2_34_ppc64le.whl", hash = "sha256:b419ae593c86b87014b9be7396b385491ad7f320bde96826d0dd174459e54665", size = 4931457, upload-time = "2025-10-15T23:17:33.478Z" },
    { url = "https://files.pythonhosted.org/packages/79/30/9b54127a9a778ccd6d27c3da7563e9f2d341826075ceab89ae3b41bf5be2/cryptography-46.0.3-cp314-cp314t-manylinux_2_34_x86_64.whl", hash = "sha256:50fc3343ac490c6b08c0cf0d704e881d0d660be923fd3076db3e932007e726e3", size = 4466074, upload-time = "2025-10-15T23:17:35.158Z" },
    { url = "https://files.pythonhosted.org/packages/ac/68/b4f4a10928e26c941b1b6a179143af9f4d27d88fe84a6a3c53592d2e76bf/cryptography-46.0.3-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:22d7e97932f511d6b0b04f2bfd818d73dcd5928db509460aaf48384778eb6d20", size = 4420569, upload-time = "2025-10-15T23:17:37.188Z" },
    { url = "https://files.pythonhosted.org/packages/a3/49/3746dab4c0d1979888f125226357d3262a6dd40e114ac29e3d2abdf1ec55/cryptography-46.0.3-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:d55f3dffadd674514ad19451161118fd010988540cee43d8bc20675e775925de", size = 4681941, upload-time = "2025-10-15T23:17:39.236Z" },
@@ -562,10 +546,8 @@ wheels = [
    { url = "https://files.pythonhosted.org/packages/26/42/fa8389d4478368743e24e61eea78846a0006caffaf72ea24a15159215a14/cryptography-46.0.3-cp38-abi3-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:15ab9b093e8f09daab0f2159bb7e47532596075139dd74365da52ecc9cb46c5d", size = 4440029, upload-time = "2025-10-15T23:17:49.837Z" },
    { url = "https://files.pythonhosted.org/packages/5f/eb/f483db0ec5ac040824f269e93dd2bd8a21ecd1027e77ad7bdf6914f2fd80/cryptography-46.0.3-cp38-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:46acf53b40ea38f9c6c229599a4a13f0d46a6c3fa9ef19fc1a124d62e338dfa0", size = 4297222, upload-time = "2025-10-15T23:17:51.357Z" },
    { url = "https://files.pythonhosted.org/packages/fd/cf/da9502c4e1912cb1da3807ea3618a6829bee8207456fbbeebc361ec38ba3/cryptography-46.0.3-cp38-abi3-manylinux_2_28_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:10ca84c4668d066a9878890047f03546f3ae0a6b8b39b697457b7757aaf18dbc", size = 4012280, upload-time = "2025-10-15T23:17:52.964Z" },
-    { url = "https://files.pythonhosted.org/packages/6b/8f/9adb86b93330e0df8b3dcf03eae67c33ba89958fc2e03862ef1ac2b42465/cryptography-46.0.3-cp38-abi3-manylinux_2_28_ppc64le.whl", hash = "sha256:36e627112085bb3b81b19fed209c05ce2a52ee8b15d161b7c643a7d5a88491f3", size = 4978958, upload-time = "2025-10-15T23:17:54.965Z" },
    { url = "https://files.pythonhosted.org/packages/d1/a0/5fa77988289c34bdb9f913f5606ecc9ada1adb5ae870bd0d1054a7021cc4/cryptography-46.0.3-cp38-abi3-manylinux_2_28_x86_64.whl", hash = "sha256:1000713389b75c449a6e979ffc7dcc8ac90b437048766cef052d4d30b8220971", size = 4473714, upload-time = "2025-10-15T23:17:56.754Z" },
    { url = "https://files.pythonhosted.org/packages/14/e5/fc82d72a58d41c393697aa18c9abe5ae1214ff6f2a5c18ac470f92777895/cryptography-46.0.3-cp38-abi3-manylinux_2_34_aarch64.whl", hash = "sha256:b02cf04496f6576afffef5ddd04a0cb7d49cf6be16a9059d793a30b035f6b6ac", size = 4296970, upload-time = "2025-10-15T23:17:58.588Z" },
-    { url = "https://files.pythonhosted.org/packages/78/06/5663ed35438d0b09056973994f1aec467492b33bd31da36e468b01ec1097/cryptography-46.0.3-cp38-abi3-manylinux_2_34_ppc64le.whl", hash = "sha256:71e842ec9bc7abf543b47cf86b9a743baa95f4677d22baa4c7d5c69e49e9bc04", size = 4940236, upload-time = "2025-10-15T23:18:00.897Z" },
    { url = "https://files.pythonhosted.org/packages/fc/59/873633f3f2dcd8a053b8dd1d38f783043b5fce589c0f6988bf55ef57e43e/cryptography-46.0.3-cp38-abi3-manylinux_2_34_x86_64.whl", hash = "sha256:402b58fc32614f00980b66d6e56a5b4118e6cb362ae8f3fda141ba4689bd4506", size = 4472642, upload-time = "2025-10-15T23:18:02.749Z" },
    { url = "https://files.pythonhosted.org/packages/3d/39/8e71f3930e40f6877737d6f69248cf74d4e34b886a3967d32f919cc50d3b/cryptography-46.0.3-cp38-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:ef639cb3372f69ec44915fafcd6698b6cc78fbe0c2ea41be867f6ed612811963", size = 4423126, upload-time = "2025-10-15T23:18:04.85Z" },
    { url = "https://files.pythonhosted.org/packages/cd/c7/f65027c2810e14c3e7268353b1681932b87e5a48e65505d8cc17c99e36ae/cryptography-46.0.3-cp38-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:3b51b8ca4f1c6453d8829e1eb7299499ca7f313900dd4d89a24b8b87c0a780d4", size = 4686573, upload-time = "2025-10-15T23:18:06.908Z" },
@@ -672,7 +654,7 @@ name = "exceptiongroup"
 version = "1.3.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "typing-extensions", marker = "python_full_version < '3.11'" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/0b/9f/a65090624ecf468cdca03533906e7c69ed7588582240cfe7cc9e770b50eb/exceptiongroup-1.3.0.tar.gz", hash = "sha256:b241f5885f560bc56a59ee63ca4c6a8bfa46ae4ad651af316d4e81817bb9fd88", size = 29749, upload-time = "2025-05-10T17:42:51.123Z" }
 wheels = [
@@ -863,7 +845,6 @@ wheels = [
    { url = "https://files.pythonhosted.org/packages/7d/ed/6bfa4109fcb23a58819600392564fea69cdc6551ffd5e69ccf1d52a40cbc/greenlet-3.2.4-cp310-cp310-macosx_11_0_universal2.whl", hash = "sha256:8c68325b0d0acf8d91dde4e6f930967dd52a5302cd4062932a6b2e7c2969f47c", size = 271061, upload-time = "2025-08-07T13:17:15.373Z" },
    { url = "https://files.pythonhosted.org/packages/2a/fc/102ec1a2fc015b3a7652abab7acf3541d58c04d3d17a8d3d6a44adae1eb1/greenlet-3.2.4-cp310-cp310-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:94385f101946790ae13da500603491f04a76b6e4c059dab271b3ce2e283b2590", size = 629475, upload-time = "2025-08-07T13:42:54.009Z" },
    { url = "https://files.pythonhosted.org/packages/c5/26/80383131d55a4ac0fb08d71660fd77e7660b9db6bdb4e8884f46d9f2cc04/greenlet-3.2.4-cp310-cp310-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:f10fd42b5ee276335863712fa3da6608e93f70629c631bf77145021600abc23c", size = 640802, upload-time = "2025-08-07T13:45:25.52Z" },
-    { url = "https://files.pythonhosted.org/packages/9f/7c/e7833dbcd8f376f3326bd728c845d31dcde4c84268d3921afcae77d90d08/greenlet-3.2.4-cp310-cp310-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:c8c9e331e58180d0d83c5b7999255721b725913ff6bc6cf39fa2a45841a4fd4b", size = 636703, upload-time = "2025-08-07T13:53:12.622Z" },
    { url = "https://files.pythonhosted.org/packages/e9/49/547b93b7c0428ede7b3f309bc965986874759f7d89e4e04aeddbc9699acb/greenlet-3.2.4-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:58b97143c9cc7b86fc458f215bd0932f1757ce649e05b640fea2e79b54cedb31", size = 635417, upload-time = "2025-08-07T13:18:25.189Z" },
    { url = "https://files.pythonhosted.org/packages/7f/91/ae2eb6b7979e2f9b035a9f612cf70f1bf54aad4e1d125129bef1eae96f19/greenlet-3.2.4-cp310-cp310-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:c2ca18a03a8cfb5b25bc1cbe20f3d9a4c80d8c3b13ba3df49ac3961af0b1018d", size = 584358, upload-time = "2025-08-07T13:18:23.708Z" },
    { url = "https://files.pythonhosted.org/packages/f7/85/433de0c9c0252b22b16d413c9407e6cb3b41df7389afc366ca204dbc1393/greenlet-3.2.4-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:9fe0a28a7b952a21e2c062cd5756d34354117796c6d9215a87f55e38d15402c5", size = 1113550, upload-time = "2025-08-07T13:42:37.467Z" },
@@ -874,7 +855,6 @@ wheels = [
    { url = "https://files.pythonhosted.org/packages/a4/de/f28ced0a67749cac23fecb02b694f6473f47686dff6afaa211d186e2ef9c/greenlet-3.2.4-cp311-cp311-macosx_11_0_universal2.whl", hash = "sha256:96378df1de302bc38e99c3a9aa311967b7dc80ced1dcc6f171e99842987882a2", size = 272305, upload-time = "2025-08-07T13:15:41.288Z" },
    { url = "https://files.pythonhosted.org/packages/09/16/2c3792cba130000bf2a31c5272999113f4764fd9d874fb257ff588ac779a/greenlet-3.2.4-cp311-cp311-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:1ee8fae0519a337f2329cb78bd7a8e128ec0f881073d43f023c7b8d4831d5246", size = 632472, upload-time = "2025-08-07T13:42:55.044Z" },
    { url = "https://files.pythonhosted.org/packages/ae/8f/95d48d7e3d433e6dae5b1682e4292242a53f22df82e6d3dda81b1701a960/greenlet-3.2.4-cp311-cp311-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:94abf90142c2a18151632371140b3dba4dee031633fe614cb592dbb6c9e17bc3", size = 644646, upload-time = "2025-08-07T13:45:26.523Z" },
-    { url = "https://files.pythonhosted.org/packages/d5/5e/405965351aef8c76b8ef7ad370e5da58d57ef6068df197548b015464001a/greenlet-3.2.4-cp311-cp311-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:4d1378601b85e2e5171b99be8d2dc85f594c79967599328f95c1dc1a40f1c633", size = 640519, upload-time = "2025-08-07T13:53:13.928Z" },
    { url = "https://files.pythonhosted.org/packages/25/5d/382753b52006ce0218297ec1b628e048c4e64b155379331f25a7316eb749/greenlet-3.2.4-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:0db5594dce18db94f7d1650d7489909b57afde4c580806b8d9203b6e79cdc079", size = 639707, upload-time = "2025-08-07T13:18:27.146Z" },
    { url = "https://files.pythonhosted.org/packages/1f/8e/abdd3f14d735b2929290a018ecf133c901be4874b858dd1c604b9319f064/greenlet-3.2.4-cp311-cp311-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:2523e5246274f54fdadbce8494458a2ebdcdbc7b802318466ac5606d3cded1f8", size = 587684, upload-time = "2025-08-07T13:18:25.164Z" },
    { url = "https://files.pythonhosted.org/packages/5d/65/deb2a69c3e5996439b0176f6651e0052542bb6c8f8ec2e3fba97c9768805/greenlet-3.2.4-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:1987de92fec508535687fb807a5cea1560f6196285a4cde35c100b8cd632cc52", size = 1116647, upload-time = "2025-08-07T13:42:38.655Z" },
@@ -885,7 +865,6 @@ wheels = [
    { url = "https://files.pythonhosted.org/packages/44/69/9b804adb5fd0671f367781560eb5eb586c4d495277c93bde4307b9e28068/greenlet-3.2.4-cp312-cp312-macosx_11_0_universal2.whl", hash = "sha256:3b67ca49f54cede0186854a008109d6ee71f66bd57bb36abd6d0a0267b540cdd", size = 274079, upload-time = "2025-08-07T13:15:45.033Z" },
    { url = "https://files.pythonhosted.org/packages/46/e9/d2a80c99f19a153eff70bc451ab78615583b8dac0754cfb942223d2c1a0d/greenlet-3.2.4-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:ddf9164e7a5b08e9d22511526865780a576f19ddd00d62f8a665949327fde8bb", size = 640997, upload-time = "2025-08-07T13:42:56.234Z" },
    { url = "https://files.pythonhosted.org/packages/3b/16/035dcfcc48715ccd345f3a93183267167cdd162ad123cd93067d86f27ce4/greenlet-3.2.4-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:f28588772bb5fb869a8eb331374ec06f24a83a9c25bfa1f38b6993afe9c1e968", size = 655185, upload-time = "2025-08-07T13:45:27.624Z" },
-    { url = "https://files.pythonhosted.org/packages/31/da/0386695eef69ffae1ad726881571dfe28b41970173947e7c558d9998de0f/greenlet-3.2.4-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:5c9320971821a7cb77cfab8d956fa8e39cd07ca44b6070db358ceb7f8797c8c9", size = 649926, upload-time = "2025-08-07T13:53:15.251Z" },
    { url = "https://files.pythonhosted.org/packages/68/88/69bf19fd4dc19981928ceacbc5fd4bb6bc2215d53199e367832e98d1d8fe/greenlet-3.2.4-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:c60a6d84229b271d44b70fb6e5fa23781abb5d742af7b808ae3f6efd7c9c60f6", size = 651839, upload-time = "2025-08-07T13:18:30.281Z" },
    { url = "https://files.pythonhosted.org/packages/19/0d/6660d55f7373b2ff8152401a83e02084956da23ae58cddbfb0b330978fe9/greenlet-3.2.4-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:3b3812d8d0c9579967815af437d96623f45c0f2ae5f04e366de62a12d83a8fb0", size = 607586, upload-time = "2025-08-07T13:18:28.544Z" },
    { url = "https://files.pythonhosted.org/packages/8e/1a/c953fdedd22d81ee4629afbb38d2f9d71e37d23caace44775a3a969147d4/greenlet-3.2.4-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:abbf57b5a870d30c4675928c37278493044d7c14378350b3aa5d484fa65575f0", size = 1123281, upload-time = "2025-08-07T13:42:39.858Z" },
@@ -896,7 +875,6 @@ wheels = [
    { url = "https://files.pythonhosted.org/packages/49/e8/58c7f85958bda41dafea50497cbd59738c5c43dbbea5ee83d651234398f4/greenlet-3.2.4-cp313-cp313-macosx_11_0_universal2.whl", hash = "sha256:1a921e542453fe531144e91e1feedf12e07351b1cf6c9e8a3325ea600a715a31", size = 272814, upload-time = "2025-08-07T13:15:50.011Z" },
    { url = "https://files.pythonhosted.org/packages/62/dd/b9f59862e9e257a16e4e610480cfffd29e3fae018a68c2332090b53aac3d/greenlet-3.2.4-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:cd3c8e693bff0fff6ba55f140bf390fa92c994083f838fece0f63be121334945", size = 641073, upload-time = "2025-08-07T13:42:57.23Z" },
    { url = "https://files.pythonhosted.org/packages/f7/0b/bc13f787394920b23073ca3b6c4a7a21396301ed75a655bcb47196b50e6e/greenlet-3.2.4-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:710638eb93b1fa52823aa91bf75326f9ecdfd5e0466f00789246a5280f4ba0fc", size = 655191, upload-time = "2025-08-07T13:45:29.752Z" },
-    { url = "https://files.pythonhosted.org/packages/f2/d6/6adde57d1345a8d0f14d31e4ab9c23cfe8e2cd39c3baf7674b4b0338d266/greenlet-3.2.4-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:c5111ccdc9c88f423426df3fd1811bfc40ed66264d35aa373420a34377efc98a", size = 649516, upload-time = "2025-08-07T13:53:16.314Z" },
    { url = "https://files.pythonhosted.org/packages/7f/3b/3a3328a788d4a473889a2d403199932be55b1b0060f4ddd96ee7cdfcad10/greenlet-3.2.4-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:d76383238584e9711e20ebe14db6c88ddcedc1829a9ad31a584389463b5aa504", size = 652169, upload-time = "2025-08-07T13:18:32.861Z" },
    { url = "https://files.pythonhosted.org/packages/ee/43/3cecdc0349359e1a527cbf2e3e28e5f8f06d3343aaf82ca13437a9aa290f/greenlet-3.2.4-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:23768528f2911bcd7e475210822ffb5254ed10d71f4028387e5a99b4c6699671", size = 610497, upload-time = "2025-08-07T13:18:31.636Z" },
    { url = "https://files.pythonhosted.org/packages/b8/19/06b6cf5d604e2c382a6f31cafafd6f33d5dea706f4db7bdab184bad2b21d/greenlet-3.2.4-cp313-cp313-musllinux_1_1_aarch64.whl", hash = "sha256:00fadb3fedccc447f517ee0d3fd8fe49eae949e1cd0f6a611818f4f6fb7dc83b", size = 1121662, upload-time = "2025-08-07T13:42:41.117Z" },
@@ -907,7 +885,6 @@ wheels = [
    { url = "https://files.pythonhosted.org/packages/22/5c/85273fd7cc388285632b0498dbbab97596e04b154933dfe0f3e68156c68c/greenlet-3.2.4-cp314-cp314-macosx_11_0_universal2.whl", hash = "sha256:49a30d5fda2507ae77be16479bdb62a660fa51b1eb4928b524975b3bde77b3c0", size = 273586, upload-time = "2025-08-07T13:16:08.004Z" },
    { url = "https://files.pythonhosted.org/packages/d1/75/10aeeaa3da9332c2e761e4c50d4c3556c21113ee3f0afa2cf5769946f7a3/greenlet-3.2.4-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:299fd615cd8fc86267b47597123e3f43ad79c9d8a22bebdce535e53550763e2f", size = 686346, upload-time = "2025-08-07T13:42:59.944Z" },
    { url = "https://files.pythonhosted.org/packages/c0/aa/687d6b12ffb505a4447567d1f3abea23bd20e73a5bed63871178e0831b7a/greenlet-3.2.4-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:c17b6b34111ea72fc5a4e4beec9711d2226285f0386ea83477cbb97c30a3f3a5", size = 699218, upload-time = "2025-08-07T13:45:30.969Z" },
-    { url = "https://files.pythonhosted.org/packages/dc/8b/29aae55436521f1d6f8ff4e12fb676f3400de7fcf27fccd1d4d17fd8fecd/greenlet-3.2.4-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:b4a1870c51720687af7fa3e7cda6d08d801dae660f75a76f3845b642b4da6ee1", size = 694659, upload-time = "2025-08-07T13:53:17.759Z" },
    { url = "https://files.pythonhosted.org/packages/92/2e/ea25914b1ebfde93b6fc4ff46d6864564fba59024e928bdc7de475affc25/greenlet-3.2.4-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:061dc4cf2c34852b052a8620d40f36324554bc192be474b9e9770e8c042fd735", size = 695355, upload-time = "2025-08-07T13:18:34.517Z" },
    { url = "https://files.pythonhosted.org/packages/72/60/fc56c62046ec17f6b0d3060564562c64c862948c9d4bc8aa807cf5bd74f4/greenlet-3.2.4-cp314-cp314-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:44358b9bf66c8576a9f57a590d5f5d6e72fa4228b763d0e43fee6d3b06d3a337", size = 657512, upload-time = "2025-08-07T13:18:33.969Z" },
    { url = "https://files.pythonhosted.org/packages/23/6e/74407aed965a4ab6ddd93a7ded3180b730d281c77b765788419484cdfeef/greenlet-3.2.4-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:2917bdf657f5859fbf3386b12d68ede4cf1f04c90c3a6bc1f013dd68a22e2269", size = 1612508, upload-time = "2025-11-04T12:42:23.427Z" },
@@ -2152,6 +2129,7 @@ wheels = [

 [[package]]
 name = "posthog"
+version = "7.9.2"
 source = { editable = "." }
 dependencies = [
    { name = "backoff" },
Author	SHA1	Message	Date
github-actions[bot]	16bc87b646	chore: Release v7.9.2	2026-02-18 22:05:00 +00:00
Rafael AudibertandGitHub	f1dc4d7391	chore: Migrate releases to `sampo` (#398 ) * chore: Migrate releases to `sampo` This is much closer to what we have in `posthog-js`, let's see if it's a good thing! There's still a lot to do before deploying this: - updating CI - updating README with instructions * Add sampo changeset * chore: Update to relase Python via Slack + sampo * Update release.yml * fix: Use pyproject.toml version as source of truth	2026-02-18 19:02:19 -03:00
Radu RaiceaandGitHub	23dae56d68	fix(ai): bind prompt reads to project token (#433 ) * fix(ai): bind prompt reads to project token * chore(release): bump version to 7.8.7	2026-02-17 16:58:59 +00:00
Anders GitHub Claude Opus 4.6	73bec043cf	chore: release v7.9.0 (#434 ) chore: bump version to 7.9.0 Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>	2026-02-17 10:32:05 +01:00
AndersandGitHub	603ed376dd	feat: Support device_id as bucketing identifier for local evaluation (#424 ) * feat: Support device_id as bucketing identifier for local evaluation Add support for `bucketing_identifier` field on feature flags to allow using `device_id` instead of `distinct_id` for hashing/bucketing in local evaluation.	2026-02-16 11:05:09 +01:00
AndersandGitHub	bb0c7b4fa8	test(flags): make wrong-key load_feature_flags deterministic (#432 )	2026-02-13 15:53:22 +01:00
Aleksander BłaszkiewiczandGitHub	499194e0c4	feat: limit max number of items in collection to scan (#430 ) * feat: limit max number of items in collection to scan * feat: changelog * fix: format * feat: test * feat: replace entire collection instead of truncating	2026-02-11 14:59:11 +01:00
Aleksander BłaszkiewiczandGitHub	ffb8e9b591	feat: further optimize code variables regex search (#429 ) * feat: initial * fix: ruff	2026-02-09 23:59:03 +01:00
Aleksander BłaszkiewiczandGitHub	7780ca8390	fix: long variables pattern matching (#428 )	2026-02-09 17:45:23 +01:00
AndersandGitHub	bca175214d	fix: Retry on 408 and respect Retry-After header (#426 ) * fix: Retry on 408 and respect Retry-After header 408 (Request Timeout) was incorrectly treated as a non-retryable client error. Retry-After response headers were ignored during backoff. Replace backoff library usage with a manual retry loop that honours Retry-After when present and falls back to exponential backoff otherwise. * fix: Parse HTTP-date Retry-After values Retry-After can be seconds or an HTTP-date per RFC 7231. Fall back to email.utils.parsedate_to_datetime when the numeric parse fails. * fix: Don't retry on unclassifiable APIError status When APIError.status is "N/A" (no HTTP status), treat it as non-retryable to avoid unexpected retry loops on errors the SDK cannot classify. * test: Add retry delay tests for Retry-After and exponential backoff Verify time.sleep is called with the Retry-After value when present, uses exponential backoff (2^attempt) when absent, and that 408 is retried.	2026-02-09 12:15:49 +00:00
Aleksander BłaszkiewiczandGitHub	fe3a9bbf75	fix: openai image sanitization (#425 )	2026-02-06 14:15:57 +01:00
Vikas Kapadiya GitHub Aleksander Błaszkiewicz	b6e66330e5	fix: openAI input image sanitization (#384 ) Co-authored-by: Aleksander Błaszkiewicz <kqmdjc8@gmail.com>	2026-02-06 13:53:02 +01:00
Gabriel GrinbergandGitHub	4f32fa4100	Fix feature flag 401 errors causing HTTP request storm (#422 ) * Fix feature flag 401 errors causing HTTP request storm Set feature_flags = [] on 401 error to prevent repeated requests. * Clear flag_cache, group_type_mapping, cohorts on 401	2026-02-04 10:31:10 -05:00
Radu RaiceaandGitHub	f5719f39da	fix(llma): default prompts url (#423 )	2026-02-04 15:10:00 +00:00
Radu RaiceaandGitHub	d4f2d6dfb0	fix(llma): small fixes for prompt management (#420 ) * fix(llma): small fixes for prompt management * fix(llma): tests * fix(llma): tests	2026-02-04 09:49:19 +02:00
José SequeiraandGitHub	72f448816c	feat: SDK Compliance (#397 ) * feat: SDK Compliance	2026-01-30 16:12:43 +01:00
Radu RaiceaandGitHub	4350389f93	feat(llma): add prompt management (#417 ) * feat(llma): add prompt management * chore(llma): bump version * fix(llma): use SDK session with retry logic for prompt fetching Use _get_session() from posthog/request.py instead of raw requests.get() to benefit from the SDK's existing retry configuration on transient network failures.	2026-01-30 08:43:04 -05:00
Richard Solomou GitHub Claude Sonnet 4.5	c32c78312f	feat(llma): pass raw provider usage metadata for backend cost calculations (#411 ) * feat: pass raw provider usage metadata for backend cost calculations Add raw_usage field to TokenUsage type to capture raw provider usage metadata (OpenAI, Anthropic, Gemini). This enables the backend to extract modality-specific token counts (text vs image vs audio) for accurate cost calculations. - Add raw_usage field to TokenUsage TypedDict - Update all provider converters to capture raw usage: - OpenAI: capture response.usage and chunk usage - Anthropic: capture usage from message_start and message_delta events - Gemini: capture usage_metadata from responses and chunks - Pass raw usage as $ai_usage property in PostHog events - Update merge_usage_stats to handle raw_usage in both modes - Add tests verifying $ai_usage is captured for all providers Backend will extract provider-specific details and delete $ai_usage after processing to avoid bloating properties. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * fix: add serialize_raw_usage helper to ensure JSON serializability Address PR review feedback from @andrewm4894: 1. Serialization: Add serialize_raw_usage() helper with fallback chain: - .model_dump() for Pydantic models (OpenAI/Anthropic) - .to_dict() for protobuf-like objects - vars() for simple objects - str() as last resort This ensures we never pass unserializable objects to PostHog client. 2. Data loss prevention: Change from replacing to merging raw_usage in incremental mode. For Anthropic streaming, message_start has input token details and message_delta has output token details - merging preserves both instead of losing input data. 3. Test coverage: Enhanced tests to verify: - JSON serializability with json.dumps() - Expected structure of raw_usage dicts - Coverage for both non-streaming and streaming modes - Fixed Gemini test mocks to return proper dicts from model_dump() Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * refactor: move raw_usage serialization from utils to converters Address PR feedback from @andrewm4894 - serialize in converters, not utils. Problem: Utils was receiving raw Pydantic/protobuf objects and serializing them, which meant provider-specific knowledge leaked into generic code. Solution: Move serialization into converters where provider context exists: Converters (NEW): - OpenAI: serialize_raw_usage(response.usage) → dict - Anthropic: serialize_raw_usage(event.usage) → dict - Gemini: serialize_raw_usage(metadata) → dict Utils (SIMPLIFIED): - Just passes dicts through, no serialization needed - Merge operations work with dicts only Benefits: 1. Type correctness: raw_usage is always Dict[str, Any] 2. Separation of concerns: converters handle provider formats 3. Fail fast: serialization errors in converters with context 4. Cleaner abstraction: utils doesn't know about Pydantic/protobuf Flow: Provider object → Converter serializes → dict → Utils → PostHog Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * fix: add type annotation for current_raw to satisfy mypy Fix mypy error: "Need type annotation for 'current_raw'" Extract value first, then apply explicit type annotation with ternary conditional to satisfy mypy's type checker. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>	2026-01-28 11:51:02 +02:00