Skip to content

feat!: posthog-python 8.0 on capture v1 - #1016

Open
eli-r-ph wants to merge 54 commits into
mainfrom
v1
Open

eli-r-ph wants to merge 54 commits into
mainfrom
v1

Conversation

@eli-r-ph

@eli-r-ph eli-r-ph commented Oct 6, 2026 •

Copy link
Copy Markdown
Contributor

💡 Motivation and Context

Long-lived integration branch for posthog 8.0: capture v1 becomes the only capture path, and capture_ai moves to an isolated v1 lane at /i/v1/ai/events. Feature PRs target v1; this PR merges to main once the series lands. Do not merge.

The major changeset bumps the next release from 7.64.1 to 8.0.0.

Series, stacked in merge order (each PR's base is the one before it, the first targets v1):

Related parity PRs: PostHog/posthog-go#359, PostHog/posthog-rs#279

main is merged into v1 periodically (no rebase, so open PRs keep their anchors).

💚 How did you test it?

Changeset only. Each series PR carries its own tests.

📝 Checklist

  • I reviewed the submitted code.
  • I added tests to verify the changes.
  • I updated the docs if needed.
  • No breaking change or entry added to the changelog.

If releasing new changes

  • Ran sampo add to generate a changeset file

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

Authored with Cursor (Claude Opus). Behavior follows the v1 branches of posthog-rs and posthog-go, adapted to this SDK's style.

@eli-r-ph eli-r-ph self-assigned this Oct 6, 2026
@github-actions

github-actions Bot commented Oct 6, 2026 •

Copy link
Copy Markdown
Contributor

posthog-python Compliance Report

Date: 2026-10-06T19:08:26.100214+00:00
Duration: 259866ms

✅ All Tests Passed!

121/121 tests passed


Capture_V1 Tests

✅ 95/95 tests passed

View Details
Test Status Duration
Endpoint And Method.Targets V1 Endpoint ✅ 516ms
Endpoint And Method.Does Not Use Legacy Endpoints ✅ 511ms
Required Headers.Has Authorization Bearer Header ✅ 511ms
Required Headers.Has Content Type Json ✅ 512ms
Required Headers.Has Posthog Sdk Info Format ✅ 511ms
Required Headers.Has Posthog Attempt Header ✅ 511ms
Required Headers.Has Posthog Request Id ✅ 511ms
Required Headers.Has Posthog Request Timestamp ✅ 511ms
Required Headers.Has User Agent ✅ 511ms
Body Format.Body Has Created At And Batch ✅ 512ms
Body Format.No Api Key In Body ✅ 511ms
Body Format.No Sent At In Body ✅ 511ms
Event Format.Event Has Required Root Fields ✅ 511ms
Event Format.Event Uuid Is Valid ✅ 512ms
Event Format.Event Timestamp Is Rfc3339 ✅ 511ms
Event Format.Non Utc Event Timestamp Is Converted To Utc ✅ 516ms
Event Format.Distinct Id Is String ✅ 511ms
Event Format.Distinct Id At Root Not Properties ✅ 511ms
Event Format.Custom Properties Preserved ✅ 511ms
Event Format.Set Properties Preserved ✅ 511ms
Event Format.Set Once Properties Preserved ✅ 512ms
Event Format.Groups Properties Preserved ✅ 510ms
Event Format.Sdk Generates Uuid If Not Provided ✅ 511ms
Event Format.Event Has Required Root Fields Batch ✅ 514ms
Event Format.Event Uuid Is Valid Batch ✅ 515ms
Event Format.Event Timestamp Is Rfc3339 Batch ✅ 514ms
Event Format.Distinct Id Is String Batch ✅ 514ms
Event Format.Distinct Id At Root Not Properties Batch ✅ 515ms
Event Format.Custom Properties Preserved Batch ✅ 514ms
Event Format.Set Properties Preserved Batch ✅ 514ms
Event Format.Set Once Properties Preserved Batch ✅ 515ms
Event Format.Groups Properties Preserved Batch ✅ 513ms
Event Format.Sdk Generates Uuid If Not Provided Batch ✅ 515ms
Batch Behavior.Multiple Events In Single Batch ✅ 517ms
Batch Behavior.Batch Envelope Smoke ✅ 516ms
Batch Behavior.Flush With No Events Sends Nothing ✅ 507ms
Batch Behavior.Flush At Triggers Batch ✅ 1012ms
Batch Behavior.Created At Reflects Batch Creation Time ✅ 512ms
Deduplication.Generates Unique Uuids ✅ 517ms
Deduplication.Different Events Same Content Different Uuids ✅ 513ms
Deduplication.Preserves Uuid On Retry ✅ 6520ms
Deduplication.Preserves Timestamp On Retry ✅ 6519ms
Deduplication.Preserves Uuid And Timestamp On Batch Retry ✅ 6522ms
Deduplication.No Duplicate Events In Batch ✅ 519ms
Header Behavior On Retry.Attempt Header Starts At One ✅ 511ms
Header Behavior On Retry.Attempt Header Increments On Retry ✅ 13526ms
Header Behavior On Retry.Request Id Preserved On Retry ✅ 6520ms
Header Behavior On Retry.Different Requests Have Different Request Ids ✅ 3021ms
Header Behavior On Retry.Request Timestamp Changes On Retry ✅ 6520ms
Response Format Validation.Success Response Has Uuid Keyed Results ✅ 512ms
Response Format Validation.Success Response Has Ok For Each Event ✅ 514ms
Response Format Validation.Success No Retry After When All Ok ✅ 513ms
Response Format Validation.Success Retry After Present When Retry Events ✅ 1517ms
Response Format Validation.Success No Retry After When Drop Only ✅ 513ms
Response Format Validation.Response Echoes Request Id ✅ 512ms
Retry Behavior.Retries On 408 ✅ 6520ms
Retry Behavior.Retries On 500 ✅ 6517ms
Retry Behavior.Retries On 503 ✅ 8523ms
Retry Behavior.Retries On 504 ✅ 6521ms
Retry Behavior.Retryable Errors Have Retry After ✅ 3517ms
Retry Behavior.Respects Retry After On Retryable Error ✅ 11523ms
Retry Behavior.Does Not Retry On 400 ✅ 2513ms
Retry Behavior.Does Not Retry On 401 ✅ 2514ms
Retry Behavior.Does Not Retry On 402 ✅ 2514ms
Retry Behavior.Does Not Retry On 413 ✅ 2515ms
Retry Behavior.Does Not Retry On 415 ✅ 2514ms
Retry Behavior.Non Retryable Errors Have No Retry After ✅ 2514ms
Retry Behavior.Implements Backoff ✅ 22526ms
Retry Behavior.Max Retries Respected ✅ 22537ms
Partial Batch Handling.Handles 200 Full Success ✅ 2512ms
Partial Batch Handling.Handles 200 With All Ok ✅ 3517ms
Partial Batch Handling.Does Not Retry Dropped Events ✅ 3515ms
Partial Batch Handling.Does Not Retry Limited Events ✅ 3516ms
Partial Batch Handling.Prunes Ok Events On Partial Retry ✅ 6521ms
Partial Batch Handling.Prunes Dropped Events On Partial Retry ✅ 6520ms
Partial Batch Handling.Retries Only Retry Events From Partial ✅ 6523ms
Partial Batch Handling.Partial Retry Preserves Uuids ✅ 6521ms
Partial Batch Handling.Partial Retry Attempt Header Increments ✅ 6518ms
Partial Batch Handling.Partial Retry Request Id Preserved ✅ 6522ms
Partial Batch Handling.Respects Retry After On Partial ✅ 8518ms
Partial Batch Handling.Unknown Result Treated As Terminal ✅ 3514ms
Partial Batch Handling.Mixed Ok Drop Limited No Retry ✅ 3519ms
Compression.Sends Gzip Content Encoding ✅ 512ms
Compression.No Content Encoding When Disabled ✅ 511ms
Compression.Compressed Body Is Decompressible ✅ 511ms
Error Handling.Does Not Retry On Unknown 4Xx ✅ 2513ms
Event Options.Cookieless Mode Override ✅ 511ms
Event Options.Disable Skew Correction Override ✅ 510ms
Event Options.Process Person Profile Override ✅ 511ms
Event Options.Product Tour Id Override ✅ 511ms
Event Options.Unset Options Omitted ✅ 511ms
Event Options.Options Override In Batch ✅ 514ms
Geoip And Historical Migration.Geoip Disable Injected Into Properties ✅ 512ms
Geoip And Historical Migration.Historical Migration Set In Body ✅ 511ms
Geoip And Historical Migration.Historical Migration Absent By Default ✅ 510ms

Capture_Ai Tests

✅ 5/5 tests passed

View Details
Test Status Duration
Routing.Capture Ai Posts To Ai Endpoint ✅ 511ms
Routing.Capture Does Not Reroute Ai Named Events ✅ 511ms
Identity.Capture Ai Event Has Uuid ✅ 510ms
Identity.Capture Ai Keeps Supplied Uuid ✅ 511ms
Timestamp Format.Non Utc Event Timestamp Is Converted To Utc ✅ 511ms

Feature_Flags Tests

✅ 17/17 tests passed

View Details
Test Status Duration
Request Payload.Request With Person Properties Device Id ✅ 12ms
Request Payload.Flags Request Uses V2 Query Param ✅ 9ms
Request Payload.Flags Request Hits Flags Path Not Decide ✅ 8ms
Request Payload.Flags Request Omits Authorization Header ✅ 9ms
Request Payload.Token In Flags Body Matches Init ✅ 9ms
Request Payload.Groups Round Trip ✅ 9ms
Request Payload.Groups Default To Empty Object ✅ 9ms
Request Payload.Disable Geoip False Propagates As Geoip Disable False ✅ 9ms
Request Payload.Disable Geoip Omitted Defaults To False ✅ 8ms
Request Payload.Flag Keys To Evaluate Contains Only Requested Key ✅ 9ms
Request Lifecycle.No Flags Request On Init Alone ✅ 3ms
Request Lifecycle.No Flags Request On Normal Capture ✅ 509ms
Request Lifecycle.Two Flag Calls Produce Two Remote Requests ✅ 15ms
Request Lifecycle.Mock Response Value Is Returned To Caller ✅ 9ms
Retry Behavior.Retries Flags On 502 ✅ 313ms
Retry Behavior.Retries Flags On 504 ✅ 314ms
Side Effect Events.Get Feature Flag Captures Feature Flag Called Event ✅ 511ms

Feature_Flags_Local_Evaluation Tests

✅ 4/4 tests passed

View Details
Test Status Duration
Versioned Boolean Matching.Matching Version Missing ✅ 65ms
Versioned Boolean Matching.Matching Version 1 ✅ 63ms
Versioned Boolean Matching.Matching Version 2 ✅ 61ms
Versioned Boolean Matching.Version Only Reload 1 2 1 2 Missing ✅ 34ms

@eli-r-ph
eli-r-ph marked this pull request as ready for review October 7, 2026 20:43
@eli-r-ph
eli-r-ph requested a review from a team as a code owner October 7, 2026 20:43
@eli-r-ph
eli-r-ph marked this pull request as draft October 8, 2026 19:40
@eli-r-ph

eli-r-ph commented Oct 8, 2026

Copy link
Copy Markdown
Contributor Author

Putting this back in draft for safety as per the other SDK v1 prep branches ✅

…nts (#1017)

feat!: send capture through v1 by default; capture_ai posts to /i/v1/ai/events

Capture v1 is now the default capture mode. The AI lane always posts capture v1 to /i/v1/ai/events, uncompressed, regardless of the analytics capture_mode. Tests move to a shared capture v1 send seam (posthog/test/capture_helpers.py); v0-only tests are pinned to CaptureMode.V0 until v0 removal.
Capture v1 is the only capture path. Remove CaptureMode, the capture_mode
kwarg and POSTHOG_CAPTURE_MODE, the gzip kwarg, request.batch_post and the
v0 endpoint constants, and the backoff dependency. Client and AsyncPosthog
constructors are keyword-only after host.
Rename CaptureV1Error to CaptureError and export it from posthog. Event
shaping moves to posthog.capture_event and transport to
posthog.capture_send. No behavior change.
…1020)

* feat!: report capture v1 results through CaptureError and on_error

* test: cover _uuid7 layout for the mutation gate

* fix: re-raise sync capture failures in debug mode after reporting them

Also document which events verdict_summary() counts.
* feat: AI lane config and size-aware batching

* chore: tighten AI lane comments

* fix: reject non-finite AI lane config values
eli-r-ph and others added 12 commits October 13, 2026 20:42
test(mcp): drop get_tool checks that fastmcp 4.0.11 invalidated

fastmcp 4.0.11 resolves middleware-injected tools through the server's
tool lookup (PrefectHQ/fastmcp#5431), so get_tool() now returns them.
The two setup assertions that expected None fail on any fresh install.
The SDK behavior the tests check is unchanged and passes on 4.0.10 and
4.0.11.


(cherry picked from commit 897cd43)

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
feat!: per-event capture options and legacy property hoisting
* feat!: super_options, context options and personless as the lowest option layer

* docs: state the option layer rules in capture docstrings

* feat!: fill context and super values after before_send

Context tags, context options, super_properties, super_options, the
derived personless option and the environment $release_id now fill only
keys an event leaves unset, after before_send runs. before_send sees
only the event's own values and SDK enrichment, and its changes beat
every default. A None option counts as unset and is filled. Legacy
properties still fill only unset options and are always removed.

* feat!: fill context and super values before before_send

Context tags, context options, super_properties, super_options, the
derived personless option and the environment $release_id now fill
before before_send runs, still only into keys the event leaves unset.
before_send sees every value and has the final say, including removing
a super property. $set, $set_once, $groups and $group_set fill one
level deep when both values are dicts. The set() and set_once() values
now win key by key over a $set or $set_once in properties, as ingestion
merges them. Hoisting still runs once, after the hook.

* feat!: SDK values are defaults that caller values override

$is_server, $geoip_disable and system context now fill last, only keys the
event, context tags and super_properties left unset, before before_send. A
super property $geoip_disable: False now wins over disable_geoip=True.

The groups argument merges into a $groups property key by key and wins,
in Client, AsyncClient and posthog.mcp events.

* fix: copy async default layers at admission; typed groups win on MCP exception events

AsyncPosthog filled defaults in the worker from references to the live
super_properties and options, so a change after capture could reach a queued
event. Each layer is now shallow-copied when the event is admitted, as 7.x
merged super_properties into a new dict at capture.

The MCP exception event now merges typed groups after custom properties, like
the primary event, so both events of a failing call get the same groups.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
feat!: AI wrappers and MCP send personless as a per-event option
* chore(ci): run compliance harness 1.13.1 once per codec

The adapter now advertises capture_ai_v1 and event_options, passes options and uuids through unchanged, and takes its codec from COMPRESSION. The workflow runs the harness containers directly in a gzip/deflate/zstd matrix, so it no longer needs pull-requests: write.

* chore(ci): install zstd extra for adapter protocol tests

* chore: describe the single capture compliance job and harness 1.13.1

The standalone adapter-test setup installs the zstd extra, matching CI.
enable_full_ai_capture now controls only truncation and media redaction. A client object without capture_ai still gets capture calls.
* feat: AsyncPosthog AI capture lane

* fix: type capture v1 send timeouts as float

* feat: AsyncPosthog privacy_mode and enable_full_ai_capture

* fix: keep async capture_ai from raising on a non-string event name
* test: capture invariant hardening

* test: assert exactly one capture loss line per failed batch
* docs: 8.0 migration guide, README and openfeature-provider pin

* chore: let openfeature provider accept posthog 7 and 8

* docs: say options beat legacy properties at every layer

* docs: name the CaptureError import and the posthoganalytics mirror

* docs: note that sync_mode failures re-raise in debug mode

* docs: describe the late fill of context and super values

* docs: describe the fill of context and super values before before_send

* docs: SDK values are defaults that caller values override

* docs: use real quota tags; a billing limit is a 402

Capture never sends billing_limit_exceeded or billing as a per-event tag.
Product quotas drop single events in a 2xx with tags such as
llm_events_over_quota; a request over the billing limit gets a 402 with
no per-event results.

* docs: complete the 8.0 migration guide

* chore(examples): run the pinned AI examples against the local SDK
* feat!: keep the posthog-python identity when instrumenting MCP

MCP instrumentation relabeled the whole client, so the app's own events and flag requests reported posthog-python-mcp. Capture v1 takes $lib from the per-request PostHog-Sdk-Info header, which is now always posthog-python.

* chore: remove the unused request identity plumbing and stale MCP client advice
fix!: follow capture redirects only to the configured origin

_post_v1 disables automatic redirects and resends the batch only on a 307 or 308 to the origin of host, at most 5 times. Any other redirect returns as a terminal failure. Covers the sync client and the async client, which lost its v0 same-origin guard when v1 became the only path.
* ci: reuse Python SDK upgrade PRs unless queued (#1021)

* ci: reuse Python SDK upgrade PRs unless queued

* ci: protect reused upgrade PRs from stale and queued writes

(cherry picked from commit 945bbbc)

* feat(error-tracking): standardize exception event metadata (#898)

* feat(error-tracking): standardize exception metadata

* fix(error-tracking): preserve explicit trace context

* fix(error-tracking): restrict exception group traversal

* fix(error-tracking): deprecate reserved property overrides

* fix(error-tracking): bound exception group traversal

(cherry picked from commit 6d6cc3d)

* chore: Release v7.65.0 [skip ci]

(cherry picked from commit c4e6bdd)

* Update generated references

(cherry picked from commit 50efa65)

* feat(asgi): add error tracking middleware (#1037)

* feat(asgi): add error tracking middleware

* chore(asgi): type exception capture metadata

* fix(asgi): make tracing header trust explicit

* refactor(asgi): expose request properties terminology

(cherry picked from commit 1c6a47a)

* feat(mcp): resolve fresh low-level tool ownership (#1009)

* feat(mcp): resolve fresh low-level tool ownership

Add resolve_original_tool for raw MCP SDK 1.x and 2.x servers. Prefer ownership learned from a served listing and keep unknown ownership fail-open.

Generated-By: PostHog Desktop
Task-Id: db1a8a08-fe03-43b6-9e49-705e8b1f375e

* fix(mcp): preserve listed argument ownership

Generated-By: PostHog Desktop
Task-Id: db1a8a08-fe03-43b6-9e49-705e8b1f375e

* fix(mcp): preserve arguments for unknown schemas

Generated-By: PostHog Desktop
Task-Id: db1a8a08-fe03-43b6-9e49-705e8b1f375e

* fix(mcp): preserve low-level tool schemas

Generated-By: PostHog Desktop
Task-Id: db1a8a08-fe03-43b6-9e49-705e8b1f375e
(cherry picked from commit 127009d)

* feat(flask): add error tracking integration (#1038)

* feat(flask): add error tracking integration

* fix(flask): use canonical exception metadata

* chore(flask): type exception capture metadata

* refactor(flask): expose request properties terminology

* fix(flask): preserve exception privacy settings

(cherry picked from commit 292855d)

* chore: Release v7.66.0 [skip ci]

(cherry picked from commit d23ef68)

* Update generated references

(cherry picked from commit c0e05b1)

* feat(drf): add automatic error tracking integration (#1036)

* feat(drf): add automatic error tracking integration

* fix(drf): align exception metadata with SDK spec

* chore(drf): type exception capture metadata

* fix(drf): honor Django request filtering

* test(drf): describe context data as properties

* feat(drf): support exception capture opt-out

* chore(drf): update public API snapshot

* fix(drf): align exception capture configuration

* test(drf): preserve omitted setting default

(cherry picked from commit 9baf8a5)

* chore: Release v7.67.0 [skip ci]

(cherry picked from commit a655c8b)

* Update generated references

(cherry picked from commit 509c852)

---------

Co-authored-by: Dustin Byrne <dustin@posthog.com>
Co-authored-by: Hugues Pouillot <hpouillot@gmail.com>
Co-authored-by: releaser-posthog-python[bot] <262414804+releaser-posthog-python[bot]@users.noreply.github.com>
Co-authored-by: Georgis Andonis <6430745+gesh@users.noreply.github.com>
eli-r-ph and others added 13 commits October 28, 2026 13:24
* chore: run the capture v1 compliance suites from harness 1.23.0

Declare the adapter capabilities harness 1.23.0 uses to gate tests:
event_uuid_override, bounded_retries, disable_geoip and historical_migration.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore: support the harness reported_drops, log_capture, generated_payloads, property_layers and fail_open features

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore: clear adapter diagnostics after the old client shuts down, size padding markers in bytes

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore: run the capture v1 compliance suites from harness 1.25.0

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(adapter): name each duplicate uuid drop after its own copy

Also point adapter docs at harness 1.25.0.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(adapter): type the duplicate drop names for mypy

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(ai): keep streamed OpenAI tool calls that have no index (#1091)

* fix(ai): keep streamed OpenAI tool calls that have no index

accumulate_openai_tool_calls skipped every delta without an index, so
tool calls from OpenAI-compatible providers that omit it never reached
$ai_output. A delta with a new id now opens a call, and one without an
id continues the latest call.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(ai): keep synthetic tool call keys apart from provider indexes

A call without an index took key max + 1, which a later call with that real index overwrote, merging two calls and dropping one. Index-less calls now get string keys of their own.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(ai): widen the streamed tool call accumulator key type

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(ai): update streamed tool call API snapshot

---------

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
(cherry picked from commit ffde3c57f26b9daf92f12fbb21edd4f359e4388b)

* chore: Release v7.75.7 [skip ci]

(cherry picked from commit 7f4c9be1c8aa66bf55cb2080cf0775ff4ec28cff)

* Update generated references

(cherry picked from commit b5eb2debabc7bc6e4e5d033b923d1eef6a6508a6)

* docs: require signed commits in contributor docs (#1111)

Contributors often open PRs with unsigned commits because signing isn't documented here.

(cherry picked from commit e65ed67b3a61122c6480c2198c019a493061845c)

* fix(flags): fall back for incompatible local regex evaluation (#1108)

* fix(flags): fall back for incompatible local regex evaluation

* fix(flags): defer nested regex quantifiers and fast-path ASCII values

* fix(flags): defer strict-end searches with final newline backtracking

* perf(flags): cache pure regex pattern metadata

(cherry picked from commit d58380e87f38b725ded6d38b870b734056a526e7)

* chore: Release v7.75.8 [skip ci]

(cherry picked from commit 72e013122e07345a827d108de74f079672220f1d)

* Update generated references

(cherry picked from commit a020e703fd89e4e35211419ad6f1ba3b656d798f)

* fix(examples): use gpt-image-2.5-flare in the OpenAI image generation example (#1109)

OpenAI shuts down gpt-image-1 on 2026-10-23 and lists gpt-image-2.5-sunburst or
gpt-image-2.5-flare as the replacement
(https://developers.openai.com/api/docs/deprecations).

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
(cherry picked from commit efc866ec4b18115518597bfa9f6b82a6f17cc270)

* fix(mcp): send one $mcp_initialize per initialize request (#1103)

* fix(mcp): send one $mcp_initialize per initialize request

The Python MCP SDK answers initialize itself, so posthog.mcp sent the
event on the first request of each session, once per process. A
stateless fleet sent one event per pod. A client that sent only
initialize sent no event.

Now MCP SDK 2.x reports the event from Server.middleware, on every
transport. MCP SDK 1.x over HTTP reports it from
PostHogMcpStatelessSessionMiddleware, which takes a new optional
server= argument. MCP SDK 1.x over stdio keeps the first-request
fallback.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* refactor(mcp): read only the initialize response head on mcp 1.x

The mcp 1.x hook now keeps only the status and the Mcp-Session-Id header
of the initialize response. It no longer parses the JSON or SSE body.
The event uses the protocol version that the client asked for.

record_initialize now reads the client name, version, and protocol from
the initialize params and reply, so both hooks pass them through. The
new tests share their request helpers and drop the SSE body case.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): address review on the initialize hooks

- SSE transport: the initialize event now lands in the session that the
  SSE calls use. On SDK 1.x the middleware leaves SSE requests to the
  first-request fallback. On SDK 2.x the hook uses the process session.
- The bound middleware marks each request that it passes on. The fallback
  now stops only for marked requests, not for every HTTP request.
- The hooks compute the handshake session without a write to the shared
  session that McpAnalytics.capture() reads.
- A server= value that cannot be bound no longer breaks requests.
- Remove the unused _mint_token_if_initialize.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): wait for the initialize capture inside its handshake task

Both handshake hooks run record_initialize in a tracked background task.
It started the capture as a second task, after flush() took its snapshot
of pending tasks. With a slow before_send, flush() returned early and
shutdown dropped $mcp_initialize. The handshake task now awaits the
capture itself.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): resolve the v2 handshake session from an incoming header

On MCP SDK 2.x, a stateless client can send initialize with an existing
Mcp-Session-Id and keep sending it. Its later calls resolve their
session from that header, but the handshake hook read only the
connection id and the minted token. The event landed in a generated
session. The hook now falls back to the incoming header.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): capture a v1 initialize that carries a session header

On MCP SDK 1.x, the middleware read the body only of a POST with no
Mcp-Session-Id, because only those need a token. A client that sent
initialize with an existing session id got no event: the middleware
marked the request, so the first-request fallback did not run either.

A middleware bound to a server now reads those bodies too, behind a
cheap check for "initialize". It still mints only without a header. The
event uses the header that the client keeps sending.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): pass a disconnect mid-body through the session middleware

When the client disconnected while it sent the body, the middleware
stopped its read and replayed the partial body as a complete request.
The app parsed half a request and never saw the disconnect. The
middleware now replays the partial body with more_body set, then hands
the app the disconnect.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): stop buffering tool calls once they are not an initialize

A bound middleware reads the body of a POST that carries a session id,
to catch an initialize that reuses one. That held every tool call in
memory, up to 256 KB, before the app saw it. The middleware now stops
after 4 KB when "initialize" is not in it, and streams the rest.
JSON-RPC clients put "method" before "params", so a handshake shows up
in that window.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* refactor(mcp): read the v2 handshake header through _resolve_ctx

The v2 handshake hook read the Mcp-Session-Id header and decoded the
token itself. Every later request gets the same pair from _resolve_ctx.
The hook now calls it, so the handshake and later requests follow one
rule and cannot drift apart into two sessions.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): run the initialize callbacks inline in the handshake request

Both handshake hooks ran identify and event_properties in a detached
background task. If a callback never returned, each handshake left one
more task pending, with no bound, and flush() waited for all of them.
The spec requires pending state to be bounded by count.

The hooks now run the callbacks inline, the same as a tool call does. A
hung callback holds that request, where the host sees it, and leaves no
task behind. The capture still runs in the background after the
callbacks, as for every other event.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): record the v1 handshake inside the request's auth context

The session middleware sits outside FastMCP's AuthContextMiddleware. It
ran identify after the app returned, when that middleware had already
reset the authenticated user. An identify that calls get_access_token()
got None, so $mcp_initialize stayed anonymous under OAuth. The
middleware now records the handshake when the response starts, while
the app still handles the request.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): stop the body peek only on a method that is not initialize

The 4 KB peek stopped when "initialize" was not in the first 4 KB. JSON
key order is arbitrary, so an initialize with large params before its
method got no event. The peek now stops only when the prefix shows a
"method" and no "initialize". Without a method yet, it reads up to the
full cap.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): drop the body byte checks from the session middleware

The middleware guessed from raw bytes whether a request was initialize.
A nested "method" key in params stopped the 4 KB peek too early, and a
JSON-escaped "initialize" failed the quick byte check. That second
case also stopped the token mint, which worked before this PR.

The middleware now parses the body as JSON, as main does, and reads
only requests with no Mcp-Session-Id. On MCP SDK 1.x, an initialize
that reuses an existing session id is not captured. The README lists
this limit. MCP SDK 2.x still captures it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 8cd8eb5b6fdcdb19db4c4d07fe911c6eb3976d5a)

* fix(ai): take cumulative usage and web searches from the Claude Agent SDK message_delta (#1112)

The tracker only read output_tokens from a message_delta. Its usage is cumulative,
and with server tools (web search, ...) the input and cache counts exceed the
message_start values, so streamed generations under-reported them, and the
number of web searches was never recorded. Keep the larger input and cache
counts, as the Anthropic wrapper does, and set $ai_web_search_count from
server_tool_use.web_search_requests.

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
(cherry picked from commit 48889b6cc8d0a9eec579684640cb74b393cdd56c)

* chore: Release v7.75.9 [skip ci]

(cherry picked from commit 5a2ea5f7e0c71d6b9dc70112ab9d5c29c6f99bb1)

* Update generated references

(cherry picked from commit e3fd9a49c6ca1beeb2d4f9b6a889ea7ee0774f57)

* fix(ai): record the HTTP status of google-genai errors (#1110)

* fix(ai): record the HTTP status of google-genai errors

The generation and embedding wrappers read the HTTP status of a failed call from
the exception's status_code, which OpenAI and Anthropic errors have. google-genai's
APIError (ClientError, ServerError) carries it as code, so every failed Gemini
call was captured with $ai_http_status 0. Read either attribute through a small
helper; a code outside the HTTP range still counts as no status.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix: remove unused import from HTTP status tests

---------

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
(cherry picked from commit 25d2767a70378da02c321879f0fa3a2e52d071b7)

* chore: Release v7.75.10 [skip ci]

(cherry picked from commit 82f948b0871ecbbf5c6fd30714481dc5877adbf1)

* Update generated references

(cherry picked from commit b0a2783b53e7a8e5546f11674318225b26139ec4)

* fix(aio): keep content out of ai events in privacy mode (#1105)

* fix: ai privacy mode leaks

* fix: address privacy mode review findings

* fix: strip remaining otel content in privacy mode

* fix: redact mcp credentials in otel tool definitions

* fix: address otel privacy review findings

* fix: always redact mcp credentials in otel tool definitions

* fix: strip otel embedding vectors in privacy mode

* fix: drop only the otel span that fails redaction

* fix: keep vercel rerank scores in otel privacy mode

* fix: strip logfire payload attributes in otel privacy mode

(cherry picked from commit 04e6a4c1c4f15e4dd0d010dee5fa83244f9b6dd8)

* chore: Release v7.75.11 [skip ci]

(cherry picked from commit fb9b66bcc1d679180f5aee9c1fe4397636074c75)

* Update generated references

(cherry picked from commit 16d618b1636e47dfcd3a530231fddf16defa6b81)

* chore(deps): bump opentelemetry-api from 1.40.0 to 1.45.1 (#1117)

Bumps [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python) from 1.40.0 to 1.45.1.
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

---
updated-dependencies:
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
(cherry picked from commit bc7b40331a4a9bf18456f6e9136147ab7118ca0f)

* chore(deps): bump djangorestframework from 3.18.1 to 3.18.2 (#1118)

Bumps [djangorestframework](https://github.com/encode/django-rest-framework) from 3.18.1 to 3.18.2.
- [Release notes](https://github.com/encode/django-rest-framework/releases)
- [Commits](https://github.com/encode/django-rest-framework/compare/3.18.1...3.18.2)

---
updated-dependencies:
- dependency-name: djangorestframework
  dependency-version: 3.18.2
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
(cherry picked from commit 125b26654ea317ce72077b7f3eb83d11461cf9b2)

* chore(deps): bump the uv group across 31 directories with 1 update (#1119)

Bumps the uv group with 1 update in the /examples/example-ai-anthropic directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-autogen directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-aws-bedrock directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-azure-openai directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-cerebras directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-cloudflare-ai-gateway directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-cohere directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-crewai directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-dedalus directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-deepseek directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-fireworks-ai directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-gemini directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-groq directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-helicone directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-hugging-face directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-instructor directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-langchain directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-langgraph directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-llamaindex directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-mirascope directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-mistral directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-ollama directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-openai directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-openrouter directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-perplexity directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-portkey directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-semantic-kernel directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-smolagents directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-together-ai directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-vercel-ai-gateway directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).
Bumps the uv group with 1 update in the /examples/example-ai-xai directory: [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python).

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.34.1 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

Updates `opentelemetry-api` from 1.40.0 to 1.45.1
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](https://github.com/open-telemetry/opentelemetry-python/compare/v1.40.0...v1.45.1)

---
updated-dependencies:
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.1
  dependency-type: indirect
  dependency-group: uv
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
(cherry picked from commit 3f6c80eb4df1ddffd9e5b9d1d8aae8facf9f83b2)

* feat(llma): support per-item offline evaluation models (#1130)

(cherry picked from commit 284340d6ba6df08f68c27eac750b13f3ec89ab19)

* chore: Release v7.76.0 [skip ci]

(cherry picked from commit 7d15127fd4e3aeac72bb1af59f49d96c1c81be7e)

* Update generated references

(cherry picked from commit 7833e95266da6130bf41028623f4a5e69a950d71)

* feat(ai): record the Anthropic web search count in the LangChain callback handler (#1097)

* feat(ai): record the Anthropic web search count in the LangChain callback handler

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* chore: update public API snapshot for ModelUsage.web_search_count

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore: regenerate LangChain ModelUsage API snapshot

* fix(ai): read the LangChain web search count from llm_output usage and omit zero

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
(cherry picked from commit b0a2b59dc09e96903cdf97ea3c8ab5298bdff7bb)

* chore: Release v7.76.1 [skip ci]

(cherry picked from commit a736b43290fc88fecc929916887a3c1c215984a7)

* Update generated references

(cherry picked from commit 3044d8ed5919b6b11b42642359651855c6f9ec38)

* feat(mcp): add a send_initialize_event option (#1115)

* feat(mcp): add a capture_initialize option

MCPAnalyticsOptions(capture_initialize=False) stops $mcp_initialize. The
option is on by default. Many clients reconnect before every few tool
calls, so the handshake event can outnumber $mcp_tool_call.

With the option off, record_initialize and the first-request fallback
return early. Nothing else changes: the middleware still mints the
session token, so tool calls keep their sessions and client names.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(mcp): keep the handshake identity with capture_initialize off

record_initialize returned before handle_identify when the option was
off. An identify callback that resolves the user only from the
initialize request then left the session anonymous, so later tool calls
lost the user. Identify now runs first, and the option stops only the
event.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* refactor(mcp): rename capture_initialize to send_initialize_event

PostHogMCP already has a capture_initialize() method that sends the
event by hand. An option with the same name next to that method makes
readers link the two. The new name says what the option controls.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs(mcp): name the PostHog screens that read $mcp_initialize

MCP onboarding and the users and sessions tiles of the default MCP
dashboard read $mcp_initialize. The README now says that they show no
data while send_initialize_event is off.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit d8a1b976d665d9f7b3249ac788ef253bb21eed50)

* chore: Release v7.77.0 [skip ci]

(cherry picked from commit 786db579d383a34fc6b4d21d1332743b5eec17b5)

* Update generated references

(cherry picked from commit 0894015a3730c8bbdf1cfc3276da12913bcdac27)

* feat(mcp): add a send_tools_list_event option (#1116)

MCPAnalyticsOptions(send_tools_list_event=False) stops $mcp_tools_list.
The option is on by default. Clients list tools on every reconnect, so
the listing event is about as common as $mcp_tool_call.

With the option off, record_tools_list returns early, which also drops a
failed listing's $exception. Nothing else changes: tools keep their
injected arguments, and the virtual tools stay listed.

The name says "event" to match send_initialize_event.
PostHogMCP.capture_tools_list() already captures the event by hand.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 351da59c252afb7bffed79d21c23aa7e2539a38e)

* chore: Release v7.78.0 [skip ci]

(cherry picked from commit 39110204fdf2aceb1e00f66f18bd5656f0783aff)

* Update generated references

(cherry picked from commit 75a17943805604a040ac2f204fd74e3fbeb501a3)

* chore(deps): bump the github-actions group with 3 updates (#1128)

Bumps the github-actions group with 3 updates: [github/codeql-action/init](https://github.com/github/codeql-action), [github/codeql-action/analyze](https://github.com/github/codeql-action) and [PostHog/posthog-sdk-test-harness/.github/workflows/test-sdk-action.yml](https://github.com/posthog/posthog-sdk-test-harness).

Updates `github/codeql-action/init` from 4.38.3 to 4.39.0
- [Release notes](https://github.com/github/codeql-action/releases)
- [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md)
- [Commits](https://github.com/github/codeql-action/compare/24c54180a607b1449ed407dd24f251e4e9147c8d...6f2ac81f5cd81fe351a197ba0b00d91d03cd2aee)

Updates `github/codeql-action/analyze` from 4.38.3 to 4.39.0
- [Release notes](https://github.com/github/codeql-action/releases)
- [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md)
- [Commits](https://github.com/github/codeql-action/compare/24c54180a607b1449ed407dd24f251e4e9147c8d...6f2ac81f5cd81fe351a197ba0b00d91d03cd2aee)

Updates `PostHog/posthog-sdk-test-harness/.github/workflows/test-sdk-action.yml` from 1.15.0 to 1.17.0
- [Release notes](https://github.com/posthog/posthog-sdk-test-harness/releases)
- [Changelog](https://github.com/PostHog/posthog-sdk-test-harness/blob/main/CHANGELOG.md)
- [Commits](https://github.com/posthog/posthog-sdk-test-harness/compare/64ffa6ab19166541880c077d349abe18a7e42395...8ca75c478d8a1b7d5be3f9b82b1a744b8ee630c3)

---
updated-dependencies:
- dependency-name: github/codeql-action/init
  dependency-version: 4.39.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: github-actions
- dependency-name: github/codeql-action/analyze
  dependency-version: 4.39.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: github-actions
- dependency-name: PostHog/posthog-sdk-test-harness/.github/workflows/test-sdk-action.yml
  dependency-version: 1.17.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: github-actions
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
(cherry picked from commit 7083998df0c26079e748f9279f92ca764ceb19ea)

* feat: report unresolved local flags on evaluate_flags snapshots (#1133)

* feat: report unresolved local flags on evaluate_flags snapshots

Implements the "Snapshot unresolved flag reporting" and "Warning for flags local evaluation never resolves" requirements of the evaluate-flags spec (PostHog/sdk-specs#85).

- Add `FeatureFlagEvaluations.unresolved_flags`, with an `UnresolvedFlagReason` for each flag and the spec's fixed precedence.
- Report `local_evaluation_inconclusive` instead of `flag_missing` when a snapshot read hits an unresolved flag.
- Do not stop a condition at the first inconclusive filter, so a filter that does not match resolves the condition locally.
- Check `active` before experience continuity, also for dependencies.
- Log one warning per flag that local evaluation can never resolve, repeated only when its definition changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: 39793b31-809e-4689-a5d8-b4359cbb22c1

* fix: address review findings for unresolved local flags

* test(flags): check flag warnings run outside the publication lock

* fix(flags): fall back to the server when a flag dependency needs server data

* fix(flags): keep server fallback when the definition reason lookup fails

(cherry picked from commit bb2b9d6bddd250f6dcf03be565af8ba94694e5db)

* chore: Release v7.79.0 [skip ci]

(cherry picked from commit 91d3188bb4dbf5ed8e59165f0dbd62836a05d384)

* Update generated references

(cherry picked from commit 130f5a1e8508a7a9005bf46658ff8d43b7457232)

* feat(metrics): metrics autocapture for HTTP, database and runtime metrics (#992)

* feat(metrics): add metrics autocapture for HTTP, database and runtime metrics

`metrics={"autocapture": True}` starts the installed OpenTelemetry instrumentors with private meter and tracer providers and exports to `/i/v1/metrics`. The packages come from the new `posthog[metrics]` extra.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: ef9b57c2-47ff-47e6-a509-8873ffe1fab4

* test: check concurrent LangChain traces without assuming order

Generated-By: PostHog Desktop
Task-Id: c0218209-575c-46f0-a201-d0c831e38096

* fix(metrics): address autocapture review findings

- Honour send=False in the autocapture export gate.
- Strip trailing slashes from the host before building the metrics endpoint.
- Set OTEL_SEMCONV_STABILITY_OPT_IN only while the instrumentors load, then
  restore the process environment.
- Limit each synchronous instrument to 2000 attribute sets; send the rest to
  an otel.metric.overflow series, so user-supplied hosts cannot grow the SDK cache.
- After fork, replace the module lock and point the runtime instrumentor at
  the child process.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: 10838116-90e2-4d23-af38-8099d7887d07

* fix(metrics): keep app-owned instrumentors intact and hold ownership through shutdown

- Skip an instrumentor whose singleton the app already instrumented before creating it. Creating it ran __init__ again and reset the app's system metrics config.
- Uninstrument while the module lock is held, so a client that starts during shutdown cannot skip libraries that are about to be uninstrumented. The last export still runs outside the lock.
- Correct the module docstring: an app cannot instrument a library again with its own OpenTelemetry setup while autocapture instruments it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: a129c1a0-57e3-4370-a4df-6542e710ab75

* fix(metrics): use stable HTTP metric names when another instrumentor read the opt-in first

OpenTelemetry reads OTEL_SEMCONV_STABILITY_OPT_IN once per process. If any instrumentor ran before the client started, autocapture sent http.client.duration (ms) and not http.client.request.duration (s). When the app has not set an opt-in, autocapture now sets the HTTP mode to stable after initialisation, so the names do not depend on import order.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: a129c1a0-57e3-4370-a4df-6542e710ab75

* fix(metrics): address the second review round of metrics autocapture

- Do not send the app's OTEL_EXPORTER_OTLP_* settings (other vendors' headers, endpoint, TLS files) to PostHog: the private exporter is built without them.
- Give HTTP instrumentors a no-op tracer, so app requests no longer carry an unsampled traceparent that makes downstream services drop their traces.
- Record nothing while the client is disabled, and apply metrics.before_send to autocaptured measurements and runtime observations (value and attributes, or drop).
- Exclude the PostHog host in any case and with or without its default port, escape commas, keep OTEL_PYTHON_*_EXCLUDED_URLS, and apply the exclusion to httpx.
- Reset the reader's export lock in a forked child, export autocaptured metrics on Client.flush() and within the client's exit budget (no 10 s exit wait), roll back on a failed start, and uninstrument and reinstrument Flask apps across a restart.
- Start autocapture on a later module-level setup(), once per client.
- Validate flush_interval like client.metrics, use the same resource as client.metrics, keep the first word of db.operation.name, and raise the minimum OpenTelemetry versions to the first ones with the runtime metric names.
- Rename the module to posthog/_metrics_autocapture.py to keep internals out of the public API.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: a129c1a0-57e3-4370-a4df-6542e710ab75

* fix(metrics): keep app middleware on shutdown, stop context propagation, and fix fork and flush issues

- Flask: shutdown uninstruments an app only when its WSGI app is still the one the instrumentation set. uninstrument_app() would otherwise remove middleware the app added later, such as authentication. Instrumentation left in place records nothing after shutdown.
- Do not propagate context: the instrumentors that autocapture starts no longer inject trace context or baggage into outgoing requests, including a trace extracted from an incoming request.
- Keep the attributes that a database span starts with (asyncpg passes db.system then).
- Export the last window at exit in sync mode.
- Replace the metrics SDK collection and aggregation locks in a forked child, where a parent thread could hold them.
- Use one flush thread at a time, with one follow-up export for calls made during a flush.
- Run metrics.before_send as an SDK callback, so a flush() it calls is deferred and cannot deadlock the collection.
- Raise the minimum OpenTelemetry instrumentation version to 0.60b0, the first in which the httpx instrumentor excludes URLs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: a129c1a0-57e3-4370-a4df-6542e710ab75

* fix(metrics): do not let the Django instrumentor configure empty settings

Without configured settings, DjangoInstrumentor calls settings.configure() with empty settings, so an app that configures its settings in code after creating the client fails with "Settings already configured". Autocapture now starts the Django instrumentor only when the settings are configured, as in AppConfig.ready().

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: a129c1a0-57e3-4370-a4df-6542e710ab75

* fix(metrics): drop raw URL attributes, normalize hook attributes, and stop flush loops from hooks

- Never record raw URLs and paths (url.full, http.url, http.target, ...), client addresses, user agents or captured headers. Instrumentors such as Tornado put them on metrics, and they can hold tokens and emails. Routes stay.
- Keep only valid OpenTelemetry attribute values, with sequences as tuples, before and after metrics.before_send. SDK 1.39 raised TypeError for a list value inside the app's request. Recording errors no longer reach the instrumented call.
- A flush() called from a metrics before_send hook still flushes events and spans, but does not start another autocapture export, which ran the hook again without end.
- Each flush request waits only for the first export that starts after it, not for later exports.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: a129c1a0-57e3-4370-a4df-6542e710ab75

* fix(metrics): isolate exporter settings and autocapture lifecycle state

* fix(metrics): address remaining autocapture lifecycle feedback

---------

Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
(cherry picked from commit febb3ed2f1584e68c05c73625becbdd18fe5f944)

* chore: Release v7.80.0 [skip ci]

(cherry picked from commit ce03e476bb0256bed4f3b6745ebb50fc3dead745)

* Update generated references

(cherry picked from commit 175fac6d6e4fdbb179ead60869ed5a864ea1f6ff)

* fix(ai): handle None and NOT_GIVEN stream_options without mutating the caller's dict (#1144)

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
(cherry picked from commit 1625f4fbc1809d7a19b34354ad2df2a0e016928f)

* chore: Release v7.80.1 [skip ci]

(cherry picked from commit 58d69e3127134967e317ee29e25ad70de3909b79)

* Update generated references

(cherry picked from commit e4f43a4bf81b7e8c6ec151f53086964c88e39854)

* fix(ai): capture an $ai_embedding event when an OpenAI embeddings call fails (#1143)

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
(cherry picked from commit ad1b1cb1e2052621d979cfdccb4870d6b4b31613)

* chore: Release v7.80.2 [skip ci]

(cherry picked from commit 9a9d8067157f922c55503b53a68e9a220d568351)

* Update generated references

(cherry picked from commit ad79bd03c268eb02f4b17ab8fc21998a1e3bf2fe)

* fix(ai): report OpenAI Agents cached tokens (#1102)

* fix(ai): report OpenAI Agents cached and reasoning tokens

The Agents SDK nests cached input and reasoning tokens under the usage
details fields: input_tokens_details/output_tokens_details on generation
span usage and on the Responses usage object of response spans. The
processor read flat keys that the SDK never sets, so both were always
dropped. Read the nested fields, keeping the flat keys for custom models.

* fix(ai): declare inclusive cache reporting for nested Agents usage

OpenAI-shaped usage counts cached tokens inside input_tokens. Declare
$ai_cache_reporting_exclusive: False when the cache count comes from the
nested usage details, as the Node Agents processor does, so the cost backend
does not infer exclusive reporting from a Claude model name served through an
OpenAI-compatible endpoint. Flat custom-model keys keep their own convention,
and a caller-supplied value still takes precedence.

* fix(ai): don't forward inclusive Agents reasoning tokens

OpenAI-compatible providers such as OpenRouter count reasoning inside
completion_tokens, while the cost backend adds reasoning on top for Gemini
model names, so forwarding the nested count priced reasoning twice. Keep
the original flat reasoning_tokens key only; cached tokens are unchanged.

(cherry picked from commit 4d6e44766175f80fbbf538ab4369b1f3916f05f7)

* chore: Release v7.80.3 [skip ci]

(cherry picked from commit 2989cd2f36607406c6d607d7033507ca69936711)

* Update generated references

(cherry picked from commit c380dc8cc731c4d1474a2427afcd3e426bf68b88)

* chore(deps): bump setuptools from 83.0.0 to 84.1.0 in /examples/example-ai-autogen in the uv group across 1 directory (#1131)

chore(deps): bump setuptools

Bumps the uv group with 1 update in the /examples/example-ai-llamaindex directory: [setuptools](https://github.com/pypa/setuptools).

Updates `setuptools` from 83.0.0 to 84.1.0
- [Release notes](https://github.com/pypa/setuptools/releases)
- [Changelog](https://github.com/pypa/setuptools/blob/main/NEWS.rst)
- [Commits](https://github.com/pypa/setuptools/compare/v83.0.0...v84.1.0)

---
updated-dependencies:
- dependency-name: setuptools
  dependency-version: 84.1.0
  dependency-type: indirect
  dependency-group: uv
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
(cherry picked from commit 6f765a3cb491de5b414e157e095c8155b59f18e2)

* chore(deps): bump authlib from 1.7.2 to 1.9.1 (#1140)

Bumps [authlib](https://github.com/authlib/authlib) from 1.7.2 to 1.9.1.
- [Release notes](https://github.com/authlib/authlib/releases)
- [Commits](https://github.com/authlib/authlib/compare/v1.7.2...v1.9.1)

---
updated-dependencies:
- dependency-name: authlib
  dependency-version: 1.9.1
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
(cherry picked from commit 871ad9e526058b889efd88957f29d0d86ccce5d0)

* chore(deps): bump the ai-providers group across 1 directory with 7 updates (#1145)

Bumps the ai-providers group with 7 updates in the / directory:

| Package | From | To |
| --- | --- | --- |
| [langchain](https://github.com/langchain-ai/langchain) | `1.4.4` | `1.4.7` |
| [langgraph](https://github.com/langchain-ai/langgraph) | `1.2.14` | `1.2.16` |
| [langchain-core](https://github.com/langchain-ai/langchain) | `1.6.9` | `1.6.13` |
| [langchain-openai](https://github.com/langchain-ai/langchain) | `1.7.0` | `1.7.2` |
| [langchain-anthropic](https://github.com/langchain-ai/langchain) | `1.7.5` | `1.7.6` |
| [google-genai](https://github.com/googleapis/python-genai) | `2.29.0` | `2.32.0` |
| [langchain-protocol](https://github.com/langchain-ai/agent-protocol) | `0.0.19` | `0.0.20` |

Updates `langchain` from 1.4.4 to 1.4.7
- [Release notes](https://github.com/langchain-ai/langchain/releases)
- [Commits](https://github.com/langchain-ai/langchain/compare/langchain==1.4.4...langchain==1.4.7)

Updates `langgraph` from 1.2.14 to 1.2.16
- [Release notes](https://github.com/langchain-ai/langgraph/releases)
- [Commits](https://github.com/langchain-ai/langgraph/compare/1.2.14...1.2.16)

Updates `langchain-core` from 1.6.9 to 1.6.13
- [Release notes](https://github.com/langchain-ai/langchain/releases)
- [Commits](https://github.com/langchain-ai/langchain/compare/langchain-core==1.6.9...langchain-core==1.6.13)

Updates `langchain-openai` from 1.7.0 to 1.7.2
- [Release notes](https://github.com/langchain-ai/langchain/releases)
- [Commits](https://github.com/langchain-ai/langchain/compare/langchain-openai==1.7.0...langchain-openai==1.7.2)

Updates `langchain-anthropic` from 1.7.5 to 1.7.6
- [Release notes](https://github.com/langchain-ai/langchain/releases)
- [Commits](https://github.com/langchain-ai/langchain/compare/langchain-anthropic==1.7.5...langchain-anthropic==1.7.6)

Updates `google-genai` from 2.29.0 to 2.32.0
- [Release notes](https://github.com/googleapis/python-genai/releases)
- [Changelog](https://github.com/googleapis/python-genai/blob/main/CHANGELOG.md)
- [Commits](https://github.com/googleapis/python-genai/compare/v2.29.0...v2.32.0)

Updates `langchain-protocol` from 0.0.19 to 0.0.20
- [Release notes](https://github.com/langchain-ai/agent-protocol/releases)
- [Commits](https://github.com/langchain-ai/agent-protocol/compare/langchain-protocol==0.0.19...langchain-protocol==0.0.20)

---
updated-dependencies:
- dependency-name: langchain
  dependency-version: 1.4.7
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: langgraph
  dependency-version: 1.2.16
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: langchain-core
  dependency-version: 1.6.13
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: langchain-openai
  dependency-version: 1.7.2
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: langchain-anthropic
  dependency-version: 1.7.6
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: google-genai
  dependency-version: 2.32.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: ai-providers
- dependency-name: langchain-protocol
  dependency-version: 0.0.20
  dependency-type: indirect
  update-type: version-update:semver-patch
  dependency-group: ai-providers
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
(cherry picked from commit c9105c5be181b3e81737c7481dc92f68f0cafcf7)

* ci: pin Python 3.11 by minor version for Ubuntu 26.04 runners (#1153)

(cherry picked from commit 50cf3573f2af46455c1a313e1649649f9d671180)

* feat: send transactional email from posthog templates (#1093)

* feat: send transactional email from posthog templates

Add posthog.emails.send() on Client and AsyncClient, plus standalone Emails and AsyncEmails in posthog.emails for processes that only send email. Sends go to the app host with a project secret API key and an Idempotency-Key, and retry network errors, rate limits, an in-flight key and 503 with the same key. Errors raise EmailError with PostHog's code and a Resend-style name.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: c62b30cd-eb2e-496c-a73d-2a6e7a52a62e

* fix: keep the email secret key off custom hosts and close email transports

client.emails only infers PostHog Cloud app hosts. A custom host raises app_host_required before any request, so an ingestion proxy never receives the project secret key. A kept emails reference now reads the client's disabled state on every send. Shutdown closes the email transport, a forked child drops the parent's email session, and the retry ceiling applies after jitter. Docs examples use client.emails.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: c62b30cd-eb2e-496c-a73d-2a6e7a52a62e

* fix: refuse email sends after shutdown and drain async sends first

client.emails now asks the parent client to admit each send. After shutdown, a kept or new client.emails raises EmailError with code client_closed and makes no request. On AsyncClient, an admitted send is an in-flight operation, so shutdown waits for it, retries included, before it closes the email transport.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: c62b30cd-eb2e-496c-a73d-2a6e7a52a62e

* feat: configure email sends on the client and document setup

Posthog and AsyncPosthog take app_host, email_request_timeout_seconds and email_request_max_retries, so client.emails works behind a custom host and a job runner can turn retries off. Emails takes a session and AsyncEmails an http_client, which gives tests a supported no-network path. The posthog.emails module docs now hold a setup quickstart: keys and scope, template ID, hosts, retries, errors to expect, cleanup and testing. The release note names client.emails.send().

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: c62b30cd-eb2e-496c-a73d-2a6e7a52a62e

* fix: match the email contract's missing_app_host and keep network causes

- A non-Cloud host without app_host now fails with missing_app_host, the code the cross-SDK contract in PostHog/posthog#120921 defines, matching posthog-node.
- A network_error keeps the requests or httpx exception as __cause__, so a failed send is easier to debug.
- Regenerate the public API snapshot after the rebase onto main.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: e72a2c10-f145-4f7c-8754-a4fcc06a18be

* fix: keep the bearer token and send timeout on injected async clients

An httpx.AsyncClient passed as http_client can carry its own auth and timeout. Its auth replaced the email's Bearer header, and its timeout replaced the send timeout, so a client built with timeout=None left sends unbounded. Each async send now passes auth=None and timeout=self.timeout.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: e72a2c10-f145-4f7c-8754-a4fcc06a18be

* fix: name 422 send checks validation_error

The template send route answers 422 for recipient_suppressed, sending_paused and sending_suspended as well as idempotency_key_reused. Only idempotency_key_reused is now invalid_idempotent_request; the send checks are validation_error, as the cross-SDK contract names them and posthog-node does.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: e72a2c10-f145-4f7c-8754-a4fcc06a18be

* fix: raise client_disabled for email sends from a disabled client

A disabled client returned SendEmailResult(id=None), so a caller that treats a returned result as sent dropped the email silently. It now sends nothing and raises EmailError with code client_disabled (name application_error), as the cross-SDK contract in PostHog/posthog#120921 defines and posthog-node does. SendEmailResult.id is always the queued message ID, so its type is str.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: e72a2c10-f145-4f7c-8754-a4fcc06a18be

* feat: send transactional email through the single send endpoint

`emails.send()` now posts to `/api/projects/@current/messaging/send/email/` with the template ID in the body as `template.id`, instead of the template-scoped route.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: e72a2c10-f145-4f7c-8754-a4fcc06a18be

* feat: read a transactional email's status with emails.get

`emails.get(message_id)` on `Emails` and `AsyncEmails` reads `/api/projects/@current/messaging/emails/{id}/` and returns an `EmailStatus` with `status`, `created_at`, `status_at` and `reason`. It retries network errors, rate limits and an unavailable service like `send`, which now shares that attempt loop.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: e72a2c10-f145-4f7c-8754-a4fcc06a18be

* docs: describe the email error name as a category

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: e72a2c10-f145-4f7c-8754-a4fcc06a18be

* feat: add a per-send max_retries to emails.send

`send(max_retries=...)` on `Emails` and `AsyncEmails` overrides the client's retry limit for one send, like posthog-node's `maxRetries`. A negative or non-integer value raises `ValueError` before any request.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: e72a2c10-f145-4f7c-8754-a4fcc06a18be
(cherry picked from commit 01e91db94185969f636d501a58a21eb9f2c3721e)

* chore: Release v7.81.0 [skip ci]

(cherry picked from commit 139161e95fc7328148a54e442f3792a3a6038da0)

* Update generated references

(cherry picked from commit af8e1aa3f000a89091fe43fe7b639449671e1529)

* fix(ai): keep the choices of an n>1 OpenAI chat stream apart (#1152)

Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
(cherry picked from commit be77b83462a690c8c303f5676a28207a9c17f68a)

* fix(ai): preserve Gemini tool call IDs in captured generations (#1150)

Gemini returns IDs on function calls and function responses, but the
captured $ai_output_choices and $ai_input dropped them, so tool results
could not be matched to their calls. Keep the call ID on output (streaming
and non-streaming) and on input history calls, and the result's ID as
tool_use_id, matching PostHog/posthog-js#5087.

Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
(cherry picked from commit a61e64d1a40be337af3e09925d53f56c78d0baba)

* chore: Release v7.81.1 [skip ci]

(cherry picked from commit f23efe9d3e0dfe68a7136f80a48e6420ada38e41)

* Update generated references

(cherry picked from commit dca22075ced7f4beaca798a8dd155169eec50b4d)

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: alexcastgonz <82875882+pcbeingused333@users.noreply.github.com>
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
Co-authored-by: releaser-posthog-python[bot] <262414804+releaser-posthog-python[bot]@users.noreply.github.com>
Co-authored-by: Anna Garcia <11654201+turnipdabeets@users.noreply.github.com>
Co-authored-by: Manoel Aranda Neto <5731772+marandaneto@users.noreply.github.com>
Co-authored-by: Georgis Andonis <6430745+gesh@users.noreply.github.com>
Co-authored-by: Carlos Marchal <carlos@posthog.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Radu Raicea <radu@raicea.com>
Co-authored-by: Gustavo H. Strassburger <gustavo@posthog.com>
Co-authored-by: Daniel Visca <daniel.v@posthog.com>
Co-authored-by: mosafariuk <mosafariuk@gmail.com>
Co-authored-by: Anders <754494+andehen@users.noreply.github.com>
Co-authored-by: Harley Alexander <43975092+mayteio@users.noreply.github.com>
Co-authored-by: Jae Bird <jaeminbird@gmail.com>
…tegrations (#1158)

* fix(ai): treat an empty or whitespace distinct id as missing in AI integrations

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: note blank AI distinct ids count as missing in the 8.0 guide

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: correct the 7.x empty ai distinct id behavior per integration

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…1159)

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* docs: close review gaps in the 8.0 migration guide

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix: always send fractional seconds in capture timestamps

Event timestamp, batch created_at and PostHog-Request-Timestamp now
use one helper that writes six fractional digits in UTC, so whole
seconds no longer lose the fraction.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* test: cover the capture timestamp formatter directly

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: link the capture v1 SDK upgrade guide from the 8.0 guide

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* test: check timestamp UTC conversion under a non-UTC local zone

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* docs(mcp): clarify unsigned session token trust (#1151)

(cherry picked from commit b4c2e47)

* feat: send a flat body from emails.send (#1155)

emails.send now posts template_id and distinct_id at the top level of the request body, for the sync and async clients. The server accepts both shapes, so this changes only the wire format.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Generated-By: PostHog Desktop
Task-Id: e72a2c10-f145-4f7c-8754-a4fcc06a18be
(cherry picked from commit f20066e)

* chore: Release v7.81.2 [skip ci]

(cherry picked from commit 0aec2d9)

* Update generated references

(cherry picked from commit 14b4d33)

* feat: add a per-read send_feature_flag_events option to snapshot reads (#1148)

(cherry picked from commit b10c38a)

* fix: send flag id and version on locally evaluated $feature_flag_called (#1149)

Locally evaluated flags in get_feature_flag(), feature_enabled() and
get_feature_flag_result() sent $feature_flag_called without
$feature_flag_id or $feature_flag_version, because both were read only
from remote /flags metadata. Read them from the local flag definition
that produced the value. evaluate_flags() already sent the id for local
flags but dropped the definition's version; send it too.

(cherry picked from commit e24100b)

* chore: Release v7.82.0 [skip ci]

(cherry picked from commit b6f8eaf)

* Update generated references

(cherry picked from commit 0a5417f)

* fix(flags): treat concurrent fallback-cache eviction as a cache miss (#1167)

* fix(flags): treat concurrent cache eviction as a cache miss

* chore: add changeset for fallback cache eviction fix

---------

Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
(cherry picked from commit 9ebfb87)

* fix(flags): stringify regex properties as compact JSON (#1157)

Local regex and not_regex matching called str() on the property.

(cherry picked from commit 5b50683)

* fix(flags): serialize UUID values in flags request bodies (#1156)

The flags request body goes straight to DatetimeSerializer without clean(), so a UUID distinct id, group key, or flag property raised TypeError. The SDK swallowed it and returned missing flags without sending a request. Serialize UUIDs as strings, matching what clean() already does for captured events.

(cherry picked from commit a239479)

* chore: Release v7.82.1 [skip ci]

(cherry picked from commit 8c5e9df)

* Update generated references

(cherry picked from commit fbbd109)

* fix(ai): count Gemini tool use prompt tokens as input tokens (#1164)

* fix(ai): count Gemini tool use prompt tokens as input tokens

Built-in tools (Google Search, URL context) report the retrieved content in
tool_use_prompt_token_count, separately from prompt_token_count, so input
tokens were under-reported on grounded responses.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Signed-off-by: Alex Castillo <alex.castillog33@gmail.com>

* chore: add changeset for Gemini tool use prompt tokens

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Signed-off-by: Alex Castillo <alex.castillog33@gmail.com>

* fix(ai): keep unbilled Google Search retrieval out of Gemini input tokens

tool_use_prompt_token_count is only added to input tokens when the response
carries url_context_metadata (billed as input); Search retrieval is not.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(ai): carry Gemini URL context flag across stream chunks

URL context metadata can arrive on a different chunk than the final
usage_metadata, so the tool-use prompt tokens were dropped in streaming.
Track the flag across chunks in the sync and async clients.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(ci): format Gemini clients and refresh API snapshot

---------

Signed-off-by: Alex Castillo <alex.castillog33@gmail.com>
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
(cherry picked from commit 57b9db8)

* chore: Release v7.82.2 [skip ci]

(cherry picked from commit 16e99f0)

* Update generated references

(cherry picked from commit 1dfca27)

---------

Signed-off-by: Alex Castillo <alex.castillog33@gmail.com>
Co-authored-by: Dustin Byrne <dustin@posthog.com>
Co-authored-by: Harley Alexander <43975092+mayteio@users.noreply.github.com>
Co-authored-by: releaser-posthog-python[bot] <262414804+releaser-posthog-python[bot]@users.noreply.github.com>
Co-authored-by: Anders <754494+andehen@users.noreply.github.com>
Co-authored-by: Sarbajit Sengupta <73594751+Sarbajit-Sengupta@users.noreply.github.com>
Co-authored-by: Manoel Aranda Neto <marandaneto@gmail.com>
Co-authored-by: Luan Taraschi <130802253+luantaraschi@users.noreply.github.com>
Co-authored-by: alexcastgonz <82875882+pcbeingused333@users.noreply.github.com>
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
* feat(flags): read the /flags?v=4 response (#1031)

(cherry picked from commit e094dc5)

* chore: Release v7.83.0 [skip ci]

(cherry picked from commit 43c7545)

* Update generated references

(cherry picked from commit 314fad0)

* fix(mcp): inject context into mcp 2.x tool objects and test without fastmcp (#1166)

* test(mcp): share the dual-major structured content reader

* test(mcp): read structured content without the fastmcp camelcase shim

* fix(mcp): inject context into mcp 2.x tool objects

* ci: run the mcp 2.x suite without fastmcp

(cherry picked from commit 0c12ca1)

* chore: Release v7.83.1 [skip ci]

(cherry picked from commit 9f5d458)

* Update generated references

(cherry picked from commit 06e5611)

* test(flags): read flag scenario events in before_send on v1

v1 passes flag properties to _enqueue separately and layers them while
building the event, so a stubbed _enqueue never saw them. Reading the
built event in before_send compares v1's real output with 7.82.1.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Anders <754494+andehen@users.noreply.github.com>
Co-authored-by: releaser-posthog-python[bot] <262414804+releaser-posthog-python[bot]@users.noreply.github.com>
Co-authored-by: Connor Grady <50903811+connor-grady@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…y context options, AI options (#1202, #1206, #1207, #1208)

Squash of the per-event $lib stack: #1202 (per-event $lib/$lib_version), #1206 (MCP event_options), #1207 (Celery context options to workers) and #1208 (capture options in the AI integrations).
…ng (#1224)

* docs: show how to keep identity calls under a global personless setting

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: scope the context tag opt-out to alias and group_identify

`set` and `set_once` add context tags to the person properties they set,
so a `$process_person_profile` context tag doesn't turn person processing
off there. Also note that events sent with person processing off are
still linked to an existing person and billed as identified.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
… errors, AI, references)

Real merge of main at f66e696 (the commit before #1154), so v1's merge
base with main moves. $experiment_exposure (#1154) and later come in a
second merge.

v1 already carries main through 7.83.1 and its reference update
(06e5611) from the squashed sync #1175, so every path is resolved
against 06e5611, not the old merge base: paths main has not changed
since keep v1's version, paths v1 has not changed take main's, and the
rest are a three-way merge of v1, 06e5611 and main. references/ ends
equal to main's (18 HogRef files; the 163 #1225 removed stay removed).

Resolutions:
- posthog/capture_v1.py, posthog/test/test_capture_v1.py: stay removed.
  main's only change is #1178 (7.x opt-in v1 retries on 404/429); v1
  retries those since #1132. AGENTS.md keeps v1's capture note.
- posthog/client.py: v1's super_properties/super_options docs plus
  main's migrated_from doc (emails).
- posthog/ai/anthropic/anthropic_async.py: keep v1 without
  `from posthog import setup`; #1146's logging import is kept.
- sdk_compliance_adapter: v1's capability list plus #1179's
  feature_flag_payloads and its /get_feature_flag_payload route; harness
  image stays 1.26.0 (the harness bump is separate); workflow keeps v1's
  jobs with main's action pins.
- examples crewai/dspy/litellm: v1's local posthog source plus #1222's
  versions (litellm 1.100.4, oauthlib 4.0.1); locks regenerated.
- references/public_api_snapshot.txt: regenerated; the only changes are
  main's email API and migrated_from.

Test-only edits to tests main added, for v1's API: AI wrapper tests read
capture_ai (#1189, #1205, #1211); #1210's test stubs _send_v1_batch.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…flag cache URLs)

Second real merge, of main at c60e7a6. #1154 sends $experiment_exposure
for config version 2 experiment splits and pins their flag attribution so
no super property changes or adds it.

Resolutions (#1154 into v1's layered pipeline):
- The pin rides in _EventDefaults and applies at the end of
  _fill_event_defaults, after every layer and the library pair, then the
  property allowlist; main applies it after its super property merge and
  before the allowlist. One place serves both clients: the sync client
  passes it to _build_event_defaults, and AsyncPosthog, which fills
  defaults later in _process_event, passes it through _event_defaults.
- main's own enrichment block (super property merge, $release_id,
  $is_server) stays out: v1's layers replace it.
- _enqueue and capture take v1's layer arguments plus #1154's
  pinned_properties and before_send_drops; v1's _report_capture_failure
  stays. args.py keeps v1's _capture_trace_context next to #1154's fields.

Test-only edits to #1154's tests, for v1:
- Senders are v1's (_send_v1_batch, async_send_v1_batch); capture_mode is
  gone.
- The parity recorders keep the event in before_send (v1 reads flag
  events there) instead of returning an event UUID, which v1's hook
  rejects; returning the event is what makes capture accept it.
- The async attribution test records events in before_send, after the
  pin, not from _prepare_event, which on v1 runs before defaults fill and
  would pass without the pin. Removing the pin fails it, and the sync
  tests.
- Config version 1 calls: super properties fill every key the event
  leaves unset and the event's own values win (8.0's rule), instead of
  1.x's override.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Run with --major 8, so the 7.x and latest pages stay as main generated
them. The 8.x page now documents 8.0 as merged here (emails, per-event
$lib, MCP and AI options, Celery context options); the migration page
matches docs/migration-7.x-to-8.0.md.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
eli-r-ph and others added 7 commits November 11, 2026 13:44
main's benchmark harness (#1134, #1135) drove 7.x's legacy sender: the
client passed gzip=False and capture_mode="v0", and the loopback server
answered /batch with {"status":1}. 8.0 removed both options and only
speaks capture v1, so every benchmark worker failed on this branch, and
would on main after 8.0 merges.

- workload.py and run.py: drop gzip, capture_mode and
  POSTHOG_CAPTURE_MODE; requests are uncompressed capture v1.
- server.py: receive /i/v1/analytics/events and answer with a result per
  event uuid, so the SDK reads each event as delivered. The delivery
  checks are unchanged.
- test_benchmarks.py posts its synthetic batch to the v1 path; README
  describes the v1 transport.

benchmarks/test_benchmarks.py: 28 tests pass (18 errored before).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
)

* docs: link the Celery headers guide by its main URL in the 8.0 guide

The migration guide is also published on posthog.com, where the relative
link resolves to a missing page.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: describe Celery eager headers without an external link

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: regenerate the migration reference for the Celery eager note

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Each generated reference page now installs the major it describes, so a
reader of the 7.x page doesn't get 8.0 after it ships.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* docs: link the Celery headers guide by its main URL in the 8.0 guide

The migration guide is also published on posthog.com, where the relative
link resolves to a missing page.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: describe Celery eager headers without an external link

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: sort the 8.0.0 migration guide by when you make each change

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: replace filler lead-ins in the 8.0.0 migration guide

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: keep the 8.0.0 guide's table cells to one value

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: title a single migration guide's page with its H1

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit f143283)

* docs: move the posthog.com Python SDK pages into the reference

The posthog.com Python SDK pages redirect to this reference once 8.0.0
ships, so their setup, how-to, and configuration sections now live in
docs/getting-started.md.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 9db7014)

* docs: link the JavaScript Web reference instead of its retired page

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 0e9f755)

* Revert "docs: link the JavaScript Web reference instead of its retired page"

This reverts commit 7abca08.

* Revert "docs: move the posthog.com Python SDK pages into the reference"

This reverts commit bc7bc9c.

* docs: add the Python facts from the retired posthog.com page that no other page has

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: move the Python one-API facts from getting-started into docstrings

The reference renders each docstring under its entry, per major, so the
tracing, ASGI, keepalive, and AsyncPosthog facts now live there.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: preserve Python reference details

* feat!: default identify-style events to person processing on

$identify, $create_alias, $merge_dangerously, $groupidentify, $set and
$set_once now resolve process_person_profile from the call alone before any
layer fills: a non-null call option wins, else a non-null call
$process_person_profile property moves into the option, else the option is
true unless the SDK generated the distinct ID. Context options and tags,
super_options and super_properties can no longer turn person processing off
for these events. before_send still sees the value and has the final say.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: describe the identify-style person processing default in the 8.0 guide

Replaces the guidance that layered process_person_profile false reaches
alias, group_identify, set and set_once and needs a per-call true, and
regenerates the 8.x and migration references.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix: show before_send one person processing value on identify-style events

After the layers fill, an identify-style event whose process_person_profile
option is set loses its $process_person_profile property, so before_send
sees one value and a hook that drops options cannot send a layered legacy
false. Event defaults now take one personless flag, which sets both the
personless option layer and the generated-id check. Option layer
docstrings say they never set process_person_profile on these events.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: fold the identify-style default into the identity release note

Moves the release note into identity-and-context-rules, notes that before_send
sees only the option, and regenerates the 8.x and migration references.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: scope the identify-style default to SDK-chosen distinct IDs

Context options and super_options still fill process_person_profile on an
identify-style event whose distinct ID the SDK generated, so the release
note, guide, AGENTS.md and option docstrings now say so, and a test pins it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: restore the self-hosted, 404 and duplicate-uuid notes in the sorted guide

The sorted 8.0.0 guide dropped facts the v1 guide states: the self-hosted
capture v1 route check, the 404 hint, the 400 duplicate_event_uuid on older
PostHog, the legacy path removal, and NaN as null since 7.75.0.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: describe the identify-style person processing default in the sorted guide

Identify-style events now default process_person_profile to true unless
the call sets it or the SDK generated the distinct ID, so layered false no
longer reaches them. Replaces the old reach rules and opt-in sample, and
updates the checklist row, precedence step 1 and the before_send section.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: correct the span limit order and the record_exception loss note

start_span evicts spans older than max_span_age first and returns an
unrecorded span only if still at max_live_spans. A full span keeps the
exception message in its status and loses only the type and stack trace.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: regenerate the 8.x and migration references

Generated with bin/docs_scripts/generate_markdown.py --major 8. The
migration page now carries the sorted guide, which it had not before.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: name the project setting for keeping identify-style events personless

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: regenerate the migration reference

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: say how to keep an identify-style event personless

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: regenerate the migration reference

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: name the calls whose $process_person_profile property becomes the option

Only capture and capture_ai take event properties. The properties of set,
set_once and group_identify are person and group properties.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs: name the calls whose $process_person_profile property becomes the option

Only capture and capture_ai take event properties. The properties of set,
set_once and group_identify are person and group properties.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat: give CaptureError.drops entries named fields

Each entry is now a CaptureDroppedEvent named tuple with uuid and details, so
code that unpacks or compares (uuid, details) tuples keeps working.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs: name the CaptureDroppedEvent entries of CaptureError.drops in the guide

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs: restore upgrade notes that the retired posthog.com page carried

Compression formats including zstd, the proxy pass-through rule, and that a
`$set` in a global or context layer reaches every event.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs: name the SDK in versions, cut self-hosted notes, and split dashes and semicolons

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: drop self-hosted lines from the migration guide

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: drop the "(experimental)" title from the next-major reference page

The 8.x page is titled "PostHog Python SDK 8.x" before and after the release.
Regenerating also picks up the self-hosted paragraph the migration guide's source
already dropped.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: say to pass process_person_profile in options on set and group_identify

"so set the option there" read as setting it in properties, where the SDK
treats it as a person property. set, set_once, and group_identify take
options, so the guide now names them.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: cut the 404 hint and self-hosted lines from the migration guide

Vincent ruled out self-hosted framing in the guides, and the endpoint and 404 sentences read as self-host framing. The retry tables and the proxy notes stay.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Vincent (Wen Yu) Ge <vincent@posthog.com>
Co-authored-by: Vincent (Wen Yu) Ge <gewenyu99@gmail.com>
Brings main 83eee80 into v1 so the release-day merge of v1 into main is
conflict-free.

#1242 (refactor, no behavior change) is left out where v1 rewrote the
same code: client.py, async_client.py, consumer.py, _async_consumer.py,
_async_request.py, capture_send.py (main's capture_v1.py),
ai/prompts.py, ai/utils.py, ai/langchain/callbacks.py,
ai/openai_agents/processor.py, ai/claude_agent_sdk/processor.py and the
MCP __init__/_instrumentation modules. It merges as is everywhere else.

Behavior changes ported onto v1's code in those files:
- #1238 failed flag records: EVALUATION_FAILED error, value_from_details
  for failed records, failed records stay in unresolved_flags.
- #1253 Anthropic stream error status (ai/utils.py, langchain).
- #1256 Claude Agent SDK tool results, every output block, subagent
  messages and retried requests.
- #1226, #1229, #1231 MCP raised-error reporting and $mcp_error_status
  (applied from main before #1242).
- test_utils.py (#1230), test_pipeline.py (#1226) and
  test_http_status.py (#1253): both sides kept.

main's new tests adapted to v1's contract:
- AI events go through capture_ai (#1249 Anthropic stream tests,
  claude_agent_sdk/test_client.py).
- Person processing is the process_person_profile option, not the
  $process_person_profile property (claude_agent_sdk/test_client.py).
- Snapshot flag properties reach _enqueue as flag_properties
  (test_failed_flag_records.py).

Kept from v1: harness 1.26.0 in sdk-compliance.yml, the compliance
docker-compose and CONTRIBUTING (HARNESS-1; main moved to 1.31.1).
references/public_api_snapshot.txt regenerated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Brings main f225118 (ExceptionCapture docstrings) into v1. No conflicts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@eli-r-ph eli-r-ph changed the title V1 release - DO NOT MERGE feat!: posthog-python 8.0 on capture v1 Nov 16, 2026
@eli-r-ph
eli-r-ph marked this pull request as ready for review November 16, 2026 20:35
@veria-ai

veria-ai Bot commented Nov 16, 2026 •

Copy link
Copy Markdown

PR overview

All previously flagged issues have been addressed. No open security concerns remain on this pull request.

Fixed/addressed: 1 · PR risk: 0/10

Reviews

@github-actions

Copy link
Copy Markdown
Contributor

SDK performance: posthog-python

Benchmark execution failed. No overall performance conclusion is available.

Flags require at least a 5% change in the median and all six pairs agreeing on direction. This is a noise filter, not proof of a performance change or a CI gate.

View all results
Test Measurement Before PR With PR Change Assessment
Capture and deliver events (32-byte payload, burst)
Shard: capture
Allocated memory per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, burst)
Shard: capture
Allocations per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, burst)
Shard: capture
Time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, burst)
Shard: capture
Idle process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, burst)
Shard: capture
Observed peak process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, burst)
Shard: capture
Managed heap after settling — — — unavailable; Source execution:​ failure; base:​ error; head:​ unsupported; head:​ CPython has no uninstrumented total managed-heap byte counter; tracing would distort this RSS pass.​; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, burst)
Shard: capture
Process memory after settling (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, burst)
Shard: capture
Process CPU time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, burst)
Shard: capture
Lifetime peak process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Allocated memory per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Allocations per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Idle process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Observed peak process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Managed heap after settling — — — unavailable; Source execution:​ failure; base:​ error; head:​ unsupported; head:​ CPython has no uninstrumented total managed-heap byte counter; tracing would distort this RSS pass.​; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Process memory after settling (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Process CPU time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Lifetime peak process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, steady)
Shard: capture
Allocated memory per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, steady)
Shard: capture
Allocations per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, steady)
Shard: capture
Time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, steady)
Shard: capture
Idle process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, steady)
Shard: capture
Observed peak process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, steady)
Shard: capture
Managed heap after settling — — — unavailable; Source execution:​ failure; base:​ error; head:​ unsupported; head:​ CPython has no uninstrumented total managed-heap byte counter; tracing would distort this RSS pass.​; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, steady)
Shard: capture
Process memory after settling (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, steady)
Shard: capture
Process CPU time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (32-byte payload, steady)
Shard: capture
Lifetime peak process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Allocated memory per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Allocations per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Idle process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Observed peak process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Managed heap after settling — — — unavailable; Source execution:​ failure; base:​ error; head:​ unsupported; head:​ CPython has no uninstrumented total managed-heap byte counter; tracing would distort this RSS pass.​; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Process memory after settling (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Process CPU time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Lifetime peak process memory (RSS) — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete; base allocation repetition 0: Worker did not complete; base memory repetition 0: Worker did not complete
Queue an event (32-byte payload)
Shard: capture
Allocated memory per operation — — — unavailable; Source execution:​ failure; base:​ error; head:​ unsupported; head:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​; base timing repetition 0: Worker did not complete
Queue an event (32-byte payload)
Shard: capture
Allocations per operation — — — unavailable; Source execution:​ failure; base:​ error; head:​ unsupported; head:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​; base timing repetition 0: Worker did not complete
Queue an event (32-byte payload)
Shard: capture
Caller-thread CPU time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete
Queue an event (32-byte payload)
Shard: capture
Time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete
Queue an event (4,096-byte payload)
Shard: capture
Allocated memory per operation — — — unavailable; Source execution:​ failure; base:​ error; head:​ unsupported; head:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​; base timing repetition 0: Worker did not complete
Queue an event (4,096-byte payload)
Shard: capture
Allocations per operation — — — unavailable; Source execution:​ failure; base:​ error; head:​ unsupported; head:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​; base timing repetition 0: Worker did not complete
Queue an event (4,096-byte payload)
Shard: capture
Caller-thread CPU time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete
Queue an event (4,096-byte payload)
Shard: capture
Time per operation — — — unavailable; Source execution:​ failure; base:​ error; base timing repetition 0: Worker did not complete
Evaluate all 10 flags
Shard: flags-scaling
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate all 10 flags
Shard: flags-scaling
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate all 10 flags
Shard: flags-scaling
Time per operation — — — unavailable; Source execution:​ failure
Evaluate all 10 flags
Shard: flags-scaling
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate all 100 flags
Shard: flags-scaling
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate all 100 flags
Shard: flags-scaling
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate all 100 flags
Shard: flags-scaling
Time per operation — — — unavailable; Source execution:​ failure
Evaluate all 100 flags
Shard: flags-scaling
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate all 1,000 flags
Shard: flags-scaling
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate all 1,000 flags
Shard: flags-scaling
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate all 1,000 flags
Shard: flags-scaling
Time per operation — — — unavailable; Source execution:​ failure
Evaluate all 1,000 flags
Shard: flags-scaling
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Idle process memory (RSS) — — — unavailable; Source execution:​ failure
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Observed peak process memory (RSS) — — — unavailable; Source execution:​ failure
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Managed heap after settling — — — unavailable; Source execution:​ failure; base:​ unsupported; head:​ unsupported; base:​ CPython has no uninstrumented total managed-heap byte counter; tracing would distort this RSS pass.​; head:​ CPython has no uninstrumented total managed-heap byte counter; tracing would distort this RSS pass.​
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Process memory after settling (RSS) — — — unavailable; Source execution:​ failure
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Lifetime peak process memory (RSS) — — — unavailable; Source execution:​ failure
Refresh 1,000 flag definitions
Shard: flags-memory
Idle process memory (RSS) — — — unavailable; Source execution:​ failure
Refresh 1,000 flag definitions
Shard: flags-memory
Observed peak process memory (RSS) — — — unavailable; Source execution:​ failure
Refresh 1,000 flag definitions
Shard: flags-memory
Managed heap after settling — — — unavailable; Source execution:​ failure; base:​ unsupported; head:​ unsupported; base:​ CPython has no uninstrumented total managed-heap byte counter; tracing would distort this RSS pass.​; head:​ CPython has no uninstrumented total managed-heap byte counter; tracing would distort this RSS pass.​
Refresh 1,000 flag definitions
Shard: flags-memory
Process memory after settling (RSS) — — — unavailable; Source execution:​ failure
Refresh 1,000 flag definitions
Shard: flags-memory
Lifetime peak process memory (RSS) — — — unavailable; Source execution:​ failure
Evaluate one boolean flag
Shard: flags-single
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate one boolean flag
Shard: flags-single
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate one boolean flag
Shard: flags-single
Time per operation — — — unavailable; Source execution:​ failure
Evaluate one boolean flag
Shard: flags-single
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate a cohort-filtered flag
Shard: flags-single
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate a cohort-filtered flag
Shard: flags-single
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate a cohort-filtered flag
Shard: flags-single
Time per operation — — — unavailable; Source execution:​ failure
Evaluate a cohort-filtered flag
Shard: flags-single
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (early match)
Shard: flags-single
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (early match)
Shard: flags-single
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (early match)
Shard: flags-single
Time per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (early match)
Shard: flags-single
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (late match)
Shard: flags-single
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (late match)
Shard: flags-single
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (late match)
Shard: flags-single
Time per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (late match)
Shard: flags-single
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (no match)
Shard: flags-single
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (no match)
Shard: flags-single
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (no match)
Shard: flags-single
Time per operation — — — unavailable; Source execution:​ failure
Evaluate a flag (no match)
Shard: flags-single
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate a property-filtered flag
Shard: flags-single
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate a property-filtered flag
Shard: flags-single
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate a property-filtered flag
Shard: flags-single
Time per operation — — — unavailable; Source execution:​ failure
Evaluate a property-filtered flag
Shard: flags-single
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate a regex-filtered flag
Shard: flags-single
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate a regex-filtered flag
Shard: flags-single
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate a regex-filtered flag
Shard: flags-single
Time per operation — — — unavailable; Source execution:​ failure
Evaluate a regex-filtered flag
Shard: flags-single
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate a rollout flag
Shard: flags-single
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate a rollout flag
Shard: flags-single
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate a rollout flag
Shard: flags-single
Time per operation — — — unavailable; Source execution:​ failure
Evaluate a rollout flag
Shard: flags-single
Process CPU time per operation — — — unavailable; Source execution:​ failure
Evaluate a multivariate flag
Shard: flags-single
Allocated memory per operation — — — unavailable; Source execution:​ failure
Evaluate a multivariate flag
Shard: flags-single
Allocations per operation — — — unavailable; Source execution:​ failure
Evaluate a multivariate flag
Shard: flags-single
Time per operation — — — unavailable; Source execution:​ failure
Evaluate a multivariate flag
Shard: flags-single
Process CPU time per operation — — — unavailable; Source execution:​ failure

Values are medians of paired measurements. Not flagged does not mean unchanged. Unavailable means a measurement could not be provided, not zero. Estimated allocation measurements retain their instrumentation uncertainty. RSS measures the whole process, not just SDK allocations; observed peaks are sampled checkpoints. Paired ranges in the diagnostics describe variability, not confidence intervals.

Each shard uses its own runner; compare Before PR against With PR within a row, not across shards.

Compared f22511891225fb15a4e6a0f889e723e9b5d0ebe4 → 018d15de54954282867f1e3ea65313a282c96419.

Full report and diagnostics

102 metric rows; 102 rows shown in all results. Omitted or shortened data remain in comparison.json, including per-runner environments, resource limits, measurement methods, samples and checkpoints.

Comment thread posthog/ai/openai/openai.py Resolved

@dustinbyrne dustinbyrne left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed the complete 8.0 change at 018d15d.

I independently reproduced the existing streaming setup-error options finding: sync and async OpenAI 401 errors emit empty options despite explicit per-call cookieless/personless controls. Equivalent 7.x property controls and the new API’s success cases preserve them. Consider incorporating the scoped forwarding correction before accepting this release; #1264 is not part of this snapshot.

The focused suite passed 877 tests (11 skipped), and real-HTTP sync/async lane, gzip and partial-retry checks passed. Two optional inline notes cover the paired benchmark receiver and an inaccurate NaN release-note example.

AI-assisted review.

Comment thread benchmarks/server.py Resolved
@dustinbyrne
dustinbyrne requested a review from a team November 16, 2026 22:57

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants