Skip to content

chore: merge main into v1 (DO NOT SQUASH, lands by fast-forward) - #1262

Merged
eli-r-ph merged 35 commits into
v1from
eli/v1-sync-main-20261116
Nov 16, 2026
Merged

eli-r-ph merged 35 commits into
v1from
eli/v1-sync-main-20261116

Conversation

@eli-r-ph

Copy link
Copy Markdown
Contributor

💡 Motivation and Context

v1 is missing main's 7.87.2-7.91.2 and later fixes. These include:

The 8.0 release PR (#1016) conflicts with main until v1 carries them.

Warning

Do not squash or merge this PR with the button. After approval it lands by fast-forwarding v1 to this exact head, which keeps both merge commits. Squashing drops them, and #1016 keeps conflicting.

Changes:

💚 How did you test it?

  • Every line main added or removed since the last sync, outside refactor: reduce SDK complexity and improve CRAP analysis #1242, is on this head, except the listed adaptations.
  • Every v1 line is kept, except one comment rewrapped to add error_status.
  • 8,312 tests pass, including main's tests for each ported fix. The 3 fork tests in test_metrics_autocapture.py fail locally on pristine v1 too (macOS fork).
  • ruff, mypy (baseline), the public API check and the utils.py CRAP gate pass.
  • Merging this head into current main has 0 conflicts, and the result equals this head.

📝 Checklist

  • All commits are signed (see Commit signing)
  • 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

No changeset: main's changes shipped in 7.87.2-7.91.2, and 8.0's changesets are on v1.

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

Claude Code (Claude Opus 5.5).

🤖 Generated with Claude Code

Ishkirat-Singh and others added 30 commits November 12, 2026 11:48
…1230)

clean() turned set, list and tuple values into lists but sent frozenset
and deque values as None, so those properties were silently lost.
…e-ai-crewai in the uv group across 1 directory (#1241)

chore(deps): bump langgraph

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


Updates `langgraph` from 1.2.12 to 1.2.14
- [Release notes](https://github.com/langchain-ai/langgraph/releases)
- [Commits](langchain-ai/langgraph@1.2.12...1.2.14)

---
updated-dependencies:
- dependency-name: langgraph
  dependency-version: 1.2.14
  dependency-type: direct:production
  dependency-group: uv
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* feat(emails): let a code send turn off open and click tracking

Adds track_opens_and_clicks to the code form of emails.send. False sends
the email without PostHog's open pixel and rewritten links. The field
goes into the request body only when the caller sets it, so a default
send keeps today's body. The template form refuses it, like
capture_engagement.

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

* docs(emails): say track_opens_and_clicks only removes the open pixel

PostHog never rewrites the links of an email sent from code, so it
counts no clicks either way. False removes the open pixel only.

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

* feat(emails): let a code send choose its tracking

Replace track_opens_and_clicks with tracking: True, False, or an
EmailTracking dict of opens, clicks and engagement. Leave tracking and
capture_engagement out of the body unless the caller sets them, so the
API can refuse a send that sets both. Deprecate capture_engagement.

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

* test(emails): prove the tracking choice over real HTTP

Send each tracking form through the public send API to a local HTTP
server and assert the body it receives, instead of a stubbed transport.

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

* feat(emails): let a code send mark its email sensitive

A code send now takes an optional `sensitive`, sent only when set. Without
it, PostHog keeps today's behavior: a send whose template_key matches a
saved template follows that template, and any other send is sensitive.
Click tracking follows it unless `tracking` sets clicks.

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

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…1226)

* test(mcp): failing repro for FastMCP tool failures losing the raised exception

Standalone FastMCP converts a raised tool exception into an isError result
before our wrapped handler returns, so the failure is reported as a generic
"Error" with no frames (posthog-python#1223). Fails on fastmcp 3/mcp 1 and
fastmcp 4/mcp 2.

* fix(mcp): report the exception behind a FastMCP isError tool result

Standalone FastMCP converts a raised tool exception into
CallToolResult(isError=True) inside the tools/call handler we wrap, so
every failure was captured from the result text: $mcp_error_type "Error",
one generic $exception_list entry, no frames (posthog-python#1223).

instrument() now wraps the server's public call_tool, which still raises
(ToolError chained from the tool's exception). The wrapper hands the
outermost call's outcome to the enclosing handler through a context
variable, and the handler records it next to the result. fastmcp's
"Error calling tool" ToolError joins the dispatch wrappers the scalars
step past, so $mcp_error_type names the tool's own exception.

Tested:
- test_fastmcp_tool_errors.py fails before, passes after, on fastmcp 3.2 +
  mcp 1.30 and fastmcp 4.1.8 + mcp 2.3
- pytest posthog/test/mcp: 694 passed (mcp 1.x), 575 passed (mcp 2.x)
- issue repro: $mcp_error_type RuntimeError, chain ToolError ->
  RuntimeError -> ConnectionError with frames
- ruff format/check, mypy clean

Not included: $mcp_error_status. The taxonomy defines it as the upstream
status of PostHog's own server calls and no SDK sets it today; setting it
from the SDK is a cross-SDK contract change.

* fix(mcp): keep another server's handled tool failure off this call

Pre-PR gate findings on the previous commit:

- The error slot was shared across servers. A tool that called another
  instrumented FastMCP's call_tool and handled its failure got that
  exception on its own successful event, whenever its own server's
  call_tool was not wrapped (a raw low-level server, or a FastMCP
  instrumented only through _mcp_server). The slot now names the server
  whose handler opened it, and only that server's call_tool writes to it.
- The "Error calling tool" wrapper match also unwrapped a tool's own
  ToolError worded that way. It now requires fastmcp's quoted tool name.

Tested: both repros added to test_fastmcp_tool_errors.py; pytest
posthog/test/mcp 697 passed (mcp 1.30 + fastmcp 3.2), 578 passed
(mcp 2.3 + fastmcp 4.1.8); ruff and mypy clean.

* fix(mcp): keep FastMCP 2.x instrumented, and freeze the slot on exit

Review on 81b3170 (PostHog Review, confirmed by ioannisj):

- FastMCP 2.x has no public call_tool, so wrapping it raised
  AttributeError and instrument() fell back to no-op analytics: no MCP
  events at all on a supported stack. 2.x dispatches tools/call through
  _call_tool_middleware, so that is the seam wrapped there. 2.x now also
  reports the raised exception.
- A background task holding the slot could settle it after the handler's
  dispatch returned. The slot now closes when the handler leaves it.
- CI gains a v1-fastmcp2 leg (mcp 1.x + fastmcp 2.x), which would have
  caught the first point.

Tested: pytest posthog/test/mcp passes on fastmcp 2.14.7 + mcp 1.30,
fastmcp 3.2 + mcp 1.30, fastmcp 4.1.8 + mcp 2.3 (at 81b3170, 63 tests
failed on fastmcp 2.14.7; at ae8f0bd all passed). The background race
did not reproduce in a 0-40 tick sweep on 3.2 or 4.1.8, so it has no
test; the close-on-exit makes it impossible by construction.

* fix(mcp): redact secrets in captured exception frame source lines

The MCP sanitizer scrubbed each exception's value but not its frames'
pre_context, context_line and post_context. A token written in the
source around a raise reached $exception verbatim. This was already true
on the official SDK path; #1223's fix made it reachable on standalone
FastMCP too, where frames were never sent before. Those three keys now
pass through the same sanitize_captured_value as the value.

Tested: test_secrets_in_frame_source_lines_are_redacted fails without
the sanitizer change and passes with it; full MCP suites pass on fastmcp
2.14.7, 3.2 and 4.1.8.

* fix(mcp): don't let a lingering background task keep a tool's traceback alive

PostHog Review on 34fbaeb: a background task started by a failing tool
copies the request's context, and with it the slot holding the raised
exception. Closing the slot left the exception attached, so its traceback
and every frame's locals lived as long as the task. The slot the context
can reach is now emptied on exit; the handler reads the outcome from a
separate object only it holds.

Tested: test_a_background_task_does_not_keep_the_failure_alive holds a
weakref to a tool local while the tool's background task is still
pending. It fails before and passes after on fastmcp 2.14.7, 3.2 and
4.1.8. The test turns logging off, because pytest's log capture keeps
fastmcp's error records and their exc_info would retain the traceback by
itself. Full MCP suites: 690 / 699 / 580 passed.

* fix(mcp): redact sensitive JSON members inside captured strings

Review on 52fc421 (PostHog Review, dustinbyrne): a failed FastMCP tool
whose source has `credentials = {"password": "hunter2"}` sent the
password in its frame's source context. The cause is wider than frames:
the MCP Payload privacy spec requires redacting, inside any captured
string, the value of a JSON member whose key is on the credential list,
and posthog-python never implemented it. posthog-go and posthog-rs do.
This ports their pattern into _redact_credentials, so it covers response
text, error messages and frame source alike.

Tested: posthog-go's TestSanitizeStringRedactsSensitiveJSONMembers table
ported as a parameterized test, plus a source-line case; 13 redaction
cases failed before, all 17 pass after. The frame test now also carries
a JSON credential next to a benign member that must survive. Full MCP
suites pass: 716 (fastmcp 3.2 + mcp 1.30), 597 (fastmcp 4.1.8 + mcp 2.3),
707 (fastmcp 2.14.7 + mcp 1.30). Ruff and mypy clean.

* fix(mcp): keep JSON-member redaction linear on malformed arrays

Review on d8a7a55 (PostHog Review, veria-ai): the array part of the
JSON-member pattern made every comma optional. Go and Rust run that
pattern on linear-time engines; Python's re backtracks, so a caller's
`{"note":[` + 40 digits + `!` split the digits into array items in
exponentially many ways and blocked the event loop. Every item after the
first now needs its comma, with an optional trailing one so an array cut
off mid-list still matches, which leaves each input one parse.

Tested: test_json_member_redaction_runs_in_linear_time_on_malformed_arrays
timed out at 5 s on two of three inputs before and passes after. Ten
adversarial 100k-character shapes (open arrays, lone quotes, unclosed
strings, escapes, number characters) each finish under 10 ms. All
JSON-member cases still pass. Full MCP suites: 719 / 600 / 710 passed on
fastmcp 3.2 / 4.1.8 / 2.14.7.

* fix(mcp): scan JSON members once instead of retrying the regex at every quote

Review on 9c1cb4f (PostHog Review, veria-ai): with no member to match,
the JSON-member regex restarted at every quote and rescanned the rest of
the text, so 16,000 escaped quote pairs took about 3 s and 64,000 took
49 s on the server's event loop. This is the second performance finding
in that regex: Go and Rust run it on linear-time engines, and Python's re
does not. It is replaced by a scanner that tries the same key starts but
caches where a string body ends and what follows each closing quote, so
overlapping candidates share their work.

Tested: the linear-time test gains escaped-quote and JSON-encoded-quote
inputs at 32,000 pairs, which timed out at 5 s before and pass after; ten
adversarial 100k-200k-character shapes each finish under 60 ms. A
differential test keeps the reference pattern as an oracle and checks the
scanner against it on 30,000 random texts (a one-off run of 200,000
longer ones found no mismatch). Full MCP suites: 722 / 603 / 713 passed
on fastmcp 3.2 / 4.1.8 / 2.14.7; ruff and mypy clean.
…ain (#1229)

* feat(mcp): set $mcp_error_status from the raised exception's cause chain

MCP Analytics groups failed calls by $mcp_error_status, but only PostHog's own server set
it, so every posthog.mcp failure landed in "unknown". Rootly asked for it in #1223.
Walk the cause chain inward (__cause__, else __context__ unless suppressed) and take the
first int 100-599 from status_code, statusCode or status on the exception or its response,
per the sdk-specs Failure capture rule. A caller's properties still win.

Tested: pytest posthog/test/mcp (720 passed, 11 skipped); new parameterized
test_error_status.py covers chain shapes and rejected values; FastMCP end-to-end row
raises httpx.HTTPStatusError(503) under mask_error_details=True.

* fix(mcp): follow __cause__ whenever set, and cap the status walk at 50 links

Pre-PR adversarial review found two gaps against the sdk-specs rule: a __cause__ left
with __suppress_context__ False was skipped for __context__, and a chain whose links are
built on read never ended. Walk the chain a traceback prints, capped at 50 links like
$exception_list.

Tested: pytest posthog/test/mcp (724 passed, 11 skipped); mypy baseline clean.

* fix(mcp): type the status walk's seen set, and record ERROR_STATUS in the API snapshot

CI on 68b8974 failed Mypy (var-annotated on `seen` in _exceptions.py) and
the Public API snapshot check (the new
PostHogMCPAnalyticsProperty.ERROR_STATUS constant), which PostHog Review
also flagged.

Tested: mypy | mypy-baseline filter clean; check_public_api.py up to date;
pytest posthog/test/mcp 725 / 606 / 716 passed on fastmcp 3.2 + mcp 1.30,
4.1.8 + mcp 2.3, 2.14.7 + mcp 1.30.
…sults (#1231)

* fix(mcp): report the raised exception on mcp 1.x low-level isError results

On mcp 1.x the low-level @server.call_tool() decorator catches a tool's raise
and returns CallToolResult(isError=True) before our handler sees it, so
$mcp_tool_call had $mcp_error_type "Error", no frames and no $mcp_error_status
(#1223). The decorator builds that result through server._make_error_result
inside its except block, so an instance wrapper reads the live exception there
and settles it into the same raised_tool_error slot standalone FastMCP uses.
Validation failures also pass through that method inside an except block; only
a message equal to str(exception) is the tool's own raise.

mcp 2.x low-level needs no change: raw handlers raise through to our wrapper,
and a scratch check showed RuntimeError, both frames and status 503.

Tested: parameterized test in test_lowlevel.py (plain and wrapped 503, plain
ValueError, input-validation guard). Red before on the three raising rows.
Removing the message guard turns the validation row into ValidationError.
pytest posthog/test/mcp: mcp 1.30 727 passed / 11 skipped; mcp 2.3 bare 551
passed / 55 skipped. CI-exact mypy: only the base branch's _exceptions.py:65.

* chore(mcp): scope the low-level raise recovery note and changeset to what it covers

Pre-PR review: the message guard admits any exception the decorator's catch-all
sees (tool body, list_tools cache refresh, result construction), not only the
tool's own raise, and _make_error_result first ships in mcp 1.10. Comment and
changeset now say so. Tested: pytest posthog/test/mcp on mcp 1.30, 727 passed.

* fix(mcp): tell validation results apart before rendering the exception

Review on #1231 (PostHog Review, dustinbyrne): the message guard called str(error)
on jsonschema ValidationErrors, which renders the whole instance and schema. On a
deeply nested payload that raised RecursionError and replaced the SDK's validation
response; on a large one it blocked the loop. A validation result is now matched by
the prefix plus the error's own short .message first, and any failure while
inspecting the exception leaves the original result untouched. A tool's own raise
worded like a validation error is still recovered.

Tested: parameterized test gains a 400-deep payload row (red before: response text
'maximum recursion depth exceeded') and a prefix-worded ValueError row.
pytest posthog/test/mcp: mcp 1.30 730 passed / 11 skipped; mcp 2.3 bare 551 passed
/ 55 skipped. CI-exact mypy clean.
* feat: send email content from code with posthog.emails.send

Squash of Silthus#2 at e5fdedf, re-signed for required signatures.

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

* feat: cut emails.send to the template and rendered code forms

Parks the local-template form (template_data on a code send, the
content hash, the upload by key and its retry) and the cc, bcc,
headers, tags and attachments fields, to match posthog-node. A code
send now takes rendered html and/or text only.

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

* test(emails): cover the async code-form send and migrated_from

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

* feat: send cc, bcc and attachments with emails.send

A code send takes cc and bcc as one address or a list, and attachments
as files with bytes or Base64 content. The client encodes bytes once,
before any retry. A file it can't send raises EmailError with the code
invalid_attachment and a field such as attachments__0__content, and
the error never repeats the file's content. A template_id send refuses
all three fields before any request.

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

* fix: refuse every unknown attachment key, a None key included

next(..., None) used None as its no-match value, so a None key hid the
unknown keys after it.

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

* test: cover a None attachment key before an unknown one

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

* docs: describe an older server's refusal without its error code

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

* fix(emails): leave reply_to out when the list is empty

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

* fix(emails): refuse an attachment string that is not base64

A text file passed as a str was sent unchecked as if it were Base64.
The client now raises invalid_attachment before any request and hints
to pass the file's bytes instead.

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

* chore(emails): name the refused-send test after what it guards

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

* chore(emails): drop a comment the name and test already state

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

* chore(emails): give cc, bcc and attachments their own changeset

The base PR owns emails-code-form.md, so this PR no longer edits it.

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

* fix(emails): refuse an attachment with an empty filename

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

* feat(emails): upload attachments by hash, then send their references

A code send hashes each file, asks PostHog which files it lacks, PUTs
those straight to PostHog's file storage with the returned headers and
without the secret key, then sends the email naming each file by its
SHA-256. The client checks 10 files and the 40 MB message estimate before
any request. A failed upload sends no email and surfaces the storage's
answer.

Part of PostHog/michaelr-context#81.

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

* fix(emails): send the storage upload exactly as signed

An injected client's defaults leaked into the presigned PUT: httpx
params replaced the URL's signature query, a requests session appended
its params, and default headers rode along. The upload is now a
standalone request with only the returned URL and headers, sent through
the same transport and retries.

Part of PostHog/michaelr-context#81.

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

* fix(emails): keep the environment's TLS settings and the signed URL private on uploads

The standalone storage PUT skipped requests' environment merge, so a
custom CA bundle or proxy applied to sends but not to uploads. It now
merges the environment's transport settings, without session headers or
auth. A storage network error is re-raised without its cause, so the
presigned URL never reaches a traceback.

Part of PostHog/michaelr-context#81.

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

* fix(emails): upload shared bytes once and retry a storage's passing faults

- The client counts attachments before it reads any file.
- Bytes that two attachments share go in the upload request once.
- The upload parser returns only the uploads still to make, each with
  its attachment index.
- A storage 500, 502 or 504 is retried like 503.
- A refused upload no longer takes the names of PostHog's API statuses.
- An unreachable storage raises network_error naming the attachment.

PostHog/michaelr-context#81

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

* fix(emails): retry a storage's 429 and name what the comments said

The previous commit stopped retrying a storage 429, which the general
rate-limit branch had retried. Storage 429 is retried again and named
rate_limit_exceeded. Names replace the comments that commit added.

PostHog/michaelr-context#81

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

* fix(emails): name attachments when PostHog has no upload route

A PostHog version without the upload route answered 404, and the send
raised a bare not_found with no field. The docs promise that an older
PostHog's refusal names the field. It now raises not_found with field
attachments, a reason and a fix, and sends nothing. Tests now cover the
names a storage fault gets once its retries run out.

PostHog/michaelr-context#81

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

* fix(emails): size attachments before copying them

The client copied and hashed every file before it checked the message
size. Ten views of one large buffer could exhaust memory before
message_too_large. Each file is now sized from its buffer or its Base64
length, the estimate is checked, and only then are files read and
hashed.

PostHog/michaelr-context#81

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

* fix(emails): size Base64 content by its data, not its padding

Extra trailing padding decodes, so it must not shrink the estimate.

Ticket: PostHog/michaelr-context#81

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

* test(emails): pin each body's count in the size estimate

The server counts html and text each like a file: UTF-8 bytes, Base64 with
line breaks. Pin both bodies, their UTF-8 length and per-body rounding.

Ticket: PostHog/michaelr-context#81

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

* fix(emails): report PostHog's own answer to an upload-route 404

A 404 can mean a PostHog without the upload route or sending from code
off for the project, so keep the server's reason on attachments instead
of naming one cause.

Ticket: PostHog/michaelr-context#81

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

* fix(emails): say which Base64 the SDK takes and where a refused upload goes wrong

A wrapped, unpadded or URL-safe string now learns it needs one padded line
from base64.b64encode. A storage refusal points at a fresh send or a proxy,
not at bytes the caller can't change.

Ticket: PostHog/michaelr-context#81

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

* fix(emails): count the subject in the message estimate like the server

The server now adds 75 bytes per 42 UTF-8 bytes of subject, so a long
subject beside a near-limit file must be refused before the upload.

Ticket: PostHog/michaelr-context#81

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

* test(emails): check uploads against their signed checksum like real storage

The fake storage now refuses a PUT whose bytes don't match the signed
checksum and length, and each upload slot is signed for its own file.
A new test asserts every file PostHog lacks lands in storage under its
own hash.

Refs PostHog/michaelr-context#81

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

* fix(emails): size padded Base64 without a copy and never follow a storage redirect

Sizing a Base64 string no longer strips its padding into a new string,
so refusing an oversized file allocates nothing the size of the file.
The async upload passes follow_redirects=False, so an injected client
that follows redirects can't turn a storage 302 into a GET that looks
like a stored file. The subject's share of the size limit is now pinned
through send instead of through a private helper.

Refs PostHog/michaelr-context#81

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

* test(emails): prove the async upload refuses a storage redirect over real sockets

The redirect test now runs against a local HTTP server that answers the
upload route, redirects the PUT and serves a login page, instead of a
scripted transport. It asserts upload_failed and that the login page and
the send route were never reached.

Refs PostHog/michaelr-context#81

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

* test(emails): test attachments against fakes that behave like PostHog and S3

The attachment tests now run against one in-memory PostHog with its file
storage. The upload and send routes need the secret key and refuse
fields outside their schema. Storage refuses a presigned PUT the way S3
does: an added query parameter, an Authorization header, an expired URL
or bytes that don't match the signed checksum. A send delivers an email
with the stored bytes and replays a repeated idempotency key. Tests
assert the delivered email, the bucket and the EmailError, not request
shapes. The CA test runs a real TLS handshake against a local storage
server with a certificate made in the test.

Refs PostHog/michaelr-context#81

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

* fix(emails): count the text PostHog derives and upload only over https

Without text, PostHog derives the plain text from the HTML and counts it
in the message size, so an HTML-only email near the limit uploaded its
files and was then refused. The client now counts the HTML twice, since
the derived text is never longer.

A plain HTTP upload URL from misconfigured storage would carry the file
and its signed URL unencrypted. The client now stops before any upload
with api_error naming the attachment, unless the client's host is plain
HTTP too, as in local development.

The fake PostHog now checks each route's method and records the content
type storage keeps. Tests assert stored files, not storage keys, and the
tests that only exercised the fake are gone.

Refs PostHog/michaelr-context#81

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

* fix(emails): name the caller's attachment in upload refusals and read storage answers whole

An upload-route refusal named the wire field (files, files__1__size),
which the caller never passed. It now names attachments, or the first
attachment holding that file. A session with stream=True left a storage
answer unread until after the upload, so a cut-off answer lost its
attachment; the storage request now always reads the answer at once.

The cc and bcc tests now assert the recipients of the delivered email,
and the template-content tests are back to main's shape.

Refs PostHog/michaelr-context#81

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

* test(emails): serve the test's TLS storage at TLS 1.2 or later

CodeQL flags the local HTTPS storage server's SSLContext for allowing
TLS 1.0 and 1.1. Pin the minimum version.

PostHog/michaelr-context#81

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

* test(emails): assert outcomes, not request traces, in the real-socket upload tests

The redirect and cut-short storage tests compared the server's request
list, which held the storage object path. They now assert the caller's
error, that no email was sent, and that the redirect was never followed.

PostHog/michaelr-context#81

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

* test(emails): run the field refusal and size tests against the PostHog fake

The cc refusal test now sends to a PostHog that doesn't take cc and
asserts the caller's error and that nothing was delivered, instead of
counting requests on a canned transport. The two oversized-email tests
assert that nothing was stored or delivered.

PostHog/michaelr-context#81

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

* test(emails): prove an upload never follows storage to another server

A client that follows a 307 or 308 on the PUT sends the file again to
the new URL, even over plain HTTP, and reports success. Both upload paths
already refuse redirects. New sync and async tests have the storage
answer 307 to a plain-HTTP server and assert upload_failed on the
attachment without either URL, that the other server received nothing,
and that no email was sent. Without the fix, requests and an httpx client
with follow_redirects deliver the file there. The docs and changeset say
the client never follows a redirect.

PostHog/michaelr-context#81

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

* test(emails): name the lost-reply retry test for what it proves

The test shows a send retried after a lost reply replays with its
idempotency key. It can't observe whether uploads are asked for again,
so the name no longer promises that.

PostHog/michaelr-context#81

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

* fix(emails): refuse extra Base64 padding before copying the string

The size check walked back over any number of trailing "=", so a short
Base64 string followed by tens of millions of "=" passed as a few bytes,
and b64decode then copied and scanned all of it, blocking the async
client's event loop. Standard Base64 has at most two "=", so a third
now raises invalid_attachment with the Base64 advice during that walk,
before any copy. The extra-padding size case is refused this way now.

PostHog/michaelr-context#81

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

* feat(emails): name a failed attachment upload attachment_upload_failed

Go, JS, Ruby and PHP raise attachment_upload_failed for a refused or
redirected upload, and the docs list one code for every SDK. The code
was upload_failed only in this unreleased change, so it takes the shared
name in the error, the tests, the docstrings and the changeset.

PostHog/michaelr-context#81

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

* test(emails): follow a storage 307 to another server over real sockets

The two cross-server redirect tests now run against a local PostHog whose
storage answers the upload with a 307 to a second local server. They assert
the second server receives nothing and no email is sent, instead of
injecting the redirect through mocked transports.

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

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* chore: remove mutation testing setup

* refactor: reduce SDK complexity and correct CRAP scoring

* test: isolate MCP cancellation from optional version helpers
* fix: treat failed feature flag records as unevaluated

* test: expect no result for failed flag attribution

* test: update released comparisons for failed flag records

* fix: use qualified PyPI package in flag changeset

* fix(flags): preserve unresolved reasons for failed fallback

* fix(flags): clarify failed-record classification and regression coverage

* fix(flags): preserve failed snapshot diagnostics

* ci: upgrade SDK compliance harness to 1.31.1
…ges (#1251)

* docs: publish the posthog-python 8.x reference and migration guide pages

Adds references/posthog-python-8.md and references/posthog-python-migration.md as
the 8.0.0 release will write them, so the Strapi sync puts them on posthog.com
before 8.0.0 ships. Both come from running bin/docs on docs/migration-guide-structure
(#1245) with posthog/version.py at 8.0.0, the version the release sets before it
generates. The generator on main writes only the 7.x page and the reference page,
and skips the migration page without docs/migration-*.md, so neither file is
rewritten before the release.

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

* docs: regenerate the migration page from the guide's latest fix

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

* docs: regenerate the migration page without the 404 hint and self-hosted lines

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

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Bumps [opentelemetry-api](https://github.com/open-telemetry/opentelemetry-python) from 1.45.1 to 1.45.2.
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

---
updated-dependencies:
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
)

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-llamaindex 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).


Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

---
updated-dependencies:
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  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>
…1255)

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-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-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-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-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 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

Updates `opentelemetry-api` from 1.45.1 to 1.45.2
- [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-python@v1.45.1...v1.45.2)

---
updated-dependencies:
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  dependency-type: indirect
  dependency-group: uv
- dependency-name: opentelemetry-api
  dependency-version: 1.45.2
  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>
…ry output block (#1256)

The Claude Code CLI sends each content block of a response as its own
AssistantMessage, and each tool result as its own UserMessage after
the generation's message_stop. The processor and PostHogClaudeSDKClient
replaced the pending output on every AssistantMessage and promoted
queued input at message_stop, so a generation kept only its last output
block and the generation after a tool call had no $ai_input.

Append output blocks per generation, append tool results to the queued
input, and promote queued input when the next main-agent generation
starts. A generation with nothing queued still has no input, as before.
A request that is restarted before its message_stop drops its partial
output. Without stream events, the result fallback reports every block
of the final assistant turn. Messages a subagent sends
(parent_tool_use_id set) stay out of the main generations' content.
)

fix(ai): record the status of an Anthropic stream error event instead of 200

Anthropic sends errors that happen after a stream starts, such as
overloaded_error, as an SSE error event, and the anthropic SDK raises them
with the stream response's 200 status. Map the error type to its documented
HTTP status (0 when unknown) so these failures are not recorded as a 200.
#1249)

messages.stream() wrapped the manager's request without a try/except, so a failed initial request (401, connection error) raised without any $ai_generation error event, unlike create(stream=True). Capture the error in both sync and async clients and re-raise.
Bumps the github-actions group with 5 updates:

| Package | From | To |
| --- | --- | --- |
| [actions/setup-python](https://github.com/actions/setup-python) | `7.0.0` | `7.1.0` |
| [astral-sh/setup-uv](https://github.com/astral-sh/setup-uv) | `10.3.0` | `10.3.3` |
| [github/codeql-action/init](https://github.com/github/codeql-action) | `4.39.2` | `4.39.3` |
| [github/codeql-action/analyze](https://github.com/github/codeql-action) | `4.39.2` | `4.39.3` |
| [peter-evans/create-pull-request](https://github.com/peter-evans/create-pull-request) | `8.1.1` | `8.1.2` |


Updates `actions/setup-python` from 7.0.0 to 7.1.0
- [Release notes](https://github.com/actions/setup-python/releases)
- [Commits](actions/setup-python@v7...6d4fd22)

Updates `astral-sh/setup-uv` from 10.3.0 to 10.3.3
- [Release notes](https://github.com/astral-sh/setup-uv/releases)
- [Commits](astral-sh/setup-uv@v10.3.0...68559e9)

Updates `github/codeql-action/init` from 4.39.2 to 4.39.3
- [Release notes](https://github.com/github/codeql-action/releases)
- [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md)
- [Commits](github/codeql-action@1fc58d0...33d7b85)

Updates `github/codeql-action/analyze` from 4.39.2 to 4.39.3
- [Release notes](https://github.com/github/codeql-action/releases)
- [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md)
- [Commits](github/codeql-action@1fc58d0...33d7b85)

Updates `peter-evans/create-pull-request` from 8.1.1 to 8.1.2
- [Release notes](https://github.com/peter-evans/create-pull-request/releases)
- [Commits](peter-evans/create-pull-request@5f6978f...15733db)

---
updated-dependencies:
- dependency-name: actions/setup-python
  dependency-version: 7.1.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: github-actions
- dependency-name: astral-sh/setup-uv
  dependency-version: 10.3.3
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: github-actions
- dependency-name: github/codeql-action/init
  dependency-version: 4.39.3
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: github-actions
- dependency-name: github/codeql-action/analyze
  dependency-version: 4.39.3
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: github-actions
- dependency-name: peter-evans/create-pull-request
  dependency-version: 8.1.2
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: github-actions
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
dependabot Bot and others added 5 commits November 16, 2026 12:07
Bumps the ai-providers group with 9 updates:

| Package | From | To |
| --- | --- | --- |
| [langchain](https://github.com/langchain-ai/langchain) | `1.4.7` | `1.4.9` |
| [anthropic](https://github.com/anthropics/anthropic-sdk-python) | `1.13.0` | `2.6.0` |
| [langgraph](https://github.com/langchain-ai/langgraph) | `1.2.16` | `1.2.20` |
| [langchain-core](https://github.com/langchain-ai/langchain) | `1.6.13` | `1.7.6` |
| [langchain-openai](https://github.com/langchain-ai/langchain) | `1.7.2` | `1.7.4` |
| [langchain-anthropic](https://github.com/langchain-ai/langchain) | `1.7.6` | `1.7.10` |
| [google-genai](https://github.com/googleapis/python-genai) | `2.32.0` | `2.33.0` |
| [langchain-protocol](https://github.com/langchain-ai/agent-protocol) | `0.0.20` | `0.0.21` |
| [openai](https://github.com/openai/openai-python) | `2.54.0` | `3.46.0` |


Updates `langchain` from 1.4.7 to 1.4.9
- [Release notes](https://github.com/langchain-ai/langchain/releases)
- [Commits](langchain-ai/langchain@langchain==1.4.7...langchain==1.4.9)

Updates `anthropic` from 1.13.0 to 2.6.0
- [Release notes](https://github.com/anthropics/anthropic-sdk-python/releases)
- [Changelog](https://github.com/anthropics/anthropic-sdk-python/blob/main/CHANGELOG.md)
- [Commits](anthropics/anthropic-sdk-python@v1.13.0...v2.6.0)

Updates `langgraph` from 1.2.16 to 1.2.20
- [Release notes](https://github.com/langchain-ai/langgraph/releases)
- [Commits](langchain-ai/langgraph@1.2.16...1.2.20)

Updates `langchain-core` from 1.6.13 to 1.7.6
- [Release notes](https://github.com/langchain-ai/langchain/releases)
- [Commits](langchain-ai/langchain@langchain-core==1.6.13...langchain-core==1.7.6)

Updates `langchain-openai` from 1.7.2 to 1.7.4
- [Release notes](https://github.com/langchain-ai/langchain/releases)
- [Commits](langchain-ai/langchain@langchain-openai==1.7.2...langchain-openai==1.7.4)

Updates `langchain-anthropic` from 1.7.6 to 1.7.10
- [Release notes](https://github.com/langchain-ai/langchain/releases)
- [Commits](langchain-ai/langchain@langchain-anthropic==1.7.6...langchain-anthropic==1.7.10)

Updates `google-genai` from 2.32.0 to 2.33.0
- [Release notes](https://github.com/googleapis/python-genai/releases)
- [Changelog](https://github.com/googleapis/python-genai/blob/main/CHANGELOG.md)
- [Commits](googleapis/python-genai@v2.32.0...v2.33.0)

Updates `langchain-protocol` from 0.0.20 to 0.0.21
- [Release notes](https://github.com/langchain-ai/agent-protocol/releases)
- [Commits](langchain-ai/agent-protocol@langchain-protocol==0.0.20...langchain-protocol==0.0.21)

Updates `openai` from 2.54.0 to 3.46.0
- [Release notes](https://github.com/openai/openai-python/releases)
- [Changelog](https://github.com/openai/openai-python/blob/main/CHANGELOG.md)
- [Commits](openai/openai-python@v2.54.0...v3.46.0)

---
updated-dependencies:
- dependency-name: langchain
  dependency-version: 1.4.9
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: anthropic
  dependency-version: 2.6.0
  dependency-type: direct:production
  update-type: version-update:semver-major
  dependency-group: ai-providers
- dependency-name: langgraph
  dependency-version: 1.2.20
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: langchain-core
  dependency-version: 1.7.6
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: ai-providers
- dependency-name: langchain-openai
  dependency-version: 1.7.4
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: langchain-anthropic
  dependency-version: 1.7.10
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: google-genai
  dependency-version: 2.33.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: ai-providers
- dependency-name: langchain-protocol
  dependency-version: 0.0.21
  dependency-type: indirect
  update-type: version-update:semver-patch
  dependency-group: ai-providers
- dependency-name: openai
  dependency-version: 3.46.0
  dependency-type: indirect
  update-type: version-update:semver-major
  dependency-group: ai-providers
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* docs: add docstrings to ExceptionCapture public methods

Add class-level docstring and per-method docstrings to ExceptionCapture,
covering __init__, close, exception_handler, thread_exception_handler,
exception_receiver, and capture_exception.

No behaviour changes — documentation only.

Signed-off-by: mi-vault <asmisharma2026@gmail.com>

* docs: clarify bucket_size capacity and exception_receiver caller scope

Signed-off-by: mi-vault <asmisharma2026@gmail.com>

---------

Signed-off-by: mi-vault <asmisharma2026@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 self-assigned this Nov 16, 2026
@eli-r-ph
eli-r-ph marked this pull request as ready for review November 16, 2026 18:39
@eli-r-ph
eli-r-ph requested a review from a team as a code owner November 16, 2026 18:39
@veria-ai

veria-ai Bot commented Nov 16, 2026 •

Copy link
Copy Markdown

PR overview

No security issues were found in this pull request.

Reviews

@github-actions

Copy link
Copy Markdown
Contributor

SDK performance: posthog-python

22 changes worth reviewing: 4 observed improvements, 18 observed regressions.

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.

Test Measurement Before PR With PR Change
Evaluate a flag (no match) Allocated memory per operation 52.16 KiB/op 59.22 KiB/op 13.5% more memory
Evaluate a flag (late match) Allocated memory per operation 53.61 KiB/op 60.59 KiB/op 13% more memory
Evaluate all 1,000 flags Allocated memory per operation 19.24 MiB/op 21.63 MiB/op 12.4% more memory
Evaluate all 100 flags Allocated memory per operation 1.93 MiB/op 2.17 MiB/op 12.4% more memory
Evaluate all 10 flags Allocated memory per operation 200.98 KiB/op 225.43 KiB/op 12.2% more memory
Evaluate a cohort-filtered flag Allocations per operation 270.27 count/op 239.28 count/op 11.5% fewer
Evaluate a flag (early match) Allocated memory per operation 35.73 KiB/op 39.81 KiB/op 11.4% more memory
Evaluate a regex-filtered flag Allocated memory per operation 19.24 KiB/op 20.66 KiB/op 7.4% more memory
Evaluate all 1,000 flags Time per operation 7.24 ms/op 7.77 ms/op 7.4% more time
Evaluate all 1,000 flags Process CPU time per operation 7.24 ms/op 7.77 ms/op 7.4% more time
Evaluate all 10 flags Process CPU time per operation 77.18 µs/op 82.8 µs/op 7.3% more time
Evaluate all 10 flags Time per operation 77.19 µs/op 82.8 µs/op 7.3% more time

Showing the 12 largest flagged changes; 10 more are in all results.

View all results
Test Measurement Before PR With PR Change Assessment
Capture and deliver events (32-byte payload, burst)
Shard: capture
Managed heap after settling — — — unavailable; 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.​
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Managed heap after settling — — — unavailable; 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.​
Capture and deliver events (32-byte payload, steady)
Shard: capture
Managed heap after settling — — — unavailable; 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.​
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Managed heap after settling — — — unavailable; 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.​
Queue an event (32-byte payload)
Shard: capture
Allocated memory per operation — — — unavailable; base:​ unsupported; head:​ unsupported; base:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​; head:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​
Queue an event (32-byte payload)
Shard: capture
Allocations per operation — — — unavailable; base:​ unsupported; head:​ unsupported; base:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​; head:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​
Queue an event (4,096-byte payload)
Shard: capture
Allocated memory per operation — — — unavailable; base:​ unsupported; head:​ unsupported; base:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​; head:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​
Queue an event (4,096-byte payload)
Shard: capture
Allocations per operation — — — unavailable; base:​ unsupported; head:​ unsupported; base:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​; head:​ Background preparation has no public completion boundary before flush; see capture.​drain for full-pipeline allocations.​
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Managed heap after settling — — — unavailable; 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
Managed heap after settling — — — unavailable; 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.​
Capture and deliver events (32-byte payload, burst)
Shard: capture
Allocated memory per operation 133.68 KiB/op 133.69 KiB/op <0.1% more memory Not flagged
Capture and deliver events (32-byte payload, burst)
Shard: capture
Allocations per operation 1,345.51 count/op 1,345.52 count/op <0.1% more Not flagged
Capture and deliver events (32-byte payload, burst)
Shard: capture
Time per operation 76.96 µs/op 76.95 µs/op <0.1% less time Not flagged
Capture and deliver events (32-byte payload, burst)
Shard: capture
Idle process memory (RSS) 37.45 MiB 37.63 MiB 0.5% more memory Not flagged
Capture and deliver events (32-byte payload, burst)
Shard: capture
Observed peak process memory (RSS) 39.88 MiB 40.08 MiB 0.5% more memory Not flagged
Capture and deliver events (32-byte payload, burst)
Shard: capture
Process memory after settling (RSS) 39.88 MiB 40.08 MiB 0.5% more memory Not flagged
Capture and deliver events (32-byte payload, burst)
Shard: capture
Process CPU time per operation 81.07 µs/op 80.01 µs/op 1.3% less time Not flagged
Capture and deliver events (32-byte payload, burst)
Shard: capture
Lifetime peak process memory (RSS) 40.05 MiB 40.18 MiB 0.3% more memory Not flagged
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Allocated memory per operation 185.76 KiB/op 185.76 KiB/op No observed change Not flagged
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Allocations per operation 1,337.51 count/op 1,337.51 count/op No observed change Not flagged
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Time per operation 100.22 µs/op 100.22 µs/op <0.1% more time Not flagged
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Idle process memory (RSS) 41.11 MiB 41.25 MiB 0.4% more memory Not flagged
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Observed peak process memory (RSS) 46.87 MiB 47.04 MiB 0.4% more memory Not flagged
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Process memory after settling (RSS) 46.87 MiB 47.04 MiB 0.4% more memory Not flagged
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Process CPU time per operation 97.64 µs/op 98.1 µs/op 0.5% more time Not flagged
Capture and deliver events (4,096-byte payload, burst)
Shard: capture
Lifetime peak process memory (RSS) 51.29 MiB 51.39 MiB 0.2% more memory Not flagged
Capture and deliver events (32-byte payload, steady)
Shard: capture
Allocated memory per operation 138.74 KiB/op 138.74 KiB/op <0.1% more memory Not flagged
Capture and deliver events (32-byte payload, steady)
Shard: capture
Allocations per operation 1,390.29 count/op 1,390.29 count/op <0.1% more Not flagged
Capture and deliver events (32-byte payload, steady)
Shard: capture
Time per operation 88.73 µs/op 88.97 µs/op 0.3% more time Not flagged
Capture and deliver events (32-byte payload, steady)
Shard: capture
Idle process memory (RSS) 37.42 MiB 37.52 MiB 0.3% more memory Not flagged
Capture and deliver events (32-byte payload, steady)
Shard: capture
Observed peak process memory (RSS) 37.59 MiB 37.67 MiB 0.2% more memory Not flagged
Capture and deliver events (32-byte payload, steady)
Shard: capture
Process memory after settling (RSS) 37.59 MiB 37.67 MiB 0.2% more memory Not flagged
Capture and deliver events (32-byte payload, steady)
Shard: capture
Process CPU time per operation 91.08 µs/op 90.94 µs/op 0.2% less time Not flagged
Capture and deliver events (32-byte payload, steady)
Shard: capture
Lifetime peak process memory (RSS) 38.98 MiB 38.98 MiB No observed change Not flagged
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Allocated memory per operation 190.92 KiB/op 190.92 KiB/op No observed change Not flagged
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Allocations per operation 1,382.19 count/op 1,382.19 count/op No observed change Not flagged
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Time per operation 105.68 µs/op 105.41 µs/op 0.3% less time Not flagged
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Idle process memory (RSS) 40.93 MiB 41.06 MiB 0.3% more memory Not flagged
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Observed peak process memory (RSS) 41.39 MiB 41.51 MiB 0.3% more memory Not flagged
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Process memory after settling (RSS) 40.98 MiB 41.11 MiB 0.3% more memory Not flagged
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Process CPU time per operation 104 µs/op 102.19 µs/op 1.7% less time Not flagged
Capture and deliver events (4,096-byte payload, steady)
Shard: capture
Lifetime peak process memory (RSS) 41.27 MiB 41.44 MiB 0.4% more memory Not flagged
Queue an event (32-byte payload)
Shard: capture
Caller-thread CPU time per operation 57.39 µs/op 57.7 µs/op 0.5% more time Not flagged
Queue an event (32-byte payload)
Shard: capture
Time per operation 57.46 µs/op 57.71 µs/op 0.4% more time Not flagged
Queue an event (4,096-byte payload)
Shard: capture
Caller-thread CPU time per operation 57.14 µs/op 57.59 µs/op 0.8% more time Not flagged
Queue an event (4,096-byte payload)
Shard: capture
Time per operation 57.14 µs/op 57.59 µs/op 0.8% more time Not flagged
Evaluate all 10 flags
Shard: flags-scaling
Allocated memory per operation 200.98 KiB/op 225.43 KiB/op 12.2% more memory Worth reviewing: observed regression
Evaluate all 10 flags
Shard: flags-scaling
Allocations per operation 2,339.33 count/op 2,279.42 count/op 2.6% fewer Not flagged
Evaluate all 10 flags
Shard: flags-scaling
Time per operation 77.19 µs/op 82.8 µs/op 7.3% more time Worth reviewing: observed regression
Evaluate all 10 flags
Shard: flags-scaling
Process CPU time per operation 77.18 µs/op 82.8 µs/op 7.3% more time Worth reviewing: observed regression
Evaluate all 100 flags
Shard: flags-scaling
Allocated memory per operation 1.93 MiB/op 2.17 MiB/op 12.4% more memory Worth reviewing: observed regression
Evaluate all 100 flags
Shard: flags-scaling
Allocations per operation 23,013.56 count/op 22,413.69 count/op 2.6% fewer Not flagged
Evaluate all 100 flags
Shard: flags-scaling
Time per operation 730.91 µs/op 776.3 µs/op 6.2% more time Worth reviewing: observed regression
Evaluate all 100 flags
Shard: flags-scaling
Process CPU time per operation 730.9 µs/op 776.33 µs/op 6.2% more time Worth reviewing: observed regression
Evaluate all 1,000 flags
Shard: flags-scaling
Allocated memory per operation 19.24 MiB/op 21.63 MiB/op 12.4% more memory Worth reviewing: observed regression
Evaluate all 1,000 flags
Shard: flags-scaling
Allocations per operation 229,654.94 count/op 223,655.44 count/op 2.6% fewer Not flagged
Evaluate all 1,000 flags
Shard: flags-scaling
Time per operation 7.24 ms/op 7.77 ms/op 7.4% more time Worth reviewing: observed regression
Evaluate all 1,000 flags
Shard: flags-scaling
Process CPU time per operation 7.24 ms/op 7.77 ms/op 7.4% more time Worth reviewing: observed regression
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Idle process memory (RSS) 43.03 MiB 43.93 MiB 2.1% more memory Not flagged
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Observed peak process memory (RSS) 43.33 MiB 44.27 MiB 2.2% more memory Not flagged
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Process memory after settling (RSS) 43.33 MiB 44.27 MiB 2.2% more memory Not flagged
Evaluate 1,000 flags under concurrent load
Shard: flags-memory
Lifetime peak process memory (RSS) 44.27 MiB 44.02 MiB 0.6% less memory Not flagged
Refresh 1,000 flag definitions
Shard: flags-memory
Idle process memory (RSS) 43.09 MiB 43.13 MiB 0.1% more memory Not flagged
Refresh 1,000 flag definitions
Shard: flags-memory
Observed peak process memory (RSS) 43.1 MiB 43.14 MiB 0.1% more memory Not flagged
Refresh 1,000 flag definitions
Shard: flags-memory
Process memory after settling (RSS) 43.1 MiB 43.14 MiB 0.1% more memory Not flagged
Refresh 1,000 flag definitions
Shard: flags-memory
Lifetime peak process memory (RSS) 43.02 MiB 43.47 MiB 1% more memory Not flagged
Evaluate one boolean flag
Shard: flags-single
Allocated memory per operation 15.06 KiB/op 15.74 KiB/op 4.5% more memory Not flagged
Evaluate one boolean flag
Shard: flags-single
Allocations per operation 173.76 count/op 162.77 count/op 6.3% fewer Worth reviewing: observed improvement
Evaluate one boolean flag
Shard: flags-single
Time per operation 13.7 µs/op 14.33 µs/op 4.6% more time Not flagged
Evaluate one boolean flag
Shard: flags-single
Process CPU time per operation 13.7 µs/op 14.32 µs/op 4.6% more time Not flagged
Evaluate a cohort-filtered flag
Shard: flags-single
Allocated memory per operation 23.14 KiB/op 24.6 KiB/op 6.3% more memory Worth reviewing: observed regression
Evaluate a cohort-filtered flag
Shard: flags-single
Allocations per operation 270.27 count/op 239.28 count/op 11.5% fewer Worth reviewing: observed improvement
Evaluate a cohort-filtered flag
Shard: flags-single
Time per operation 21.65 µs/op 22.98 µs/op 6.2% more time Worth reviewing: observed regression
Evaluate a cohort-filtered flag
Shard: flags-single
Process CPU time per operation 21.65 µs/op 22.98 µs/op 6.2% more time Worth reviewing: observed regression
Evaluate a flag (early match)
Shard: flags-single
Allocated memory per operation 35.73 KiB/op 39.81 KiB/op 11.4% more memory Worth reviewing: observed regression
Evaluate a flag (early match)
Shard: flags-single
Allocations per operation 387.27 count/op 388.27 count/op 0.3% more Not flagged
Evaluate a flag (early match)
Shard: flags-single
Time per operation 27.53 µs/op 28.72 µs/op 4.3% more time Not flagged
Evaluate a flag (early match)
Shard: flags-single
Process CPU time per operation 27.53 µs/op 28.72 µs/op 4.3% more time Not flagged
Evaluate a flag (late match)
Shard: flags-single
Allocated memory per operation 53.61 KiB/op 60.59 KiB/op 13% more memory Worth reviewing: observed regression
Evaluate a flag (late match)
Shard: flags-single
Allocations per operation 576.27 count/op 584.27 count/op 1.4% more Not flagged
Evaluate a flag (late match)
Shard: flags-single
Time per operation 38.27 µs/op 39.86 µs/op 4.2% more time Not flagged
Evaluate a flag (late match)
Shard: flags-single
Process CPU time per operation 38.26 µs/op 39.85 µs/op 4.2% more time Not flagged
Evaluate a flag (no match)
Shard: flags-single
Allocated memory per operation 52.16 KiB/op 59.22 KiB/op 13.5% more memory Worth reviewing: observed regression
Evaluate a flag (no match)
Shard: flags-single
Allocations per operation 556.77 count/op 567.77 count/op 2% more Not flagged
Evaluate a flag (no match)
Shard: flags-single
Time per operation 34.84 µs/op 36.16 µs/op 3.8% more time Not flagged
Evaluate a flag (no match)
Shard: flags-single
Process CPU time per operation 34.84 µs/op 36.16 µs/op 3.8% more time Not flagged
Evaluate a property-filtered flag
Shard: flags-single
Allocated memory per operation 17.85 KiB/op 19.04 KiB/op 6.7% more memory Worth reviewing: observed regression
Evaluate a property-filtered flag
Shard: flags-single
Allocations per operation 198.27 count/op 192.27 count/op 3% fewer Not flagged
Evaluate a property-filtered flag
Shard: flags-single
Time per operation 16.58 µs/op 17.03 µs/op 2.7% more time Not flagged
Evaluate a property-filtered flag
Shard: flags-single
Process CPU time per operation 16.58 µs/op 17.03 µs/op 2.7% more time Not flagged
Evaluate a regex-filtered flag
Shard: flags-single
Allocated memory per operation 19.24 KiB/op 20.66 KiB/op 7.4% more memory Worth reviewing: observed regression
Evaluate a regex-filtered flag
Shard: flags-single
Allocations per operation 210.77 count/op 197.77 count/op 6.2% fewer Worth reviewing: observed improvement
Evaluate a regex-filtered flag
Shard: flags-single
Time per operation 17.29 µs/op 17.94 µs/op 3.8% more time Not flagged
Evaluate a regex-filtered flag
Shard: flags-single
Process CPU time per operation 17.29 µs/op 17.93 µs/op 3.8% more time Not flagged
Evaluate a rollout flag
Shard: flags-single
Allocated memory per operation 14.05 KiB/op 14.89 KiB/op 6% more memory Worth reviewing: observed regression
Evaluate a rollout flag
Shard: flags-single
Allocations per operation 160.76 count/op 152.77 count/op 5% fewer Not flagged
Evaluate a rollout flag
Shard: flags-single
Time per operation 12.94 µs/op 13.58 µs/op 4.9% more time Not flagged
Evaluate a rollout flag
Shard: flags-single
Process CPU time per operation 12.94 µs/op 13.58 µs/op 4.9% more time Not flagged
Evaluate a multivariate flag
Shard: flags-single
Allocated memory per operation 15.22 KiB/op 15.9 KiB/op 4.5% more memory Not flagged
Evaluate a multivariate flag
Shard: flags-single
Allocations per operation 177.77 count/op 166.77 count/op 6.2% fewer Worth reviewing: observed improvement
Evaluate a multivariate flag
Shard: flags-single
Time per operation 14.4 µs/op 14.86 µs/op 3.2% more time Not flagged
Evaluate a multivariate flag
Shard: flags-single
Process CPU time per operation 14.4 µs/op 14.86 µs/op 3.2% more time Not flagged

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 3e804b9d481f1546a3542100be2e18d429e7c5e0 → 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.

@eli-r-ph eli-r-ph added do not merge Lands by fast-forward, never by the merge button and removed do not merge Lands by fast-forward, never by the merge button labels Nov 16, 2026
@eli-r-ph
eli-r-ph merged commit 018d15d into v1 Nov 16, 2026
51 checks passed
@eli-r-ph
eli-r-ph deleted the eli/v1-sync-main-20261116 branch November 16, 2026 20:33

@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.

Review of the merged snapshot 018d15de54954282867f1e3ea65313a282c96419: initial Anthropic stream-request failures discard explicit per-call capture options, allowing global processing settings to win. Both sync and async cases reproduced through the public wrappers and actual SDK transport. Consider forwarding those options in both error handlers and adding the regression in a follow-up.

AI-assisted review.

time.time() - start_time,
accumulator.content_blocks,
accumulator.accumulated_content,
error=exc,

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.

F1 — [P1] Preserve per-call options when the initial Anthropic stream request fails

IMPORTANT · REQUIRED · Introduced event path

Location: posthog/ai/anthropic/anthropic.py:171, with the same omission in anthropic_async.py:171.

When entering messages.stream() fails before iteration—for example, with an initial 401—the new error-capture path emits $ai_generation without the caller’s posthog_options. With an explicit identity and permissive global defaults, that changes requested cookieless/personless processing back to the global settings. The expected result is an error event retaining the same per-call controls as a successful stream.

A supported caller can trigger this with:

with provider.messages.stream(
    model=model,
    max_tokens=8,
    messages=[{"role": "user", "content": "Hello"}],
    posthog_distinct_id="user",
    posthog_options={
        "cookieless_mode": True,
        "process_person_profile": False,
    },
):
    ...

Both new _capture_streaming_event(..., error=exc) calls omit options=posthog_options; the adjacent successful paths forward it. Public stream() documentation and .sampo/changesets/ai-posthog-options.md promise that these per-call options win over context and super_options. The backend recognizes both controls in rust/capture/src/v1/analytics/types.rs:102–110 at the pinned backend revision.

Executed through both public sync/async wrappers and real SDK HTTP delivery:

Scenario Base Reviewed head
Initial synthetic 401 No event One error event uses process_person_profile: true, cookieless_mode: false from conflicting globals
Normal stream / mid-stream SSE error Requested options retained Requested options retained

The base’s missing error event is the intended fix—not evidence that the base captured correctly. The regression is that this newly emitted event bypasses existing per-call restrictions. Content privacy_mode itself is not shown to fail, and real backend person creation was not executed. Bedrock/Vertex reuse these wrapped classes by source inspection; execution covered direct Anthropic wrappers. The similar create(stream=True) omission predates this diff and is not counted separately.

Consider passing options=posthog_options in both new exception handlers, retaining the telemetry guard and original exception re-raise, and adding this regression:

from unittest.mock import AsyncMock, Mock
import pytest
from posthog.ai.anthropic import Anthropic, AsyncAnthropic

@pytest.mark.asyncio
@pytest.mark.parametrize("asynchronous", [False, True])
async def test_initial_request_error_preserves_options(asynchronous):
    ph = Mock(privacy_mode=False, enable_full_ai_capture=False)
    options = {"cookieless_mode": True, "process_person_profile": False}
    failure = RuntimeError("synthetic connect failure")
    wrapper = AsyncAnthropic if asynchronous else Anthropic
    client = wrapper(api_key="synthetic-key", posthog_client=ph)
    client.post = (
        AsyncMock(side_effect=failure)
        if asynchronous else Mock(side_effect=failure)
    )
    try:
        manager = client.messages.stream(
            model="claude-synthetic", max_tokens=1,
            messages=[{"role": "user", "content": "synthetic prompt"}],
            posthog_distinct_id="synthetic-user", posthog_options=options,
        )
        with pytest.raises(RuntimeError) as caught:
            if asynchronous:
                async with manager:
                    pass
            else:
                with manager:
                    pass
        assert caught.value is failure
        ph.capture_ai.assert_called_once()
        assert ph.capture_ai.call_args.kwargs.get("options", {}) == options
    finally:
        if asynchronous:
            await client.close()
        else:
            client.close()

Both cases were executed and fail on the last assertion: {} instead of the requested options. Original exception identity and exactly one capture pass first. The separate transport regression returns base exit 0/head exit 1 specifically for initial-error option precedence. Environment: CPython 3.11.13, pytest 8.4.1, Anthropic 0.120.0 and HTTPX 0.28.1. No fix was applied or tested.

@dustinbyrne
dustinbyrne requested a review from a team November 16, 2026 21:25
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.