fix: track OpenAI chat completions parse calls

[marandaneto]

💡 Motivation and Context

Fixes #614.

OpenAI chat.completions.parse calls were falling through to the raw OpenAI SDK object, so PostHog tracking was skipped and posthog_* arguments could be forwarded to OpenAI and crash. Unsupported fallback wrapper access now also logs a warning when tracking is bypassed.

💚 How did you test it?

• uv run ruff check posthog/ai/openai/openai.py posthog/ai/openai/openai_async.py posthog/ai/openai/wrapper_utils.py posthog/test/ai/openai/test_openai.py
• uv run --extra test pytest posthog/test/ai/openai/test_openai.py -q

📝 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

[marandaneto]

@PostHog/team-ai-observability pls double check this since i dont know much about those wrappers, i just let an agent figure it out but i think it does what's supposed to do

[greptile-apps]

Prompt To Fix All With AI

Fix the following 2 code review issues. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 2
posthog/test/ai/openai/test_openai.py:1196-1204
**Non-parameterized fallback warning test**

The change adds `warn_on_fallback` to eight wrapper classes (`WrappedResponses`, `WrappedChat`, `WrappedCompletions`, `WrappedEmbeddings`, `WrappedBeta`, `WrappedBetaChat`, `WrappedBetaCompletions` — in both sync and async variants), but the test only exercises `WrappedChat`. Per the team's preference for parameterised tests, this should cover a representative set of wrapper classes so a typo or missing call in any one of them is caught.

### Issue 2 of 2
posthog/ai/openai/wrapper_utils.py:4-5
**Module-level mutable set leaks across test runs**

`_fallback_warnings` is initialised once at import time and never cleared. `test_fallback_logs_warning` passes today because `"posthog_unwrapped_test_attribute"` is a unique key, but if the same test is re-executed in the same process (e.g. `pytest-repeat`, `pytest-xdist` with `--looponfail`, or a future test that happens to reuse the same combination), the set already contains the key, the `log.warning` call is skipped, and the caplog assertion fails. Exposing a way to reset the set (or keying it per-client instance) would make the deduplication safe to test.

_{Reviews (1): Last reviewed commit: "fix openai chat completions parse tracki..." | Re-trigger Greptile}

[github-actions]

posthog-python Compliance Report

Date: 2026-05-22 12:08:39 UTC
Duration: 4674917ms

⚠️ Some Tests Failed

29/45 tests passed, 16 failed

---

Capture Tests

✅ 29/29 tests passed

View Details

Test	Status	Duration
Format Validation.Event Has Required Fields	✅	518ms
Format Validation.Event Has Uuid	✅	1507ms
Format Validation.Event Has Lib Properties	✅	1507ms
Format Validation.Distinct Id Is String	✅	1508ms
Format Validation.Token Is Present	✅	1506ms
Format Validation.Custom Properties Preserved	✅	1507ms
Format Validation.Event Has Timestamp	✅	1507ms
Retry Behavior.Retries On 503	✅	9515ms
Retry Behavior.Does Not Retry On 400	✅	3510ms
Retry Behavior.Does Not Retry On 401	✅	3507ms
Retry Behavior.Respects Retry After Header	✅	9510ms
Retry Behavior.Implements Backoff	✅	23525ms
Retry Behavior.Retries On 500	✅	7509ms
Retry Behavior.Retries On 502	✅	7512ms
Retry Behavior.Retries On 504	✅	7512ms
Retry Behavior.Max Retries Respected	✅	23533ms
Deduplication.Generates Unique Uuids	✅	1497ms
Deduplication.Preserves Uuid On Retry	✅	7513ms
Deduplication.Preserves Uuid And Timestamp On Retry	✅	14515ms
Deduplication.Preserves Uuid And Timestamp On Batch Retry	✅	7512ms
Deduplication.No Duplicate Events In Batch	✅	1508ms
Deduplication.Different Events Have Different Uuids	✅	1507ms
Compression.Sends Gzip When Enabled	✅	1508ms
Batch Format.Uses Proper Batch Structure	✅	1507ms
Batch Format.Flush With No Events Sends Nothing	✅	1005ms
Batch Format.Multiple Events Batched Together	✅	1506ms
Error Handling.Does Not Retry On 403	✅	3509ms
Error Handling.Does Not Retry On 413	✅	3506ms
Error Handling.Retries On 408	✅	7511ms

Feature_Flags Tests

⚠️ 0/16 tests passed, 16 failed

View Details

Test	Status	Duration
Request Payload.Request With Person Properties Device Id	❌	509ms
Request Payload.Flags Request Uses V2 Query Param	❌	300872ms
Request Payload.Flags Request Hits Flags Path Not Decide	❌	301031ms
Request Payload.Flags Request Omits Authorization Header	❌	300945ms
Request Payload.Token In Flags Body Matches Init	❌	301024ms
Request Payload.Groups Round Trip	❌	300998ms
Request Payload.Groups Default To Empty Object	❌	301026ms
Request Payload.Person Properties Distinct Id Auto Populated When Caller Omits It	❌	300990ms
Request Payload.Disable Geoip False Propagates As Geoip Disable False	❌	300962ms
Request Payload.Disable Geoip Omitted Defaults To False	❌	301024ms
Request Payload.Flag Keys To Evaluate Contains Only Requested Key	❌	300992ms
Request Lifecycle.No Flags Request On Init Alone	❌	301019ms
Request Lifecycle.No Flags Request On Normal Capture	❌	301020ms
Request Lifecycle.Two Flag Calls Produce Two Remote Requests	❌	300997ms
Request Lifecycle.Mock Response Value Is Returned To Caller	❌	300926ms
Side Effect Events.Get Feature Flag Captures Feature Flag Called Event	❌	301062ms

Failures

request_payload.request_with_person_properties_device_id

Field 'token' not found in /flags request body at path 'token'. Available keys: ['distinct_id', 'groups', 'person_properties', 'group_properties', 'geoip_disable', 'device_id', 'flag_keys_to_evaluate', 'sentAt', 'api_key']

request_payload.flags_request_uses_v2_query_param

No error message

request_payload.flags_request_hits_flags_path_not_decide

No error message

request_payload.flags_request_omits_authorization_header

No error message

request_payload.token_in_flags_body_matches_init

No error message

request_payload.groups_round_trip

No error message

request_payload.groups_default_to_empty_object

No error message

request_payload.person_properties_distinct_id_auto_populated_when_caller_omits_it

No error message

request_payload.disable_geoip_false_propagates_as_geoip_disable_false

No error message

request_payload.disable_geoip_omitted_defaults_to_false

No error message

request_payload.flag_keys_to_evaluate_contains_only_requested_key

No error message

request_lifecycle.no_flags_request_on_init_alone

No error message

request_lifecycle.no_flags_request_on_normal_capture

No error message

request_lifecycle.two_flag_calls_produce_two_remote_requests

No error message

request_lifecycle.mock_response_value_is_returned_to_caller

No error message

side_effect_events.get_feature_flag_captures_feature_flag_called_event

No error message

[dustinbyrne]

change makes sense to me 👍