feat: dedicated AI endpoint routing#656
Merged
carlos-marchal-ph merged 3 commits intoJun 10, 2026
Merged
Conversation
Contributor
Prompt To Fix All With AIFix the following 3 code review issues. Work through them one at a time, proposing concise fixes.
---
### Issue 1 of 3
posthog/consumer.py:147-152
**AI events silently dropped when analytics send fails**
If `_send(analytics_events, EVENTS_ENDPOINT)` exhausts its retries and raises, the exception propagates immediately — the `_send(ai_events, AI_EVENTS_ENDPOINT)` call is never reached. Both sub-lists were dequeued from the shared queue and `task_done()` will be called for every item in `upload()`, so the AI events are permanently lost without any send being attempted. The docstring says "a failure on one does not re-send the other" which implies symmetric independence, but the ordering creates a dependency: analytics failure silently kills AI delivery too.
### Issue 2 of 3
posthog/consumer.py:147-148
`is_ai_event` is evaluated for every item in `batch` twice — once per list comprehension. A single partitioning loop halves the predicate calls and says the intent once.
```suggestion
ai_events, analytics_events = [], []
for item in batch:
(ai_events if is_ai_event(item.get("event")) else analytics_events).append(item)
```
### Issue 3 of 3
posthog/test/test_dedicated_ai_endpoint.py:46-74
**Sync-mode tests could be parameterised**
The three `TestDedicatedAiEndpointSyncMode` methods differ only in `(flag_enabled, event_name, expected_path)` — a table-driven `subTest` or `parameterized.expand` would express all three cases in one method, which aligns with the team's preference for parameterised tests.
Reviews (1): Last reviewed commit: "feat: dedicated ai endpoint routing" | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
Contributor
posthog-python Compliance ReportDate: 2026-06-10 11:12:15 UTC ✅ All Tests Passed!45/45 tests passed Capture Tests✅ 29/29 tests passed View Details
Feature_Flags Tests✅ 16/16 tests passed View Details
|
19 tasks
marandaneto
reviewed
Jun 9, 2026
marandaneto
reviewed
Jun 9, 2026
Comment on lines
+344
to
+345
| EVENTS_ENDPOINT = "/batch/" | ||
| AI_EVENTS_ENDPOINT = "/i/v0/ai/batch/" |
Member
There was a problem hiding this comment.
is the whole HTTP contract the very same? as in retry logic, request and response payload, etc?
Contributor
Author
There was a problem hiding this comment.
Actually I just noticed that beyond capping batch sizes on the backend, we also pre-emptively do it on the SDK. Just bumped the AI event size to the 8MB we're gonna be trialing.
Didn't touch batching size for the moment because it was a more comprehensive change, affecting non-AI events too. This means that for events larger than the batch size, we'll just send them as a 1 event batch, this seems reasonable for now.
…dicated_ai_endpoint Generated-By: PostHog Code Task-Id: 142d3917-41e8-4bb5-ad6c-5778e13041c2
marandaneto
approved these changes
Jun 10, 2026
marandaneto
left a comment
marandaneto
left a comment
Member
There was a problem hiding this comment.
approving to unblock, but still this comment left #656 (comment)
marandaneto
approved these changes
Jun 10, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
AI Observability wants
$ai_*events isolated into their own batch to a dedicated capture endpoint so AI and analytics workloads stop affecting each other, and different settings can be applied to each.Changes
Sends
$ai_prefixed events to a dedicated endpoint.Gated behind
_internal_dedicated_ai_endpoint. When off, behaviour is byte-identical to today.💚 How did you test it?
I'm an agent (Claude Code), paired with Carlos. Added
posthog/test/test_dedicated_ai_endpoint.py(6 cases): consumer routing of mixed batches to the two endpoints, AI-only batches, the disabled no-op, and thesync_modepath for both AI and non-AI events. New tests pass; the touched suites (test_consumer,test_request,test_client_fork) also pass. No manual testing claimed.📝 Checklist
If releasing new changes
sampo addto generate a changeset file🤖 Agent context
Authored with Claude Code (agent), paired with Carlos, as the python port of the posthog-js change. Design note: python's consumer materializes each batch in memory before POSTing (no persisted contiguous-removal invariant like posthog-node), so the clean approach here is to partition the batch at flush in the consumer rather than maintain separate queues. We target the existing analytics
/batch/handler via a dedicated path (/i/v0/ai/batch/) — not the dead multipart/i/v0/aiendpoint. Still pending before this is live: the backend route alias + charts Contour ingress entry, and confirmation the AI pipeline's per-message size limit is raised (WarpStream, per RFC #1111) so oversized events stop being dropped. Requires human review.