feat(llma): add $ai_stop_reason property to LLM analytics#54037
Merged
Conversation
Captures the LLM's reason for stopping generation (e.g. "stop", "end_turn", "tool_use", "SAFETY") as a top-level queryable event property. Previously, finish_reason was stripped from OTel events and never extracted from SDK wrappers.
Contributor
Prompt To Fix All With AIThis is a comment left during a code review.
Path: nodejs/src/ingestion/ai/otel/middleware/vercel-ai.test.ts
Line: 97
Comment:
**Missing test for `gen_ai.response.finish_reasons` array path**
The `gen_ai.response.finish_reasons` fallback (the `else if` branch in `vercel-ai.ts`) is untested. That branch has its own logic — array type check plus `[0]` indexing — and currently any breakage there would go undetected. A parametrised test covering `ai.response.finishReason` (primary), `gen_ai.response.finish_reasons` array (fallback), and neither present (undefined) would cover the three cases OnceAndOnlyOnce.
How can I resolve this? If you propose a fix, please make it concise.
---
This is a comment left during a code review.
Path: nodejs/src/ingestion/ai/otel/middleware/traceloop.test.ts
Line: 207-208
Comment:
**`finish_reason`-only path is not exercised**
The test has both `llm.response.stop_reason` and `llm.response.finish_reason` present. The case where only `llm.response.finish_reason` is provided (no `stop_reason`) is untested, so a regression in the `??` fallback would go undetected. A parametrised test covering both-present (stop_reason wins), finish_reason-only, and neither-present (undefined) would match the preference for parameterised tests.
How can I resolve this? If you propose a fix, please make it concise.Reviews (1): Last reviewed commit: "feat(llma): add $ai_stop_reason property..." | Re-trigger Greptile |
Contributor
|
Size Change: +8.45 kB (+0.01%) Total Size: 128 MB
ℹ️ View Unchanged
|
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| expect(key).not.toBe('operation.name') | ||
| expect(key).not.toBe('resource.name') | ||
| } | ||
| expect(event.properties!['$ai_stop_reason']).toBe('stop') |
Contributor
There was a problem hiding this comment.
gen_ai.response.finish_reasons array path
The gen_ai.response.finish_reasons fallback (the else if branch in vercel-ai.ts) is untested. That branch has its own logic — array type check plus [0] indexing — and currently any breakage there would go undetected. A parametrised test covering ai.response.finishReason (primary), gen_ai.response.finish_reasons array (fallback), and neither present (undefined) would cover the three cases OnceAndOnlyOnce.
Prompt To Fix With AI
This is a comment left during a code review.
Path: nodejs/src/ingestion/ai/otel/middleware/vercel-ai.test.ts
Line: 97
Comment:
**Missing test for `gen_ai.response.finish_reasons` array path**
The `gen_ai.response.finish_reasons` fallback (the `else if` branch in `vercel-ai.ts`) is untested. That branch has its own logic — array type check plus `[0]` indexing — and currently any breakage there would go undetected. A parametrised test covering `ai.response.finishReason` (primary), `gen_ai.response.finish_reasons` array (fallback), and neither present (undefined) would cover the three cases OnceAndOnlyOnce.
How can I resolve this? If you propose a fix, please make it concise.
Comment on lines
+207
to
+208
| // stop_reason takes priority over finish_reason | ||
| expect(event.properties!['$ai_stop_reason']).toBe('end_turn') |
Contributor
There was a problem hiding this comment.
finish_reason-only path is not exercised
The test has both llm.response.stop_reason and llm.response.finish_reason present. The case where only llm.response.finish_reason is provided (no stop_reason) is untested, so a regression in the ?? fallback would go undetected. A parametrised test covering both-present (stop_reason wins), finish_reason-only, and neither-present (undefined) would match the preference for parameterised tests.
Prompt To Fix With AI
This is a comment left during a code review.
Path: nodejs/src/ingestion/ai/otel/middleware/traceloop.test.ts
Line: 207-208
Comment:
**`finish_reason`-only path is not exercised**
The test has both `llm.response.stop_reason` and `llm.response.finish_reason` present. The case where only `llm.response.finish_reason` is provided (no `stop_reason`) is untested, so a regression in the `??` fallback would go undetected. A parametrised test covering both-present (stop_reason wins), finish_reason-only, and neither-present (undefined) would match the preference for parameterised tests.
How can I resolve this? If you propose a fix, please make it concise.
andrewm4894
approved these changes
Apr 10, 2026
andrewm4894
approved these changes
Apr 10, 2026
Member
andrewm4894
left a comment
andrewm4894
left a comment
There was a problem hiding this comment.
oh interesting - we can do some stuff with this one i think
Contributor
|
🎭 Playwright report · View test results →
These issues are not necessarily caused by your changes. |
Contributor
Author
|
Companion PRs:
|
This was referenced Apr 10, 2026
- Test finish_reason-only path (no stop_reason) in traceloop middleware - Test gen_ai.response.finish_reasons array path in vercel-ai middleware - Test undefined when neither reason is present
Contributor
|
This PR hasn't seen activity in a week! Should it be merged, closed, or further worked on? If you want to keep it open, please remove the |
…top-reason # Conflicts: # frontend/src/taxonomy/core-filter-definitions-by-group.json
Contributor
|
⏭️ Skipped snapshot commit because branch advanced to The new commit will trigger its own snapshot update workflow. If you expected this workflow to succeed: This can happen due to concurrent commits. To get a fresh workflow run, either:
|
inkeep Bot
added a commit
that referenced
this pull request
Apr 20, 2026
Reflects PR #54037 which adds $ai_stop_reason to LLM analytics, capturing why model responses ended (e.g. stop, end_turn, max_tokens, tool_use).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
LLM analytics users cannot filter or analyze why model responses ended. The finish/stop reason from providers (e.g.
stop,end_turn,tool_use,max_tokens,SAFETY) was either stripped from OTel events or never extracted by SDK wrappers. This was reported as a user feature request on the LLM analytics docs page.Changes
llm.response.finish_reason,llm.response.stop_reason,ai.response.finishReason, andgen_ai.response.finish_reasonsto$ai_stop_reasoninstead of stripping them$ai_stop_reasonproperty definition intaxonomy.py,hogql/ai.py, and regenerated frontend JSONCompanion PRs:
@posthog/aiextraction for OpenAI, Anthropic, Gemini, Vercel AI SDK, LangChainHow did you test this code?
Publish to changelog?
no