fix(ai): add message truncation to openai–agents #4968
Conversation
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
| SPANDATA.GEN_AI_REQUEST_MESSAGES, | ||
| messages_data, | ||
| unpack=False, | ||
| ) |
There was a problem hiding this comment.
Bug: Missing Data Normalization in Agent Span
The invoke_agent_span function is missing a call to _normalize_data() before passing normalized_messages to truncate_and_annotate_messages(). This differs from utils.py and means the truncation function receives Python objects instead of the expected serialized data, preventing it from working correctly. _normalize_data also needs to be imported.
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## master #4968 +/- ##
==========================================
- Coverage 83.96% 83.94% -0.03%
==========================================
Files 165 178 +13
Lines 17073 17842 +769
Branches 2999 3172 +173
==========================================
+ Hits 14336 14978 +642
- Misses 1839 1900 +61
- Partials 898 964 +66
|
| # Just verify that messages are being set and truncation is applied | ||
| # The actual truncation behavior is tested in the ai_monitoring tests | ||
| ai_client_span = ai_client_spans[0] | ||
| if "gen_ai.request.messages" in ai_client_span["data"]: | ||
| messages_data = ai_client_span["data"]["gen_ai.request.messages"] | ||
| assert isinstance(messages_data, str) | ||
|
|
||
| parsed_messages = json.loads(messages_data) | ||
| assert isinstance(parsed_messages, list) | ||
| # Verify messages were processed | ||
| assert len(parsed_messages) >= 1 |
There was a problem hiding this comment.
Is this checking that truncation is applied?
I would have thought len(parsed_messages) >= 1 is always true, even without truncation.
2 participants
Closes TET-1208