fix(haystack): add input and output #1202
Conversation
|
@tibor-reiss want to rebase and fix the lint issues? we can merge this one as well (although it won't solve #1169 completely - it's a much needed work in this direction) |
tibor-reiss
force-pushed
the
1169-haystack-input-output
branch
2 times, most recently
from
3842155 to
7837e22
Compare
@nirga Rebased, fixed linting issue. |
| kwargs_to_serialize[key] = value | ||
| args_to_serialize = [arg for arg in args if not isinstance(arg, dict)] | ||
| input_entity = {"args": args_to_serialize, "kwargs": kwargs_to_serialize} | ||
| return json.dumps(input_entity, cls=EnhancedJSONEncoder) |
There was a problem hiding this comment.
@tibor-reiss almost missed this - should be behind if should_send_prompts()
There was a problem hiding this comment.
Good catch, thanks!
There was a problem hiding this comment.
@nirga Just looked at it a second time: should this check rather be in wrap_pipeline.py? Otherwise both spans will be set to None even if should_send_prompts would be False.
I was using the implementation in langchain as blueprint, however I did not like that much that the process_request / process_response functions change state (it's kinda a side effect, and also harder to test).
Let me know your preference!
There was a problem hiding this comment.
I'd keep it separate since we may be adding more things to process_request / process_response
tibor-reiss
force-pushed
the
1169-haystack-input-output
branch
from
920e41e to
b5f7825
Compare
Towards #1169
feat(instrumentation): ...orfix(instrumentation): ....