feat(python): OpenAI Agents Tracing Interoperability #2070
Conversation
angus-langchain
changed the title
angus-langchain
changed the title
python/langsmith/__init__.py
Outdated
| # Avoid calling into importlib on every call to __version__ | ||
|
|
||
| __version__ = "0.4.34" | ||
| __version__ = "0.4.35rc1" |
| new_run.post() | ||
|
|
||
| ctx = copy_context() | ||
| ctx.run(_context._PARENT_RUN_TREE.set, new_run) |
There was a problem hiding this comment.
Why is storing context necessary? We're just pulling the parent out of the span maps aren't we?
There was a problem hiding this comment.
modifying the context is what enables @traceable and wrap_openai to automatically find the agent span as a parent without us having to pass it explicitly
There was a problem hiding this comment.
via setting _PARENT_RUN_TREE
| project_name: Optional[str] | ||
| name: Optional[str] | ||
| run_tree: rt.RunTree | ||
| context: contextvars.Context |
There was a problem hiding this comment.
See below, do we need to store context here?
There was a problem hiding this comment.
This is correct though, we don't need to actually store it in the RunData map
| parent_run_id: Optional[str] | ||
| project_name: Optional[str] | ||
| name: Optional[str] | ||
| run_tree: rt.RunTree |
There was a problem hiding this comment.
nit: why not just store RunTrees directly
There was a problem hiding this comment.
we are, just importing from langsmith import run_trees as rt
2 participants
Description
Use RunTree API for openai agents sdk to improve tracing devex by allowing for input/output processing, nested traces with
@traceableandwrap_openai.Test Plan
Resolves #2063