feat(core): automatically count and store meta for tool call count #33756
Conversation
CodSpeed Performance ReportMerging #33756 will not alter performanceComparing
|
| def _persist_run(self, run: Run) -> None: | ||
| """Persist a run.""" | ||
|
|
||
| def _store_tool_call_metadata(self, run: Run) -> None: |
There was a problem hiding this comment.
I'd put this all in _complete_llm_run in tracers/core. We don't need to do this on all runs IMO
| ) | ||
| llm_run.end_time = datetime.now(timezone.utc) | ||
| llm_run.events.append({"name": "end", "time": llm_run.end_time}) | ||
| store_tool_call_count_in_run(llm_run) |
There was a problem hiding this comment.
nit: I really dislike side-effects but I understand they're a thing in Python
| return tool_call_count | ||
|
|
||
|
|
||
| def store_tool_call_count_in_run(run: Run, *, always_store: bool = False) -> int: |
There was a problem hiding this comment.
Does this really need to be its own function
Adds automatic tool call counting to tracing by means of a new
store_tool_call_count_in_run(), which calls on newly addedcount_tool_calls_in_run().Runs on successful LLM completion. Does not run on errored runs.