Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions py/src/braintrust/integrations/openai/test_openai.py
Original file line number Diff line number Diff line change
Expand Up @@ -525,6 +525,10 @@ def test_openai_responses_web_search_tool_spans(memory_logger):
tool_spans = _find_spans_by_type(spans, SpanTypeAttribute.TOOL)

assert len(llm_spans) == 1
if version_satisfies(openai.__version__, ">=3.41.0"):
tool_usage = response.tool_usage.model_dump(exclude_none=True)
assert tool_usage["web_search"]["num_requests"] == 1
assert _try_to_dict(llm_spans[0]["metadata"]["tool_usage"]) == tool_usage
tool_span = _find_span_by_name(tool_spans, "web_search_call")
assert tool_span["span_parents"] == [llm_spans[0]["span_id"]]
assert tool_span["metadata"]["tool_type"] == "web_search_call"
Expand Down Expand Up @@ -566,6 +570,10 @@ def test_openai_responses_web_search_tool_spans_stream(memory_logger):
tool_spans = _find_spans_by_type(spans, SpanTypeAttribute.TOOL)

assert len(llm_spans) == 1
if version_satisfies(openai.__version__, ">=3.41.0"):
tool_usage = final_response.tool_usage.model_dump(exclude_none=True)
assert tool_usage["web_search"]["num_requests"] == 1
assert _try_to_dict(llm_spans[0]["metadata"]["tool_usage"]) == tool_usage
tool_span = _find_span_by_name(tool_spans, "web_search_call")
assert tool_span["span_parents"] == [llm_spans[0]["span_id"]]
assert tool_span["metadata"]["tool_type"] == "web_search_call"
Expand Down
5 changes: 5 additions & 0 deletions py/src/braintrust/integrations/openai/tracing.py
Original file line number Diff line number Diff line change
Expand Up @@ -122,6 +122,7 @@ def start_span(*args, **kwargs):
"object",
"background",
"incomplete_details",
"tool_usage",
)

_EMBEDDING_METADATA_PARAMS = (
Expand Down Expand Up @@ -1735,6 +1736,10 @@ def _postprocess_streaming_results(cls, all_results: list[Any]) -> dict[str, Any

for result in all_results:
usage = getattr(result, "usage", None)
if getattr(result, "type", None) in ("response.completed", "response.incomplete", "response.failed"):
tool_usage = getattr(getattr(result, "response", None), "tool_usage", None)
if tool_usage is not None:
metadata["tool_usage"] = tool_usage
if (
not usage
and hasattr(result, "type")
Expand Down
Loading