Add span_id and trace_id to EvaluationReport (#2627)

Author: Kludex

pydantic_evals/pydantic_evals/dataset.py

@@ -297,6 +297,12 @@ async def _handle_case(case: Case[InputsT, OutputT, MetadataT], report_case_name
                         progress_bar.update(task_id, advance=1)
                     return result
 
+            if (context := eval_span.context) is None:  # pragma: no cover
+                trace_id = None
+                span_id = None
+            else:
+                trace_id = f'{context.trace_id:032x}'
+                span_id = f'{context.span_id:016x}'
             report = EvaluationReport(
                 name=name,
                 cases=await task_group_gather(
@@ -305,6 +311,8 @@ async def _handle_case(case: Case[InputsT, OutputT, MetadataT], report_case_name
                         for i, case in enumerate(self.cases, 1)
                     ]
                 ),
+                span_id=span_id,
+                trace_id=trace_id,
             )
             # TODO(DavidM): This attribute will be too big in general; remove it once we can use child spans in details panel:
             eval_span.set_attribute('cases', _REPORT_CASES_ADAPTER.dump_python(report.cases))
@@ -929,8 +937,8 @@ async def _run_task_and_evaluators(
 
         context = case_span.context
         if context is None:  # pragma: no cover
-            trace_id = ''
-            span_id = ''
+            trace_id = None
+            span_id = None
         else:
             trace_id = f'{context.trace_id:032x}'
             span_id = f'{context.span_id:016x}'

pydantic_evals/pydantic_evals/reporting/__init__.py

@@ -68,8 +68,8 @@ class ReportCase(Generic[InputsT, OutputT, MetadataT]):
     total_duration: float  # includes evaluator execution time
 
     # TODO(DavidM): Drop these once we can reference child spans in details panel:
-    trace_id: str
-    span_id: str
+    trace_id: str | None
+    span_id: str | None
 
 
 ReportCaseAdapter = TypeAdapter(ReportCase[Any, Any, Any])
@@ -158,9 +158,16 @@ class EvaluationReport(Generic[InputsT, OutputT, MetadataT]):
 
     name: str
     """The name of the report."""
+
     cases: list[ReportCase[InputsT, OutputT, MetadataT]]
     """The cases in the report."""
 
+    span_id: str | None = None
+    """The span ID of the evaluation."""
+
+    trace_id: str | None = None
+    """The trace ID of the evaluation."""
+
     def averages(self) -> ReportCaseAggregate:
         return ReportCaseAggregate.average(self.cases)
 

tests/evals/test_dataset.py

@@ -561,6 +561,8 @@ async def my_task(inputs: TaskInput) -> TaskOutput:
                     span_id='0000000000000007',
                 ),
             ],
+            span_id='0000000000000001',
+            trace_id='00000000000000000000000000000001',
         )
     )