fix(batch-evaluation): pass trace ID for score creation if scope = observations (#1504)

Author: hassiebp

langfuse/batch_evaluation.py

@@ -20,6 +20,7 @@
     Protocol,
     Tuple,
     Union,
+    cast,
 )
 
 from langfuse.api.resources.commons.types import (
@@ -1220,6 +1221,9 @@ async def _process_batch_evaluation_item(
             self._create_score_for_scope(
                 scope=scope,
                 item_id=item_id,
+                trace_id=cast(ObservationsView, item).trace_id
+                if scope == "observations"
+                else None,
                 evaluation=evaluation,
                 additional_metadata=metadata,
             )
@@ -1242,6 +1246,9 @@ async def _process_batch_evaluation_item(
                     self._create_score_for_scope(
                         scope=scope,
                         item_id=item_id,
+                        trace_id=cast(ObservationsView, item).trace_id
+                        if scope == "observations"
+                        else None,
                         evaluation=composite_eval,
                         additional_metadata=metadata,
                     )
@@ -1361,8 +1368,10 @@ async def _run_composite_evaluator(
 
     def _create_score_for_scope(
         self,
+        *,
         scope: str,
         item_id: str,
+        trace_id: Optional[str] = None,
         evaluation: Evaluation,
         additional_metadata: Optional[Dict[str, Any]],
     ) -> None:
@@ -1371,6 +1380,7 @@ def _create_score_for_scope(
         Args:
             scope: The type of entity ("traces", "observations").
             item_id: The ID of the entity.
+            trace_id: The trace ID of the entity; required if scope=observations
             evaluation: The evaluation result to create a score from.
             additional_metadata: Additional metadata to merge with evaluation metadata.
         """
@@ -1393,6 +1403,7 @@ def _create_score_for_scope(
         elif scope == "observations":
             self.client.create_score(
                 observation_id=item_id,
+                trace_id=trace_id,
                 name=evaluation.name,
                 value=evaluation.value,  # type: ignore
                 comment=evaluation.comment,

tests/test_batch_evaluation.py

@@ -25,7 +25,7 @@
 # ============================================================================
 
 
-pytestmark = pytest.mark.skip(reason="Github CI runner overwhelmed by score volume")
+# pytestmark = pytest.mark.skip(reason="Github CI runner overwhelmed by score volume")
 
 
 @pytest.fixture
@@ -67,6 +67,32 @@ def simple_evaluator(*, input, output, expected_output=None, metadata=None, **kw
 # ============================================================================
 
 
+def test_run_batched_evaluation_on_observations_basic(langfuse_client):
+    """Test basic batch evaluation on traces."""
+    result = langfuse_client.run_batched_evaluation(
+        scope="observations",
+        mapper=simple_trace_mapper,
+        evaluators=[simple_evaluator],
+        max_items=1,
+        verbose=True,
+    )
+
+    # Validate result structure
+    assert isinstance(result, BatchEvaluationResult)
+    assert result.total_items_fetched >= 0
+    assert result.total_items_processed >= 0
+    assert result.total_scores_created >= 0
+    assert result.completed is True
+    assert isinstance(result.duration_seconds, float)
+    assert result.duration_seconds > 0
+
+    # Verify evaluator stats
+    assert len(result.evaluator_stats) == 1
+    stats = result.evaluator_stats[0]
+    assert isinstance(stats, EvaluatorStats)
+    assert stats.name == "simple_evaluator"
+
+
 def test_run_batched_evaluation_on_traces_basic(langfuse_client):
     """Test basic batch evaluation on traces."""
     result = langfuse_client.run_batched_evaluation(