fix(scores): parse session ID correctly

langfuse/_client/client.py

@@ -1818,7 +1818,7 @@ def create_score(
         try:
             new_body = ScoreBody(
                 id=score_id,
-                session_id=session_id,
+                sessionId=session_id,
                 datasetRunId=dataset_run_id,
                 traceId=trace_id,
                 observationId=observation_id,

langfuse/_utils/request.py

@@ -48,9 +48,8 @@ def generate_headers(self) -> dict:
 
     def batch_post(self, **kwargs: Any) -> httpx.Response:
         """Post the `kwargs` to the batch API endpoint for events"""
-        logger.debug("uploading data: %s", kwargs)
-
         res = self.post(**kwargs)
+
         return self._process_response(
             res, success_message="data uploaded successfully", return_json=False
         )

langfuse/langchain/CallbackHandler.py

@@ -1057,10 +1057,10 @@ def _convert_message_to_dict(self, message: BaseMessage) -> Dict[str, Any]:
                 and len(message.tool_calls) > 0
             ):
                 message_dict["tool_calls"] = message.tool_calls
-            
+
             if (
-                hasattr(message, "invalid_tool_calls") 
-                and message.invalid_tool_calls is not None 
+                hasattr(message, "invalid_tool_calls")
+                and message.invalid_tool_calls is not None
                 and len(message.invalid_tool_calls) > 0
             ):
                 message_dict["invalid_tool_calls"] = message.invalid_tool_calls

tests/test_core_sdk.py

@@ -118,6 +118,49 @@ def test_invalid_score_data_does_not_raise_exception():
     # We can't assert queue size in OTEL implementation, but we can verify it completes without exception
 
 
+def test_create_session_score():
+    langfuse = Langfuse()
+
+    session_id = "my-session"
+
+    # Create a span and set trace properties
+    with langfuse.start_as_current_observation(name="test-span"):
+        with propagate_attributes(
+            trace_name="this-is-so-great-new",
+            user_id="test",
+            metadata={"test": "test"},
+            session_id=session_id,
+        ):
+            pass
+
+    # Ensure data is sent
+    langfuse.flush()
+    sleep(2)
+
+    # Create a numeric score
+    score_id = create_uuid()
+
+    langfuse.create_score(
+        score_id=score_id,
+        session_id=session_id,
+        name="this-is-a-score",
+        value=1,
+    )
+
+    # Ensure data is sent
+    langfuse.flush()
+    sleep(2)
+
+    # Retrieve and verify
+    score = langfuse.api.scores.get_by_id(score_id)
+
+    # find the score by name (server may transform the id format)
+    assert score is not None
+    assert score.value == 1
+    assert score.data_type == "NUMERIC"
+    assert score.session_id == session_id
+
+
 def test_create_numeric_score():
     langfuse = Langfuse()
     api_wrapper = LangfuseAPI()