Merge pull request #34067 from MUSE-CODE-SPACE/fix/batch-logging-null-output-file

fix(batches): don't crash logging when a completed batch has no output file

litellm/batches/batch_utils.py

@@ -59,6 +59,17 @@ async def _handle_completed_batch(
         model_name: Optional model name
         litellm_params: Optional litellm parameters containing credentials (api_key, api_base, etc.)
     """
+    # A completed batch whose request lines all failed has no output file - the
+    # results are written to a separate error_file_id and output_file_id is None.
+    # There is nothing to price or measure, so report an empty result set instead
+    # of calling _fetch_batch_output_file_content, which raises on a missing
+    # output file. Without this guard the logging worker crashes on every
+    # aretrieve_batch poll and the completed batch's zero-cost accounting is lost.
+    # The generic retrieval helper keeps raising for callers that explicitly ask
+    # for a missing output file.
+    if batch.output_file_id is None:
+        return 0.0, Usage(prompt_tokens=0, completion_tokens=0, total_tokens=0), []
+
     file_content = await _fetch_batch_output_file_content(batch, custom_llm_provider, litellm_params=litellm_params)
 
     if (

tests/test_litellm/batches/test_batch_utils.py

@@ -977,6 +977,27 @@ async def test_handle_completed_batch_orchestration(monkeypatch):
     assert models == ["gpt-4o"]
 
 
+@pytest.mark.asyncio
+async def test_handle_completed_batch_no_output_file_is_zero(monkeypatch):
+    """
+    Regression: an all-error batch completes with output_file_id=None (results go
+    to a separate error_file_id). _handle_completed_batch must report an empty
+    result set - zero cost, zero usage, no models - instead of letting the file
+    fetch raise "Output file id is None" on every aretrieve_batch logging poll.
+    """
+    # The output-file fetch must not even be attempted when there is no output file.
+    async def _must_not_fetch(*args, **kwargs):
+        pytest.fail("_fetch_batch_output_file_content should not be called")
+
+    monkeypatch.setattr(bu, "_fetch_batch_output_file_content", _must_not_fetch)
+
+    cost, usage, models = await bu._handle_completed_batch(_batch(None), custom_llm_provider="openai")
+
+    assert cost == 0.0
+    assert (usage.prompt_tokens, usage.completion_tokens, usage.total_tokens) == (0, 0, 0)
+    assert models == []
+
+
 @pytest.mark.asyncio
 async def test_handle_completed_batch_vertex_disable_transform_path(monkeypatch):
     raw_rows = [{"response": {"usageMetadata": {"promptTokenCount": 1, "candidatesTokenCount": 2}}}]