fix(check_batch_cost): mark completed batches with no output file as processed

An OpenAI batch where every request fails completes with status=completed
but output_file_id=None. It matched neither the completed-with-output nor the
failed/expired/cancelled branch, so batch_processed never flipped and the row
was re-polled forever, saturating the oldest-N poll window and starving newer
batches. Mark such batches complete + batch_processed=True (no cost to track).

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

enterprise/litellm_enterprise/proxy/common_utils/check_batch_cost.py

@@ -627,10 +627,15 @@ class CheckBatchCost:
                         f"CheckBatchCost: failed to mark job {job.id} complete in DB: {db_err}"
                     )
 
-            elif response.status in ("failed", "expired", "cancelled"):
+            elif response.status in ("failed", "expired", "cancelled") or (
+                response.status == "completed" and response.output_file_id is None
+            ):
+                terminal_status = (
+                    "complete" if response.status == "completed" else response.status
+                )
                 try:
                     update_data = {
-                        "status": response.status,
+                        "status": terminal_status,
                         "file_object": response.model_dump_json(),
                     }
                     if self._has_batch_processed_column:
@@ -640,11 +645,12 @@ class CheckBatchCost:
                         data=update_data,
                     )
                     verbose_proxy_logger.info(
-                        f"CheckBatchCost: marked job {job.id} as {response.status} in DB"
+                        f"CheckBatchCost: marked job {job.id} as {terminal_status} in DB "
+                        f"(provider status={response.status}, output_file_id={response.output_file_id})"
                     )
                 except Exception as db_err:
                     verbose_proxy_logger.error(
-                        f"CheckBatchCost: failed to mark job {job.id} as {response.status} in DB: {db_err}"
+                        f"CheckBatchCost: failed to mark job {job.id} as {terminal_status} in DB: {db_err}"
                     )
 
         # Record polling run metrics (always, even if nothing was processed)

tests/proxy_unit_tests/test_check_batch_cost.py

@@ -556,6 +556,83 @@ class TestCheckBatchCost:
             update_data["batch_processed"] is True
         ), "terminal-status update() must set batch_processed=True so polling stops"
 
+    @pytest.mark.asyncio
+    async def test_completed_with_no_output_file_marks_job_processed(
+        self,
+        check_batch_cost_instance,
+        mock_prisma_client,
+        mock_llm_router,
+    ):
+        """A batch that OpenAI reports as completed but with output_file_id=None (every
+        request failed, only an error file is produced) must be marked complete and
+        batch_processed=True. Otherwise it matches neither the completed-with-output nor
+        the failed/expired/cancelled branch and gets re-polled forever, eventually
+        saturating the oldest-N poll window and starving newer batches.
+        """
+        from unittest.mock import patch
+
+        mock_prisma_client.db.litellm_managedobjecttable.update_many = AsyncMock(
+            return_value=0
+        )
+        mock_prisma_client.db.litellm_managedobjecttable.update = AsyncMock()
+        mock_prisma_client.db.litellm_usertable.find_unique = AsyncMock(
+            return_value=None
+        )
+
+        mock_job = MagicMock()
+        mock_job.id = "job-completed-no-output-1"
+        mock_job.unified_object_id = "dW5pZmllZF9iYXRjaF9pZA=="
+        mock_job.created_by = "user-1"
+
+        assert check_batch_cost_instance._has_batch_processed_column is True
+        mock_prisma_client.db.litellm_managedobjecttable.find_many = AsyncMock(
+            return_value=[mock_job]
+        )
+
+        mock_response = MagicMock()
+        mock_response.status = "completed"
+        mock_response.output_file_id = None
+        mock_response.error_file_id = "file-error-123"
+        mock_response.model_dump_json.return_value = (
+            '{"id":"batch-1","status":"completed","output_file_id":null}'
+        )
+
+        mock_llm_router.aretrieve_batch = AsyncMock(return_value=mock_response)
+
+        decoded_id = "llm_model_id,model-123;llm_batch_id,batch-456;"
+
+        with (
+            patch(
+                "litellm.proxy.openai_files_endpoints.common_utils._is_base64_encoded_unified_file_id",
+                side_effect=[decoded_id, None],
+            ),
+            patch(
+                "litellm.proxy.openai_files_endpoints.common_utils.get_model_id_from_unified_batch_id",
+                return_value="model-123",
+            ),
+            patch(
+                "litellm.proxy.openai_files_endpoints.common_utils.get_batch_id_from_unified_batch_id",
+                return_value="batch-456",
+            ),
+            patch(
+                "litellm.files.main.afile_content",
+                new_callable=AsyncMock,
+            ) as mock_afile_content,
+        ):
+            await check_batch_cost_instance.check_batch_cost()
+
+        mock_afile_content.assert_not_called()
+        assert (
+            mock_prisma_client.db.litellm_managedobjecttable.update.call_count == 1
+        ), "Expected update() to be called once for a completed-with-no-output batch"
+        update_data = mock_prisma_client.db.litellm_managedobjecttable.update.call_args[
+            1
+        ]["data"]
+        assert update_data["status"] == "complete"
+        assert (
+            update_data["batch_processed"] is True
+        ), "completed-with-no-output update() must set batch_processed=True so polling stops"
+
     @pytest.mark.asyncio
     async def test_raw_output_file_id_converted_to_managed_id(
         self, check_batch_cost_instance, mock_prisma_client, mock_llm_router