fix(spend): fall back from null response ids

litellm/proxy/spend_tracking/spend_tracking_utils.py

@@ -216,7 +216,7 @@ def get_spend_logs_id(call_type: str, response_obj: dict, kwargs: dict) -> str |
         kwargs.get("litellm_call_id"),
     )
     resolved_id: Final = next(
-        (candidate for candidate in candidate_ids if isinstance(candidate, str) and candidate), None
+        (candidate for candidate in candidate_ids if isinstance(candidate, str) and candidate not in ("", "None")), None
     )
     if resolved_id is not None and call_type == CallTypes.aretrieve_batch.value:
         return f"{resolved_id}{BATCH_COST_REQUEST_ID_SUFFIX}"

tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py

@@ -3399,6 +3399,19 @@ def test_get_spend_logs_id_prefers_the_response_id_over_the_standard_logging_id(
     )
 
 
+def test_get_spend_logs_id_falls_back_when_the_standard_logging_id_is_none():
+    ids = [
+        get_spend_logs_id(
+            "acompletion",
+            {"id": None},
+            {"litellm_call_id": call_id, "standard_logging_object": {"id": "None"}},
+        )
+        for call_id in ("call-id-1", "call-id-2")
+    ]
+
+    assert ids == ["call-id-1", "call-id-2"]
+
+
 @pytest.mark.asyncio
 async def test_spend_log_request_id_is_the_message_id_a_bridged_streaming_caller_was_streamed():
     """A streaming /v1/messages call against a non-Anthropic model is served a msg_ id the