From d7406750741a96f663093e46dbf96d17ceca140c Mon Sep 17 00:00:00 2001 From: mateo-berri <277851410+mateo-berri@users.noreply.github.com> Date: Thu, 29 Oct 2026 04:55:41 -0700 Subject: [PATCH] fix(batches): log no cost for a batch retrieve that is not final yet A retrieve now reads OpenAI's batch usage into the chat counts, so a retrieve that saw the batch completed before its output file arrived priced those counts for the logging callbacks, and the final retrieve priced the whole batch again. A batch that is not final now logs no cost, as it did before. Batch usage dumps also pass exclude_unset and exclude_defaults through to the usage keys. --- litellm/litellm_core_utils/litellm_logging.py | 3 ++ litellm/types/utils.py | 13 +++++-- .../test_litellm_logging.py | 36 ++++++++++++------- tests/unit/types/test_types_utils.py | 13 +++++++ 4 files changed, 51 insertions(+), 14 deletions(-) diff --git a/litellm/litellm_core_utils/litellm_logging.py b/litellm/litellm_core_utils/litellm_logging.py index 7f4b069ded6..1e622ddd696 100644 --- a/litellm/litellm_core_utils/litellm_logging.py +++ b/litellm/litellm_core_utils/litellm_logging.py @@ -2071,6 +2071,9 @@ class Logging(LiteLLMLoggingBaseClass): elif router_model_id is None and "model_id" in hidden_params: # use model_id if not already set router_model_id = hidden_params["model_id"] + if isinstance(priced_result, LiteLLMBatch) and not batch_cost_is_final(priced_result): + return 0.0 + # Fallback: extract router_model_id from litellm_params when not available # from the result object. ResponsesAPIResponse objects (used by /v1/responses # streaming) don't carry _hidden_params["model_id"] like ModelResponse does. diff --git a/litellm/types/utils.py b/litellm/types/utils.py index 167c6311804..f488cc8d853 100644 --- a/litellm/types/utils.py +++ b/litellm/types/utils.py @@ -4872,10 +4872,19 @@ class LiteLLMBatch(Batch): return usage return { **usage.model_dump( - mode=info.mode, include=info.include, exclude=info.exclude, exclude_none=info.exclude_none + mode=info.mode, + include=info.include, + exclude=info.exclude, + exclude_unset=info.exclude_unset, + exclude_defaults=info.exclude_defaults, + exclude_none=info.exclude_none, ), **_batch_usage_from_chat_usage(usage).model_dump( - mode=info.mode, include=info.include, exclude=info.exclude + mode=info.mode, + include=info.include, + exclude=info.exclude, + exclude_unset=info.exclude_unset, + exclude_defaults=info.exclude_defaults, ), } diff --git a/tests/unit/litellm_core_utils/test_litellm_logging.py b/tests/unit/litellm_core_utils/test_litellm_logging.py index 299088a6c4f..34eec8fe87e 100644 --- a/tests/unit/litellm_core_utils/test_litellm_logging.py +++ b/tests/unit/litellm_core_utils/test_litellm_logging.py @@ -1519,22 +1519,33 @@ class TestRetrieveBatchPricesOnlyFinalBatches: litellm_call_id="batch-call-2", function_id="f", ) - obj.custom_llm_provider = "openai" + obj.update_environment_variables( + model="gpt-5.6-luna", user=None, optional_params={}, litellm_params={}, custom_llm_provider="openai" + ) return obj @staticmethod def _batch(status: str, output_file_id: str | None): from litellm.types.utils import LiteLLMBatch - return LiteLLMBatch( - id="batch_6a9c99e185588190877d391f8b9d7f8a", - completion_window="24h", - created_at=1, - endpoint="/v1/chat/completions", - input_file_id="file-in", - object="batch", - status="validating", - output_file_id=output_file_id, + return LiteLLMBatch.model_validate( + { + "id": "batch_6a9c99e185588190877d391f8b9d7f8a", + "completion_window": "24h", + "created_at": 1, + "endpoint": "/v1/chat/completions", + "input_file_id": "file-in", + "object": "batch", + "status": "validating", + "output_file_id": output_file_id, + "usage": { + "input_tokens": 29, + "input_tokens_details": {"cached_tokens": 0}, + "output_tokens": 48, + "output_tokens_details": {"reasoning_tokens": 0}, + "total_tokens": 77, + }, + } ).model_copy(update={"status": status}) @pytest.mark.asyncio @@ -1548,11 +1559,12 @@ class TestRetrieveBatchPricesOnlyFinalBatches: handle_completed_batch = AsyncMock() monkeypatch.setattr(logging_module, "handle_completed_batch", handle_completed_batch) batch = self._batch(status, output_file_id) + logging_obj = self._logging_obj() - await self._logging_obj()._async_success_handler_body(result=batch, start_time=None, end_time=None) + await logging_obj._async_success_handler_body(result=batch, start_time=None, end_time=None) handle_completed_batch.assert_not_awaited() - assert "response_cost" not in batch._hidden_params + assert logging_obj.model_call_details["response_cost"] == 0.0 @pytest.mark.asyncio async def test_completed_batch_with_output_is_priced(self, monkeypatch) -> None: diff --git a/tests/unit/types/test_types_utils.py b/tests/unit/types/test_types_utils.py index e12109e10f7..aaf338600b1 100644 --- a/tests/unit/types/test_types_utils.py +++ b/tests/unit/types/test_types_utils.py @@ -1154,3 +1154,16 @@ def test_batch_dump_applies_nested_include_and_exclude_to_both_usage_shapes() -> assert {"prompt_tokens_details", "input_tokens_details"}.isdisjoint(excluded) assert (excluded["prompt_tokens"], excluded["input_tokens"]) == (21, 21) assert included == {"prompt_tokens": 21, "input_tokens": 21} + + +def test_batch_dump_applies_exclude_defaults_to_the_chat_usage_keys() -> None: + batch: Final = LiteLLMBatch.model_validate( + {**_BATCH_FIELDS, "usage": Usage(prompt_tokens=10, completion_tokens=5, total_tokens=15)} + ) + + python_usage: Final = batch.model_dump(exclude_defaults=True)["usage"] + json_usage: Final = json.loads(batch.model_dump_json(exclude_defaults=True, warnings=False))["usage"] + + assert python_usage == json_usage + assert {"prompt_tokens_details", "completion_tokens_details"}.isdisjoint(python_usage) + assert (python_usage["prompt_tokens"], python_usage["input_tokens"]) == (10, 10)