scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-08 03:08:45 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 115

fix(streaming): keep partial-failure spend on the logging cost pipeline

Browse source
This commit is contained in:
mateo-berri 2026-08-28 13:13:37 -07:00
parent a17bc1f22c
commit 1754a0a33c
2 changed files with 16 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
litellm/litellm_core_utils/streaming_handler.py
View file

@ -2338,6 +2338,7 @@ class CustomStreamWrapper:
partial_response: Final = litellm.stream_chunk_builder(
chunks=self.chunks,
messages=self.messages if isinstance(self.messages, list) else None,
logging_obj=self.logging_obj,
)
if partial_response is None:
return

15
tests/test_litellm/litellm_core_utils/test_streaming_handler.py
View file

@ -3469,6 +3469,21 @@ def test_record_partial_usage_for_failure_backfills_missing_cache_fields():
assert stashed.prompt_tokens_details.cached_tokens == 0
def test_record_partial_usage_for_failure_prices_corrected_model_not_chunk_model():
wrapper, logging_obj = _wrapper_with_partial_chunks(
chunk_model="claude-opus-5",
usage=Usage(prompt_tokens=40, completion_tokens=5, total_tokens=45),
model="gpt-4o-mini",
custom_llm_provider="openai",
)
wrapper._record_partial_usage_for_failure()
rates = litellm.model_cost["gpt-4o-mini"]
expected = 40 * rates["input_cost_per_token"] + 5 * rates["output_cost_per_token"]
assert logging_obj.model_call_details["response_cost"] == pytest.approx(expected)
def test_record_partial_usage_for_failure_carries_up_openai_style_cached_tokens():
recovered = Usage(
prompt_tokens=1000,