scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-11-05 07:37:50 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 233

fix(nvidia_nim): sort typed retrieval results before response construction

Browse source
This commit is contained in:
Ivan Konstantinov 2026-10-19 11:52:29 +03:00
parent c56f093ccc
commit 46472fcd99
1 changed files with 9 additions and 12 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

21
litellm/llms/nvidia_nim/rerank/transformation.py
View file

@ -348,23 +348,20 @@ class NvidiaNimRerankConfig(BaseRerankConfig):
meta: Final[RerankResponseMeta] = {"billed_units": billed_units}
response: Final = RerankResponse(
# The hosted /v1/retrieval/{model}/reranking API rejects top_k.
# Sort before limiting in case the provider returns unsorted rankings.
# Work with the typed local results, not the Pydantic response attribute.
top_n: Final[int | None] = self._client_side_top_n
if top_n is not None and len(results) > top_n:
results.sort(key=lambda result: result["relevance_score"], reverse=True)
del results[top_n:]
return RerankResponse(
id=_STR.validate_python(payload.get("id") or str(uuid.uuid4())),
results=results,
meta=meta,
)
# The hosted /v1/retrieval/{model}/reranking API rejects top_k.
# Sort before limiting in case the provider returns unsorted rankings.
top_n: Final[int | None] = self._client_side_top_n
if top_n is not None and response.results is not None and len(response.results) > top_n:
response.results = sorted(
response.results,
key=lambda result: result["relevance_score"],
reverse=True,
)[:top_n]
return response
def get_error_class(self, error_message: str, status_code: int, headers: dict | httpx.Headers) -> BaseLLMException:
return BaseLLMException(
status_code=status_code,