mirror of
https://github.com/BerriAI/litellm.git
synced 2026-09-14 23:21:35 +00:00
Two zero-error gate regressions from PR #32857: - completion_response_id relied on json.loads (returns Any) and dict.get on an unnarrowed dict, which triggered reportAny and reportUnknownVariableType. Route the parse through a TypeAdapter[dict[str, object]] so parsed and raw are typed and validation errors funnel through ValidationError. - LangfuseListParams(...) was constructed with the field-name kwargs trace_id and from_start_time, but pydantic's Field alias narrows the generated __init__ to traceId and fromStartTime; basedpyright flagged those calls as reportCallIssue. Pass the aliased names; populate_by_name plus by_alias serialization keep runtime identical. Co-authored-by: Krrish Dholakia <krrish-berri-2@users.noreply.github.com> |
||
|---|---|---|
| .. | ||
| conftest.py | ||
| logging_client.py | ||
| test_langfuse_e2e.py | ||
| test_prometheus_cardinality_e2e.py | ||