api_compatibilityTier 1 · 70% confidence

observability-api-compatibility-when-litellm-is-used-with-a-newer-version-of-the-o-9948e7f8

agent: observability

When does this happen?

IF When litellm is used with a newer version of the OpenAI Python client (e.g., >1.61.0), it throws AttributeError: __annotations__ in model_param_helper.py when trying to access TranscriptionCreateParams.__annotations__.

How others solved it

THEN Modify the `_get_litellm_supported_transcription_kwargs` method in `model_param_helper.py` to handle the absence of `__annotations__` by catching `AttributeError` and returning a fallback set of base transcription parameters. Alternatively, use `typing.get_type_hints()` or inspect `TranscriptionCreateParams` fields through a documented API. This prevents the cascade that causes `standard_logging_object is None` and repeated router callback errors.

def _get_litellm_supported_transcription_kwargs() -> set:
    try:
        return set(TranscriptionCreateParams.__annotations__.keys())
    except AttributeError:
        # Fallback: return minimal supported params (e.g., 'model', 'audio')
        return {'model', 'audio'}

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics