observability-pydantic-validation-langfuse-s-updategenerationbody-raises-validation--bb5b8695

THEN Before passing usage data to langfuse's update method, ensure that nested fields under usageDetails (e.g., completion_tokens_details, prompt_tokens_details) are provided as complete objects with all integer fields set to 0 if empty. For example, if the OpenAI response returns empty `{}` for these details, replace them with a dict containing all expected keys set to 0. Alternatively, check pydantic version compatibility and update the model schema if needed.

if 'completion_tokens_details' in usage and not usage['completion_tokens_details']:
    usage['completion_tokens_details'] = {
        'reasoning_tokens': 0,
        'audio_tokens': 0,
        'accepted_prediction_tokens': 0,
        'rejected_prediction_tokens': 0
    }