[Bug]: Usage Tracking with OpenAI streaming fails with pydantic validation

[mayankchutani]

What component(s) are affected?

• Opik Python SDK
• Opik Typescript SDK
• Opik Agent Optimizer SDK
• Opik UI
• Opik Server
• Documentation

Opik version

• Opik version: 1.9.51

Describe the problem

My langgraph agent uses OpenAI streaming, and for usage tracking I get a pydantic validation error.

OPIK: Failed to extract token usage from presumably OpenAI LLM langchain run. The run dictionary...

Reproduction steps and code snippets

No response

Error logs or stack trace

Traceback (most recent call last):
  File "/.venv/lib/python3.14/site-packages/opik/integrations/langchain/provider_usage_extractors/openai_usage_extractor.py", line 74, in _try_get_token_usage
    return llm_usage.OpikUsage.from_openai_completions_dict(token_usage)
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
  File "/.venv/lib/python3.14/site-packages/opik/llm_usage/opik_usage.py", line 74, in from_openai_completions_dict
    provider_usage = openai_chat_completions_usage.OpenAICompletionsUsage.from_original_usage_dict(
        usage
    )
  File ".venv/lib/python3.14/site-packages/opik/llm_usage/openai_chat_completions_usage.py", line 61, in from_original_usage_dict
    return cls(
        **usage_dict,
        completion_tokens_details=completion_tokens_details,
        prompt_tokens_details=prompt_tokens_details,
    )
  File ".../.venv/lib/python3.14/site-packages/pydantic/main.py", line 250, in __init__
    validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
pydantic_core._pydantic_core.ValidationError: 3 validation errors for OpenAICompletionsUsage
completion_tokens
  Field required [type=missing, input_value={'model_name': {'type': '...t_tokens_details': None}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/missing
prompt_tokens
  Field required [type=missing, input_value={'model_name': {'type': '...t_tokens_details': None}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/missing
total_tokens
  Input should be a valid integer [type=int_type, input_value={'default': None, 'type': 'integer'}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/int_type

Healthcheck results

No response

[alexkuzmik]

Hi @mayankchutani!
Could you please check if you've enabled streaming usage metadata? Usage is not returned by default for openai models when streaming mode is on, it must be explicitly enabled.
https://docs.langchain.com/oss/python/integrations/chat/openai#streaming-usage-metadata

If you verified it, I'd appreciate the reproduction script and pip freeze output so we could take a look at your issue.