openai[patch]: fix dropping response headers while streaming / Azure

[joshy-deshaw]

• Currently, while streaming with OpenAI with include_response_headers=True, the response headers are set as generation_info in the first chunk here.
• However, the _convert_chunk_to_generation_chunk drops generation_info because it returns early. This PR makes it so that the info is preserved.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Skipped Deployment

Name	Status	Preview	Comments	Updated (UTC)
langchain	⬜️ Ignored (Inspect)	Visit Preview		Jun 12, 2025 1:19pm

[codspeed-hq]

CodSpeed Walltime Performance Report

Merging #31580 will not alter performance

_{Comparing joshy-deshaw:usage-metadata-fix (1c1dbc4) with master (446a9d5)}

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

Summary

✅ 13 untouched benchmarks

[codspeed-hq]

CodSpeed Instrumentation Performance Report

Merging #31580 will not alter performance

_{Comparing joshy-deshaw:usage-metadata-fix (1c1dbc4) with master (446a9d5)}

Summary

✅ 13 untouched benchmarks

[joshy-deshaw]

@baskaryan tagging for review

[ccurme]

Can you clarify what this does with a test? I don't see any update to response metadata here:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="gpt-4.1-mini",
    include_response_headers=True,
    stream_usage=True,
)

full = None
for chunk in llm.stream("hello"):
    full = chunk if full is None else full + chunk

full.response_metadata

{'headers': {'date': 'Fri, 20 Jun 2025 23:18:14 GMT',
  'content-type': 'text/event-stream; charset=utf-8',
  ...},
 'finish_reason': 'stop',
 'model_name': 'gpt-4.1-mini-2025-04-14',
 'system_fingerprint': '...',
 'service_tier': 'default'}

[joshy-deshaw]

An example with AzureChatOpenAI below, though it seems like you're already getting these headers with ChatOpenAI

llm = AzureChatOpenAI(
    **azure_kwargs,
    model_name="gpt-4o-mini",
    streaming=True,
    stream_options={"include_usage": True},
    include_response_headers=True,
)
full = None
for chunk in llm.stream("hello"):
    full = chunk if full is None else full + chunk

print(full.response_metadata)

Without the changes in this PR,

{'finish_reason': 'stop', 'model_name': 'gpt-4o-mini-2024-07-18', 'system_fingerprint': '...'}

With the changes

{
  "headers": {
    "transfer-encoding": "chunked",
    "content-type": "text/event-stream; charset=utf-8",
    "apim-request-id": "...",
    "strict-transport-security": "...",
    "x-content-type-options": "nosniff",
    "x-ms-region": "...",
    "x-ratelimit-remaining-requests": "19999",
    "x-ratelimit-limit-requests": "20000",
    "x-ratelimit-remaining-tokens": "1999998",
    "x-ratelimit-limit-tokens": "2000000",
    "azureml-model-session": "...",
    "cmp-upstream-response-duration": "126",
    "x-accel-buffering": "no",
    "x-ms-rai-invoked": "true",
    "x-request-id": "...",
    "x-ms-client-request-id": "Not-Set",
    "x-ms-deployment-name": "...",
    "date": "Mon, 23 Jun 2025 05:33:39 GMT"
  },
  "finish_reason": "stop",
  "model_name": "gpt-4o-mini-2024-07-18",
  "system_fingerprint": "..."
}