langchain - 💡(How to fix) Fix ChatOpenAI.with_structured_output(..., method="json_schema").stream() drops response headers / x-request-id in response_metadata

Submission checklist

• This is a bug, not a usage question.
• I added a clear and descriptive title that summarizes this issue.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).
• This is not related to the langchain-community package.
• I posted a self-contained, minimal, reproducible example. A maintainer can copy it and run it AS IS.

Package (Required)

• langchain
• langchain-openai
• langchain-anthropic
• langchain-classic
• langchain-core
• langchain-model-profiles
• langchain-tests
• langchain-text-splitters
• langchain-chroma
• langchain-deepseek
• langchain-exa
• langchain-fireworks
• langchain-groq
• langchain-huggingface
• langchain-mistralai
• langchain-nomic
• langchain-ollama
• langchain-openrouter
• langchain-perplexity
• langchain-qdrant
• langchain-xai
• Other / not sure / general

Related Issues / PRs

No response

Reproduction Steps / Example Code (Python)

from pathlib import Path
from typing import Any

from langchain_core.callbacks import BaseCallbackHandler
from langchain_openai import ChatOpenAI
from pydantic import BaseModel


class SimpleAnswer(BaseModel):
    answer: str


class HeaderCaptureHandler(BaseCallbackHandler):
    """Records whether ``x-request-id`` arrived via on_llm_end / on_llm_new_token."""

    def __init__(self) -> None:
        self.x_request_id: str | None = None

    def _grab(self, message: Any) -> None:
        if message is None:
            return
        headers = (getattr(message, 'response_metadata', None) or {}).get('headers') or {}
        rid = headers.get('x-request-id') or headers.get('X-Request-Id')
        if rid and self.x_request_id is None:
            self.x_request_id = rid

    def on_llm_new_token(self, token: str, *, chunk: Any = None, **kwargs: Any) -> None:
        self._grab(chunk)

    def on_llm_end(self, response: Any, **kwargs: Any) -> None:
        for gen_list in getattr(response, 'generations', None) or []:
            for gen in gen_list:
                self._grab(getattr(gen, 'message', None))


def run_case(label: str, runnable: Any) -> None:
    handler = HeaderCaptureHandler()
    for _ in runnable.stream('Say hello in one word.', config={'callbacks': [handler]}):
        pass
    status = handler.x_request_id or '<MISSING>'
    print(f'{label:50s} x-request-id={status}')


llm = ChatOpenAI(model='gpt-4.1-mini', streaming=True, include_response_headers=True)

run_case('A) plain streaming                              ', llm)
run_case('B) streaming + with_structured_output(json_schema)', llm.with_structured_output(SimpleAnswer, method='json_schema'))

Error Message and Stack Trace (if applicable)

Description

Hi team,

We’re seeing inconsistent response header propagation when using LangChain structured output + streaming with OpenAI models.

Summary

When using:

chat_model.with_structured_output(
    SimpleAnswer,
    method="json_schema"
).stream(...)

the streamed AIMessageChunk / final message objects do not contain HTTP response headers (specifically x-request-id) inside response_metadata.

However, the same observability flow works correctly for:

non-structured streaming AzureChatOpenAI structured streaming regular create(stream=True, include_response_headers=True) paths

This makes it difficult to correlate streamed requests with provider-side logs/incidents.

System Info

Package Information

langchain_core: 1.2.29 langchain: 1.2.15 langchain_community: 0.4.1 langsmith: 0.3.45 langchain_aws: 1.4.3 langchain_classic: 1.0.3 langchain_google_genai: 4.2.1 langchain_google_vertexai: 2.1.2 langchain_openai: 0.3.34 langchain_text_splitters: 1.1.1 langgraph_sdk: 0.3.13