langchain - 💡(How to fix) Fix `ChatOpenAI` Responses API streaming drops reasoning summary when provider sends `response.output_item.done` instead of `response.reasoning_summary_text.delta`

## NO CREDS REPRO - MOCKING CHUNK CONTENT
from unittest.mock import MagicMock
from langchain_openai.chat_models.base import _convert_responses_chunk_to_generation_chunk

# Simulate the event Azure sends: response.output_item.done, item.type == "reasoning"
item = MagicMock()
item.type = "reasoning"
item.id = "rs_abc123"
item.model_dump.return_value = {
    "type": "reasoning",
    "id": "rs_abc123",
    "summary": [{"type": "summary_text", "text": "Let me think step by step..."}],
}

chunk = MagicMock()
chunk.type = "response.output_item.done"
chunk.item = item
chunk.output_index = 0

_, _, _, result = _convert_responses_chunk_to_generation_chunk(chunk, 0, -1, -1)

# Bug: result is None — reasoning text silently dropped
assert result is None, "Already fixed!"
print("Bug confirmed: result is None, reasoning text is lost")

## E2E REPRO

from langchain_openai import AzureChatOpenAI

llm = AzureChatOpenAI(
    azure_deployment="o4-mini",
    azure_endpoint="https://<your-resource>.openai.azure.com/",
    api_key="<your-key>",
    api_version="2025-04-01-preview",
    use_responses_api=True,
    model_kwargs={"reasoning": {"effort": "medium", "summary": "auto"}},
)

for chunk in llm.stream("Explain step by step why 0.1 + 0.2 != 0.3"):
    reasoning = [
        b for b in chunk.message.content
        if isinstance(b, dict) and b.get("type") == "reasoning"
    ]
    if reasoning:
        print(reasoning)
        # Prints {'id': 'rs_...', 'type': 'reasoning'} — summary field is absent

---



---

# track which output indices have been streamed via delta events
elif chunk.type == "response.reasoning_summary_text.delta":
    if streamed_reasoning_output_indices is not None:
        streamed_reasoning_output_indices.add(chunk.output_index)
    ...  # existing code unchanged

# new handler
elif (
    chunk.type == "response.output_item.done"
    and chunk.item.type == "reasoning"
    and chunk.output_index not in (streamed_reasoning_output_indices or set())
):
    _advance(chunk.output_index)
    item = chunk.item.model_dump(exclude_none=True, mode="json")
    text = "".join(
        part.get("text", "")
        for part in (item.get("summary") or [])
        if part.get("type") == "summary_text"
    )
    if text:
        content.append({
            "type": "reasoning",
            "id": item.get("id", ""),
            "summary": [{"type": "summary_text", "text": text}],
            "index": current_index,
        })

---

System Information
------------------
> OS:  Darwin
> OS Version:  Darwin Kernel Version 25.3.0: Wed Jan 28 20:56:35 PST 2026; root:xnu-12377.91.3~2/RELEASE_ARM64_T6030
> Python Version:  3.12.0 (main, Mar 17 2026, 16:14:47) [Clang 17.0.0 (clang-1700.6.3.2)]

Package Information
-------------------
> langchain_core: 1.4.0
> langchain_community: 0.4.1
> langsmith: 0.8.3
> langchain_classic: 1.0.7
> langchain_google_genai: 4.1.3
> langchain_openai: 1.1.6
> langchain_protocol: 0.0.15
> langchain_text_splitters: 1.1.2

Optional packages not installed
-------------------------------
> deepagents
> deepagents-cli

Other Dependencies
------------------
> aiohttp: 3.13.5
> async-timeout: 5.0.1
> dataclasses-json: 0.6.7
> filetype: 1.2.0
> google-genai: 1.75.0
> httpx: 0.28.1
> httpx-sse: 0.4.3
> jsonpatch: 1.33
> numpy: 2.4.4
> openai: 1.109.1
> opentelemetry-api: 1.36.0
> opentelemetry-exporter-otlp-proto-http: 1.36.0
> opentelemetry-sdk: 1.36.0
> orjson: 3.11.9
> packaging: 24.2
> pydantic: 2.13.4
> pydantic-settings: 2.14.1
> pytest: 8.4.2
> pyyaml: 6.0.3
> PyYAML: 6.0.3
> requests: 2.34.0
> requests-toolbelt: 1.0.0
> rich: 15.0.0
> SQLAlchemy: 2.0.49
> sqlalchemy: 2.0.49
> tenacity: 9.1.4
> tiktoken: 0.12.0
> typing-extensions: 4.15.0
> uuid-utils: 0.15.0
> websockets: 16.0
> wrapt: 1.17.3
> xxhash: 3.7.0
> zstandard: 0.25.0