fix: resolve issue #29513 - cannot retrieve reasoning_content while s…

[codergma]

Extract reasoning_content in the _convert_delta_to_message_chunk function

• PR title: "langchain_openai: add reasoning_content to stream response when using deepseek-reasoner"

• PR message:

  • Description: Add reasoning_content to stream response
  • Issue: the issue Cannot retrieve reasoning_content while streaming #29513
  • Dependencies: no dependencies

[vercel]

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

Name	Status	Preview	Comments	Updated (UTC)
langchain	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Feb 2, 2025 0:49am

[ccurme]

Can we do this on the Deepseek integration, similar to how it's done for invoke?

langchain/libs/partners/deepseek/langchain_deepseek/chat_models.py

Lines 210 to 218 in e8b9128

	rtn = super()._create_chat_result(response, generation_info)
	if not isinstance(response, openai.BaseModel):
	return rtn
	if hasattr(response.choices[0].message, "reasoning_content"): # type: ignore
	rtn.generations[0].message.additional_kwargs["reasoning_content"] = (
	response.choices[0].message.reasoning_content # type: ignore
	)

[liuruibin]

anything new?

[codergma]

Can we do this on the Deepseek integration, similar to how it's done for invoke?

langchain/libs/partners/deepseek/langchain_deepseek/chat_models.py

Lines 210 to 218 in e8b9128

	rtn = super()._create_chat_result(response, generation_info)
	if not isinstance(response, openai.BaseModel):
	return rtn
	if hasattr(response.choices[0].message, "reasoning_content"): # type: ignore
	rtn.generations[0].message.additional_kwargs["reasoning_content"] = (
	response.choices[0].message.reasoning_content # type: ignore
	)

Yes, you are right, but it seems that there isn't a suitable way to do it at the moment.

[codergma]

Hi @baskaryan , could you please take a look at this PR?