Could .with_structured_output() work together with the parameter "seed"?

[Junjie-Chu]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

agent_llm = ChatOpenAI(
    model=AGENT_LLM, 
    temperature=0.0, 
    seed=SEED
    )

structured_agent_llm = agent_llm.with_structured_output(
        schema=ScoringPointJudgement,
        method="json_schema",
        tools=tools,
        strict=True,
)
result = structured_agent_llm.invoke(input_messages)

Description

I load gpt4o with both the following:

agent_llm = ChatOpenAI(
    model=AGENT_LLM, 
    temperature=0.0, 
    model_kwargs={
        "seed": SEED
        }
    )

OR

agent_llm = ChatOpenAI(
    model=AGENT_LLM, 
    temperature=0.0, 
    seed=SEED
    )

However, when I run:

    structured_agent_llm = agent_llm.with_structured_output(
        schema=ScoringPointJudgement,
        method="json_schema",
        tools=tools,
        strict=True,
        )
    result = structured_agent_llm.invoke(input_messages)

I both meet:

TypeError("Responses.parse() got an unexpected keyword argument 'seed'")

I think ChatOpenAI already supports "seed," but it seems the structure output not?

System Info

langchain==0.3.12
langchain-anthropic==0.3.10
langchain-community==0.3.12
langchain-core==0.3.63
langchain-experimental==0.3.4
langchain-milvus==0.1.8
langchain-openai==0.3.19
langchain-text-splitters==0.3.7

MAC

Python 3.10.16