How have non streaming nodes in a graph I call stream() on?

[itay-moav]

I have a very simple graph.
First node calls an llm and get back structured data. I need that result as returned all at once.
that is passed in the state, the second and last node is a simple LLM invoke with the human message.
(I did not add any of the real logic I actually need to have there to keep this simple).

First node:

def chatbot(state: State):
    llm = AzureChatOpenAI(
        api_version=config["apiVersion"],
        streaming=True,
        azure_deployment=config["model"],
        temperature=config["temprature"],
    ).with_structured_output(schema=ClassificationRouterStruct, strict=True)
    res = llm.invoke(state["messages"]
    return {"route": "operations","prefetch_tools":[]}. # <--------- NOTICE I PUT HARD CODED RETURN TO ELIMINATE ISSUES HERE

Second/last node:

def chatbot(state: State):
    llm = AzureChatOpenAI(
        api_version=config["apiVersion"],
        streaming=True,
        azure_deployment=config["model"],
        temperature=config["temprature"],
    )
    return {"messages": [llm.invoke(state["messages"])]}

This still streams my first node (I print each chunk so I see it)
This result in getting:
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/itay/.local/share/virtualenvs/backend-XCG2kbnH/lib/python3.12/site-packages/openai/lib/streaming/chat/_completions.py", line 393, in _accumulate_chunk
cast("dict[object, object]", choice.delta.to_dict()),
^^^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'to_dict'