ai-agents-azure-openai-streami-azurechatopenai-with-use-responses-api-true-and-re-ec886af7

THEN Upgrade langchain-openai to the latest stable version. The bug was resolved in a recent release; ensure your package version includes the fix for the Responses API streaming compatibility.

# Example that reproduces the bug (before fix)
from langchain_openai import AzureChatOpenAI
from langchain.prompts.chat import ChatPromptTemplate, HumanMessagePromptTemplate

reasoning = {"effort": "medium", "summary": "auto"}
az_llm_stream = AzureChatOpenAI(
    openai_api_key="...",
    azure_endpoint="...",
    model_name="o4-mini",
    azure_deployment="o4-mini",
    api_version="2025-04-01-preview",
    verbose=True,
    use_responses_api=True,
    model_kwargs={"reasoning": reasoning},
)

prompt = ChatPromptTemplate.from_messages([HumanMessagePromptTemplate.from_template("{question}")])
chain = prompt | az_llm_stream
for chunk in chain.stream({"question": "Tell me..."}):
    print(chunk)
# After upgrading langchain-openai, this code works.