langgraph/cloud/how-tos/stream_updates/

[giscus[bot]]

langgraph/cloud/how-tos/stream_updates/

Build language agents as graphs

https://langchain-ai.github.io/langgraph/cloud/how-tos/stream_updates/

[VijayDeyTR]

I have a requirement to stream only the updates - but capture the final full state (with all messages). How can that be done?

[vbarda]

you can stream multiple streaming modes (and filter) -- would that help w/ your use case? https://langchain-ai.github.io/langgraph/cloud/how-tos/stream_multiple/

[gautam-mantracare]

Hi! I am new to langgraph and I am trying to create a chatbot with retrieval and tools with streaming. I have defined my graph as provided below. When I am streaming the text, for each tool call, two entries are generated in for ToolMessage checkpoint in the database. I am using PostgresSaver. Can you please guide me, what am I doing wrong here.

workflow = StateGraph(state_schema=MessagesState)

        existing_tools = get_tools()
        tools = ToolNode(existing_tools + [retrieve])

        workflow.add_node(call_model)
        workflow.add_node(tools)
        workflow.add_node(generate)

        workflow.set_entry_point("call_model")
        workflow.add_conditional_edges(
            "call_model",
            tools_condition,
            {END: END, "tools": "tools"}
        )

        workflow.add_edge("tools", "generate")
        workflow.add_edge("generate", END)

This is how I am streaming

config = self.__get_config(thread_id)
        events = self.app.stream(
            {"messages": input_messages}, config, stream_mode="messages")

        for message, _ in events:
            # print(message)
            if isinstance(message, BaseMessage):
                yield message.content