Langgraph summarization node is never invoked

[romano-neosec]

I'm using langgraph for running llm workflows and sometimes I stumble upon:

This model maximum context length is 128000 tokens.

To resolve it I tried summarization according to langgraph docs. However the summarization node is never invoked even for checking the number of messages ( i tried in debug).

def create_simple_agent_graph(state, agent_node, all_tools, router=agent_router):
    agent_node_name = "Agent"
    tool_node_name = "call_tool"
    tool_node = ToolNode(all_tools)
    builder = StateGraph(state)

    # Add nodes
    builder.add_node(agent_node_name, agent_node)
    builder.add_node(SUMMARIZE_NODE_NAME, summarize_node)
    builder.add_node(tool_node_name, tool_node)

    # Start with the agent node
    builder.add_edge(START, agent_node_name)

    # Conditional edges from the agent node
    builder.add_conditional_edges(
        agent_node_name,
        router,
        {"continue": END, tool_node_name: tool_node_name, END: END},
    )
    builder.add_conditional_edges(
        tool_node_name,
        should_summarize
    )

    builder.add_conditional_edges(
        SUMMARIZE_NODE_NAME,
        router,
        {"continue": END, tool_node_name: tool_node_name, END: END}
    )
    return builder

def summarize_node(state: type[FrameworkState]):
    model = GPT_4o_LLM
    summary_message = "Create a summary of the conversation above:"
    summary = state.get("summary", "")
    if summary:
        summary_message = (
            f"This is summary of the conversation to date: {summary}\n\n"
            "Extend the summary by taking into account the new messages above:"
        )
    messages = state["messages"] + [HumanMessage(content=summary_message)]
    response = model.invoke(messages)
    # Keeping system message and last 2 messages
    delete_messages = [RemoveMessage(id=m.id) for m in state["messages"][1:-2]]
    return {
        "summary": response.content,
        "messages": delete_messages,
    }

def should_summarize(state: TypedDict) -> str:
    SUMMARIZE_MAX = 6
    if len(state.get("messages", [])) > SUMMARIZE_MAX:
        return SUMMARIZE_NODE_NAME
    return "Agent"

def agent_router(state) -> Literal["call_tool", "__end__", "continue"]:
    # This is the router
    messages = state["messages"]
    last_message = messages[-1]
    if last_message.tool_calls:
        # The previous agent is invoking a tool
        return "call_tool"
    if "FINAL ANSWER" in last_message.content:
        return END

    return "continue"

Is there a way to resolve this issue and cause summarization node to run?

[Jrcordal]

what did you do at the end? @romano-neosec