LangGraph中ToolNode调用工具提示tool_call_id缺失，如何实现工具+Command模式？

错误原因 #

tool_call_id参数未被LangGraph的ToolNode正确注入，导致调用工具时缺少必填参数。这通常是因为注入机制需要配合正确的工具定义或版本支持。

方案一：正确使用InjectedToolCallId（推荐） #

确保使用最新版的LangChain和LangGraph，然后按照以下方式调整代码：

1. 导入必要依赖

from typing import Annotated
from pydantic import BaseModel, Field
from langchain_core.tools import tool, InjectedToolCallId
from langchain_core.messages import ToolMessage
from langgraph.prebuilt import ToolNode
from langgraph.graph import StateGraph, MessagesState

2. 定义对话状态

class ChatState(MessagesState):
    active_chat: bool = True

3. 修正工具定义
   注意tool_call_id不需要加入args_schema，它会被自动注入：

class EndConversationSchema(BaseModel):
    """Call this tool when the user says goodbye"""
    reason: str = Field(description="Why does the user want to end the conversation?")

@tool("end_conversation", args_schema=EndConversationSchema)
def end_conversation(
    reason: str,
    tool_call_id: Annotated[str, InjectedToolCallId]
) -> dict:
    """End the conversation with the given reason."""
    return {
        "active_chat": False,
        "messages": [ToolMessage(f"Conversation finished. Reason: {reason}", tool_call_id=tool_call_id)]
    }

4. 创建并配置图

# 初始化工具节点
tool_node = ToolNode([end_conversation])

# 定义结束判断逻辑
def should_continue(state: ChatState) -> str:
    return "__end__" if not state["active_chat"] else "tools"

# 构建图
graph_builder = StateGraph(ChatState)
graph_builder.add_node("tools", tool_node)
graph_builder.set_entry_point("tools")
graph_builder.add_conditional_edges("tools", should_continue)

graph = graph_builder.compile()

方案二：手动获取tool_call_id（兼容旧版本） #

如果升级版本后仍有问题，可以直接从工具调用消息中提取tool_call_id，无需依赖注入机制：

from langchain_core.messages import ToolMessage
from langgraph.graph import StateGraph, MessagesState
from langchain_core.runnables import RunnableConfig

class ChatState(MessagesState):
    active_chat: bool = True

def end_conversation(state: ChatState, config: RunnableConfig):
    last_msg = state["messages"][-1]
    # 提取工具调用信息
    tool_call = last_msg.tool_calls[0]
    reason = tool_call["args"]["reason"]
    tool_call_id = tool_call["id"]
    
    return {
        "active_chat": False,
        "messages": [ToolMessage(f"Conversation finished. Reason: {reason}", tool_call_id=tool_call_id)]
    }

# 构建图
graph_builder = StateGraph(ChatState)
graph_builder.add_node("end_convo", end_conversation)
graph_builder.set_entry_point("end_convo")
graph = graph_builder.compile()

额外注意事项 #

• 先执行升级命令确保依赖版本正确：pip install --upgrade langchain langgraph
• 返回的更新内容会被LangGraph自动合并到对话状态中，无需手动处理状态修改

内容的提问来源于stack exchange，提问作者Facundo