Search code examples
nlplangchainlanggraph

not able END function using "add_conditional_edges" in lang graph

this is my code:

import os
from dotenv import load_dotenv
load_



dotenv()
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
from langgraph.graph import Graph, MessagesState
from typing import Annotated, Any, Dict, Optional, List,Sequence, TypedDict
from langgraph.graph.message import add_messages
 
class AgentState(TypedDict):
    # The `add_messages` function within the annotation defines
    # *how* updates should be merged into the state.
    messages: Annotated[list, add_messages]
 
def function1(state):
    return {"messages": "Hi"}
 
def function2(state):
    return {"messages": "Hello"}
 
def my_condition(state):
    return "end"
 
workflow=StateGraph(AgentState)
 
workflow.add_node("agent", function1)
workflow.add_node("tool", function2)
 
workflow.add_edge('agent','tool')
workflow.set_entry_point("agent")
 
workflow.add_conditional_edges("agent", my_condition,{ "end": END})
app=workflow.compile()
print(app.invoke({"messages": "tell me about you"}))

In above code I want to END function at "function1" and get this result:

{'messages': [HumanMessage(content='tell me about you', id='70a7cb55-4cb2-4d0b-9623-79cb06bcabf3'), HumanMessage(content='Hi', id='d95bd56d-93b6-44b1-ae05-3449472d8463')]}

But I am getting this below result:

{'messages': [HumanMessage(content='tell me about you', id='70a7cb55-4cb2-4d0b-9623-79cb06bcabf3'), HumanMessage(content='Hi', id='d95bd56d-93b6-44b1-ae05-3449472d8463'), HumanMessage(content='Hello', id='7ea9ab2a-635f-46eb-8f17-d9a6af79688e')]}

Solution

  • Edges tell lang graph where to look after a node. If it's "conditional edges," it will go to different next nodes in different conditions. If it's just a normal edge, it is a direct line from that node to the next node.

    In your code, you are adding both conditional edges and a normal edge from the same "agent" node.

    workflow.add_edge('agent','tool')
...
workflow.add_conditional_edges("agent", my_condition,{ "end": END})

    You need to decide, after the "agent" function is called, which thing do you want to happen next? If you want to reach END, you could put the conditional edges from "tool" instead of "agent," which already has an edge, and all of the functions will be called.