[React] agent scratch 부터 만들기

---

• 개요
• Graph 상태 정의
• Tools 정의
• Node , Edge 정의
• Graph 정의
• ReAct 에이전트 사용하기
• Reference

개요

Langgraph 를 사용하여 처음부터 ReAct 에이전트를 구현하는 방법을 적도록 하겠다.

Graph 상태 정의

from typing import Annotated, Sequence, TypedDict
from langchain_core.messages import BaseMessage
from langgraph.graph.message import add_messages

class AgentState(TypedDict):
    """The state of the agent"""
    ## add_messags is a reducer
    messages: Annotated[Sequence[BaseMessage], add_messages]

Tools 정의

from langchain_core.tools import tool

@tool
def get_weather(location: str):
    """Call to get the weather from a specific location."""
    # This is a placeholder for the actual implementation
    # Don't let the LLM know this though 😊
    if any([city in location.lower() for city in ["sf", "san francisco"]]):
        return "It's sunny in San Francisco, but you better look out if you're a Gemini 😈."
    else:
        return f"I am not sure what the weather is in {location}"

tools = [get_weather]
model = model.bind_tools(tools)

Node , Edge 정의

기본 ReAct 에이전트는 모델을 호출하는 노드와 툴을 사용하는 노드, 이렇게 두 개의 노드만 포함한다. 하지만 이 구조는 사용 사례에 따라 수정 가능하다.

예: 구조화된 출력을 추가하는 노드, 외부 작업(이메일 보내기, 캘린더 이벤트 추가 등)을 실행하는 노드 등을 추가하거나, 모델 호출 방식 및 툴 호출 여부를 판단하는 조건 등을 커스터마이징 할 수 있다.

import json
from langchain_core.messages import ToolMessage, SystemMessage
from langchain_core.runnables import RunnableConfig

tools_by_name = {tool.name: tool for tool in tools}

# Define our tool node
def tool_node(state:AgentState):
    outputs = []
    for tool_call in state['messages'][-1].tool_calls:
        tool_result = tools_by_name[tool_call['name']].invoke(tool_call['args'])
        outputs.append(
            ToolMessage(
                content=json.dumps(tool_result),
                name=tool_call['name'],
                tool_call_id=tool_call['id']
            )
        )
  
    return {'messages' : outputs}

# Define the node that calls the model
def call_model(
    state: AgentState,
    config: RunnableConfig,
):
    system_prompt = SystemMessage(
        "You are a helpful AI assistant, please respond to the users query to the best of your ability!"
    )
    response = model.invoke([system_prompt] + state['messages'] , config)
    return {'messages' : [response]}
  
# Define the conditional edge that determines wheter to continue or not
def should_continue(state:AgentState):
    messages = state['messages']
    last_message = messages[-1]
  
    if not last_message.tool_calls:
        return "end"
    else:
        return "continue"

Graph 정의

from langgraph.graph import StateGraph, END

workflow = StateGraph(AgentState)
workflow.add_node('agent', call_model)
workflow.add_node('tools', tool_node)
workflow.set_entry_point('agent')

workflow.add_conditional_edges(
    'agent',
    should_continue,
    {
        'continue': 'tools',
        'end' : END
    }
)

workflow.add_edge('tools','agent')
graph = workflow.compile()

ReAct 에이전트 사용하기

# Helper function for formatting the stream nicely
def print_stream(stream):
    for s in stream:
        message = s["messages"][-1]
        if isinstance(message, tuple):
            print(message)
        else:
            message.pretty_print()

inputs = {"messages": [("user", "what is the weather in sf")]}
print_stream(graph.stream(inputs, stream_mode="values"))

================================ Human Message ================================= 
what is the weather in sf 

================================== Ai Message ================================== Tool Calls: get_weather (call_6f7py2q8) Call ID: call_6f7py2q8 Args: location: sf 

================================= Tool Message ================================= Name: get_weather "It's sunny in San Francisco, but you better look out if you're a Gemini \ud83d\ude08." 

================================== Ai Message ================================== It's sunny in San Francisco, but you better look out if you're a Gemini \ud83d\ude08. Is there anything else you would like to know about the weather?

• 코드 주소

---

Reference

• How to create a ReAct agent from scratch