pip install langgraph langchain langchain_google_genai langfuse python-dotenv

# .env 파일 내용.env file content
GOOGLE_API_KEY='your-google-api-key-here'

# Langfuse 설정 (4단계에서 채울 예정)Langfuse settings (to be filled in step 4)
LANGFUSE_SECRET_KEY=""
LANGFUSE_PUBLIC_KEY=""
LANGFUSE_HOST="http://localhost:3000"

my-langgraph-agent/
├── .env
└── run_agent.py

import os
from typing import TypedDict
from dotenv import load_dotenv
from langchain_google_genai import ChatGoogleGenerativeAI
from langgraph.graph import StateGraph, END
from langfuse.callback import CallbackHandler

# --- 0. 환경 변수 로드 (.env에 KEY들을 저장) ------ 0. Load environment variables (store KEYS in .env) ---
load_dotenv()

# --- 1. Langfuse 콜백 핸들러 초기화 ------ 1. Initialize Langfuse Callback Handler ---
# .env 파일에 저장된 키를 사용하여 핸들러를 생성합니다.Create the handler using keys stored in the .env file.
langfuse_handler = CallbackHandler(
    public_key=os.getenv("LANGFUSE_PUBLIC_KEY"),
    secret_key=os.getenv("LANGFUSE_SECRET_KEY"),
    host=os.getenv("LANGFUSE_HOST")
)

# --- 2. 그래프용 상태 정의 ------ 2. Define State for the Graph ---
class AgentState(TypedDict):
    topic: str
    research_result: str

# --- 3. 노드 함수 정의 ------ 3. Define Node Functions ---
def researcher(state: AgentState):
    print("--- 리서처 노드 실행 ------ Executing Researcher Node ---")
    topic = state["topic"]
    llm = ChatGoogleGenerativeAI(model="gemini-1.5-flash", temperature=0)
    research_result = llm.invoke(
        f"'{topic}'에 대한 핵심적인 사실 3가지를 알려줘.Tell me 3 key facts about '{topic}'."
    ).content
    print(f"리서치 결과: Research Result: {research_result}")
    return {"research_result": research_result}

def writer(state: AgentState):
    print("--- 작가 노드 실행 ------ Executing Writer Node ---")
    research_result = state['research_result']
    llm = ChatGoogleGenerativeAI(model="gemini-1.5-flash", temperature=0.7)
    final_output = llm.invoke(
        f"다음 정보를 바탕으로 흥미로운 단락을 작성해줘:\n\nWrite an interesting paragraph based on the following information:\n\n{research_result}"
    ).content
    print(f"최종 결과물:\nFinal Output:\n{final_output}")
    return {}

# --- 4. 그래프 구축 ------ 4. Build the Graph ---
workflow = StateGraph(AgentState)
workflow.add_node("researcher", researcher)
workflow.add_node("writer", writer)
workflow.set_entry_point("researcher")
workflow.add_edge("researcher", "writer")
workflow.add_edge("writer", END)
app = workflow.compile()

# --- 5. 실행 ------ 5. Execute ---
inputs = {"topic": "인공지능의 역사The History of Artificial Intelligence"}
print("Langfuse와 함께 그래프 실행을 시작합니다…Starting graph execution with Langfuse…")
app.invoke(inputs, config={"callbacks": [langfuse_handler]})

# --- 6. 남은 이벤트 전송 ------ 6. Flush remaining events ---
langfuse_handler.flush()
print("완료! 대시보드에서 확인하세요 →Done! Check it out on the dashboard →", os.getenv("LANGFUSE_HOST"))