{
  "2020": {"Z_generation": {"avg_sentiment": 0.7, "box_office_share": 0.3},
  "2023": {"Z_generation": {"avg_sentiment": 0.4, "box_office_share": 0.2}}
}

from langgraph.graph import StateGraph, State

class IPAnalysisState(State):
    ip_name: str
    current_year: int
    data_sources: dict
    generation_insights: dict
    failures: list

# 定义节点
async def plan_tasks(state):
    # 使用LLM生成子问题列表
    tasks = await llm.ainvoke(f"为IP {state.ip_name} 生成分析代际差异所需的任务列表")
    state["tasks"] = tasks
    return state

async def execute_tools(state):
    # 并行或顺序调用工具
    results = {}
    for task in state["tasks"]:
        tool = get_tool(task["type"])
        try:
            data = await tool.execute(task["params"])
            results[task["id"]] = data
        except Exception as e:
            state["failures"].append((task["id"], str(e)))
            # 降级：使用缓存或近似数据
            results[task["id"]] = await fallback_tool(task)
    state["data_sources"] = results
    return state

async def analyze(state):
    # 合并数据，计算代际差异
    prompt = f"根据以下数据，对比Z世代、千禧一代的共鸣度差异：{state['data_sources']}"
    state["generation_insights"] = await llm.ainvoke(prompt)
    return state

async def report(state):
    # 输出可读报告
    print(state["generation_insights"])
    return state

graph = StateGraph(IPAnalysisState)
graph.add_node("plan", plan_tasks)
graph.add_node("execute", execute_tools)
graph.add_node("analyze", analyze)
graph.add_node("report", report)
graph.set_entry_point("plan")
graph.add_edge("plan", "execute")
graph.add_edge("execute", "analyze")
graph.add_edge("analyze", "report")

app = graph.compile()

import os
import tweepy
from openai import OpenAI

# 配置（需要自己的 Twitter Developer账号）
client = tweepy.Client(bearer_token=os.environ["TWITTER_BEARER_TOKEN"])
llm = OpenAI()

def fetch_tweets(query="Star Wars generation Z", max_results=10):
    response = client.search_recent_tweets(query=query, max_results=max_results, 
                                           tweet_fields=["created_at", "public_metrics"])
    return [t.text for t in response.data]

def analyze_sentiment(tweets):
    prompt = f"""分析以下Tweets的情感（1非常正面，5非常负面），并给出对Z世代的吸引力评分（1-10）：
    {tweets}
    返回JSON格式：{{"avg_sentiment": 分数, "appeal_to_gen_z": 分数}}"""
    response = llm.chat.completions.create(model="gpt-4o-mini", messages=[{"role":"user","content":prompt}])
    return response.choices[0].message.content

def agent():
    tweets = fetch_tweets()
    if not tweets:
        print("无数据")
        return
    result = analyze_sentiment(tweets)
    print(result)
    # 保存到记忆文件（简化版用文本文件）
    with open("memory.txt", "a") as f:
        f.write(f"{datetime.now()}: {result}\n")

if __name__ == "__main__":
    agent()