class ClaimAgent:
    def __init__(self, llm, tools, memory):
        self.llm = llm
        self.tools = tools
        self.memory = memory

    def process_claim(self, claim_id):
        # 1. 从消息队列获取初始索赔数据
        claim_data = self.tools.get_claim_by_id(claim_id)

        # 2. 规划当前步骤（由LLM动态生成）
        plan = self.llm.plan([
            {"step": "verify_member", "args": ["member_id"]},
            {"step": "lookup_plan_coverage", "args": ["plan_id", "cpt_code"]},
            {"step": "calculate_payment", "args": ["claim_amount", "deductible", "coinsurance"]},
            {"step": "update_status", "args": ["claim_id", "approved"]}
        ], context=claim_data)

        # 3. 按计划执行，每一步调用对应工具
        for step in plan:
            result = self.execute_step(step)
            if result.status == "failed":
                # 动态重规划：尝试修复或标记人工
                new_plan = self.llm.replan(step, result.error, context=claim_data)
                self.execute_plan(new_plan)

        # 4. 保存记忆并通知相关方
        self.memory.save(claim_id, {"final_status": "approved", "payment_amount": result.amount})
        self.tools.send_notification(member_email, result)

    def execute_step(self, step):
        tool = self.tools.get(step.name)
        response = tool.invoke(step.args)
        return response

from langchain.agents import AgentExecutor, create_openai_tools_agent
from langchain.tools import tool
from langchain_openai import ChatOpenAI
from langchain.prompts import ChatPromptTemplate

@tool
def verify_member(member_id: str) -> str:
    """验证会员ID是否存在并返回保险计划ID"""
    # 模拟调用会员数据库
    if member_id == "12345":
        return json.dumps({"status": "active", "plan_id": "GOLD-2026"})
    else:
        return json.dumps({"status": "not_found"})

@tool
def lookup_coverage(cpt_code: str, plan_id: str) -> str:
    """查询CPT代码是否覆盖以及自付额"""
    if cpt_code == "99213" and plan_id == "GOLD-2026":
        return json.dumps({"covered": True, "deductible": 500, "coinsurance": 0.2})
    else:
        return json.dumps({"covered": False})

@tool
def calculate_payment(claim_amount: float, deductible: float, coinsurance: float) -> str:
    """计算赔付金额"""
    if claim_amount <= deductible:
        payment = 0
    else:
        payment = (claim_amount - deductible) * (1 - coinsurance)
    return json.dumps({"payment": round(payment, 2)})

# 创建Agent
llm = ChatOpenAI(model="gpt-4", temperature=0)
tools = [verify_member, lookup_coverage, calculate_payment]
prompt = ChatPromptTemplate.from_messages([
    ("system", "你是一个医疗索赔处理助手。请按顺序调用工具：先验证会员，再查询覆盖范围，最后计算赔付。如果会员不存在或覆盖不通过，立即返回拒绝。"),
    ("human", "处理索赔：会员ID {member_id}，CPT代码 {cpt_code}，索赔金额 {claim_amount}")
])
agent = create_openai_tools_agent(llm, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

# 运行示例
result = agent_executor.invoke({"member_id": "12345", "cpt_code": "99213", "claim_amount": 800})
print(result["output"])