import json
from typing import List, Dict, Callable
from dataclasses import dataclass

# 模拟LLM调用（实际项目接入OpenAI/Claude等）
def call_llm(prompt: str) -> str:
    return f"模拟LLM响应: 基于'{prompt[:30]}...'的分析结果"

class Tool:
    def __init__(self, name: str, func: Callable, description: str):
        self.name = name
        self.func = func
        self.description = description
    
    def execute(self, **kwargs):
        return self.func(**kwargs)

class ReActAgent:
    def __init__(self, tools: List[Tool], max_steps: int = 10):
        self.tools = {t.name: t for t in tools}
        self.max_steps = max_steps
        self.memory: List[Dict] = []
    
    def _build_prompt(self, task: str) -> str:
        tool_desc = "\n".join([
            f"- {name}: {tool.description}" 
            for name, tool in self.tools.items()
        ])
        
        history = "\n".join([
            f"Step {i+1}: {step['thought']} -> {step['action']}"
            for i, step in enumerate(self.memory)
        ])
        
        return f"""你是一个智能助手，使用ReAct范式解决任务。

可用工具：
{tool_desc}

历史步骤：
{history if history else '无'}

任务：{task}

请按以下格式回复：
Thought: [你的思考过程]
Action: [工具名称, 参数JSON]
或者任务完成时：
Final Answer: [最终答案]"""
    
    def run(self, task: str) -> str:
        for step in range(self.max_steps):
            prompt = self._build_prompt(task)
            response = call_llm(prompt)
            
            # 解析响应
            if "Final Answer:" in response:
                return response.split("Final Answer:")[1].strip()
            
            # 提取Thought和Action
            thought = "继续分析..."
            action_str = "search"
            action_params = {"query": task}
            
            # 执行工具
            if action_str in self.tools:
                result = self.tools[action_str].execute(**action_params)
            else:
                result = f"未知工具: {action_str}"
            
            # 记录记忆
            self.memory.append({
                "step": step + 1,
                "thought": thought,
                "action": f"{action_str}({action_params})",
                "observation": result
            })
            
            print(f"Step {step+1}: {thought}")
        
        return "达到最大步数限制"

# 定义工具函数
def search_web(query: str) -> str:
    return f"搜索结果: 关于'{query}'的相关信息..."

def calculate(expression: str) -> str:
    try:
        result = eval(expression)  # 生产环境使用安全计算
        return str(result)
    except:
        return "计算错误"

# 初始化Agent并运行
if __name__ == "__main__":
    tools = [
        Tool("search", search_web, "搜索网络信息"),
        Tool("calc", calculate, "执行数学计算")
    ]
    
    agent = ReActAgent(tools)
    result = agent.run("查询2026年AI Agent市场规模并计算增长率")
    print(f"\n✅ 最终结果: {result}")