import openai
from typing import List, Dict, Any
import chromadb   向量数据库

 1. RAG实现：从向量数据库检索相关知识
class RAGEngine:
    def __init__(self):
        self.client = chromadb.Client()
        self.collection = self.client.create_collection("knowledge_base")
    
    def retrieve(self, query: str, top_k: int = 3) -> List[str]:
        """根据用户问题检索最相关的知识片段"""
        results = self.collection.query(query_texts=[query], n_results=top_k)
        return results['documents'][0] if results['documents'] else []
    
    def add_knowledge(self, docs: List[str]):
        """向知识库添加文档"""
        self.collection.add(documents=docs, ids=[f"doc_{i}" for i in range(len(docs))])

 2. Function Calling实现：定义可被AI调用的工具
def get_weather(city: str, date: str) -> Dict[str, Any]:
    """获取指定城市指定日期的天气信息"""
     实际实现中调用天气API
    return {"city": city, "date": date, "temperature": 22, "condition": "晴"}

def send_message(user: str, content: str) -> Dict[str, Any]:
    """向指定用户发送消息"""
     实际实现中调用消息API
    return {"status": "success", "to": user, "content": content}

 工具定义（JSON Schema格式，供模型识别）
tools = [
    {
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "获取指定城市的天气信息",
            "parameters": {
                "type": "object",
                "properties": {
                    "city": {"type": "string", "description": "城市名称"},
                    "date": {"type": "string", "description": "日期，格式YYYY-MM-DD"}
                },
                "required": ["city"]
            }
        }
    },
    {
        "type": "function",
        "function": {
            "name": "send_message",
            "description": "向指定用户发送消息",
            "parameters": {
                "type": "object",
                "properties": {
                    "user": {"type": "string", "description": "接收者"},
                    "content": {"type": "string", "description": "消息内容"}
                },
                "required": ["user", "content"]
            }
        }
    }
]

 3. Agent主循环：整合RAG + Function Calling
class XiaoXiangAgent:
    def __init__(self, api_key: str, model: str = "gpt-4o"):
        openai.api_key = api_key
        self.model = model
        self.rag = RAGEngine()
        self.conversation_history = []
    
    def chat(self, user_input: str) -> str:
         Step 1: RAG检索相关知识
        relevant_knowledge = self.rag.retrieve(user_input)
        context = "\n".join(relevant_knowledge) if relevant_knowledge else ""
        
         Step 2: 构建消息（包含检索到的知识）
        messages = [
            {"role": "system", "content": f"你是香香AI助手。相关知识：{context}"},
            self.conversation_history[-10:],
            {"role": "user", "content": user_input}
        ]
        
         Step 3: 调用LLM，同时提供工具定义
        response = openai.chat.completions.create(
            model=self.model,
            messages=messages,
            tools=tools,
            tool_choice="auto"
        )
        
         Step 4: 处理响应
        message = response.choices[0].message
        
        if message.tool_calls:
             模型要求调用工具，执行工具调用
            for tool_call in message.tool_calls:
                func_name = tool_call.function.name
                args = eval(tool_call.function.arguments)
                if func_name == "get_weather":
                    result = get_weather(args)
                elif func_name == "send_message":
                    result = send_message(args)
                 将工具执行结果回传给模型
                messages.append(message)
                messages.append({
                    "role": "tool",
                    "tool_call_id": tool_call.id,
                    "content": str(result)
                })
             再次调用模型生成最终回复
            final_response = openai.chat.completions.create(
                model=self.model, messages=messages
            )
            reply = final_response.choices[0].message.content
        else:
            reply = message.content
        
         Step 5: 记录对话历史
        self.conversation_history.append({"role": "user", "content": user_input})
        self.conversation_history.append({"role": "assistant", "content": reply})
        return reply

 使用示例
agent = XiaoXiangAgent(api_key="your-api-key")
response = agent.chat("帮我查一下北京明天的天气，然后告诉我同事小王")
print(response)