AI0814_jiaoan_public/w1_d3/7_toolcalls5.py

from openai import OpenAI
import os
import json
from dotenv import load_dotenv

load_dotenv()

client = OpenAI(
    api_key=os.environ.get("DEEPSEEK_API_KEY"),
    base_url="https://api.deepseek.com",
)

# 1. 定义一个获取天气的工具
tools = [
    {
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "获取指定城市的天气",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {"type": "string", "description": "城市名，如杭州"},
                },
                "required": ["location"],
            },
        },
    }
]

# 工具实现
def get_weather(location: str) -> str:
    return f"{location}的天气为 8℃"

tools_map = {'get_weather': get_weather}

# ====================== 【新增：工具执行函数】======================
def execute_tool_calls(tool_calls, tools_map: dict) -> list:
    """
    执行模型返回的tool_calls，返回tool角色消息列表，用于追加到messages
    :param tool_calls: response.choices[0].message.tool_calls
    :param tools_map: 工具名字 -> 函数 的映射字典
    :return: list[dict] tool消息数组
    """
    tool_messages = []
    for tool_call in tool_calls:
        func_name = tool_call.function.name
        print(f"调用工具：{func_name}")
        # 解析参数
        args = json.loads(tool_call.function.arguments)
        print(f"参数：{args}")
        # 获取工具函数并执行
        func = tools_map[func_name]
        result = func(**args)
        # 组装tool消息
        tool_msg = {
            "role": "tool",
            "tool_call_id": tool_call.id,
            "content": result
        }
        tool_messages.append(tool_msg)
    return tool_messages
# ==================================================================

# 2. 用户提问
messages = [{"role": "user", "content": "你好,杭州和深圳和北京的天气怎么样"}]

# 第一次调用模型：模型会返回 tool_calls
response = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=messages,
    tools=tools,
)

# 获取模型的回复
assistant_msg = response.choices[0].message
print(assistant_msg)
print(assistant_msg.model_dump_json(indent=2))
print(assistant_msg.content)
print("模型第一次回复：", assistant_msg.content)

# 把assistant消息放进对话上下文
messages.append(assistant_msg)

# 如果存在工具调用，执行工具
if assistant_msg.tool_calls:
    # 调用封装好的函数，拿到tool消息列表
    tool_msg_list = execute_tool_calls(assistant_msg.tool_calls, tools_map)
    # 批量追加tool结果到对话
    messages.extend(tool_msg_list)

    # 第二次请求模型，传入工具返回结果
    response = client.chat.completions.create(
        model="deepseek-v4-pro",
        messages=messages,
        tools=tools,
    )
    print("模型第二次回复：", response.choices[0].message.content)

# 思考题：如何让你的程序 实现自己判断是否还需要再次请求大模型的调用
# 参考思路：当大模型不再需要工具调用的时候就结束大模型请求