AI Function Calling / Tool Use 完全指南：2026年让大模型真正替你干活

Function Calling（函数调用）是 AI Agent 的核心能力。通过 Function Calling，AI 可以查询实时数据、操作数据库、调用第三方 API——不再是"聊天"，而是真正"替你工作"。本文用 2026 年最新的 GPT-4o 和 Claude 3.5 讲解。

Function Calling 能做什么？

• 🔍 实时查询：查天气、查股票、查汇率
• 🗄️ 数据库操作：查询 CRM、订单系统、用户数据
• 🌐 网页搜索：实时搜索最新信息
• 📤 发送消息：发邮件、发微信、post 到 API
• 🔧 执行操作：创建工单、更新记录、触发流程

GPT-4o Function Calling 示例

import anthropic

client = anthropic.Anthropic(
    api_key="sk-holysheep-xxx",
    base_url="https://api.holysheep.ai/v1"
)

# 定义可用工具
tools = [
    {
        "name": "get_weather",
        "description": "获取指定城市的天气",
        "input_schema": {
            "type": "object",
            "properties": {
                "city": {"type": "string", "description": "城市名称"}
            },
            "required": ["city"]
        }
    }
]

# 发送支持工具的消息
message = client.messages.create(
    model="gpt-4o",
    max_tokens=1024,
    tools=tools,
    messages=[
        {"role": "user", "content": "北京今天天气怎么样？"}
    ]
)

# 处理工具调用
for tool_use in message.tool_calls:
    if tool_use.name == "get_weather":
        city = tool_use.input["city"]
        weather = fetch_weather(city)  # 调用真实天气 API
        print(f"{city}天气：{weather}")

# 输出最终回复
print(message.content)

Claude Function Calling（Tool Use）

import anthropic

client = anthropic.Anthropic(
    api_key="sk-holysheep-xxx",
    base_url="https://api.holysheep.ai/v1"
)

# Claude 使用 tools 字段
message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    tools=[
        {
            "name": "search_database",
            "description": "Search customer database by email",
            "input_schema": {
                "type": "object",
                "properties": {
                    "email": {"type": "string"}
                },
                "required": ["email"]
            }
        }
    ],
    messages=[
        {"role": "user", "content": "帮我查一下 [email protected] 的订单"}
    ]
)

# Claude 会返回 tool_use 结果
for tool_use in message.tool_calls:
    if tool_use.name == "search_database":
        result = query_orders(tool_use.input["email"])
        print(result)

多轮 Tool Use 最佳实践

实际应用中，AI 可能需要连续调用多个工具。以下是多轮对话的框架：

def chat_with_tools(messages):
    while True:
        response = client.messages.create(
            model="gpt-4o",
            max_tokens=1024,
            tools=tools,
            messages=messages
        )

        # 收集所有工具调用
        tool_results = []
        for tool_call in response.tool_calls:
            result = execute_tool(tool_call.name, tool_call.input)
            tool_results.append({
                "tool_call_id": tool_call.id,
                "output": str(result)
            })
            messages.append({
                "role": "user",
                "content": None,
                "tool_call_id": tool_call.id,
                "type": "tool_result"
            })

        if not response.tool_calls:
            # 没有更多工具调用，返回最终回复
            return response.content

        # 添加工具结果到对话
        for tr in tool_results:
            messages.append({
                "role": "user",
                "content": tr["output"],
                "tool_call_id": tr["tool_call_id"]
            })

常见坑与解决方案

一个完整示例：AI 客服助手

# 完整 AI 客服：查订单 + 发邮件
tools = [
    {
        "name": "get_order",
        "description": "根据订单号查询订单状态",
        "input_schema": {
            "type": "object",
            "properties": {"order_id": {"type": "string"}},
            "required": ["order_id"]
        }
    },
    {
        "name": "send_email",
        "description": "发送邮件给客户",
        "input_schema": {
            "type": "object",
            "properties": {
                "to": {"type": "string"},
                "subject": {"type": "string"},
                "body": {"type": "string"}
            },
            "required": ["to", "subject", "body"]
        }
    }
]

# 用户：我的订单什么时候发货？
# AI 调用 get_order(order_id="12345")
# 收到结果：预计3天后发货
# AI 回复：您的订单预计3天后发货，已发送邮件通知您