OpenAI函数调用报错求助及tools/functions参数疑问

错误修正 #

你遇到的'type' is a required property - 'tools.0'错误，是因为使用tools参数时，每个工具对象必须明确指定type字段（当前函数调用场景下固定为"function"），同时原函数定义需要嵌套在function子字段中。另外，建议补充参数的required列表，帮助模型明确必填参数。修正后的完整代码如下：

import openai

# 假设llm_config已提前定义，示例：
# llm_config = {
#     "model_name": "gpt-3.5-turbo-1106",
#     "deployment_name": "your-deployment-id"  # Azure OpenAI需指定该参数
# }

def fetch_weather_with_openai(location, unit='celsius'):
    """
    Fetch weather using OpenAI's ChatCompletion.create with a simulated function call.

    Args:
    location (str): The location for which weather information is requested.
    unit (str): The unit of temperature (celsius or fahrenheit).

    Returns:
    str: The generated response or function call details.
    """

    function_payload = {
        "type": "function",
        "function": {
            "name": "get_current_weather",
            "description": "Get the current weather in a given location",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {
                        "type": "string",
                        "description": "The city and state, e.g. San Francisco, CA"
                    },
                    "unit": {
                        "type": "string",
                        "enum": ["celsius", "fahrenheit"]
                    }
                },
                "required": ["location"]  # 明确必填参数
            }
        }
    }

    # 定义输入消息
    messages = [
        {"role": "user", "content": f"What's the weather like in {location}?"}
    ]

    response = openai.ChatCompletion.create(
        model=llm_config['model_name'],
        deployment_id=llm_config.get('deployment_name'),  # 用get避免未定义时出错
        messages=messages,
        tools=[function_payload],
        tool_choice="auto"  # 让模型自动决定是否调用函数
    )

    response_message = response.choices[0].message
    # 检查模型是否返回函数调用请求
    if response_message.get("tool_calls"):
        function_call = response_message["tool_calls"][0]["function"]
        return f"模型请求调用函数：{function_call['name']}，参数：{function_call['arguments']}"
    else:
        # 模型直接返回自然语言回答
        return response_message.get("content", "无返回内容")

# 示例调用
response = fetch_weather_with_openai("Paris, France")
print(response)

tools与functions参数的差异 #

• functions参数：旧版API参数，仅支持函数调用这一种交互类型，参数值直接是函数元数据的数组。
• tools参数：新版API扩展参数，设计目的是支持更多工具类型（未来可能包含检索、代码执行等）。当前函数调用场景下，每个工具对象必须包含type字段（固定为"function"），并将原函数元数据嵌套在function子字段中。

建议优先使用tools参数，它是API的未来演进方向，兼容性更好。

内容的提问来源于stack exchange，提问作者user3104352