OpenAI ChatGPT-4开发笔记2024-04：Chat之Tool之2：multiple functions

上一篇解决了调用一个函数的问题。这一篇扩展为调用3个。n个自行脑补。

1 设定目标

#1.设定目标
what_i_want_to_know = [{"role": "user", "content": f"汇总3个function的aiXpert的结果"}]

2 自定义function,3个

#2.自定义function,3个
def search_baidu(keyword):
    return f"{keyword}是一个技术博主"

def search_google(keyword):
    return f"{keyword}是一个后端工程师"

def search_bing(keyword):
    return f"{keyword}是一个Python爱好者"

3 接口。自定义function—>ChatGPT

采用Tool的标准写法：

tools = [
    {
        "type": "function",
        "function": {
            "name": "search_baidu",
            "parameters": {
                "type": "object",
                "properties": {
                    "keyword": {
                        "type": "string",
                    }
                },
                "required": ["keyword"],
            },
        }
    },    
    {
        "type": "function",
        "function": {
            "name": "search_google",
            "parameters": {
                "type": "object",
                "properties": {
                    "keyword": {
                        "type": "string",
                    }
                },
                "required": ["keyword"],
            },
        }
    },        
    {
        "type": "function",
        "function": {
            "name": "search_bing",
            "parameters": {
                "type": "object",
                "properties": {
                    "keyword": {
                        "type": "string",
                    }
                },
                "required": ["keyword"],
            },
        }
    }
]

available_functions = { "search_baidu": search_baidu, "search_google": search_google, "search_bing": search_bing } 

4 define function to call ChatGPT

def run_chat(messages):
    response = openai.chat.completions.create(
        model   ="gpt-3.5-turbo-1106",
        messages=messages,
        tools   =tools,
        tool_choice="auto", 
    )
    return response.choices[0].message

5 发起首次请求，告诉gpt要做什么，已经有哪些函数可以调动

first_response = run_chat(what_i_want_to_know)
tool_calls = first_response.tool_calls

6 大结局

# 检查是否需要调用函数
if tool_calls:
        # 解析所有需要调用的函数及参数
        what_i_want_to_know.append(first_response)  # 注意这里要将openai的回复也拼接到消息列表里
        # 将所有函数调用的结果拼接到消息列表里
        for tool_call in tool_calls:
            function_name = tool_call.function.name
            function_to_call = available_functions[function_name]
            function_args = json.loads(tool_call.function.arguments)
            function_response = function_to_call(**function_args)
            what_i_want_to_know.append(
                {
                    "tool_call_id": tool_call.id,
                    "role": "tool",
                    "name": function_name,
                    "content": function_response,
                }
            ) 
        print(run_chat(what_i_want_to_know))

7 参考资料

1. OpenAI的多函数调用（Multiple Function Calling)简介
2. OpenAI API