用OpenAI Assistant API构建智能助手的3种模式

我跟你们讲，OpenAI那个Assistant API是真香。不过这玩意儿用起来也讲究点技巧，我琢磨了好一阵子，总结出三种模式，今儿个跟你们好好唠唠。

基础模式：单轮对话

这种模式最简单，就是你问一句我答一句。我们先来看看代码长啥样：

from openai import OpenAI

client = OpenAI()

assistant = client.beta.assistants.create(
    name="Python助手",
    instructions="你是个Python专家，帮助解答Python相关问题。",
    model="gpt-3.5-turbo"
)

thread = client.beta.threads.create()

message = client.beta.threads.messages.create(
    thread_id=thread.id,
    role="user",
    content="Python中如何定义一个函数？"
)

run = client.beta.threads.runs.create(
    thread_id=thread.id,
    assistant_id=assistant.id
)

while run.status != "completed":
    run = client.beta.threads.runs.retrieve(
        thread_id=thread.id,
        run_id=run.id
    )

messages = client.beta.threads.messages.list(thread_id=thread.id)
print(messages.data[0].content[0].text.value)

这段代码先创建了个Assistant，然后建了个Thread，往里面丢了个问题。run一下，等它跑完，最后把回答打印出来。

温馨提示：这里用的是gpt-3.5-turbo模型，如果你想要更高级的，可以换成gpt-4，不过那玩意儿可贵了，小心钱包受不了。

进阶模式：多轮对话

单轮对话太low了，来点高级的，整个多轮对话。看代码：

from openai import OpenAI

client = OpenAI()

assistant = client.beta.assistants.create(
    name="Python助手",
    instructions="你是个Python专家，帮助解答Python相关问题。
 
![]( https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/695a2e7270874a558162b2ddcdcf4843~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1738653788&x-signature=6rHSWOEoBs9CUMsZDGRJiSLSjEk%3D)  
 ",
    model="gpt-3.5-turbo"
)

thread = client.beta.threads.create()

def ask_question(question):
    message = client.beta.threads.messages.create(
        thread_id=thread.id,
        role="user",
        content=question
    )

    run = client.beta.threads.runs.create(
        thread_id=thread.id,
        assistant_id=assistant.id
    )

    while run.status != "completed":
        run = client.beta.threads.runs.retrieve(
            thread_id=thread.id,
            run_id=run.id
        )

    messages = client.beta.threads.messages.list(thread_id=thread.id)
    return messages.data[0].content[0].text.value

print(ask_question("Python中如何定义一个函数？"))
print(ask_question("能给个具体例子吗？"))
print(ask_question("如何给函数添加参数？"))

这段代码把问答过程封装成了一个函数，你可以连着问好几个问题，它会记住前面的对话内容，回答的时候带上上下文。

高级模式：带工具的对话

这种模式最牛逼，能让AI调用一些工具。比如说，你可以让它帮你执行Python代码。看好了啊：

import json
from openai import OpenAI

client = OpenAI()

def run_python_code(code):
    try:
        exec(code)
    except Exception as e:
        return str(e)
    return"代码执行成功"

assistant = client.beta.assistants.create(
    name="Python助手",
    instructions="你是个Python专家，帮助解答Python相关问题并执行代码。
 
![]( https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/ed8de30c9d1f4ffeb6d595fe9d356311~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1738653788&x-signature=qiPCv66c05CHowNaJWb3m2CIfhw%3D)  
 ",
    model="gpt-3.5-turbo",
    tools=[{
        "type": "function",
        "function": {
            "name": "run_python_code",
            "description": "执行Python代码",
            "parameters": {
                "type": "object",
                "properties": {
                    "code": {
                        "type": "string",
                        "description": "要执行的Python代码"
                    }
                },
                "required": ["code"]
            }
        }
    }]
)

thread = client.beta.threads.create()

message = client.beta.threads.messages.create(
    thread_id=thread.id,
    role="user",
    content="帮我写个Python函数，计算斐波那契数列的第n项，然后用n=10测试一下。
 
![]( https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/ccf17943c42d406384e5d635738f3685~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1738653788&x-signature=4D46ys8ep3IJd7Y%2Bh8wkmbBtTLs%3D)  
 
 
![]( https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/f90e0cbfb10e4df79d598accf93a72d6~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1738653788&x-signature=H3qj9bXrHI5ofbhesg2J1NmcoqY%3D)  
 "
)

run = client.beta.threads.runs.create(
    thread_id=thread.id,
    assistant_id=assistant.id
)

while run.status != "completed":
    run = client.beta.threads.runs.retrieve(
        thread_id=thread.id,
        run_id=run.id
    )
    if run.status == "requires_action":
        for tool_call in run.required_action.submit_tool_outputs.tool_calls:
            if tool_call.function.name == "run_python_code":
                code = json.loads(tool_call.function.arguments)["code"]
                output = run_python_code(code)
                client.beta.threads.runs.submit_tool_outputs(
                    thread_id=thread.id,
                    run_id=run.id,
                    tool_outputs=[{
                        "tool_call_id": tool_call.id,
                        "output": output
                    }]
                )

messages = client.beta.threads.messages.list(thread_id=thread.id)
for message in messages.data:
    print(f"{message.role}: {message.content[0].text.value}")

这段代码给Assistant加了个工具，能执行Python代码。你让它写个函数，它不光能给你代码，还能帮你跑一遍，看看结果对不对。

好了，今天就唠到这儿吧。这三种模式你们可以慢慢琢磨，用好了真是开发神器。不过话说回来，这API用起来还是挺烧钱的，你们悠着点用啊，别把钱包搞空了。