Use a line of Python code to connect ChatGPT to WeChat robots:

1. Implementation Principle

Perhaps all intelligent chat robots have the following implementation ideas?

First, obtain the chat content through code (no language limit), then transfer the chat content to the AI ​​background, and finally transfer the obtained AI reply to the chat interface via the original route.

In this process, the difficulties are:

• How to obtain the chat content;

• How to call AI reply;

• How to automatically reply correctly.

In the process of development, you will find that the last item: how to correct every dialogue is the most difficult.

2. Upload the code

Download PyOfficeRobot

pip install PyOfficeRobot

1 line of code to implement WeChat robot ChatGPT

import PyOfficeRobot
PyOfficeRobot.chat.chat_by_gpt(who='程序员', api_key='你的api_key')

3. Parameter description

Only 2 parameters are listed above:

• who='Programmer': required, the person you want to reply intelligently;

• api_key='your api_key': required, your own ChatGPT

In fact, there are a series of parameters inside the method, but our project is mainly used by novices, so I filled them all in If you are a professional developer, you can refer to the instructions below to modify it. (from OpenAI official website)

    completions = openai.Completion.create(
        engine=model_engine,
        prompt=prompt,
        max_tokens=max_tokens,  # 生成结果时的最大 tokens 数。平均一个汉字是 2 个 tokens，text-davinci-003 最多是 4000 个 tokens，也就是 2000 个汉字左右
        n=n,
        stop=stop,
        temperature=temperature,  # 控制结果的随机性，如果希望结果更有差异性 0.9，或者希望有固定结果可以尝试 0.0
        top_p=top_p,  # 一个可用于代替 temperature 的参数，对应机器学习中 nucleus sampling，如果设置 0.1 意味着只考虑构成前 10% 概率质量的 tokens
        frequency_penalty=frequency_penalty,  # 控制字符的重复度，取值为 -2.0 ~ 2.0 之间的数字
        presence_penalty=presence_penalty  # 控制主题的重复度，取值为 -2.0 ~ 2.0 之间的数字
    )

The above is the detailed content of Use a line of Python code to connect ChatGPT to WeChat robots:. For more information, please follow other related articles on the PHP Chinese website!