README.md

MocoGPT

MocoGPT is an GPT mock server, which can be used to test your GPT API calls.

• Prompt and Response Generation: Setup response generation based on prompt as you wish.
• Steaming Support: Support streaming response generation.
• OpenAI API Support: Work with OpenAI API.
• Standalone Server: Run MocoGPT as a standalone server.

Usage

MocoGPT can be used as a library, or a standalone server.

As a Library

You can use MocoGPT as a library as following:

from openai import OpenAI
from mocogpt import gpt_server

def test_reply():
    server = gpt_server(12306)
    server.chat.completions.request(prompt="Hi").response(content="How can I assist you?")

    with server:
        client = OpenAI(base_url="http://localhost:12306/v1", api_key="sk-123456789")
        response = client.chat.completions.create(
            model="gpt-4",
            messages=[{"role": "user", "content": "Hi"}]
        )

        assert response.choices[0].message.content == "How can I assist you?"

In this case, we setup gpt_server to listen on port 12306. When prompt is "Hi", response content will be "How can I assist you?".

We use OpenAI library to send a chat request to our mock GPT server. Here we should set base_url to http://localhost:12306/v1 and verify the response.

As a standalone server

MocoGPT can be run as a standalone server.

You can write your own configuration:

{
    "chat.completions": [
      {
        "request": {
          "prompt": "Hi"
        },
        "response": {
          "content": "How can I assist you?"
        }
      }
    ]
}

(config.json)

Then run MocoGPT server:

$ pip install mocogpt[cli]
$ mocogpt start config.json --port 12306

Now, you can send a chat request to http://localhost:12306/v1.