chatgpt-agent/README.md

Client for the unofficial ChatGPT API with progressive responses and more.

# ChatGPT Agent

Client for the unofficial ChatGPT API. With progressive responses and more.

[Features](#features) | [Usage](#usage) | [Configuration](#configuration)

## Features

- Progressive responses
- Multiple sessions
- Conversation history
- Automatic token refresh
- Request timeout
- Request retry

## Usage

```ts
// refresh_token is optional
const agent = new Agent(token, refresh_token);

// multiple sessions can be created
const session = agent.session();

const first_conv = session.talk("Hello");
first_conv.on("partial", partial);
first_conv.once("complete", (complete) => {
  console.log("complete (event)", complete);
});
console.log("completed (promise)", await first_conv.response);
first_conv.off("partial", partial);

const second_conv = session.talk("Tell me a joke");
second_conv.on("partial", partial);
second_conv.once("complete", (complete) => {
  console.log("complete (event)", complete);
});
console.log("completed (promise)", await second_conv.response);
second_conv.off("partial", partial);

console.log("history", session.history);
```

`Agent` is the main class for the library. It represents a single user, and handles the token (and refresh token), and can create multiple sessions.

`Session` is a class that represents a single context just like the new ChatGPT tab you opened in the browser. It can be created with `Agent.session()`. It has a `history` property that is an array of messages, and a `talk` method that takes a message and returns a `Conversation` object.

`Conversation` is a class that represents a single conversation. It has a `response` property that is a promise that resolves to the full response, and an `on` method that takes an event name and a callback. You can listen to `partial` events, which are fired when the response is updated, and `complete` events, which are fired when the response is complete. Also, `error` events are fired when an error occurs.

## Configuration

You can configure the backend URL and the request timeout for the agent.

```ts
const agent = new Agent(token, refresh_token, {
  backend: "https://chat.openai.com/backend-api",
  timeout: 60_000,
});
```

The default backend URL is `https://chat.openai.com/backend-api`, and the default timeout is 60 seconds.

You can also configure them using environment variables.

```sh
CHATGPT_BACKEND="https://chat.openai.com/backend-api"
CHATGPT_TIMEOUT=60000
CHATGPT_RETRY=1
```

See [the documentation](https://jacoblincool.github.io/ChatGPT-Agent/) for more information.