auto-GPT-ts

my take of auto-GPT in typescript

To Be Added / Fixed

• More memory options
• Text to speech
• Plugins support
• Add commands
• Add a general tokenizer
• Improve user interface

How to run:

• install node
• first create a .env file
• install tha package :

$ npm i

• start the program:

$ npm run start

or:

$ npm run start:continuous

for continues mode.

Install as a package:

$ npm i auto-gpt-ts

import { startCli } from 'auto-gpt-ts';
startCli();

Customizations

import { Agent } from 'auto-gpt-ts'

export class MyAgent extends Agent {
    protected onInteractionLoopEnd(): Promise<void> {
        
    }

    protected onInteractionLoopStart(): Promise<void> {
    }

    protected shouldContinue(): Promise<boolean> {
        
    }
}

register custom commands:

import { CommandDecorator } from 'auto-gpt-ts';


@CommandDecorator({
    name: 'myCommandFunction',
    description: 'my command description',
    signature: '"myParam": string',
    register: true
})
export class MyCommand {
    static async myCommandFunction(myParam: string) {
        console.log(myParam);
    }
}

create custom prompt generator:

import { PromptGenerator } from 'auto-gpt-ts';


export class MyPromptGenerator extends PromptGenerator {
    protected async generatePrompt(): Promise<string> {
        return 'my prompt'
    }
}

a custom permanent memory provider:

import { MemoryProvider, Singleton, addMemoryTypes } from 'auto-gpt-ts';

@Singleton
export class MyMemoryProvider extends MemoryProvider {
    static memoryName: string = 'myMemoryName';
    /// implement abstract methods
}

addMemoryTypes(MyMemoryProvider);

and more.

to configure th config, you can create use the singleton Config class:

import { Config } from 'auto-gpt-ts';


const config = new Config();

config.openaiApiKey = 'my key';

Start a Agent

you can see an example (here)[src/main.ts]