Skip to main content

@robojs/ai

This plugin morphs your Robo into a sophisticated AI chatbot. By simply mentioning it or replying to its messages on Discord, you initiate a natural conversation - as if you're chatting with a fellow human!

🔑 To bring this magic to life, secure an OpenAI API key. Set it as an environment variable named OPENAI_API_KEY, and your Robo is good to go.

Installation​

To add this amazingness to your Robo, just use this spell (I mean, command):

Terminal
npx robo add @robojs/ai

And there you go! Your Robo, now powered by AI, is ready to have a chat!

New to Robo.js? Start your project with this plugin pre-installed:

Terminal
npx create-robo <project-name> -p @robojs/ai

Your Robo's Personality​

Desire a Robo with a backstory, specific character traits, or a distinctive personality? Tell it a story with the systemMessage config:

// config/plugins/robojs/ai.mjs
export default {
	systemMessage: 'You are Batman, protector of the Weeb City Discord server.'
}

Let's Talk Commands ️​

Command your AI Robo in everyday language! Got a /ban command? Just say, "Robo, ban @naughtyAlien for space spamming!", and it'll happen. And don't worry, your Robo will only listen to folks with the right permissions (roles).

Always ensure your slash commands' permissions are configured correctly, whether you use this or not! 🛡️

💡 By the way... If you ever want to turn this off, just set the commands config field to false.

Special Channels​

You can make special channels where Robo will chat freely, without always being called out. Just use the whitelist config:

export default {
	// Other configurations...
	whitelist: {
		channelIds: ['channelID1', 'channelID2']
	}
}

Insights​

insights is a special config that lets you inject your own custom knowledge into your Robo. It's a great way to make your Robo smarter and more capable!

All you have to do is put files in a folder called documents in your Robo's root directory, and your Robo will automatically learn from them.

- /robo
	- /src
	- /documents
		- /my-first-document.txt
		- /my-second-document.txt

What's more, your Robo will remember what it learns, so it can use it in future conversations! It's like long-term memory for your Robo.

💡 By the way... If you ever want to turn this off, just set the insights config field to false.

Config file​

Here's a quick look at all the settings you can play with:

// config/plugins/robojs/ai.mjs
export default {
	// Model for your AI. You might stick with the default. (string)
	model: 'gpt-3.5-turbo',

	// Tell your Robo a backstory or role. (string)
	systemMessage: 'You are a wise wizard of Webland!',

	// Let Robo use or ignore specific commands. true for all commands, false for no commands, array for only specific commands. (boolean or string array)
	commands: ['ban', 'kick', 'dev logs'],

	// Knowledge injection & long-term memory. (boolean; default: true)
	insights: true,

	// Assistant API delay for OpenAI rate limiting (number; default: 1_000)
	pollDelay: 4_000,

	// If specified, your AI will only respond to messages in these channels. (object with array of string IDs)
	restrict: {
		channelIds: ['channelID3']
	},

	// Special channels where Robo talks freely. (object with array of string IDs)
	whitelist: {
		channelIds: ['channelID1', 'channelID2']
	}
}

JavaScript API​

For those who like to tinker and build, this plugin also exposes a JavaScript API to enhance things even further!

import { AI } from '@robojs/ai'

// Prompt a chat query and obtain the AI's response.
AI.chat()

// Be the puppeteer! Control AI events or alter certain values.
AI.on()

// Disconnect from an AI event.
AI.off()

We've also got some handy AI tools for you to use:

  • selectOne: Match strings semantically from a range of choices.
  • chat: A low-level variant of AIEngine.chat() for direct OpenAI API interactions.

And guess what? With tools like these, other cool plugins like @robojs/ai-voice let your Robo chat in voice channels too!

Voice Capabilities ️​

If you've added the @robojs/ai-voice to your Robo, you're in for a treat! This AI plugin will let your Robo talk in voice channels!

Terminal
npx robo add @robojs/ai-voice

How cool is that? Your Robo can now talk to you in voice channels, just like a real person!

Web API​

Now, if you've added the @robojs/server to your Robo, there's another treat waiting. This AI plugin will unveil a shiny new /api/ai/chat path, letting your Robo have fun chats on websites!

Terminal
npx robo add @robojs/server

Imagine a chat window with your Robo on your favorite webpage. Pretty cool, right?

Custom Models​

You can also use your own custom AI models with this plugin. Just set the model config your custom model instance and you're good to go!

// config/plugins/robojs/ai.mjs
import LlamaModel from '../../../llama.js'

export default {
	// ... other configurations
	model: new LlamaModel()
}

Custom models must extend the BaseEngine class from @robojs/ai. Here's a quick example:

// llama.js
import { BaseEngine } from '@robojs/ai'

export default class LlamaModel extends BaseEngine {
	async chat(query) {
		return `Llama says: ${query}`
	}
}

Warning: The custom model API is still in beta. It may change in the future.

Troubleshooting ️​

If you encounter any issues with this plugin, feel free to reach out to us on Discord. We're always here to help you out!

Common Issues​

  • OpenAI API Key: Ensure you've set your OpenAI API key as an environment variable named OPENAI_API_KEY and set up billing on the OpenAI platform.
  • Permissions: Make sure your Robo has the necessary permissions to read and send messages in the channels you've configured, including the Message Content intent.
  • Rate Limiting: If you're hitting rate limits, try increasing the pollDelay config in your config/plugins/robojs/ai.mjs file.
  • Insights: Ensure your /documents folder is correctly set up and contains valid text files. You may need to explicitly tell your Robo to use the uploaded files in your systemMessage.

More on GitHub​

🔗 GitHub Repository

Explore source code on GitHub.

🔗 API Reference

Check out the entire API.