BeastBot is a Discord bot for our gym group. He has a range of functions, including:
- Providing motivation images (silly AI-generated memes)
- Choosing a random workout and allow for exclusions if you want to skip leg day
- Will react to messages (based on other user's reactions)
- Optional integration with InWorld.ai or OpenAI large language models (LLMs) to create dynamic and flexible in-character responses (with OpenAI option able to interpret user images and generate images of it's own through DALL-E), OR
- Very basic and limited compliment/insult system based on sentiment analysis
Best of all, he will stay out of your way unless you invoke one of his aliases in chat. Be aware that he can be rude and may not be appropriate for users <18.
Feel free to fork and host your own BeastBot. I am currently running him on a Raspberry Pi 2B (running Raspberry Pi OS) with minimal lag.
Firstly, fork and clone this repo onto your hosting platform. Next, you will need to generate a discord bot token - best to follow this guide.
Next, if you wish to integrate an LLM for more natural responses:
- Create an InWorld.ai character (optional - leave
UseInworldAIChatbot = Falseif not). This used to be free provided you don't exceed 5000 API calls per month and is very simple to set up, however they have now implemented a mandatory pricing system. Follow the instructions to generate your character, source the API key and workspace ID (by clicking on the 'More' option for your character). A character sheet is available in this repo if you wish to copy the options used for this project or feel free to make your own. - Create an OpenAI developer account, load money onto your profile and generate an API key.
Take all these settings and create a Settings.py file in the root directory of the project with the following contents:
BotToken = 'BOT_TOKEN_HERE' # Paste token here
AdminID = ['YOUR_DISCORD_ID'] # Collect this from developer mode in Discord itself
# InWorld.AI settings
UseInworldAIChatbot = True # Optional - set to false
## Only need to set the below if linking to InWorld
IW_WORKSPACE = 'workspaces/{WORKSPACE_ID}/characters/{CHARACTER_NAME}'
IW_APIURL = 'https://studio.inworld.ai/v1/' + IW_WORKSPACE + ':simpleSendText'
IW_APIKEYAuth = 'Basic {API_KEY_HERE}'
# ChatGPT API settings
UseChatGPTAPI = True # Optional - set to false
ChatGPT_Key = '{API_KEY_HERE}'You will then need to install Python and pip (usually comes packaged with Python). You will also need the libraries specified in requirements.txt. Run the requirements.txt file in the console:
pip install -r requirements.txt
Next, in python, download some necessary data:
nltk.download('punkt')
nltk.download('wordnet')
nltk.download('averaged_perceptron_tagger')Every time you wish to change the slash commands that BeastBot uses, you will need to run the below to sync the commands to Discord. Note that there is a usage limit on this command so you should only run it once every few hours or it will be ignored.
SyncSlashCommands = await bot.tree.sync()Finally run the code in a terminal (e.g. python BeastBotInitiator.py) and invite the bot to your server! If you see the message Beasty is ready for action! in your terminal, then it's working!
The bot currently supports the use of the InWorld.ai or OpenAI LLM to add conversational responses to BeastBot's other functions.
The testing of ChatGPT is ongoing but is performing well. ChatGPT implementation supports text interaction, image understanding and image generation. See below example.
InWorld.ai supports text interaction but is quite stale and repetitive in responses but performs adequately.
If you encounter a bug or crash, please file an issue with a reproducible example if possible. You may also submit requests to improve the experience through the enhancements tag.
You are welcome to fork and submit a pull request. Ensure you update the requirements.txt file by running the below code after navigating to the root directory:
pipreqs --encoding=utf8 . --force