ATTENTION - THIS PROJECT HAS BEEN DEPRECATED AND SUPERCEDED BY https://github.com/Meatfucker/metatron2 It will still work if you use the older version of ooba specified below but I highly recommend moving to metatron2.

metatron

A discord.py based machine learning bot. It provides a LLM chatbot via the oobabooga API, stable diffusion generation via the AUTOMATIC1111 API, and a text to speech command via a Bark API located at https://github.com/Meatfucker/bark-apiserver.

I wanted a bot to provide LLM, imagegen, and audiogen stuff, but all of the ones out there were quite large and overcomplicated for what I needed. I wrote this in an attempt to provide myself a nice simple base with which to work with.

To chat with the bot just tag it or reply to something it says. It keeps a separate chat history of 11 question/answer pairs for each user, which is lost on each restart or can be manually cleared by a user by telling the bot "forget". It can also see the contents of links and links to images.

Image generation is handled via the /imagegen command. It provides very basic image functionality. Mandatory negatives are handled via the settings.cfg file. Any negatives in it are applied to all gens and also stripped from prompts, useful for banning unwanted keywords. It also has a reroll button, to make a new gen with the same settings and a new seed, a DM button to dm a gen to yourself, a edit button to edit the current prompt, and a delete button which can only be used by the person who made the gen.

Audio generation is handled via the /speakgen command.

Llava multimodal model support.

/impersonate command allowing for easy one-shot or few-shot prompting of the llm.

INSTALLATION INSTRUCTIONS

REQUIREMENTS

A Working A1111 instance with --api enabled.

---ATTENTION!!!!---

Ooba has broken compatibility with its old format in favor of the OPENAI prompt format. Until I can spend the time to do the rework for it, you must make sure to use a version of ooba from before NOV 13th 2023.

A Working oobabooga instance with --api enabled AND A MODEL ALREADY LOADED. Loading LLMs is somewhat complex and VERY slow so you must manually load a model via the ooba webui first.

Python

Discord Bot Setup

Go to the Discord Developer portal and create a new bot and generate a token for it. Write this token down or else youll have to generate a new one, it only shows you once.

Go to the Bot tab on the Developer portal site and enable Privileged Gateway Intents. You need Presence, Server Members, and Message Content enabled.

Go to the URL Generator on the OAuth2 tab and select the bot scope. Then select these permissions "Read Messages/View Channels, Send Messages, Manage Messages, Attach Files, Read Message History, Use Slash Commands" then use the link to invite the bot to your server. I may have missed one, if something is missing you can enable it later in server permissions

Conda Install (OPTIONAL, Recommended) - This can be skipped if you dont mind if pip installs things globally. This can sometimes cause problems with other ML stuff so I always use conda envs.

Install miniconda if you dont already have conda.

Activate your base conda enviroment

Create a new enviroment conda create -n metatron python

Activate your new environment conda activate metatron

Install - Start here if you dont want conda or have already activated and are in your metatron conda env.

Download the repo git clone https://github.com/Meatfucker/metatron.git

Install the bots requirements pip install -r requirements.txt

Install the nltk tokenizer for English python -c "import nltk; nltk.download('punkt')" URL parsing for the LLM side will not work without this, though image parsing will.

Read and edit settings-example.cfg, Make your required changes and save as settings.cfg

Run the bot, if all goes well itll say it has logged in. python metatron.py

settings.cfg

settings.cfg provides all of the settings for the bot. If the example file has more than one line with the same first value, that means you can have multiple. For example you can have multiple models= lines to list all of your model defaults.

See https://github.com/oobabooga/text-generation-webui/tree/main/api-examples For some info on Oobas API and https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/3734 for info on the A1111 API.