"Use Linux" discord bot

Sends a random ai-generated message that is supposed to tell you to use Linux every 4 hours. I am not responsible for any nonsense it may generate.

You can use the /add-channel command to opt-in, /remove-channel to opt-out.

You can invite the bot to your guild with via the following link: https://discord.com/api/oauth2/authorize?client_id=1102611086756290560&permissions=68608&scope=bot

Message generator

This bot uses a recurrent neural network model which generates text character-by-character. For obvious reasons, you must train the said model before it can be used (see below).

Preparation

• Clone the repository with git clone --depth 1 https://github.com/mnemotechnician/use-linux-discord-bot use-linux and navigate to the directory.
• Run pipenv lock to refresh the dependencies and also ensure you have pipenv in installed and in you $PATH as the generator depends on it.

Training the model

The module message-generator-dl includes a CLI tool that allows you to train the model on your machine. It uses two datasets:

• train-learning.txt contains random phrases meant to make the network learn relationships between words and occasionally phrases. During training, 33% phrases in it are postfixed with a message terminator character, the rest is postfixed with a space.
• train-main - contains actual phrases telling people to use Linux. During training, all messages here are prefixed with a starting token and postfixed with a message terminator.

In order to train the model, you must first compile the module:

./gradlew :message-generator-dl:jar

After that, you can run the cli tool (assuming you have a jdk installed and configured):

java -jar message-generator-dl/builds/libs/message-generator-*.jar

After launching the tool, you need to type a letter to choose the mode:

• Typing 't' starts the training process. You will be prompted a few more times for training configuration. After that, a training process will begin. It may take several hours or more.
• Typing 'g' starts the generation mode, In which you can use the model to generate messages (you can provide custom starting phrases). Obviously, this requires you to train a model first.

During the training the model is saved once per 10 training epochs (1 superepoch). Its files are stored in the directory ~/use-linux/deepl/. You can transfer the contents of this directory to the same folder on a different machine and use it there if you need to.

Hosting the bot

Before you can launch the bot, you need to compile the project:

./gradlew jar

Additionally, you need to create a .env file and provide the token of your discord bot there. The file must be located in the same directory from which you're going to run the bot.

echo "TOKEN=YOUR_DISCORD_BOT_TOKEN_HERE" > .env

After that, you can simply launch the bot:

java -jar app/build/libs/app.jar

In order for the main functionality of the bot to work, you must train the message generator model first. See the "training" section.