/llamacord

A Discord bot, written in Rust, that generates responses using the LLaMA language model.

Primary LanguageRustGNU General Public License v3.0GPL-3.0

llamacord

llamacord logo: a vaguely Discord Clyde-looking llama

A Discord bot, written in Rust, that generates responses using the LLaMA language model.

Built on top of llama-rs.

Setup

Model

  • Obtain the LLaMA weights from a reputable source (like Meta).
  • Convert and quantize them to GGML-q4 format using llama.cpp.

Bot

Discord

  • Create a Discord application and fill it out with your own details.
  • Go to Bot and create a new Bot.
    • Hit Reset Token, and copy the token it gives you somewhere.
  • Go to OAuth2 > URL Generator, select bot, then select Send Messages and Use Slash Commands.
    • Go to the URL it generates, and then invite it to a server of your choice.

Application

  • Install Rust 1.68 or above using rustup.
  • Run cargo run --release to start llamacord. This will auto-generate a configuration file, and then quit.
  • Fill in the configuration file with the required details, including the path to the quantized model.
  • You can then run llamacord to your heart's content.