Skip to content

Latest commit

 

History

History
48 lines (30 loc) · 1.77 KB

README.md

File metadata and controls

48 lines (30 loc) · 1.77 KB
Raw

sussy_ducky_bot

Overview

sussy_ducky_bot is a serverless telegram bot written in Rust that interacts with models from OpenRouter. It handles commands from users and processes messages with various AI models.

Note

NEW! Now for three months the bot also uses the free flux[schnell] model from together.ai

Features

  • Serverless: Made for AWS Lambda.
  • Free: Hosted on AWS Lambda, uses free OpenRouter models.
  • Image Handling: Supports vision models.

Setup

  1. Environment Variables: Ensure you have the following environment variables set:

    • TELEGRAM_BOT_TOKEN: Your Telegram bot token
    • OPENROUTER_KEY: API key for OpenRouter
    • TOGETHER_KEY: API key for together.ai

  2. Cargo Lambda: Install the cargo-lambda tool. More information can be found here.

Important

Don't install cargo lambda using cargo install cargo-lambda, as it doesn't support ARM64. Because of this, the brew installation is recommended.

  1. AWS CLI: Ensure you have the AWS CLI installed and configured.

Commands

  • /help & /start - Displays help text
  • /caveman - Uses Llama3.1 8b model from OpenRouter with a custom system prompt
  • /llama - Uses Llama 3.1 70b or 3.2 12b vision model from OpenRouter
  • /lobotomy - Uses Llama 3.2 1b model from OpenRouter (for fun)
  • /flux - Uses the free flux[schnell] model from together.ai

Running the Bot

To run the bot, ensure all environment variables are set, then execute the deploy.sh script. It compiles the project in arm64 and deploys it to AWS Lambda.

Contributing

Feel free to open issues or submit pull requests for improvements or bug fixes.

License

This project is licensed under the MIT License. See LICENSE for details.