dilidin2/lexis — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a Python Twitch chat bot called Lexis that connects to my channel and replies with a local AI model. I want it to work with llama.cpp or any local server that has an OpenAI compatible chat API.

The bot should answer when someone types a command like !bot hello, or when they mention the bot name like @lexis. It should remember a small amount of recent chat so replies feel connected, but not save chat history forever. Add a cooldown and a simple queue so it does not spam or talk over itself when several people trigger it.

Make it easy to configure the bot name, channel name, command prefix, cooldown, model server host and port, max response length, and chat memory size. I also want a text file where I can change the bot personality, plus a place to add custom command prompt templates.

Include Twitch OAuth setup, env examples, logging, and clear setup instructions. Look up current Twitch docs online if needed.

Want more depth? Deep Reverse