# Cowpilot
This is a joke project I made at the request of a friend. It's a Discord bot that mocks Microsoft's Copilot. I may eventually split the personality from the architecture.

## Things to do
- Implement some form of short term memory.
- Split personality from architecture.

## Requirements
Python 3.10+ (Possibly older?)

## Usage
1. Clone the project
2. Make a Python virtual environment and install the requirements with `pip install -r requeirements.txt`
3. Look in `app/` and copy `cfg.py.example` to `cfg.py`
4. Add a Discord bot token to `DSC_TOKEN` and add an API URL to `LLM_API`.

### Compatible LLM APIs
I personally run a local LLM server. I can't say I've used a big language model provider like OpenAI or Claude. Personally I use [`llama.cpp`](https://github.com/ggml-org/llama.cpp)