Requirements:
* llama.cpp or another OpenAI-compatible server
* bash
* curl
* jq

Run llama.cpp server with a command line like
`./llama-server --models-preset ~/models/preset.ini --models-dir ~/models/ --models-max 1`