lgidon/Onitama-agent ? reverse-engineered prompt

Build me a terminal version of Onitama that I can play against a local AI, with both a simple command line mode and a nicer live terminal dashboard.

I want the game to let me play human vs AI by default, but also support human vs human and AI vs AI. The AI should connect to a local chat endpoint, make a legal move, and return a short explanation of why it chose that move. If the AI response is broken or unclear, please make it fail gracefully and keep the game playable.

Please include a clean way to configure the server URL, model name, timeout, and logging in a config file. I also want good board rendering in the terminal so the pieces and cards stay easy to read, plus a separate debug log file so I can see what the AI and game engine are doing without messing up the UI. If you need to check current docs for the terminal UI library, feel free to look them up online.