lachameleon/epstein-llm ? reverse-engineered prompt

Build me a simple Python app that lets me chat with the fine tuned Epstein themed LLM locally from the terminal.

I want it to load the provided model, start an interactive prompt, and keep going back and forth until I quit. Make the experience easy to run with a single command like the README shows, and include a small inference helper so I can test the model without having to edit code. If there are any missing pieces, set sensible defaults and make the script work out of the box as much as possible.

Please keep it lightweight and straightforward, with clear messages while it is loading and when it is generating a reply. If you need to check anything, feel free to look up current docs online. Also make sure the code is clean enough that I could extend it later into a nicer chat app.