Seems like a really cool project. Lowering the barrier to entry of locally run models. As llamacpp supports a ton of models, I imagine it be easy to adapt this for other models other than the prebuilt ones.