Been working on Otis, an open-source ai agent that gives you one minimal experience across local and hosted open-weight models, privacy-focused by design.
On setup it recommends a local model based on the hardware Otis is running on, downloads it and runs it through llama.cpp for you. Ollama, LM Studio and Nvidia PAIR are supported too.
Excited for everyone to try it and all feedback is welcome!