pretty cool I like it. could you support more (small) models in the future, e.g. from the qwen or gemma family?