@icedquinn @kaia think there are some multimodal models you can use with ollama? still want to look into that