pull down to refresh

This follows offerings from OpenAI and Claude and has support for multimodal prompting