Depends what you want to achieve exactly, models are not trained on your data but they are able to read images so you can start experimenting already by giving access let's say to our Design system from your codebase (if you have any) and then share screenshots to ask question about it and see how it goes