Thanks, that's a great idea! I was originally going to go the vision model route because I'd also like people to be able to send instructions to sign in with some credentials (like when visiting the nytimes or something).