Might want to do a minimum test first to make sure this works, because the documentation is unclear on whether vision is available with assistants.