Is anyone familiar with the specifics of how Circle to Search (linked) works? I was looking for a FOSS alternative to the google circle to search tool and came across it. It seems to describe its implementation in a clear way on the GitHub page, but I have not validated that it does actually operate using that architecture, or that it keeps to its promises about not caching info that is pulled from its third party OCR tools. On android it asks for "advanced permissions" which I presume are needed to take screenshots and share these with the third party OCR service. Since I havent validated how it works though, I wanted to check if anyone else has before I provide it those permissions.