Albums group your whole library by what the index already knows about each photo: your tags, the things in it, where and with what it was taken, and the year. Nothing is read by AI to make them, and there is nothing to set up.
- Tap Albums in the header of the photos screen.
- The albums are loaded again every time the screen opens, and when you swipe down on it, so they always match the index as it is now.
- Albums are the whole library: the current search and filters do not apply to them.
One album per tag you gave your photos.
The 12 words CLIP used most across your photos, never the whole vocabulary.
Cities first, then countries.
One album per camera model, shown with its make.
One album per year the photos were taken in.
An album appears as soon as one photo carries its value. A section with no albums is not shown.
- The cover is the album's 3 newest photos by date taken, stacked and slightly turned, the newest on top. The thumbnails are drawn from the photos on the phone.
- The name shows the first letter of every word capitalised: a tag written my wife shows as My Wife. Only the display changes; the value stored in the index stays exactly as written.
- The number of photos in the album is shown under its name.
Tap an album and the photos grid opens filtered to that one value only: nothing typed, no other filter. The filter shows as a removable pill, like any other; remove it to go back to all your photos.
All the albums come from one request to the phone's index: a /select with q=*:*, rows=0 and a JSON facet, with one terms facet per section (custom_tags, labels limited to 12, city, country, camera_model, year) and, inside each bucket, the three newest photos for its cover. Only counts and photo ids come back; nothing is sent to Opensolr's AI services.