KiFinder now has a Mac app around the matching engine. I can enroll a person by dropping in five to twelve reference photos, choose an album or the preschool's zip files, and review the scan without typing photo identifiers into a command.
The app splits the review into three groups. Confident matches start selected, borderline scores appear under “Worth a look,” and the remaining photos stay out of the main path. I can change every decision before saving. In a group photo, the inspector draws a box around each detected face and lets me choose the intended person. That avoids treating the highest-scoring face as the answer when several people are in the photo.
The kept photos go into a library grouped by person and month, ready to export to Photos. The app still writes the same manifest as the command-line engine, so the UI did not need a second data format. The UI presents decisions the engine already records.
For the screenshots I enrolled myself and used synthetic faces for everyone else. That kept the demo useful without putting a preschool album on a public page.