Self-hosted face and object recognition — a clean HTTP API for other apps and a full human curation UI in one container. Enroll people by name, match them across photos, review uncertain matches, and keep everything on your LAN.
Docker quickstart, first-run setup, native Python run, GPU configuration.
Review queue, suggested people (face clustering), manual face tagging, object detection models, and environments.
Every endpoint with curl examples — detect, enroll, identify, verify, test, bulk, webhooks, merge, reprocess, search, filtering, export and import.
Correlating your own records with external_ref, staying in sync via the change
feed, and batch operations.
Every live setting — face thresholds, GPU, image quality, crop padding, YOLO vocabulary — with defaults and descriptions.
GPU not detected, CUDA library errors, face model not loading, macOS segfault.
Running tests, linting, and cutting a release.
http://localhost:8100/docs