InsightFace embedding
Detects one face and generates a normalized 512-dimensional vector in server memory.
Private vector search
Find visually similar faces
Upload a single portrait. InsightFace converts it into a temporary numerical embedding, then Qdrant compares it with an approved reference gallery.
Upload one clear face
InsightFace creates a temporary facial embedding and Qdrant searches the approved reference gallery.
Private by design: the uploaded photo and query embedding are held only for this request. They are not saved to Cloudinary, PostgreSQL, local disk, or Qdrant.
Your search preview
Your image appears here locally while the one-time search runs.
Qdrant similarity
Runs cosine similarity against reference vectors intentionally indexed by administrators.
No query retention
The uploaded query photo and its embedding are discarded when the request finishes.