Changes From Face SDK v1

What Is New

• API v2: clearer answers (every decision with its score and threshold), photo-quality reasons, problem documents with stable error codes, per-key permissions.
• Dashboard: Sandbox, API reference, request log, audit log, people, live view and settings, with an authenticator app at every sign-in.
• Cameras and live video (Prod): RTSP cameras with recognition events, webhooks and live video in the dashboard.
• Thresholds measured per false accept rate, and identification thresholds that follow your gallery size.
• Privacy: encrypted templates, retention settings, audit log, no photos stored.
• One-command install and new licensing.

Your v1 Clients Keep Working

The v1 API is served at /api/v1 with the same paths, headers (x-key-secret, x-face-uid), form fields and answers. Use a v2 API key as x-key-secret. Differences:

• age is always null (age estimation is not offered); gender and expression are null unless turned on.
• recognition_confidence is a similarity score; matching uses the measured thresholds, not a fixed 0.5.
• Enrolment is stricter and refuses duplicates.
• The v1 active-liveness video API answers 410 Gone; use v2 liveness.

New integrations should use v2.

Moving To v2

1. Re-enrol everyone. v2 uses a different recognition model and v1 face data cannot be converted. Enrol people again from their photos with the same IDs. Keep v1 running until this is done.
2. New image and key. The Demo image is arsatech/face-sdk (v1 was arsatech/face-sdk-demo), and v2 needs a v2 licence key.
3. New API keys. Create v2 keys in the dashboard for each system.