Advanced Configuration
Beyond the basics, configuration splits along the architecture: corridor settings (Hub, CIGS, CBS, MCT) live in the package's environment files, while biometric settings belong to Face Matcher and are managed through its own configuration and Station. Knowing which layer owns a setting is half the work — if it concerns detection, matching, liveness, or streams, it is Face Matcher's; if it concerns units, clearance, events, zones, or retention, it is the corridor's.
Retention tuning
With storage enabled, retention is enforced per data type by a scheduled job:
STORAGE_ENABLED=true
RETENTION_TTL_EVENTS_HOURS=48
RETENTION_TTL_IMAGES_HOURS=24
RETENTION_JOB_INTERVAL_MINUTES=30
Event records and image data (face crops, full frames) expire independently — a common production pattern keeps events for audit slightly longer than images. Size the PostgreSQL volume and the S3 store for your TTLs and traffic, not the defaults.
Zone behavior
CBS raises occupancy violations against configurable per-zone person limits, and derives entry/exit and count events per unit. Tune the limits to the physical corridor — a threshold that suits a wide hall floods a narrow passage with violations. When the optional MCT overlay is deployed, these signals switch source transparently; topics and payloads do not change. MCT itself is configured in the mct/ folder and needs a per-site calibration.
Face Matcher-side configuration
Detection sensitivity, matching thresholds, liveness, stream parameters, and camera management all belong to Face Matcher — the corridor uses it unchanged, so its documentation applies directly:
- Detection and Matching Settings — detection modes, extraction algorithms, the match threshold
- Liveness — passive liveness on matched faces
- Camera Settings Reference — per-camera stream and detection parameters in Station
- Tuning Identification and Performance Tuning — which thresholds matter operationally and how to approach them
- Authentication and HTTPS — securing Station and the Face Matcher APIs