// Generated from the feature catalog
Occupancy & Behavior Learning
A Home That Pays Attention · Shipped · home
Multi-signal presence correlation and pattern prediction — learns routines without explicit schedules.
The occupancy model correlates multiple presence signals — Frigate face detection, smart lock events, WiFi device presence, Bluetooth proximity — to determine who is home, where they are, and how confident the estimate is. The behavior inferrer learns recurring patterns from timeline history and predicts upcoming routines.
The OccupancyModel ingests PresenceSignal events with confidence weights and TTL decay. It maintains a PersonPresence record for each identified individual. The PatternInferrer scans the SQLite-backed timeline to learn time-of-day, day-of-week, device usage, and guest patterns, then predicts upcoming actions.
Limits and invariants
Section titled “Limits and invariants”Occupancy data stays strictly local — no presence or behavior patterns are ever transmitted externally.
Where this lives
Section titled “Where this lives”halbert_core/halbert_core/home/occupancy.pyhalbert_core/halbert_core/home/behavior.pyhalbert_core/halbert_core/home/timeline.py