Anomaly Detector agent

anomaly-detector — learns normal patterns from time-series data (Home Assistant sensors and Sinergym) and flags anomalies in real time using statistical (z-score), percentile range, rate-of-change, and absence detection. Works with both real-world HA devices and simulated building data.

Dependencies (installed on first spawn): aiomqtt, numpy.

Spawn

@catalog spawn anomaly-detector

It subscribes to live data, builds baselines over a learning period, then reports anomalies. Pair it with the Time-Series Collector if you want history retained for training.

Commands

@anomaly-detector status       # detection state + baseline readiness
@anomaly-detector report       # recent anomalies
@anomaly-detector baselines    # learned baselines
@anomaly-detector train        # (re)build baselines from history
@anomaly-detector reset

Configuration

Returns anomalies_detected, baselines_ready, detection_active, and last_anomaly.