Changelog

Multivariate Anomaly Detection -- detect anomalies across correlated sensor streams (2-100 channels). Cross-channel correlation detection identifies when sensor relationships break down. Per-channel contribution scores show which sensors are anomalous. Azure multivariate API compatible. Streaming baseline accumulation for progressive data collection.

Azure Anomaly Detector API compatibility -- drop-in replacement with identical request and response JSON schema

State-of-the-art time-series AI engine -- superior to ARIMA, Prophet, Isolation Forest, and legacy ML approaches

Regime classification -- automatic labeling of data points into HEALTHY, ACTIVE, TRANSITION, and SHOCK states

Batch and streaming detection endpoints (/detect/entire and /detect/last)

Multi-resolution models -- sensor model (2048-point window) and metrics model (128-point window) with automatic selection

Per-machine fine-tuning on Pro and Enterprise plans for improved detection accuracy

Web dashboard with API key management, usage analytics, and machine monitoring

Free tier with 10,000 data points per month and 1 machine