Anomaly Detection & KPI Alerting · Documentation
How to Use This Demo
Driving Anomaly Detection & KPI Alerting in the host, the Cloud/OSS inspector, and the operator console — with honest labels.
← Anomaly Detection & KPI Alerting- 1
Signing in
The ecosystem is gated. Sign in as the owner for full access, or as the in-scope viewer for the operator experience. This app is viewer-visible: an operator can read the board, acknowledge alerts, and request an explanation; tuning the detection threshold is owner-only.
- 2
Reading the board
OPERATE shows six KPI tiles with sparklines (cancellation rate uses inverted good/bad colouring — down is good). Below, the alert board lists open anomalies newest-first, tagged spike, drop, or drift, with observed-vs-expected and the z-score. Acknowledge an anomaly to clear it from your view, or click Explain for an AI-assist root-cause narrative.
Owners open CONFIGURE to move a KPI's z-score threshold — lower is more sensitive. Viewers see the same panel locked. The Audit tab records threshold changes and acknowledgements.
- 3
The Cloud/OSS inspector
Explain is the metered step. Flip the mode switch and open "Behind the scenes" to see the trace (gather · explain), the mode badge, and the cost. Cloud runs `claude-haiku-4-5` cost-capped and fail-closed; OSS is recorded at $0. Detection runs the same in both modes — only the explanation is metered. Instrumentation is inspector-only.