Quiet Sentinel: let the server watch itself, quietly

Description and use

Quiet Sentinel consolidates server metrics, detects anomalies, and adds AI explanations while keeping the system private.

Problem

Monitoring a server without constant manual babysitting and providing understandable alerts with AI explanations.

Outcome

Quiet Sentinel provides a single, private dashboard that consolidates system metrics, detects anomalies with robust statistical baselines, suppresses duplicate alerts, and adds AI‑generated explanations without exposing the server to external inference services.

Workflow

• Collect CPU, RAM, disk metrics every 30 s
• Maintain rolling history per host/metric
• Compute median and median absolute deviation for baseline
• Detect anomalies exceeding baseline thresholds
• Suppress repeated alerts for 15 min
• If anomaly detected, send data to Qwen model via private SSH tunnel
• Receive explanation and verification step
• Display alert in scrollable feed on localhost dashboard

Evidence

Quiet Sentinel is a small control panel for a server I would rather not babysit. It collects CPU, RAM and disk usage every 30 seconds, learns a separate baseline for each measurement, and brings unusual changes into a scrollable alert feed.

Source

When an anomaly is worth showing, an open-weight Qwen model adds a short explanation and one verification step. The model does not get a shell or permission to change the server.

Source

The app stays private on localhost rather than exposing an unauthenticated monitoring API publicly.

Source

The first live run caught a useful mistake: normal CPU fluctuations near idle looked anomalous against a narrow baseline. I added a 15‑percentage‑point CPU deviation guard...

Source

Eight automated tests cover baseline behavior, independent series, CPU noise filtering, repeat suppression, provider parsing, and AI failures.

Source