"AI Monitoring" isn't one feature — it's three that work together: ETA prediction warns before a threshold is crossed, anomaly detection flags a deviation from a resource's own normal, and RCA explains what's actually happening in plain language. This page is the map; the two deep-dive pages below are the detail.
A trend line fit on recent history, projected forward — a short-range "about to tip over" warning before a threshold is crossed.
See Predictive Alerts →A self-learning EMA baseline per resource, per metric — flags a real deviation from what's normal for that specific server, container, or pod.
See Predictive Alerts →Deterministic root-cause explanations for every signature hit, plus a chat assistant that answers questions grounded in live data.
See AI Chat →| Asset type | ETA Prediction | Anomaly Detection | RCA | AI Chat |
|---|---|---|---|---|
| Servers | CPU, RAM, disk | CPU, RAM, disk | Full signature RCA | Yes |
| Websites | Not applicable — uptime, not a trending metric | Not yet | Via underlying server, if self-hosted | Yes |
| Containers | CPU, memory | CPU, memory, restart rate | Full signature RCA | Yes |
| Kubernetes pods | Not yet — categorical metrics only | Restart rate | No log scanning yet | Yes |
Full methodology for ETA prediction and anomaly detection lives on the Predictive Alerts page.
The ETA prediction is a least-squares trend fit projected a few minutes forward — powerful for "is this about to become a problem," useless for "will we need more capacity next quarter." The anomaly detection is a per-resource statistical baseline, not a deep model trained on your whole fleet's behavior. The RCA is deterministic pattern-matching for known signatures, with an LLM stepping in only for the lines that don't match anything known. None of that is a knock — it's what makes each layer explainable and trustworthy, which is the actual point.
Offices in India and the US — reach out and we'll get back to you.