Kizashi
AI-based failure prediction for network monitoring on Prometheus. KIZASHI (kizashi, "early sign") brings IT-Service-Intelligence-style adaptive thresholds and predictive alerts to a Prometheus stack. It learns what "normal" looks like for every interface and device, flags deviations, and estimates how long until a degrading metric (optical power, error rate, utilization) crosses a limit, so you can act before the link fails. It needs no changes to Prometheus. KIZASHI reads metrics through the Prometheus HTTP API and exposes predictions as ordinary metrics on /metrics. Prometheus scrapes them back, so your existing alert rules, Alertmanager routing, and Grafana dashboards keep working.
More from members
All projectsSee what other members are building and what each project is looking for.
Comments · 0