Self-contained observability console for nodevitals. Own time-series engine, own PromQL subset, own UI — no Prometheus, no VictoriaMetrics, no Grafana at runtime.
One binary and one Helm chart that scrape nodevitals agents, store the samples,
answer queries, raise alerts, and draw the screens. Install the agent and this
console and hardware observability is finished — there is no second stack to run.
go.mod has no require block, and that is a product decision rather than an
accident. The moment this console needs a Prometheus server, a VictoriaMetrics
cluster, or a Grafana instance to be useful, "one chart, one pod, done" stops
being true — and that claim is the entire reason to choose it over wiring three
exporters to a metrics stack.
The same rule applies to Go libraries: importing prometheus/promql would make
the query language exactly right, and would also drag in hundreds of transitive
modules. make all fails the build if a require line appears.
It is not a general-purpose metrics platform. It scrapes nodevitals, not
your applications. It does not replace VictoriaMetrics or kube-prometheus-stack
for cluster-wide monitoring — those collect from dozens of targets this console
never looks at. Reaching that scope is a separate, later question.
Rebuilt from scratch in August 2026. The previous prototype proved the pieces (time-series storage, scraping, HTTP API, UI) but grew a demo engine that took 44% of the codebase and crowded out the product; this repository keeps the proven design decisions and leaves that structure behind.
See docs/superpowers/specs/ for design specs and
docs/superpowers/plans/ for implementation plans.
MIT