Grade service vitals (metrics, SLOs, error budgets) into a pass/warn/fail verdict. Use when asked to check service health, gate a release/deploy on error-budget burn, evaluate an SLO, watch for a latency regression, or decide whether a rollout is healthy enough to call "done". Works on JSON/CSV series, a Prometheus/PromQL server, a /metrics endpoint, OTLP exports, Datadog/CloudWatch, or the process's own psutil vitals.
2026-06-23