pyrra

Pyrra turns a service level objective into working Prometheus alerts. An SLO is a promise like "99% of API requests succeed over two weeks," and doing it right by hand means writing a stack of recording rules and multi-burn-rate alerts. With Pyrra you write a short YAML object with the target, the window and the error and total metrics, and it generates the rules, four burn-rate alerts and a UI that sorts every SLO by remaining error budget. Apache 2.0, free, now part of the prometheus-community org.

It's one binary that serves the UI, the API and the rule generator. On Kubernetes it runs as an operator watching SLO objects; anywhere else it reads YAML files from disk. It works with Thanos too, and can emit generic rules for its bundled Grafana dashboards. There's no paid tier, no cloud and no account.

Solo: skip it, SLOs are overkill for side projects. Small teams with production services on Prometheus: free and worth an afternoon. Larger teams: free, and a clean way to standardize SLOs across services. Sloth is the closest alternative if you only want rule generation without a UI.

The catch: it's only useful on top of Prometheus or Thanos, and alerts only help if someone acts on them. Pyrra does the math; the on-call discipline is yours.