// the find
robusta-dev/robusta
Better Prometheus alerts for Kubernetes - smart grouping, AI enrichment, and automatic remediation
Robusta is a Prometheus alert enrichment layer for Kubernetes — it sits between Alertmanager and your notification destinations, adding pod logs, graphs, and remediation actions to alerts that would otherwise arrive as a bare metric name and a label set. It's for platform/SRE teams who are already running kube-prometheus-stack and are tired of paging oncall with alerts that contain no context. The AI investigation piece is a separate project (HolmesGPT) that integrates optionally.
The smart grouping into Slack threads is the most immediately useful feature — raw Alertmanager notification floods are genuinely painful and this solves a real problem. The playbook system for Kubernetes-native triggers (OOMKills, failing jobs) means you can get actionable alerts for things that Prometheus doesn't natively model well. The breadth of sink integrations (Slack, PagerDuty, Jira, ServiceNow, Kafka, etc.) means it fits into whatever notification pipeline you already have rather than replacing it. Helm-based install with a CLI wizard keeps the initial setup tractable.
The SaaS platform push is aggressive — the README funnels you toward a free account early and often, and the most interesting features (alert timeline, change tracking correlation) appear to live behind the cloud product rather than the OSS install. The rule-based enrichment and the AI investigation are split across two separate projects (Robusta Classic vs HolmesGPT), which means the repo you're looking at is increasingly the 'dumb' half of the product. Custom playbooks are Python functions you deploy as part of the runner image, which is fine in principle but means any enrichment change requires a new container build and rollout. No obvious way to test playbooks locally without a real cluster.