ongrid

Why we included this project

Ongrid takes a different route to incident response: instead of paging a human with another alert, it starts digging on its own. Once it is wired into your Prometheus, Grafana, OpenTelemetry, and Kubernetes setup, an alert turns into an automated sweep across metrics, logs, traces, and topology, with a likely root cause surfaced alongside relevant runbooks pulled from a RAG knowledge base. It can also search code and run fixes remotely, and specialist agents and skills cover distinct domains. You drive it from Slack, Telegram, Lark, or DingTalk, and because it is written in Go and self-hosted, an SRE or platform team can keep the whole thing inside its own infrastructure instead of depending on a hosted vendor. That is what separates it from a thin chat wrapper around your dashboards.