Prometheus collects metrics from every Spark (node exporter), its GPU (NVIDIA’s DCGM exporter), LiteLLM and the engine. Grafana at grafana.c4e.ai shows them to admins behind the same sign-in, and emails admins when a Spark stops reporting, a disk fills, a GPU runs hot or the models stay down.
Prometheus and Grafana
Monitoring for admins at grafana.c4e.ai.