I have setup the full Grafana stack: Grafana, Loki, Prometheus, Tempo on the Pi before building anything else on top of it. Right now: ~1% CPU, under 50°C, clean idle baseline.

Genuinely curious what others use here for hardware monitoring, and separately for the LLMs/agents themselves. MLflow has some LLM monitoring built in; is that the go-to, or is there something better? Drop what you're running below.

#Grafana #Prometheus #SelfHosted #Observability #MLOps #Homelab