I've #OpenSourced my custom #Prometheus exporter for NVIDIA GPU telemetry via nvidia-smi.

I built it because I wanted straightforward GPU #observability for homelab and hobby infrastructure (especially for local AI inference), without the operational overhead of DCGM or Kubernetes-oriented tooling.

It also reflects a broader principle I care about: observability should not stop at production. My own infrastructure should be measurable and understandable too.

https://github.com/attilagyorffy/prometheus-exporter-nvidia-gpu