See your power bill by the hour — and which GPU is burning it
Three new questions, answered at a glance: when is your lab most expensive, which GPU is doing the work (whoever made it), and is anything down?
🗓️ Busy-hours cost heatmap
A 7×24 grid — local day-of-week × hour — of your lab's typical total draw, shaded by cost per hour once you've set a tariff (by power otherwise). Months of history become one picture of when your rig actually costs you money, with the busiest and quietest slots called out. No setup; it fills in after about a day of samples.
🎮 Any-vendor GPU support — no longer NVIDIA-only
AMD GPUs are read on Linux straight from the kernel's amdgpu interface — no ROCm, no vendor tools — and AMD and Intel GPUs (including integrated) are read on Windows hosts, so the card finally shows up with its name, utilisation and VRAM. Strictly additive: NVIDIA and GPU-less hosts behave exactly as before.
🟢 Built-in uptime monitoring
Watch any HTTP endpoint or TCP port from the same container — know the moment anything stops responding, with a heartbeat strip, 24h/7d uptime %, latency, and smart per-check alerts (anti-flap confirm, recovery quoting the downtime, optional slow-response warning). Nothing else to self-host.
Still one container, pure Python + Flask, no new runtime dependencies. Rolls up everything shipped through next since 0.17.3.
Get it:
curl -fsSLO https://raw.githubusercontent.com/SikamikanikoBG/homelab-monitor/main/docker-compose.yml
docker compose up -dFull changelog → CHANGELOG.md