SikamikanikoBG/homelab-monitor v0.22.1

v0.22.1 — patch: hero GPU gauge reflects a full card

on GitHub

Patch release — bug fix, rolled up into the next minor. Shipped quietly: no announcement, but the in-app updater will offer it.

Fixed

• The Overview engine's GPU gauge now reflects VRAM saturation, not just compute util. A GPU with models loaded but no inference running (low util, ~full VRAM) read as a calm low number on the hero "Lab's Engine" gauge — while the GPU tab ("94% full") and the Overview status tile (crit, "94% VRAM") both correctly flagged it as effectively maxed. The gauge now reads the binding constraint — max(util%, VRAM%), capped at 100 — so a memory-pinned card lights up to match the rest of the UI, with both VRAM and % util shown on the line beneath.

Update: docker compose pull && docker compose up -d, or use the dashboard's one-click self-update.