Highlights
📦 Container disk usage that's actually right
The Containers tab now reports each container's true on-disk footprint — writable layer plus its volumes and read-write bind mounts — instead of just the thin writable layer. So model/data stores finally show their real size:
- Ollama: 22 KB → ~74 GB (its model volume)
- Immich server: 0 → ~293 GB (its photo library bind mount)
Mounts shared by several containers (e.g. a common /srv/share) are excluded so they aren't double-counted against each one. Measured in a background thread (never blocks the dashboard), published incrementally, with a generous timeout + last-known-value fallback so a slow cold du of a huge library degrades to "stale" rather than wrong. Falls back to the writable layer when the host root isn't mounted.
🤖 AI Models — broader server coverage
- Now detects WhisperX /
whisper-asr-webservice, which exposes/asr(not/v1/models) and previously never appeared on the tab. - Plus many more model servers: SGLang, NVIDIA Triton, Wyoming (Home-Assistant voice: faster-whisper / Piper / openWakeWord), OpenLLM, LiteLLM, GPUStack, Cortex/Jan, Ramalama, Nexa, LoRAX, OpenedAI-Speech.
🎛️ UI
- The time-range selector (1h/6h/…/All) now shows on every tab (also fixes the AI Models VRAM timelines, which use the range but had no picker).
- The sidebar brand no longer truncates — "HomeLab Monitor" shows in full.
🐳 Docker image sikamikaniko123/homelab-monitor:0.12.0 / :latest is published automatically by the release workflow.