The nvidia driver installer now writes only `nvidia` and `nvidia_uvm` into `/etc/modules-load.d/nvidia-vfio.conf`, leaving the vfio-pci modules for `switch_gpu_mode.sh` to add and remove when the user explicitly transitions between LXC and VM mode. A host with residual `/etc/modprobe.d/vfio.conf` entries from an earlier VM passthrough setup no longer has vfio-pci racing nvidia at boot and capturing the GPU by ID.
All three GPU-facing entry points also detect a stale `vfio.conf` entry whose vid:did belongs to a GPU currently bound to a non-vfio driver, and offer to clean it in place. `nvidia_installer.sh` runs the check as a pre-flight before install. `switch_gpu_mode.sh` presents an interactive dialog and rebuilds initramfs. `switch_gpu_mode_direct.sh` performs the same cleanup transparently for the Monitor web variant. Once cleaned, "Add GPU to LXC", the driver install and Switch Mode all proceed without any manual intervention.
The two supporting helpers `_pci_driver_of` and `_clean_vfio_conf_ids` previously local to `add_gpu_lxc.sh` move to `scripts/global/pci_passthrough_helpers.sh`, so every GPU-related script resolves them from a single source without duplication.
- **VM/LXC modal** — PVE tags (dots on list cards, editable pills in modal with click-to-edit) using NVIDIA-style hash colour and SAPC contrast; Status tab redesign (single card, always-visible subsections, Edit button, autostart toggle, blue subsection icons); Backups and Firewall tabs now fill the full modal height with sticky headers/notes; stopped VMs no longer shift the metrics grid; mount-point card brightness unified across breakpoints.
- **Disks modal** — Overview / SMART / History / Schedule tabs adopt the VM/LXC modal size and the mobile icon-only tab pattern; SMART attributes table drops the 15-row cap and gains a sticky "View full SMART report" footer; Print/Save-as-PDF collapses to two icons in the report; loose i18n and layout follow-ups.
- **NVIDIA driver installer (#298)** — version picker cross-checks kernel + NVIDIA's Production/New Feature/Legacy branch classification (scraped from `nvidia.com/en-us/drivers/unix/`) + the PCI Device IDs of every host GPU, with a release-count heuristic to keep superseded production branches selectable while dropping Vulkan-beta ones; Recommended follows same-branch head when a driver is installed, Production Branch head on a fresh install; Hardware card now shows installed alongside available driver version.
- **Custom notifications (#297)** — `event_type: "custom"` accepts `title`/`message` at the root or nested under `data`; defensive strip of stray `[TITLE]`/`[BODY]` markers echoed by the AI enhancer.
- **App tab** — new "Exclude from the LXC updates counter" toggle; the CT's aggregate updates badge now sums OS packages plus registered apps (respecting the flag); Docs page updated; App suggestion no longer treats bare OS helper slugs (alpine/ubuntu/debian…) as installable apps.
- **i18n and copy** — Monitor UI available in EN / ES / DE / FR / IT / PT / SV / SK (thanks @vaso73) surfaced as the first entry in the What's New modal with a link to the contributor's profile; ES cleanup pass (`Historial`, `Velocidad de rotación`, `Consumo actual`, `Ejecutar`, `Eliminar`, `Activar`, `Ver contenido`, `Repuesto disp.`, `Registrar`, `Ocultar`, `Descartar`); redundant "Tip: search any Linux/Proxmox command" line removed from the terminal command search across all locales.
Modal caches now refresh on events only — the periodic prewarmer runs one-shot at startup, mount points split into static/runtime endpoints, and backups get a client 6-hour gate. Updates tab shows post-apply feedback and the script terminal no longer closes the parent modal. Settings adds edit gating on 3 cards with the 3-level contrast rule applied consistently. ES translation batch (~25 fixes) and What's New for 1.2.4.1-beta refreshed.
Backend keeps every guest's modal payload (details, backups, apps, schedule, mount points) warm in-memory and exposes them through a single `/api/vms/modal-cache-all` endpoint. The dashboard hydrates its entire modal cache from that one request on page load, so opening any guest — first click or after coming back later — renders instantly. Replaces ~84 per-guest fetches with 1.
MacRimi
github-actions[bot]
github-actions[bot]
ProxMenuxBot