Detect the host's GPU vendor(s) via PCI sysfs (no lspci dependency) and
append vendor-matched encoding-package steps to the install plan:
Intel -> intel-media-driver + vpl-gpu-rt + intel-compute-runtime,
AMD -> mesa + libva-mesa-driver. NVIDIA NVENC ships with the driver's
own userspace, so nothing is added when a driver is loaded, and a
warning step (never an automatic kernel-driver install) when it isn't.
Without these packages Jellyfin installs fine and direct-plays fine,
but any client that needs a transcode fails instantly with the opaque
"FFmpeg exited with code 251" / "fatal playback error" pair.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Drives on a headless Jellyfin box ran hot 24/7: NAS-class drives (WD
Red, IronWolf) ship with no idle timer at all. v0.1.2 stopped tuistream
from keeping them awake; this makes them actually go to sleep.
Spin-down is a DEFAULT, not a feature: when the inventory shows
spinning media drives, the TUI silently installs/syncs the mechanism
and flashes what it did. [s] on Setup is the opt-OUT, remembered via
/etc/tuistream/spindown-off so the default never fights the user.
The mechanism is a tiny systemd service (tuistream --spindown-watch)
that samples /proc/diskstats and issues `hdparm -y` to any target drive
idle past 3 minutes. We deliberately do NOT use the drive's own standby
timer (hdparm -S): the 10TB helium WD Reds advertise it and then ignore
it — verified on real hardware (fresh -S 36, zero I/O, four minutes,
still active/idle). Forcing standby from outside works on everything.
An rc shipped the udev+hdparm -S approach; the plans clean its rule up.
Safety: targets are spinning (sysfs rotational), top-level, real disks,
never a system disk — same classifier the rest of Setup trusts. The
watcher's own probes can't disturb drives: /proc/diskstats and sysfs are
kernel memory, and hdparm -C (CHECK POWER MODE) doesn't wake or reset
anything. The unit's ExecStart points at the running binary and the
auto-sync rewrites it when the content goes stale (e.g. binary moved
from a dev path to /usr/local/bin).
hdparm joins the dependency pre-flight. Verified end-to-end on moviebox
(2x WD Red 10TB SATA + IronWolf 10TB USB): auto-applied on launch, all
three drives reached standby after 3 idle minutes, Monitor shows them
as "asleep" without waking them.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Three-tab Bubble Tea TUI (Setup / Manage / Monitor) for setting up and
running a headless Jellyfin server over SSH:
- Setup: install/uninstall Jellyfin, add media drives (single or btrfs
RAID pool), import an existing detached btrfs pool non-destructively,
firewall, move Jellyfin storage onto a media drive
- Boot-drive-safe drive classifier (walks LUKS/LVM/RAID to the physical
disk; never offers a disk hosting /, /boot or swap)
- Keep-existing-filesystem path previews on-disk content, skips starter
folders when content exists, and grants Jellyfin recursive read access
- Manage: copy from external drives, mount/eject, rename/delete
- Monitor: per-core CPU, load/mem/uptime, btrfs error counters, SMART
Single static binary; installs to /usr/local/bin so sudo can find it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
