SparkBox 1.6.642: GPU transcoding, Seerr quality profiles, and stalled download detection
Posted by tomspark
I've been working through the telemetry and forum reports, and this release fixes some real pain points — Jellyfin now uses your GPU instead of burning CPU, Seerr quality profiles are locked down so you don't accidentally pull massive files, and downloads that go nowhere get flagged so you can retry them. The release also tackles a few lurking issues: qBittorrent's stop point actually works now (one box had 10 GB of dead seeds since May), Byparr's browser processes no longer pile up and kill your sources, and the checkup alerts got a lot quieter and more honest. Here's what changed in 1.6.642: • Jellyfin hardware transcoding: On any box with an Intel, AMD or NVIDIA GPU, SparkBox now measures what your chip can do and switches transcoding on — plus HDR tone-mapping where supported. The old behavior left the GPU sitting there while Jellyfin converted video on the CPU. A setting you changed yourself is never touched, and the checkup now reports the real state. • Seerr quality profiles: Requests now pull from the curated profile (WEB-1080p / HD Bluray + WEB for HD, the 4K pair for 4K) instead of "Any", which is how a 1080p box could end up with 15 GB files. Existing boxes re-point once on next start; a profile you picked yourself stays put. • Stalled downloads: The Media pipeline card now spots downloads with no connections for hours or the app's own "stalled" flag, instead of counting them as in progress. You get a "Try a different copy" button that removes it, blocks that copy, and searches again. • qBittorrent stop point now really applies: The wiring step was giving up on the login answer and seeding finished downloads forever. Fixed. • Byparr browser cleanup: The health check was running a full solve every 30 seconds and never cleaning up, so boxes went quiet after a day. Now it checks every 15 minutes and reaps properly. • Other fixes: Multi-language releases start in the right language, stuck requests now say why they're stuck, SABnzbd accepts your hostname and Tailscale name, missing sources get flagged in the pipeline, leftover downloads older than a week can be cleaned up, and the checkup RAM warning only fires on real problems. Head to the Updates tab in your dashboard to pull it down. Any issues, post them here in d/sparkbox. --- Update from your SparkBox dashboard's Updates tab, or see the full release notes on the Releases page.