Payload-free catalog refresh
Research whether Jellyfin can preserve full remote namespace visibility while avoiding media-payload reads during routine VFS catalog refresh.
Project story
Useful work, closer to the media.
Slipmat Companion exists to move expensive, repetitive or browser-constrained work nearer to Jellyfin without moving product authority there. The project grows by adding optional capabilities that can disappear without making Slipmat stop being Slipmat.
History
Rapid separation and hardening.
The milestones below are drawn from published releases and merged repository work, not a retrospective roadmap.
Standalone Companion
Audio Gateway shipped as its own Jellyfin plugin with self-contained storage and integrated FFmpeg analysis rather than a separate analyzer service.
Server-side capability layer
The plugin added bounded directory and XMLTV acquisition, Live TV publication, remote-media relay, and a provider-neutral rclone adapter.
Bounded cloud projection and richer metadata
Cloud media moved to read-only rclone VFS projection with bounded cache while Companion added artwork palettes and stronger provider diagnostics.
A public explanation surface
The repository gained a human-facing GitHub Pages site alongside the machine-consumed Jellyfin repository manifest.
One simpler presentation toolchain
The presentation site moved to Vite 8 with semantic multi-page HTML, native browser behavior, and progressive View Transitions, removing Astro from the build path.
North star
Accelerate more. Own less.
Companion is not another media server or playback engine. It is an optional capability layer for work that benefits from server proximity: analysis, batching, protected acquisition bridges, provenance-rich metadata, bounded cloud projection and diagnostics.
- Keep playback authority in Slipmat. Companion can prepare evidence and sources, but it does not own the queue, transport clock, renderer or Playback Rail.
- Keep media authority in Jellyfin and source providers. Server-side acceleration must not create a second canonical catalog or silently take ownership of remote media.
- Make absence ordinary. If Companion is missing, unhealthy or lacks an optional module, Slipmat should degrade to canonical fallback paths rather than fail as a product.
- Prefer measured, provenance-bearing evidence. Analysis and metadata should be reproducible and attributable instead of turning server inference into hidden policy.
Open ambitions
Current research, not shipping claims.
The next useful work is tracked in GitHub rather than promised here.
Bound reproducible artifacts
Add independent retention for regenerable analysis artifacts without coupling it to media-cache eviction or deleting authoritative media/catalog state.
Complete managed VFS migration
Continue hardening reversible local-to-cloud migration and mount lifecycle without weakening fail-closed ownership and rollback guarantees.
Follow the implementation.
For exact current behavior, use the repository and release history as authority. This page explains the trajectory; it does not replace technical documentation or shipping evidence.