Fixed
- Metadata refresh retries unmatched items again instead of skipping them forever. v0.9.85 permanently marked failed lookups so refresh converged, but that also stranded items a later lookup improvement could resolve (e.g. v0.9.89's TMDB-id support) — so refresh became an instant no-op and TMDB-tagged items stayed unmatched. Refresh now re-tries all still-heuristic items each run; the metadata cache throttles real API load (cached results instant, failed lookups re-hit TMDB only after 24h), and no-id items return instantly. Speed still comes from keeping the api-row re-classification behind
deep.
Docker images
Pinned to this release:
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.91 # portable (amd64 + arm64, CPU)
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.91-nvenc # x86_64 + NVIDIA GPU (NVENC)
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.91-edge # as above but ffmpeg master
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.91-edge-nvencFloating tags (:latest, :edge, :nvenc, :edge-nvenc) track the most recent release of each lineage.