Sort Unsorted's destination picker still listed a course whose real
video content sat two levels deep inside its own "display name"
folder (a common release-bundle extraction shape), alongside empty
leftover folders from the extraction that neither of the two earlier
exclusion checks could safely rule out - an empty folder is
indistinguishable from a legitimate freshly-created category. Detect
this case by comparing token similarity between a folder's own name
and its single course child's name: high overlap means they're naming
the same thing (a redundant wrapper), not a deliberate category
holding one course.
Also bake the build's git commit + build time into the Docker image
(git rev-parse at build time, .git deleted again immediately after) so
a push's effect on the running container is actually visible - shown
on the Settings page and in /health - instead of having to guess
whether a webhook rebuild picked up the latest commit.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Add ffmpeg to the Docker image so ffprobe can read each video/audio
file's duration server-side. Results are cached to a small persistent
per-course JSON file keyed by (relative path, size, mtime), so a file
only gets probed once - a container restart or the library scan's
in-memory cache expiring never re-runs ffprobe on unchanged files.
Course cards in the library browser (grid, list, and search) now show
total runtime alongside the media-file count, so you know the time
commitment before opening a course. Durations under a minute are
suppressed rather than showing a noisy "0m".
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
rmsitzandClaude Sonnet 5