Release 0.4.7
Bugfixes and code cleanup
# Docker Hub
$ docker pull jareddantis/blanco-bot:0.4.7
# GHCR
$ docker pull ghcr.io/jareddantis-bots/blanco-bot:0.4.7
For users
- Blanco will try to fall back to Deezer first before YouTube when ISRC matching fails, if the Lavalink server supports it
- This should address some instances where Blanco chooses a suboptimal match just because the ISRC isn't matched, like some tracks from Genshin Impact's OST
- The enhanced track-query matching logic now also applies to metadata fallbacks for tracks with non-matching ISRCs, not just direct
/play
invocations- This should address some instances where tracks from enqueued Spotify playlists are matched against the wrong YouTube versions
- Fixed a bug where long queue item names exceed 50 characters
- Fixed a bug where Blanco would scrobble tracks that weren't playing for long enough
For admins and developers
Breaking changes
- Changes to the track-query matching weights
- The naïve weight (i.e.,
difflib.get_close_matches
) now uses a threshold of 90% (from 80%) - Blanco now turns Spotify ranks into exponential weights with a factor of 0.8
- Instead of
[100, 90, 80, ...]
Blanco now uses[100, 80, 64, ...]
- This gives better priority to Spotify's search ranking and lessens the number of ties in confidence values
- Instead of
- The naïve weight (i.e.,
- The
staging
image is now built on pushes to thenext
branch, not pull requests- Images are still published for pull requests under the tag format
pr-N
- The
next
branch will now be restricted to collaborators
- Images are still published for pull requests under the tag format
Other changes
- Linting is now performed automatically on pull requests
- Some pylint checks are disabled (too-few/too-many)
- Blanco's release version is now inserted on build and displayed in the logs