Release Notes - 7.12.0

BookBridge speaks Russian, Settings are reorganized and apply without a
restart, and each reader sets their own reading preferences. A book your
KOReader device syncs now finds its Audiobookshelf copy on its own, and a book
you start reading in BookOrbit is matched to its audiobook when the match is
certain.

Audiobooks and ebooks match more reliably across Suggestions, Up Next and
Continue Reading, Add all exact only takes matches nobody needs to check,
and KOReader positions are better protected from out-of-date devices.

What's New

• Russian interface (#516). Each reader picks a language on their Account
  page; the Library, Settings, buttons, dialogs and statistics follow it, and
  anything not yet translated stays in English. Contributed by
  @chuzhba1-art.

• KOReader books find their Audiobookshelf copy (#576). With Sync with
  Audiobookshelf's own ebook reader on, a new EPUB that is the same file as an
  Audiobookshelf book's ebook is linked to that book and kept in step with
  Audiobookshelf's reader, with no matching by hand.

• Settings, reorganized. The sidebar is grouped by purpose (Services,
  Reading & Sync, Library & Matching, Alignment, AI, Server, Maintenance, Users),
  settings are described in plain words, search covers every tab, and
  Different from defaults lists what you've changed. Most saves apply
  immediately; a setting marked restart still restarts BookBridge, and the
  Save bar says so first.

• One setting per choice. Polling, highlight sync, the "Up Next" shelf and
  the tracker posting wait each apply to every app you use. Find shelves
  lists collections from every library app. More options that used to be
  environment variables only are now in Settings.

• Per-reader reading preferences. Under My Integrations → Reading
  preferences, each reader can turn highlight, status and History sharing,
  out-of-date device protection, finish-everywhere, Audiobookshelf reading
  checkpoints, Storyteller listening time and the series view on or off for
  themselves. Everything starts at Server default.

• Continue Reading auto-match. With Auto-match books you start reading in
  BookOrbit on, a book whose audiobook scores 95 or higher is linked straight
  away; weaker matches still wait in Suggestions.

Fixed

• Matching is more reliable. Letter case, "(Unabridged)", accents, "&" or
  "and", a leading "The", a missing author and subtitles such as "Stories" no
  longer lower a correct match, while other volumes, similar titles by other
  authors and reordered titles are no longer offered as sure matches. AI
  re-ranking keeps exact matches at 100, and a tie is left for you to choose.
• Add all exact is safe to press. It skips ties, other editions, unconfirmed
  authors, same-title volumes and loose files on disk, rechecks on the server,
  and each skipped card says why.
• Suggestions page: progress shows until a scan finishes (including the AI
  step), a second scan request joins the running one, suggestions from listening
  progress skip other authors, and the wrong volume of a "Series: Book N" title
  or a multi-series Audiobookshelf book is no longer offered (#563, #564).
  Contributed by @chemicalsno.
• Up Next and Continue Reading: Review opens with the right books selected,
  approving another book no longer removes the shelved one, and an audiobook you
  just added is found.
• KOReader positions are better protected. A device replaying an old
  position no longer rewinds your other apps, pages read right after a sync are
  kept, a removed book no longer returns on its own, and books keep their own
  KOReader hashes.
• Multi-user: a book found from a reader's KOReader device belongs to that
  reader and is found with their own logins (#575).
• BookOrbit: a page you just turned to is no longer overwritten (BookOrbit
  3.6+), a brief download hiccup no longer forces slow Whisper transcription, and
  matching a BookOrbit or Grimmory audiobook links Hardcover and StoryGraph.
• Listening: Storyteller listening shows as one Audiobookshelf session per
  sitting (#566), and the stats page's listening books add up to its total with
  Audiobookshelf (#560). Contributed by @chemicalsno.
• Alignment: the content-match guard accepts correct pairings scored by
  OpenAI embeddings (#556), and the English-only CTC aligner recognizes other
  alphabets (#569). Contributed by @chuzhba1-art.
• Startup: a DATA_DIR other than /data works (#573), and the "no ebook
  source" warning names your real books folder (#574).

Upgrading

Back up your /data directory, then pull the new image and restart:

docker compose pull && docker compose up -d

There are no database changes in this release. If you build your own image,
rebuild it: translations are now compiled into the image. BridgeSync is
unchanged (0.9.10).

Operational Notes

• Run Full Refresh on the Suggestions page once after updating. Suggestions
  found before the update are marked Manual choice until they're rescanned
  with the new checks.
• DATA_DIR is set only in the environment (for example in
  docker-compose.yml). Settings shows it but no longer offers to change it.
• Your settings carried over. If an option seems to have moved, use the
  search box at the top of Settings. Reading preferences start at Server
  default, so nothing changes for a reader until they override one.
• Books a reader's KOReader device added to the admin's dashboard before this
  update stay there. Remove them from the admin's dashboard; the reader keeps
  the book and its progress.
• AI re-ranking needs Ollama's embedding model. If the log shows
  Ollama /api/embeddings returned 404, pull the model named in Embedding
  Model (by default ollama pull nomic-embed-text); until then re-ranking is
  skipped.

Thanks to @chemicalsno,
@chuzhba1-art, and everyone who contributed to this release.