Skip to content

v1.52.0

Latest

Choose a tag to compare

@github-actions github-actions released this 18 Nov 19:42
· 1 commit to main since this release
6e9717e

BookOrbit support lands: Bindery can import an existing BookOrbit library, link to a book in BookOrbit, and there is a guide for connecting the two. Auto approve gets a system wide default, External import mode got safer, and there are three database migrations (131 to 133), so take a backup first. Behaviour changes are listed below.

The migrations run on their own at startup and nothing needs doing by hand. 131 makes per account auto approve tri-state, 132 adds a flag on held downloads, and 133 repairs Audiobookshelf edition provenance and can rewrite rows. Auto approve now covers books already in your catalogue, automatic grabs skip dead torrents, Reorganize can write metadata.opf for a whole library, and multi user installs get two ownership fixes. There are two new settings, requests.auto_approve_default and metadata.audnex_region, plus a default for monitoring new books.

Upgrade notes

  • Auto approve now covers books already in your catalogue, which changes what v1.51.1 told you (#3954). v1.51.1 said a request for a book already in the library waits for an admin even for an auto approving account. That is no longer true. If Bindery already has a row for the book but not the file, an auto approving account's request approves itself: the requested format is monitored on that one book and a search starts. That means such an account can monitor and search a book an admin unmonitored by hand, up to the daily auto approve limit (requests.max_pending_per_user, 25 by default, counted per UTC day). A book an admin Excluded or marked Skipped is the exception: its request still waits for a person and the book is not touched.
  • Migration 131 changes how per account auto approve is stored (#3900). Every existing account keeps exactly the setting it has today, as an explicit Always or Never. Only accounts created after the upgrade follow the new system wide default (requests.auto_approve_default, off), so nothing changes until you turn it on. The Users page shows Use default, Always or Never for each requester. Admin and user accounts are not covered by the default.
  • External import mode now parks a download instead of dropping it when the language or content check says no (#3945). If the file's language is not allowed by the author's metadata profile, or the content check is set to block, nothing is placed in the drop folder, the download is parked as blocked and the release is blocklisted. On warn, the default, the file is still dropped and the finding shows in History. Migration 132 adds the flag this needs for downloads held by pair gating, and a download held before the upgrade is checked like an automatic import.
  • Reorganize asks for confirmation before touching External mode files (#3946). The Rename files dialog keeps its button off until you tick a confirmation for formats in External mode, and the apply call answers 409 without it. If you script Reorganize against the API, POST /api/v1/reorganize/apply now answers 409 with the code reorganize_external_mode and the list of formats, unless the body sends "confirmExternalMode": true. It checks every file in the request, so one request that mixes an External format with managed ones is refused whole. Send the managed files in their own request, or confirm.
  • API clients: the users list and the auto approve call changed shape (#3900). GET /api/v1/auth/users now returns autoApproveRequests as true, false or null, where null means the account follows the system default, and a new autoApproveEffective boolean says what a request from that account gets today. PUT /api/v1/auth/users/{id}/auto-approve is now strict. The body must be {"enabled": true}, {"enabled": false} or {"enabled": null}, with null clearing the account's own setting. A missing enabled, a value that is not a bool or null, or any extra field answers 400 instead of being guessed at.
  • Automatic grabs skip a torrent that reports zero seeders (#3953). If every candidate is dead nothing is grabbed and the book stays Wanted. Manual search still lets you grab one by hand. Usenet and torrents whose indexer reports no seeder count behave as before.
  • Reorganize can now write metadata.opf for every book, not only renamed ones (#4008). It only runs when import.write_opf_sidecar is on. It writes a missing sidecar or one Bindery wrote itself, and leaves Calibre, Audiobookshelf and hand edited ones alone. It skips a folder holding files of more than one book, and in such a folder it removes a stale Bindery sidecar left by an earlier run. External mode formats are skipped. With the setting off, nothing changes.
  • Migration 133 repairs Audiobookshelf edition provenance (#3978). It repoints edition rows in abs_provenance at the edition they name. It is idempotent, leaves correct rows alone and does not touch import run history.
  • With multi user ownership on, a non admin can no longer take over a stale failed download that is not theirs (#4027). The grab answers 409, so an admin clears the stale row. A grab into a shared book the caller cannot edit now answers 404 with "That book is shared, so only an admin can grab for it." (a book the caller cannot see at all still answers "book not found"), and bulk search on a shared book reports it as not found per book. Admins, API key calls and single user installs are unchanged.
  • With multi user ownership on, a series is visible only to accounts that can see one of its books (#4025). A hidden series answers 404, the series list total follows, and a series with no books at all is visible to admins only.
  • New setting metadata.audnex_region (#3982). Default us, so nothing changes unless you pick another marketplace.
  • Rolling back to v1.51.1 is not clean, so take a backup before upgrading. The binary has no downgrade guard, so v1.51.1 starts without complaint on the newer database (migrations 131 to 133 stay applied). v1.51.1 reads the old per account auto approve column, so an Always or Never you set on this version is not seen after a rollback, and an account created on this version reads as off. Migration 133 is not undone, which is fine because the ids it repaired are the correct ones. A download that handed off several files, or one with a file name that is not valid UTF-8, now records its sources in import_path in a form (a JSON array, with {"b64":...} entries for the odd names) that v1.51.1 does not read.
  • Helm users: the chart gains opt in extraVolumes, extraVolumeMounts and affinity. They are empty by default and the rendered manifests are unchanged.

Added

  • BookOrbit import (#1927). Bindery can now learn what you already own from a BookOrbit server. Under Settings, BookOrbit, enter the server URL and a read only API key, tick the libraries to read, and add a path remap if BookOrbit and Bindery mount the files at different paths. A preview shows what an import would do without changing anything. The import then matches or creates each author, book and series and registers every ebook and audiobook file where it already sits. Files Bindery cannot see are counted and left unregistered, and running it again creates nothing twice. It only reads: nothing is written to BookOrbit and no file is moved, renamed or deleted. BookOrbit's API is not promised to be stable, so the whole selection is read and checked before anything is written, and a response the importer does not understand stops the run with nothing imported. Setup and known limits are in docs/BookOrbit-Import.md. Thanks to gavinmcfall and gerwitz for the request and for the details of their libraries.
  • Open in BookOrbit link on the book page (#1927). A book that a BookOrbit import matched or created now remembers its BookOrbit id, and its page shows an "Open in BookOrbit" link that opens the book in BookOrbit in a new tab. Only admins see it, since it comes from the BookOrbit settings. Books handed off through External mode but never imported from BookOrbit have no link, and neither does a book whose import match was ambiguous. If your browser reaches BookOrbit at a different address than Bindery does, set the new "BookOrbit address for your browser" field on Settings, BookOrbit. Leave it empty and the link uses the server URL.
  • BookOrbit integration guide (#3956). A new docs page covers three ways to connect Bindery and BookOrbit: Bindery files the library and BookOrbit reads it, External mode with the Book Dock, and importing an existing BookOrbit library. It also covers converting audiobooks in place and sharing a volume on Kubernetes, and says plainly what was tested and what was not.
  • A default for auto approving requests (#3900). There is now a system wide setting, requests.auto_approve_default in Settings, Advanced, that decides whether requests approve themselves for every requester account without a setting of its own. Admin and user accounts are not covered, so requests made through the API key as the first admin still wait for a person, and an admin account set to Always through the API still auto approves. It is off, so nothing changes on upgrade, and accounts that already had auto approve on or off keep exactly what they had (migration 131). On the Users page each requester now picks Use default, Always or Never, and you can see what that account gets today. The daily ceiling still applies. Thanks to jviall for the request.
  • Install wide default for monitoring newly discovered books (#3958). If you curate a few books per author and do not want a refresh pulling in the rest, set Settings, Metadata Profiles, Library Defaults, Default for monitoring newly discovered books to Don't add them (author.default_monitor_new_items). New authors start with that choice, and the Add Author dialog preselects it. Nothing changes unless you set it, existing authors keep their own value, and the Audiobookshelf and Calibre imports and Hardcover list sync still keep their own. One small change applies to everyone: adding an author that matches one already in your library now keeps that author's own value when you do not pick one, instead of quietly resetting it to the follow monitor mode choice.
  • Pick your audnex marketplace (#3982). audnex always looked ASINs up in the US Audible store, so audiobooks sold only elsewhere, like German ones, came back as not found. Settings, Metadata now has an audnex marketplace dropdown (US, Canada, UK, Australia, France, Germany, Japan, Italy, India, Spain), US by default so nothing changes unless you pick one. It takes effect on the next lookup without a restart, and a cached not found from one marketplace is not served for another. Thanks Kyraira for the report.
  • Helm chart: extraVolumes, extraVolumeMounts and affinity (#3955). The chart can now mount an existing claim that another app also mounts, for example one tree holding downloads, a drop folder and the library, and can pin the pod next to that app with pod affinity. All three are empty by default and the rendered manifests are unchanged unless you set them.

Changed

  • Auto approve now covers books already in your catalogue (#3954). A request from an auto approving account for a book Bindery already has a row for, but not the file, used to sit in the queue waiting for an admin. Now it approves itself: the requested format gets monitored on that one book and a search starts, the same as an admin pressing Approve. For existing installs this means requests from auto approving accounts for unmonitored or partly monitored catalogue books no longer wait. This also means an auto approving account can monitor and search a book an admin unmonitored by hand, up to the daily auto approve limit (requests.max_pending_per_user, 25 by default, counted per UTC day). A book an admin Excluded or marked Skipped keeps the request waiting in the queue and the book is not touched, so an admin's call to leave a book out still holds. The Exclude case was added right after the first version, because the web UI has no Skip button and Exclude is how an admin says no. Thanks to flaevers for the report.
  • Edition wrappers no longer become series cards (#3317). Hardcover sometimes files an omnibus, a boxed set, a coloring book or a run of single comic issues as a series of one, and Bindery used to show each one as a series. Those references are now skipped when books are synced, so the book just isn't linked to the fake series and keeps its real one if it has one. Series you already have stay until you merge or leave them, and rebinding a book to an omnibus keeps the real series it already had. A wrapper you own on its own (a complete collection) is no longer linked to the series, and it no longer auto monitors through series mode. Thanks to condensed69 for the report.
  • An Audiobookshelf import and a BookOrbit import cannot run at the same time (#1927). They share one import engine, so the second is refused with a 409 until the first finishes.

Fixed

  • External mode now checks the file before it drops it (#3945). If you use External import mode with a drop folder, the language check and the ebook and audiobook content checks used to be skipped, so a wrong book went straight into the folder your other tool watches, with no warning and nothing in History. They now run first, with the same settings as a normal import. On warn, the default, the file is still dropped and the finding shows on the drop's History entry. On block, or when the language is not allowed, nothing is dropped and the download is parked as blocked with the same message as a normal import. External mode without a drop folder is unchanged.
  • Reorganize now asks before touching files in External import mode (#3946). If ebooks or audiobooks are set to External, another tool owns that layout, and Reorganize used to plan moves for those files anyway. The preview now shows a warning, the Rename files dialog keeps its button off until you tick a confirmation, and the apply call answers 409 unless it is confirmed. Formats in a managed mode, like audiobooks in Copy mode, reorganize as before. The warning names exactly which files are in External mode, like "ebook files" or "ebook files and audiobook files" (#3970), and stays out of the way when nothing in the plan is in External mode.
  • Reorganize now writes metadata.opf for books that are already in the right place (#4008). With the metadata.opf sidecar setting on, Reorganize used to write one only for a book it renamed or moved. It now covers every book, so a whole library ends up with matching sidecars. It only writes a file that is missing or that Bindery wrote itself (new ones carry a bindery:generator marker, and older Bindery sidecars are recognised by their shape), leaves one from Calibre, Audiobookshelf or a hand edit alone, and does not rewrite a file that is already up to date. A folder that holds files of more than one book gets none, since two books would rewrite each other's file every run, and a stale Bindery sidecar left there by an earlier run is removed. A metadata.opf symlink is never read through or replaced. The result shows how many were written and how many skipped. Formats in External import mode are skipped, and with the setting off nothing changes. Thanks to coppit for the report.
  • Accented titles now match releases written without the accent (#3952). A book like "Vallée du silicium" no longer has its best release dropped as not relevant just because the release says "Vallee", and the same goes for accented author names and the other way round. Thanks to n1coh for the report. Nothing stored changes, since dedup and sort keys are unchanged, so no rescan or migration is needed.
  • Automatic grabs no longer pick dead torrents (#3953). Bindery now reads the seeder count your torznab indexer reports. A torrent known to have 0 seeders is skipped by automatic grabs in favour of the next release, and if every candidate is dead nothing is grabbed and the book stays Wanted so a later search can find a seeded copy. A torrent whose indexer does not report seeders is treated as unknown and behaves as before, and so does usenet. Seeders also now break ties between otherwise similar results, but only as a small tie-breaker. The manual search page still lets you grab a zero seeder release by hand. Thanks n1coh for the report.
  • Approving a request now searches for a book that was already wanted (#3961). If a request was approved for a book the library already monitored but had no file for, the request resolved and nothing went looking for it, even with search on add ticked, which was most visible when the author was not monitored. Approving now starts the search whenever the requested format is monitored and not on disk, and a format that is already on disk is still left alone. Thanks to flaevers for the report.
  • Audiobookshelf re-imports no longer record the wrong edition id (#3978). Re-importing an item whose edition already existed could save another edition's id in the import history, so provenance pointed at the wrong row and a rollback could aim at the wrong edition. SQLite's last_insert_rowid() does not reset on a conflict update, so the id was whatever the connection inserted last. Upserts now read the id back with RETURNING id, and the same fix covers the Audiobookshelf run entity and Calibre snapshot records. Migration 133 repoints existing Audiobookshelf edition provenance at the edition it names. Thanks to akouwenberg for the report and the reproduction.
  • A file already handed off in External mode is no longer offered again (#3975). With a drop folder and copy or hardlink placement, importing a file from a folder copies it into the drop folder and leaves the original where it was. Rescanning that folder then listed the original under Matched, already ticked, so one click dropped a second copy. It is now left out while its download is waiting on your import tool, and Show already imported lists it as already handed off, unticked. It comes back as a normal match once the drop fails or is blocked, or you remove the row from the Queue.
  • History now shows which file an External drop placed (#3975). The Folder Imported row for an ebook named only the drop folder, so there was no telling which file was dropped. It now names the file itself, and for an audiobook the folder it was dropped into.
  • Handed off files with odd characters in their names no longer reappear in the folder scan (#4016). If a file you handed to an external import tool had a name that isn't valid UTF-8, like an old Latin-1 name from an archive, Bindery recorded it slightly wrong and the scan offered it again, which could drop a second copy. The exact name is now kept.
  • Pair gating no longer drops unrelated files from a shared download folder (#3985). When a book wanted in both formats was held until its sibling arrived, a Transmission download was remembered by the shared category folder instead of its own files, so releasing it could hand off every other book sitting in that folder. A hold now remembers only what the download owns, and a hold made before this update still releases as it did.
  • Held downloads no longer look handed off or have a fake ETA (#3986). When a book is wanted as both formats and the first one finishes early, History now labels that row "Held for pair" with a short note, instead of "Folder Imported" with the download path. The Queue also stops showing "2400h 00m remaining" on a completed torrent that is waiting; qBittorrent sends a 100 day placeholder there, and finished downloads now show no time left at all.
  • Import from a folder reads release folder names properly (#3974). A folder like "The War of the Worlds - H. G. Wells [m4b]" used to come out with the author "H G", so the suggested metadata search found nothing. Initials now keep their dots, tags such as [epub], (Unabridged), a year in brackets or an ASIN in braces come off the title and author, and a release folder with no author in its name, like "TEST-bare [epub]", no longer has its whole name shown as the author. An author folder that carries a tag, like "Brandon Sanderson [M4B]", still counts as the author. The same name parser also feeds automatic import and the library scan, where the only visible change is that initials keep their dots; matching there works as before.
  • Import previews counted things the import does not (#1927). An Audiobookshelf dry run reported "authors created" as the number of books by authors Bindery did not have yet. It now counts each new author once. A title listed as an ebook and again as an audiobook is counted as one book, its series once, and an author spelled two ways, like Jane Austen and Austen, Jane, as one author, which is how the real import counts them (#3972).
  • No more scary relink message on every book (#3972). "book relink skipped: upstream book match was ambiguous" showed up on books and authors that already had their metadata link, where nothing needed fixing. It now only appears when a book or author has no provider link yet and several equally good matches were found, and it says so in plain words: not linked to a metadata provider, none was picked.
  • BookOrbit settings page polish (#3971). The ticked library and the import history showed "Library 1" until you pressed Load libraries; Bindery now looks the names up when the page opens, and only falls back to "Library 1" when BookOrbit cannot be reached. An old error from Load libraries no longer hangs around after you test the connection, import or save. A wrong host, a refused connection, a timeout, a certificate problem, http and https mixed up, or an unexpected status now gets a short explanation instead of the raw Go error, with the full detail in the server log. The setup help and docs now say "Read only", which is what BookOrbit calls that API key option.
  • Preferred narrators no longer match Japanese names that differ only by a voicing mark (#4003, the #2452 bug class). For example, a preference for かっこう used to also rank がっこう's releases first. Narrator names now fold the same way author names do.
  • Clearer History, request and grab messages (#3999). The History section on a book page now names every event instead of showing a raw name like downloadFolderImported, marks a held drop as held, and shows the reason on failed imports and downloads. The request dialog hint no longer promises that an admin always decides, since an account with auto approve on adds the book on its own. A manual grab that fails now says so next to the results, with the server's message or the status code, instead of doing nothing.
  • No more admin errors on Settings for regular users (#3973). Opening Settings as a non admin no longer asks the server for the indexer, download client, Prowlarr and storage details it is not allowed to see, so the browser console stays clean. Admins see no change.
  • A series page no longer shows an empty book list when loading its books fails (#4002). The error now reaches the page instead of being hidden. Series looked up by provider id also keep their genre override.
  • Grimmory settings now say when the server URL is missing http:// (#4004). Entering grimmory:6060 used to produce a misleading complaint about the scheme; the message now names the corrected value, as the Audiobookshelf settings already did (#2056).
  • Indexer error messages no longer end in a broken character (#4004) when a non-English error page is cut to fit.

Security

  • Grab and bulk search now respect who owns what (#4027). With multi user tenancy on, a regular user can no longer take over a failed download that belongs to someone else or to nobody, grab a release into a shared book they cannot edit, or queue a bulk search on a shared book. Those requests now get the same refusal the rest of the app gives, and admins and single user installs are unchanged.
  • Series now follow the books you can see (#4025). With per user ownership on, a series that only linked to another account's books still appeared in everyone else's series list and opened by id with its title and Hardcover link, even though none of its books were visible. Now an account only sees a series that has at least one of its own books or an unowned shared one, and a series it cannot see answers not found, the same as a series that does not exist. The list total follows the same rule. A series with no books at all is only visible to admins. Admins and installs with ownership off see everything as before.

Removed

  • Telemetry /stats/preview dashboard (#4021). The second, beta telemetry dashboard is gone. It repeated the queries behind the main /stats page, so /stats is now the only public dashboard.