Paginate Supabase select queries in the storage stats API to avoid
the 1000 row default limit. Align StorageManager's formatFileSize
with useQuotaStats calculation so quota values display consistently.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add "Full sync all annotations" menu item that pushes and pulls all
annotations regardless of the last sync timestamp, enabling users to
sync old highlights that were created before the plugin was installed.
Changes:
- Add full_sync parameter to push/pull that bypasses timestamp filter
and uses since=0 for pulling all server annotations
- Deduplicate by annotation ID alongside position-based dedup
- Store server ID on pulled annotations and reuse it when pushing
- Parse ISO 8601 timestamps from server to preserve original
created/updated dates instead of using current time
- Resolve KOReader page numbers from xpointers via getPageFromXPointer
- Resolve Readest page numbers from CFI via getCFIProgress on pull
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(library): always sort series books by index ascending, closes#3709
When grouping by series, the sort direction (asc/desc) was applied to
within-series book sorting, causing books to appear in reverse series
index order when the library was sorted descending. Now series index
sort is always ascending (1, 2, 3…) regardless of the global sort
direction. Also sort series groups by name when "Sort by Series" is
selected instead of falling back to updatedAt.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(library): skip merge sort when inside a series group
The allItems.sort at the end of sortedBookshelfItems was re-sorting
books using the regular bookSorter (e.g. Date Read), which overrode
the within-group sort that correctly prioritized series index. Now
when inside a group, the already-sorted books are returned directly.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add ~290 new unit tests covering utils (lru, diff, css, a11y, walk,
usage, txt-worker), services (EdgeTTSClient, transformers), and
suppress noisy console output across all test files. Rename 27
camelCase test files to dash-case for consistency.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a new TTS tab in settings with media metadata update frequency control
(sentence/paragraph/chapter) to reduce Bluetooth notification spam, and move
TTS highlight settings from Color tab to the new TTS tab. Also add a highlight
opacity setting with live preview in the Color tab.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When TOC entries use fragment-suffixed hrefs (e.g. ch01.xhtml#ch01),
the sectionsMap lookup matched subitems with cfi: undefined instead of
parent sections with valid CFIs. This caused findTocItemBS to map every
annotation to the last TOC entry. Now skip subitems lacking a CFI and
fall back to the base section.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two issues caused embedded EPUB fonts to fail:
1. Adobe font deobfuscation (http://ns.adobe.com/pdf/enc#RC) derived the
XOR key from the wrong UUID. getUUID() picked the first UUID across all
dc:identifier elements (e.g. Calibre's internal UUID) instead of the
book's unique-identifier. Also fixed getElementById not working on
DOMParser-parsed XML (no DTD), and made UUID regex case-insensitive.
2. transformStylesheet replaced generic font families (sans-serif, serif,
monospace) with CSS var() references without fallback values. In
fixed-layout iframes where setStyles doesn't inject the variables,
the undefined var() invalidated the entire font-family declaration
per CSS spec, causing all fonts to fall back to system defaults.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(shortcuts): change bookmark shortcut from Ctrl+D to Ctrl+B, closes#3675
Ctrl+D was bound to both Toggle Bookmark (General) and Dictionary
Lookup (Selection). When text was selected and Ctrl+D pressed, both
actions fired — opening the dictionary AND adding an unwanted bookmark.
Changed bookmark to Ctrl+B/Cmd+B to resolve the conflict. Added a
unit test that detects identical keybinding lists across actions to
prevent this class of bug in the future.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* i18n: mark shortcut section names for translation
Wrap SHORTCUT_SECTIONS values with stubTranslation() so i18next-scanner
picks them up. Translate the 4 new keys (General, Text to Speech, Zoom,
Window) across all 29 locales.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a keyboard shortcuts help dialog toggled with '?' that displays all
shortcuts grouped by section with platform-appropriate key rendering.
Restructure DEFAULT_SHORTCUTS to include i18n descriptions and section
metadata. Translate 37 new keys across 29 locales.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add bidirectional annotation/highlight sync between KOReader and Readest:
- Add xpointer0/xpointer1 fields to BookNote and DBBookNote types for
KOReader XPointer positions alongside Readest's CFI format
- Extend transform layer to pass through xpointer fields to/from DB
- Convert CFI→XPointer on push and XPointer→CFI on pull in useNotesSync,
discarding notes that fail conversion
- Support KOReader's text()[K].N indexed text node format in xcfi.ts for
paragraphs with inline elements (e.g. <a> page anchors)
- Generate KOReader-compatible XPointers: text().N for single text nodes,
text()[K].N only when multiple direct text nodes exist
- Skip cfi-inert elements (injected by Readest at runtime) in XPointer
path building and resolution
- Map highlight colors between KOReader and Readest color systems
- Implement KOReader plugin annotation push/pull with deterministic IDs,
auto-sync on document open/close, and UIManager refresh on pull
- Refactor koplugin into focused modules: syncauth, syncconfig,
syncannotations, selfupdate
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(reader): import Foliate annotations on book open (Linux only), closes#2180
Automatically imports annotations, bookmarks, and reading progress from
Foliate's data files when opening a book on Linux. Uses foliateImportedAt
flag to prevent re-importing on subsequent opens.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* refactor(annotation): move Foliate import into annotation provider pattern
Restructure annotation import as a multi-provider service following the
translators pattern. Foliate becomes a provider under
services/annotation/providers/, making it easy to add more import sources.
Each provider controls its own availability check and skip-if-imported guard.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a "Refresh Metadata" option under Advanced Settings that re-parses
book files to update metadata (series info, language, etc.) for books
imported before these fields were supported.
- Add refreshBookMetadata() in bookService that opens a book file,
re-parses with DocumentLoader, and updates metadata/metaHash/series
without touching updatedAt or user-edited fields
- Add menu item in SettingsMenu with progress display
- Add unit tests for single-book metadata refresh
- Translate new i18n strings across all 29 locales
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Three bugs in createChapterRegexps for English txt-to-epub conversion:
1. numberPattern had a capturing group (\d+|...) that leaked single
digits into split() results instead of chapter titles. Changed to
non-capturing (?:\d+|...).
2. combinedPattern wrapped the heading in (?:...) so split() consumed
the title without returning it. Changed to (...) capturing group,
matching the Chinese regex structure.
3. The prefix (?:^|\n|\s) allowed matching Chapter/Section after any
whitespace, causing mid-sentence false positives. Changed to
(?:^|\n) for line-start only.
Added a second-tier regex for bare numbered headings (e.g. "1.1The
Elements of Programming", "1Building Abstractions") commonly found
in academic texts. Dotted numbers allow optional space before title;
single bare digits require immediate title to avoid footnote matches.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Huang Xin