Releases · 0xzerolight/anki_miner

25 Apr 10:49

github-actions

v2.3.3

4bce37d

v2.3.3 Latest

Latest

Easier updates

Smart Download button. Banner deep-links to the asset matching your install: .deb, .AppImage, Windows installer, or archive. pip/source installs see "View release".
Skip this version. Dismiss a release permanently from the banner. Re-toggling "Check for updates" in Settings clears the skip.
Post-update confirmation. First launch after upgrading shows a dialog with a link to release notes. Crash-safe: new version recorded before dialog opens.
Auto-check setting is now in the Settings tab.
Version comparison upgraded to packaging.Version so prerelease tags (-rc1, .post1) work.

Better mining

Pronouns mined. 彼, 誰, 何, 我々, 貴様 now appear in vocab. Hiragana-only pronouns (これ, それ) stay filtered.
Compound nouns reassembled. 刑務所, 爆発的, 死傷者, 入院中 mine as one word instead of being split and filtered.
Prefix compounds mined. 不可能, 無関心, 非常識, 反社会, 超能力 merge into single words.
Verb-stem nominalizations mined. 言い方, 読み方, やり方 mine as one word.
Expression field uses subtitle surface form. When the dictionary lemma differs from what was on screen (豪腕 vs 剛腕), cards show what you actually saw. ExpressionFurigana matches.
Auto-migration for existing users. Upgrading from 2.3.2 adds 代名詞 to the default allowed_pos so the pronoun fix takes effect. Custom POS lists preserved.

Full Changelog: v2.3.2...v2.3.3

Assets 8

24 Apr 17:40

github-actions

v2.3.2

9764e98

v2.3.2

Fixed #4

Full Changelog: v2.3.1...v2.3.2

Assets 8

23 Apr 20:56

github-actions

v2.3.1

d7f13f0

v2.3.1

Fixed Issue #4

Full Changelog: v2.3.0...v2.3.1

Assets 8

23 Apr 13:17

github-actions

v2.3.0

a67f54c

v2.3.0

Added

YouTube mining: new YouTube tab. Paste a URL, probe the video, mine vocabulary through the same pipeline as file-based sources. Supports manual Japanese subtitles and real native Japanese auto-captions (translated-from-English tracks rejected).
Cancellable YouTube jobs: cancel button kills the yt-dlp and ffmpeg processes and cleans up the temp workspace.
Missing resource warnings: clear startup warning when optional files are absent (JMdict XML, pitch accent CSV, frequency CSV) instead of failing silently later.

Removed

CLI interface: anki_miner and anki_miner_folder commands removed. GUI (anki_miner_gui) is the only entrypoint. Pin to v2.2.x if you relied on the CLI.

Fixed

Wheel install: GUI themes and icons now ship correctly in the wheel.
Temp media cleanup: concurrent runs no longer delete each other's files.

Notes

PyInstaller binaries are 50 to 100 MB larger because yt-dlp bundles ~1600 extractors.

Full Changelog: v2.2.0...v2.3.0

Assets 8

16 Apr 21:43

github-actions

v2.2.0

54e0d59

v2.2.0

Anki Miner v2.2.0

Modular JSON Theme System

New Tokyo Night theme, matching the VS Code version
Light, Dark, and Sakura themes migrated to JSON format
Color variables in common.qss replace duplicated style rules
Dynamic theme discovery - no more hardcoded theme registration

Card & Mining Improvements

Known words filter fixed - words were not being filtered due to mismatched card field names
Pitch accent fields updated to match Lapis card type
Frequency and pitch accent errors now surface clearly instead of failing silently

UI Polish

UI element consistency improved across all themes
Fixed all themes incorrectly rendering with dark backgrounds
Dead code removed (IconProvider)

Internal

Full mypy pass - stale type: ignore comments cleaned up
Theme class rewritten to use JSON discovery instead of static enum mapping

Full Changelog: v2.1.0...v2.2.0

Assets 8

01 Apr 13:12

github-actions

v2.1.0

ca88474

v2.1.0

What's New in v2.1.0

Richer Cards

Pitch accent patterns: Cards now include pitch accent data, sourced from a local pitch accent CSV file
Word frequency rankings: Frequency data is included on cards to help prioritize high-value vocabulary
Multiple dictionary backends: Choose between offline JMdict XML lookups and the online Jisho API, with automatic fallback between providers

Smarter Word Filtering

Known word database: A persistent SQLite cache of words already in your Anki deck, so you never get duplicate cards across sessions
Blacklist / whitelist: Fine-grained control over which words to include or exclude from mining
Sentence deduplication: Automatically skips redundant example sentences across cards
Cross-episode frequency: When batch processing, words that appear across multiple episodes are prioritized first

Export System

CSV/TSV export: Export mined vocabulary with all fields (lemma, reading, sentence, definition, furigana, pitch accent, frequency rank, timestamps) and optional media references
Vocab list export: Export deduplicated word lists in multiple formats: plain text, Takoboto, and JPDB

Analytics Dashboard

Mining statistics: Visualize your mining sessions and track progress over time
Series difficulty rankings: Compare how vocab-dense different anime series are
Comprehension %: See what percentage of words in a given episode you already know
Progress tracking: Monitor your vocabulary growth across sessions

GUI Workflow Improvements

Word curation dialog: Preview and hand-pick words before card creation, with keyboard shortcuts (Space to toggle, Ctrl+A / Ctrl+D for bulk select/deselect)
Cancel button: Gracefully cancel long-running mining operations mid-process
Subtitle viewer: Built-in subtitle display with timing offset adjustment
Drag-and-drop: Drop video/subtitle files directly onto the app to load them
Recent files: Quick access to recently processed file pairs
History and undo: SQLite-backed session history with the ability to resume interrupted batch jobs
Configurable card fields: Map your own Anki note type fields directly in settings, with auto-detection
Auto-update notifications: The app checks GitHub for new releases and notifies you in-app
Issue reporting: One-click link to open a GitHub issue for bug reports