recite_quran 1.0.5

Real-time on-device Quran recitation tracking, dynamic DTW word alignment, and deterministic Tajweed rule verification engine for Flutter.

1.0.5

• FIX (Static Analysis): Fixed flow control block syntax in ErrorExplainer (curly_braces_in_flow_control_structures), conditional assignment and super parameter lints across engine and models.
• FIX (Web JS Interop): Updated Web runtime type check to res.isA<JSBoolean>() conforming to modern Dart 3 JS interop.
• DOC: Updated all documentation guide links in README.md to absolute repository references for full link resolution on pub.dev.
• CHORE: Package quality, static analysis and pana score enhancements.

1.0.4

• FEAT (Dictation & Loss-of-Tracking Recovery): Added enableAutoReanchor and enableEarlyMatching to TrackerConfig with Multi-Stage Window Probing and the Anti-Ambiguity Guard ($\Delta\text{dist} < 3$) preventing false jumps across repeated refrains (Surah Ar-Rahman).
• FEAT (Tajweed Engine): Complete overhaul of the acoustic verification engine with a strict compile-time AcousticThreshold matrix across all recitation speeds (Tahqiq/Tartil, Tadweer, Hadr).
• FEAT (Tajweed): Multi-Madd span matching support (e.g. Lazem + Aared in "الضَّآلِّينَ") and legitimate Aared Madd Qasr/Tawassut/Tool allowance.
• FEAT (Tajweed): Pro-rated character duration distribution for multi-phoneme ASR CTC spikes, zero-duration safety shields, and glyph equivalence matching in PhoneticCostEngine.
• FEAT (UI & DX): Enhanced ReciterError with localized message & advice helpers (messageAr, messageEn, adviceAr, adviceEn), clean JSON serialization (toMap/fromMap), while preserving full raw telemetry for custom UIs.
• FEAT (UI): Added built-in showTajweedErrorSheet and TajweedErrorSheet widget for 1-line drop-in Islamic modal bottom sheets with automatic dark/light theme adaptation.
• FEAT (Multi-Riwayah & Qira'at): Comprehensive cross-Riwayah and cross-madhhab ayah mapping via QiraatAyahMapper, RiwayaRegistry, RiwayaDescriptor, and AyahMappingDownloader, supporting all 6 canonical counting madhhabs (6,236 down to 6,204 ayahs) and 20 mutawatir rawis sourced from Quranpedia.
• FEAT (Omission Detection): Added LcsOmissionDetector and OmissionResult using the 2-row dynamic-programming Best-Drop LCS algorithm (derived from tasmee3-muaalem-findings benchmark research) for precise word omission localization.
• FEAT (Assets & Networking): Built-in ModelDownloader for streaming on-demand neural model download (~72 MB), keeping initial app download size under 25 MB.
• FEAT (Voice Navigation): Overhauled VoiceSearchController with real-time candidate Ayah streaming (onSearchResult, currentResult), progressive narrowing auto-jump, Mutashabihat (متشابهات) disambiguation, and AyahSearchMatch metadata enrichment.
• DOC: Added 12 comprehensive modular engineering guides in doc/ covering Dictation, Tajweed, Voice Search, Tarteel-style Memorization Mode, Production Audio Pipeline, Mus'haf UI & Typography, Multi-Riwayah, Neural Model Deployment, API Reference Cheat Sheet, and Troubleshooting FAQ.
• TEST: Added regression test suite test/auto_reanchor_test.dart covering 5 critical scenarios.

1.0.3

• FEAT: Enabled official platform classification for Windows and Linux on pub.dev.
• REFACTOR: Streamlined AudioProcessor by removing redundant mobile-only audio_session dependency in favor of direct cross-platform recording via record.

1.0.2

• FIX: Restored baseline DTW endpoint alignment in QuranDictationMatcher to ensure complete word phonetic consumption and eliminate trailing phoneme bleed.
• FIX: Removed experimental lookahead guard in DictationSequencer.
• DOC: Documented streaming alignment and fast-committing mechanisms in QuranDictationMatcher.

1.0.1

• FIX: Drastically improved DTW alignment stability on Waqf for short words containing Shaddah (e.g. رَبِّ) and Madd (e.g. الرَّحِيمِ).
• FIX: Improved ErrorExplainer to correctly honor Waqf (Sukoon), preventing false NORMAL -> DELETE and TASHKEEL errors when pausing at the end of a word.

1.0.0

• Initial open-source release of recite_quran.
• Real-time continuous recitation tracking with Semi-Global Dynamic Time Warping (DTW).
• Deterministic Tajweed rule verification engine:
  • Madd duration evaluation (2, 4, 6 Harakat).
  • Mushaddad Ghunnah verification (Noon & Meem).
  • Shaddah consonant closure & timing verification.
• Cross-platform support (Android, iOS, Windows, macOS, Linux, and Web).
• Automated on-demand model downloader and caching via ModelLoader.
• Developer CLI tool for pre-bundling models: dart run recite_quran:model_loader.
• Configurable difficulty matrix (TrackerConfig.normal(), .easy(), .strict()).