- 6-Stage Producer-Consumer Scanning Pipeline:
Re-architected entire photo discovery, metadata extraction, hardware decode, AI
inference, and clustering stages using bounded high-throughput
System.Threading.Channels.
- Multimodal Semantic Search & Vector
Indexing: Embedded on-device CLIP-based image embedding engine with DirectML
acceleration and HNSW similarity search for natural language queries (e.g., "sunset at
beach", "birthday party", "snow mountain").
- InsightFace AI Face Recognition: High-accuracy
facial feature detection, ArcFace embedding vectors, identity clustering, face grouping,
and people tagging.
- Hardware-Accelerated WIC/DXVA Image Decoding:
Direct GPU decoding for rapid thumbnail generation, zero-allocation memory pooling, and
ultra-flat RAM usage.
- Interactive Offline Leaflet Maps: Reverse
geocoding of GPS EXIF coordinates with offline spatial clustering and map exploration.
- Burst Grouping & Quality Scoring: Automated
identification of burst shots with sharpest frame recommendation, blur/exposure
analysis, and duplicate detection.
- Contact Synchronization: Optional profile
linking with Google and Microsoft Contacts with local DPAPI-encrypted token protection.
- Introduced MainViewModel unified event clustering engine with progressive background
scanning.
- Added deep face recognition service infrastructure with direct ONNX runtime bindings.
- Updated Store packaging and Obfuscar security hardening rules.
- Restructured core services and UI components to establish full WinUI 3 modern desktop
architecture.
- Enhanced custom TitleBar with Mica backdrop and system theme synchronization.
- Integrated StoreService in-app purchase verification and license state caching.
- Implemented SceneClassifierService with ONNX model support for automatic scene
categorization (Nature, Urban, Documents, Water, Portraits).
- Enhanced FacesView UI with real-time face search, identity merging, and bulk tagging
controls.
- Added robust asynchronous cache persistence with SQLite WAL mode.
- Implemented centralized UI Dispatcher and initialized modern MVVM architecture for
high-performance photo organization.
- Added comprehensive unit test suites covering edge cases in Store purchase flows, face
recognition, and database concurrency.
- Implemented dynamic grid layout with UI virtualization supporting 50,000+ photo
collections smoothly.
- Added About dialog with real-time Hardware.Info system classification (CPU, GPU, RAM,
NPU tier).
- Implemented Google & Microsoft OAuth 2.0 authentication flows with DPAPI local token
encryption.
- Added offline Leaflet map integration and robust activity logging.
- Initial Microsoft Store release for Photo Organizer AI.
- Core perceptual hashing for duplicate detection and burst culling.
- EXIF metadata extraction and fast two-tier cache engine.
