Face Intelligence Releases

1.0 TBD

Feature

Initial release — on-device face enrollment and recognition split across a dependency-free core plus swappable embedder, detector and store packages, composed with AddFaceIntelligence.

Feature

IFaceIntelligence — Enroll / Recognize (with or without a face box), GetAll, and Forget, keyed on a caller-chosen PersonIdentifier.

Feature

Shiny.FaceIntelligence.Onnx — ArcFace embedder (UseOnnxEmbedder) and UltraFace detector (UseOnnxDetector), both loading their model lazily on first use so a missing model never breaks startup.

Feature

Shiny.FaceIntelligence.DocumentDb / .DocumentDb.Sqlite — vector storage and nearest-neighbour search over Shiny.DocumentDb, with the vector dimension read from the embedder so it always matches the model. sqlite-vec is registered as a SQLite auto-extension so it works on iOS.

Feature

Shiny.FaceIntelligence.Maui — FaceRecognitionView for continuous live identification and FaceEnrollmentView, a guided multi-shot wizard with a face-hole overlay, quality gates and an embedding-novelty check.

Feature

Detection gating on the no-box Enroll — FaceDetectionException for no face, low confidence, multiple faces or a face too small, and FaceEnrollmentConflictException when the face already matches a different enrolled identity.

Feature

AOT- and trim-compatible across every package, with a source-generated JsonSerializerContext for the stored document type.

Feature

Automatic iOS / Mac Catalyst linker fix for ONNX Runtime’s _RegisterCustomOps symbol, shipped as buildTransitive MSBuild targets so consuming app heads need no configuration.