stat_ort_plugin 0.0.4

On-device speech recognition (Vaani ASR by IISc Bangalore) for Flutter via dart:ffi — file and streaming transcription with Silero VAD and speaker diarization, powered by ONNX Runtime.

0.0.4

• Switched the FFI layer to the ffigen-generated bindings (StatOrtPluginBindings) as the single source of truth; removed the duplicate hand-written typedefs.
• Reorganized lib/: native library loading now lives in src/native_library.dart, and the generated bindings moved to src/ (internal, no longer importable from the package root).
• Added NativeFinalizer backstops to Vaani and VaaniStream: native resources are freed on garbage collection if dispose()/close() is not called. Explicit disposal still works and is preferred.
• Vaani.transcribe and createStream now throw StateError if called after dispose() instead of using a freed pointer.
• Documentation and formatting: documented VaaniStream's public API and ran dart format. Fixed broken model-download and example links in the README.
• No public API changes (Vaani, VaaniStream).

0.0.3

Updated read me

0.0.2

Updated repo & issue links

0.0.1

Initial release.

• On-device Vaani ASR for Flutter via dart:ffi, powered by ONNX Runtime.
• File transcription API (Vaani.transcribe) for 16 kHz mono PCM WAV.
• Real-time streaming API (VaaniStream) with Silero VAD segmentation and optional speaker diarization.
• Android and iOS support. ONNX Runtime is provided by the onnxruntime-android Gradle dependency and the onnxruntime-c CocoaPod; no native binaries are committed to the repository.
• Heavy work (model init, file transcription) runs on a background isolate.