earshot_classifier_platform_interface 1.0.0
earshot_classifier_platform_interface: ^1.0.0 copied to clipboard
The pure Dart platform interface for earshot_classifier, with the shared types for on-device sound classification.
Changelog #
1.0.0 #
- First stable release. The public API follows semantic versioning from now on: breaking changes only in a major version (ADR 0021).
- Breaking: remove
ClassifierBackend.auto.ClassifierBackendnow names engines only, and the vocabulary APIs no longer throw anArgumentErrorfor a backend (ADR 0023). - Add
ClassifierExceptionandClassifierErrorCode(unsupported,modelUnavailable,unknown). Backends deliver their own failures as aClassifierException, and forward errors from the PCM unchanged (ADR 0024). - Document that
EarshotSound,SoundGroupandClassifierBackendcan gain values in a minor version (ADR 0021).
0.2.0 #
- Add
EarshotClassifierPlatform.classify(Stream<PcmChunk>, ClassifierConfig), so native backends classify PCM fromearshot_captureinstead of opening their own microphone session (ADR 0018). It defaults to a stream that errors withUnimplementedError. detectionsnow captures withearshot_capturetoo, and detection timestamps follow the PCM clock.- Depend on
earshot_capture_platform_interfaceforPcmChunk.
0.1.0 #
- Add
EarshotSound, the cross-backend sound vocabulary. Each sound lists the Apple and YAMNet labels it maps to, validated against the official label lists. - Add
SoundVocabularyto map raw labels to sounds and groups. Unmapped labels fall back to their group, and silence is never reported. - Add
SoundDetection.sound. - BREAKING:
EarshotClassifierPlatform.detectionstakes aClassifierConfigand reports raw, unsmoothed detections from the backend's own microphone session. - Add
EarshotClassifierPlatform.isAvailable.
0.0.1 #
- Initial API preview.