react-native-sherpa-onnx 0.2.0 → 0.3.0

package/lib/typescript/src/enhancement/index.d.ts

@@ -10,15 +10,16 @@
  * // Future usage:
  * import { initializeEnhancement, enhanceAudio } from 'react-native-sherpa-onnx/enhancement';
  *
- * await initializeEnhancement({ modelPath: 'models/enhancement-model' });
+ * await initializeEnhancement({ modelPath: { type: 'auto', path: 'models/enhancement-model' } });
  * const enhancedPath = await enhanceAudio('path/to/noisy-audio.wav');
  * ```
  */
+import type { ModelPathConfig } from '../types';
 /**
  * Enhancement initialization options (placeholder)
  */
 export interface EnhancementInitializeOptions {
-    modelPath: string;
+    modelPath: ModelPathConfig;
 }
 /**
  * Enhancement result

package/lib/typescript/src/enhancement/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../src/enhancement/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;GAeG;AAEH;;GAEG;AACH,MAAM,WAAW,4BAA4B;IAC3C,SAAS,EAAE,MAAM,CAAC;CAEnB;AAED;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,UAAU,EAAE,MAAM,CAAC;CAEpB;AAED;;;;GAIG;AACH,wBAAsB,qBAAqB,CACzC,QAAQ,EAAE,4BAA4B,GACrC,OAAO,CAAC,IAAI,CAAC,CAIf;AAED;;;;GAIG;AACH,wBAAgB,YAAY,CAAC,SAAS,EAAE,MAAM,GAAG,OAAO,CAAC,iBAAiB,CAAC,CAI1E;AAED;;;;GAIG;AACH,wBAAgB,iBAAiB,IAAI,OAAO,CAAC,IAAI,CAAC,CAIjD"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../src/enhancement/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;GAeG;AAEH,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,UAAU,CAAC;AAEhD;;GAEG;AACH,MAAM,WAAW,4BAA4B;IAC3C,SAAS,EAAE,eAAe,CAAC;CAE5B;AAED;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,UAAU,EAAE,MAAM,CAAC;CAEpB;AAED;;;;GAIG;AACH,wBAAsB,qBAAqB,CACzC,QAAQ,EAAE,4BAA4B,GACrC,OAAO,CAAC,IAAI,CAAC,CAIf;AAED;;;;GAIG;AACH,wBAAgB,YAAY,CAAC,SAAS,EAAE,MAAM,GAAG,OAAO,CAAC,iBAAiB,CAAC,CAI1E;AAED;;;;GAIG;AACH,wBAAgB,iBAAiB,IAAI,OAAO,CAAC,IAAI,CAAC,CAIjD"}

package/lib/typescript/src/index.d.ts

@@ -1,7 +1,31 @@
-export type { InitializeOptions, ModelPathConfig, ModelType } from './types';
-export { assetModelPath, autoModelPath, fileModelPath, getDefaultModelPath, listAssetModels, resolveModelPath, } from './utils';
+import type { AccelerationSupport } from './NativeSherpaOnnx';
+export type { AccelerationSupport } from './NativeSherpaOnnx';
+export type { ModelPathConfig } from './types';
+export { assetModelPath, autoModelPath, fileModelPath, getAssetPackPath, getDefaultModelPath, getPlayAssetDeliveryModelsPath, listAssetModels, listModelsAtPath, resolveModelPath, } from './utils';
 /**
  * Test method to verify sherpa-onnx native library is loaded.
  */
 export declare function testSherpaInit(): Promise<string>;
+/**
+ * QNN support (Android). Optional modelBase64 for canInit (session test); if omitted, SDK uses embedded test model.
+ */
+export declare function getQnnSupport(modelBase64?: string): Promise<AccelerationSupport>;
+/**
+ * Return the list of available ONNX Runtime execution providers
+ * (e.g. "CPU", "NNAPI", "QNN", "XNNPACK").
+ * Requires the ORT Java bridge from the onnxruntime AAR.
+ */
+export declare function getAvailableProviders(): Promise<string[]>;
+/**
+ * NNAPI support (Android). Optional modelBase64 for canInit (session test). On iOS returns all false.
+ */
+export declare function getNnapiSupport(modelBase64?: string): Promise<AccelerationSupport>;
+/**
+ * XNNPACK support. hasAccelerator = true when providerCompiled (CPU-optimized). Optional modelBase64 for canInit. On iOS returns all false.
+ */
+export declare function getXnnpackSupport(modelBase64?: string): Promise<AccelerationSupport>;
+/**
+ * Core ML support (iOS). providerCompiled = true (Core ML on iOS 11+), hasAccelerator = Apple Neural Engine. Optional modelBase64 for canInit. On Android returns all false.
+ */
+export declare function getCoreMlSupport(modelBase64?: string): Promise<AccelerationSupport>;
 //# sourceMappingURL=index.d.ts.map

package/lib/typescript/src/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/index.tsx"],"names":[],"mappings":"AAGA,YAAY,EAAE,iBAAiB,EAAE,eAAe,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAC7E,OAAO,EACL,cAAc,EACd,aAAa,EACb,aAAa,EACb,mBAAmB,EACnB,eAAe,EACf,gBAAgB,GACjB,MAAM,SAAS,CAAC;AAUjB;;GAEG;AACH,wBAAgB,cAAc,IAAI,OAAO,CAAC,MAAM,CAAC,CAEhD"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/index.tsx"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,oBAAoB,CAAC;AAG9D,YAAY,EAAE,mBAAmB,EAAE,MAAM,oBAAoB,CAAC;AAC9D,YAAY,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAC/C,OAAO,EACL,cAAc,EACd,aAAa,EACb,aAAa,EACb,gBAAgB,EAChB,mBAAmB,EACnB,8BAA8B,EAC9B,eAAe,EACf,gBAAgB,EAChB,gBAAgB,GACjB,MAAM,SAAS,CAAC;AAWjB;;GAEG;AACH,wBAAgB,cAAc,IAAI,OAAO,CAAC,MAAM,CAAC,CAEhD;AAED;;GAEG;AACH,wBAAgB,aAAa,CAC3B,WAAW,CAAC,EAAE,MAAM,GACnB,OAAO,CAAC,mBAAmB,CAAC,CAE9B;AAED;;;;GAIG;AACH,wBAAgB,qBAAqB,IAAI,OAAO,CAAC,MAAM,EAAE,CAAC,CAEzD;AAED;;GAEG;AACH,wBAAgB,eAAe,CAC7B,WAAW,CAAC,EAAE,MAAM,GACnB,OAAO,CAAC,mBAAmB,CAAC,CAE9B;AAED;;GAEG;AACH,wBAAgB,iBAAiB,CAC/B,WAAW,CAAC,EAAE,MAAM,GACnB,OAAO,CAAC,mBAAmB,CAAC,CAE9B;AAED;;GAEG;AACH,wBAAgB,gBAAgB,CAC9B,WAAW,CAAC,EAAE,MAAM,GACnB,OAAO,CAAC,mBAAmB,CAAC,CAE9B"}

package/lib/typescript/src/separation/index.d.ts

@@ -10,15 +10,16 @@
  * // Future usage:
  * import { initializeSeparation, separateSources } from 'react-native-sherpa-onnx/separation';
  *
- * await initializeSeparation({ modelPath: 'models/separation-model' });
+ * await initializeSeparation({ modelPath: { type: 'auto', path: 'models/separation-model' } });
  * const sources = await separateSources('path/to/mixed-audio.wav');
  * ```
  */
+import type { ModelPathConfig } from '../types';
 /**
  * Separation initialization options (placeholder)
  */
 export interface SeparationInitializeOptions {
-    modelPath: string;
+    modelPath: ModelPathConfig;
 }
 /**
  * Separated audio source

package/lib/typescript/src/separation/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../src/separation/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;GAeG;AAEH;;GAEG;AACH,MAAM,WAAW,2BAA2B;IAC1C,SAAS,EAAE,MAAM,CAAC;CAEnB;AAED;;GAEG;AACH,MAAM,WAAW,eAAe;IAC9B,QAAQ,EAAE,MAAM,CAAC;IACjB,UAAU,EAAE,MAAM,CAAC;CAEpB;AAED;;;;GAIG;AACH,wBAAsB,oBAAoB,CACxC,QAAQ,EAAE,2BAA2B,GACpC,OAAO,CAAC,IAAI,CAAC,CAIf;AAED;;;;GAIG;AACH,wBAAgB,eAAe,CAAC,SAAS,EAAE,MAAM,GAAG,OAAO,CAAC,eAAe,EAAE,CAAC,CAI7E;AAED;;;;GAIG;AACH,wBAAgB,gBAAgB,IAAI,OAAO,CAAC,IAAI,CAAC,CAIhD"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../src/separation/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;GAeG;AAEH,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,UAAU,CAAC;AAEhD;;GAEG;AACH,MAAM,WAAW,2BAA2B;IAC1C,SAAS,EAAE,eAAe,CAAC;CAE5B;AAED;;GAEG;AACH,MAAM,WAAW,eAAe;IAC9B,QAAQ,EAAE,MAAM,CAAC;IACjB,UAAU,EAAE,MAAM,CAAC;CAEpB;AAED;;;;GAIG;AACH,wBAAsB,oBAAoB,CACxC,QAAQ,EAAE,2BAA2B,GACpC,OAAO,CAAC,IAAI,CAAC,CAIf;AAED;;;;GAIG;AACH,wBAAgB,eAAe,CAAC,SAAS,EAAE,MAAM,GAAG,OAAO,CAAC,eAAe,EAAE,CAAC,CAI7E;AAED;;;;GAIG;AACH,wBAAgB,gBAAgB,IAAI,OAAO,CAAC,IAAI,CAAC,CAIhD"}

package/lib/typescript/src/stt/index.d.ts

@@ -1,62 +1,50 @@
-import type { STTInitializeOptions } from './types';
-import type { InitializeOptions } from '../types';
+import type { STTInitializeOptions, STTModelType, SttEngine } from './types';
+import type { ModelPathConfig } from '../types';
 /**
- * Initialize Speech-to-Text (STT) with model directory.
+ * Detect STT model type and structure without initializing the recognizer.
+ * Uses the same native file-based detection as createSTT. Stateless; no instance required.
  *
- * Supports multiple model source types:
- * - Asset models (bundled in app)
- * - File system models (downloaded or user-provided)
- * - Auto-detection (tries asset first, then file system)
- *
- * @param options - STT initialization options or model path configuration
- * @returns Object with success status and array of detected models (each with type and modelDir)
+ * @param modelPath - Model path configuration (asset, file, or auto)
+ * @param options - Optional preferInt8 and modelType (default: auto)
+ * @returns Object with success, detectedModels (array of { type, modelDir }), and modelType (primary detected type)
  * @example
  * ```typescript
- * // Simple string (auto-detect)
- * const result = await initializeSTT('models/sherpa-onnx-model');
- * console.log('Detected models:', result.detectedModels);
- * // result.detectedModels = [{ type: 'transducer', modelDir: '/path/to/model' }]
- *
- * // Asset model
- * const result = await initializeSTT({
- *   modelPath: { type: 'asset', path: 'models/sherpa-onnx-model' }
- * });
- *
- * // File system model with preferInt8 option
- * const result = await initializeSTT({
- *   modelPath: { type: 'file', path: '/path/to/model' },
- *   preferInt8: true  // Prefer quantized int8 models (smaller, faster)
- * });
- *
- * // With explicit model type
- * const result = await initializeSTT({
- *   modelPath: { type: 'asset', path: 'models/sherpa-onnx-nemo-parakeet-tdt-ctc-en' },
- *   modelType: 'nemo_ctc'
- * });
+ * const path = { type: 'asset' as const, path: 'models/sherpa-onnx-whisper-tiny-en' };
+ * const result = await detectSttModel(path);
+ * if (result.success && result.detectedModels.length > 0) {
+ *   console.log('Detected type:', result.modelType, result.detectedModels);
+ * }
  * ```
  */
-export declare function initializeSTT(options: STTInitializeOptions | InitializeOptions['modelPath']): Promise<{
+export declare function detectSttModel(modelPath: ModelPathConfig, options?: {
+    preferInt8?: boolean;
+    modelType?: STTModelType;
+}): Promise<{
     success: boolean;
     detectedModels: Array<{
         type: string;
         modelDir: string;
     }>;
+    modelType?: string;
 }>;
 /**
- * Transcribe an audio file.
+ * Create an STT engine instance. Call destroy() on the returned engine when done to free native resources.
  *
- * @param filePath - Path to WAV file (16kHz, mono, 16-bit PCM)
- * @returns Promise resolving to transcribed text
+ * @param options - STT initialization options or model path configuration
+ * @returns Promise resolving to an SttEngine instance
  * @example
  * ```typescript
- * const transcription = await transcribeFile('path/to/audio.wav');
- * console.log('Transcription:', transcription);
+ * const stt = await createSTT({
+ *   modelPath: { type: 'asset', path: 'models/whisper-tiny' },
+ * });
+ * const result = await stt.transcribeFile('/path/to/audio.wav');
+ * console.log(result.text);
+ * await stt.destroy();
  * ```
  */
-export declare function transcribeFile(filePath: string): Promise<string>;
-/**
- * Release STT resources.
- */
-export declare function unloadSTT(): Promise<void>;
-export type { STTInitializeOptions, TranscriptionResult } from './types';
+export declare function createSTT(options: STTInitializeOptions | ModelPathConfig): Promise<SttEngine>;
+export type { STTInitializeOptions, STTModelType, SttModelOptions, SttRecognitionResult, SttRuntimeConfig, SttEngine, SttInitResult, } from './types';
+export { STT_MODEL_TYPES, STT_HOTWORDS_MODEL_TYPES, sttSupportsHotwords, } from './types';
+export { getWhisperLanguages, WHISPER_LANGUAGES, getSenseVoiceLanguages, SENSEVOICE_LANGUAGES, getCanaryLanguages, CANARY_LANGUAGES, getFunasrNanoLanguages, FUNASR_NANO_LANGUAGES, getFunasrMltNanoLanguages, FUNASR_MLT_NANO_LANGUAGES, } from './sttModelLanguages';
+export type { SttModelLanguage, WhisperLanguage } from './sttModelLanguages';
 //# sourceMappingURL=index.d.ts.map

package/lib/typescript/src/stt/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../src/stt/index.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,oBAAoB,EAAE,MAAM,SAAS,CAAC;AACpD,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,UAAU,CAAC;AAGlD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAkCG;AACH,wBAAsB,aAAa,CACjC,OAAO,EAAE,oBAAoB,GAAG,iBAAiB,CAAC,WAAW,CAAC,GAC7D,OAAO,CAAC;IACT,OAAO,EAAE,OAAO,CAAC;IACjB,cAAc,EAAE,KAAK,CAAC;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;CAC3D,CAAC,CAkBD;AAED;;;;;;;;;;GAUG;AACH,wBAAgB,cAAc,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC,CAEhE;AAED;;GAEG;AACH,wBAAgB,SAAS,IAAI,OAAO,CAAC,IAAI,CAAC,CAEzC;AAGD,YAAY,EAAE,oBAAoB,EAAE,mBAAmB,EAAE,MAAM,SAAS,CAAC"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../src/stt/index.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EACV,oBAAoB,EACpB,YAAY,EACZ,SAAS,EAIV,MAAM,SAAS,CAAC;AACjB,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,UAAU,CAAC;AA2BhD;;;;;;;;;;;;;;;GAeG;AACH,wBAAsB,cAAc,CAClC,SAAS,EAAE,eAAe,EAC1B,OAAO,CAAC,EAAE;IAAE,UAAU,CAAC,EAAE,OAAO,CAAC;IAAC,SAAS,CAAC,EAAE,YAAY,CAAA;CAAE,GAC3D,OAAO,CAAC;IACT,OAAO,EAAE,OAAO,CAAC;IACjB,cAAc,EAAE,KAAK,CAAC;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;IAC1D,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,CAAC,CAOD;AAED;;;;;;;;;;;;;;GAcG;AACH,wBAAsB,SAAS,CAC7B,OAAO,EAAE,oBAAoB,GAAG,eAAe,GAC9C,OAAO,CAAC,SAAS,CAAC,CAsIpB;AAGD,YAAY,EACV,oBAAoB,EACpB,YAAY,EACZ,eAAe,EACf,oBAAoB,EACpB,gBAAgB,EAChB,SAAS,EACT,aAAa,GACd,MAAM,SAAS,CAAC;AACjB,OAAO,EACL,eAAe,EACf,wBAAwB,EACxB,mBAAmB,GACpB,MAAM,SAAS,CAAC;AACjB,OAAO,EACL,mBAAmB,EACnB,iBAAiB,EACjB,sBAAsB,EACtB,oBAAoB,EACpB,kBAAkB,EAClB,gBAAgB,EAChB,sBAAsB,EACtB,qBAAqB,EACrB,yBAAyB,EACzB,yBAAyB,GAC1B,MAAM,qBAAqB,CAAC;AAC7B,YAAY,EAAE,gBAAgB,EAAE,eAAe,EAAE,MAAM,qBAAqB,CAAC"}

package/lib/typescript/src/stt/sttModelLanguages.d.ts

@@ -0,0 +1,52 @@
+/**
+ * STT model language codes and display names.
+ * Per-model lists for Whisper, SenseVoice, and others. Use these for language-hint
+ * dropdowns so users only pick valid codes (invalid codes can crash the app, e.g. Whisper).
+ */
+export interface SttModelLanguage {
+    /**
+     * Value to pass as language (e.g. "en" for Whisper, "中文" for FunASR Nano).
+     * Use as modelOptions.<model>.language (or srcLang/tgtLang where applicable).
+     */
+    id: string;
+    /** Display name in English (e.g. "english", "chinese"). */
+    name: string;
+}
+/** @deprecated Use SttModelLanguage. Kept for backward compatibility. */
+export type WhisperLanguage = SttModelLanguage;
+/** Ordered list of all Whisper-supported language codes and names. */
+export declare const WHISPER_LANGUAGES: readonly SttModelLanguage[];
+/**
+ * Returns the list of Whisper-supported language codes and display names.
+ * Use for building a language-hint dropdown so users only pick valid codes (invalid codes can crash the app).
+ */
+export declare function getWhisperLanguages(): readonly SttModelLanguage[];
+/** Ordered list of SenseVoice-supported language codes and names. */
+export declare const SENSEVOICE_LANGUAGES: readonly SttModelLanguage[];
+/**
+ * Returns the list of SenseVoice-supported language codes and display names.
+ * Use for modelOptions.senseVoice.language so users only pick valid codes.
+ */
+export declare function getSenseVoiceLanguages(): readonly SttModelLanguage[];
+/** Canary: en, es, de, fr. */
+export declare const CANARY_LANGUAGES: readonly SttModelLanguage[];
+/**
+ * Returns the list of Canary-supported language codes and display names.
+ * Use for modelOptions.canary.srcLang and modelOptions.canary.tgtLang.
+ */
+export declare function getCanaryLanguages(): readonly SttModelLanguage[];
+/** Fun-ASR-Nano-2512: Chinese, English, Japanese. */
+export declare const FUNASR_NANO_LANGUAGES: readonly SttModelLanguage[];
+/** Fun-ASR-MLT-Nano-2512: multilingual list. */
+export declare const FUNASR_MLT_NANO_LANGUAGES: readonly SttModelLanguage[];
+/**
+ * Returns languages for Fun-ASR-Nano-2512 (中文, 英文, 日文).
+ * Id is the value for modelOptions.funasrNano.language (e.g. "中文").
+ */
+export declare function getFunasrNanoLanguages(): readonly SttModelLanguage[];
+/**
+ * Returns languages for Fun-ASR-MLT-Nano-2512 (multilingual).
+ * Id is the value for modelOptions.funasrNano.language (e.g. "中文").
+ */
+export declare function getFunasrMltNanoLanguages(): readonly SttModelLanguage[];
+//# sourceMappingURL=sttModelLanguages.d.ts.map

package/lib/typescript/src/stt/sttModelLanguages.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"sttModelLanguages.d.ts","sourceRoot":"","sources":["../../../../src/stt/sttModelLanguages.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,MAAM,WAAW,gBAAgB;IAC/B;;;OAGG;IACH,EAAE,EAAE,MAAM,CAAC;IACX,2DAA2D;IAC3D,IAAI,EAAE,MAAM,CAAC;CACd;AAED,yEAAyE;AACzE,MAAM,MAAM,eAAe,GAAG,gBAAgB,CAAC;AAK/C,sEAAsE;AACtE,eAAO,MAAM,iBAAiB,EAAE,SAAS,gBAAgB,EAqG/C,CAAC;AAEX;;;GAGG;AACH,wBAAgB,mBAAmB,IAAI,SAAS,gBAAgB,EAAE,CAEjE;AAKD,qEAAqE;AACrE,eAAO,MAAM,oBAAoB,EAAE,SAAS,gBAAgB,EAOlD,CAAC;AAEX;;;GAGG;AACH,wBAAgB,sBAAsB,IAAI,SAAS,gBAAgB,EAAE,CAEpE;AAOD,8BAA8B;AAC9B,eAAO,MAAM,gBAAgB,EAAE,SAAS,gBAAgB,EAK9C,CAAC;AAEX;;;GAGG;AACH,wBAAgB,kBAAkB,IAAI,SAAS,gBAAgB,EAAE,CAEhE;AAMD,qDAAqD;AACrD,eAAO,MAAM,qBAAqB,EAAE,SAAS,gBAAgB,EAInD,CAAC;AAEX,gDAAgD;AAChD,eAAO,MAAM,yBAAyB,EAAE,SAAS,gBAAgB,EAgCvD,CAAC;AAEX;;;GAGG;AACH,wBAAgB,sBAAsB,IAAI,SAAS,gBAAgB,EAAE,CAEpE;AAED;;;GAGG;AACH,wBAAgB,yBAAyB,IAAI,SAAS,gBAAgB,EAAE,CAEvE"}

package/lib/typescript/src/stt/types.d.ts

@@ -1,4 +1,88 @@
-import type { InitializeOptions, ModelType } from '../types';
+import type { ModelPathConfig } from '../types';
+/**
+ * Supported STT model types.
+ * Must match ParseSttModelType() in android/.../sherpa-onnx-model-detect-stt.cpp.
+ */
+export type STTModelType = 'transducer' | 'nemo_transducer' | 'paraformer' | 'nemo_ctc' | 'wenet_ctc' | 'sense_voice' | 'zipformer_ctc' | 'ctc' | 'whisper' | 'funasr_nano' | 'fire_red_asr' | 'moonshine' | 'dolphin' | 'canary' | 'omnilingual' | 'medasr' | 'telespeech_ctc' | 'auto';
+/** Model types that support hotwords (contextual biasing). Transducer and NeMo transducer support hotwords in sherpa-onnx (NeMo: see k2-fsa/sherpa-onnx#3077). */
+export declare const STT_HOTWORDS_MODEL_TYPES: readonly STTModelType[];
+/**
+ * Returns true only for model types that support hotwords (transducer, nemo_transducer).
+ * Use this to show/hide hotword options in the UI or to validate before init/setSttConfig.
+ */
+export declare function sttSupportsHotwords(modelType: STTModelType | string): boolean;
+/** Runtime list of supported STT model types (must match ParseSttModelType in native). */
+export declare const STT_MODEL_TYPES: readonly STTModelType[];
+/** Result of initializeSTT(). decodingMethod is set when init succeeds (e.g. "greedy_search" or "modified_beam_search"; auto-set when hotwords are used). */
+export interface SttInitResult {
+    success: boolean;
+    detectedModels: Array<{
+        type: string;
+        modelDir: string;
+    }>;
+    modelType?: string;
+    decodingMethod?: string;
+}
+/** Options for Whisper models. Applied only when modelType is 'whisper'. */
+export interface SttWhisperModelOptions {
+    /** Language code (e.g. "en", "de"). Used with multilingual models. Default: "en". */
+    language?: string;
+    /** "transcribe" or "translate". Default: "transcribe". With "translate", result text is English. */
+    task?: 'transcribe' | 'translate';
+    /** Padding at end of samples. Kotlin default 1000; C++ default -1. */
+    tailPaddings?: number;
+    /** Token-level timestamps. Android only; ignored on iOS. */
+    enableTokenTimestamps?: boolean;
+    /** Segment-level timestamps. Android only; ignored on iOS. */
+    enableSegmentTimestamps?: boolean;
+}
+/** Options for SenseVoice models. Applied only when modelType is 'sense_voice'. */
+export interface SttSenseVoiceModelOptions {
+    /** Language hint. */
+    language?: string;
+    /** Inverse text normalization. Default: true (Kotlin), false (C++). */
+    useItn?: boolean;
+}
+/** Options for Canary models. Applied only when modelType is 'canary'. */
+export interface SttCanaryModelOptions {
+    /** Source language code. Default: "en". */
+    srcLang?: string;
+    /** Target language code. Default: "en". */
+    tgtLang?: string;
+    /** Use punctuation. Default: true. */
+    usePnc?: boolean;
+}
+/** Options for FunASR Nano models. Applied only when modelType is 'funasr_nano'. */
+export interface SttFunAsrNanoModelOptions {
+    /** System prompt. Default: "You are a helpful assistant." */
+    systemPrompt?: string;
+    /** User prompt prefix. Default: "语音转写：" */
+    userPrompt?: string;
+    /** Max new tokens. Default: 512. */
+    maxNewTokens?: number;
+    /** Temperature. Default: 1e-6. */
+    temperature?: number;
+    /** Top-p. Default: 0.8. */
+    topP?: number;
+    /** Random seed. Default: 42. */
+    seed?: number;
+    /** Language hint. */
+    language?: string;
+    /** Inverse text normalization. Default: true. */
+    itn?: boolean;
+    /** Hotwords string. */
+    hotwords?: string;
+}
+/**
+ * Model-specific STT options. Only the block for the actually loaded model type is applied;
+ * others are ignored (e.g. whisper options have no effect when a paraformer model is loaded).
+ */
+export interface SttModelOptions {
+    whisper?: SttWhisperModelOptions;
+    senseVoice?: SttSenseVoiceModelOptions;
+    canary?: SttCanaryModelOptions;
+    funasrNano?: SttFunAsrNanoModelOptions;
+}
 /**
  * STT-specific initialization options
  */
@@ -6,7 +90,7 @@ export interface STTInitializeOptions {
     /**
      * Model directory path configuration
      */
-    modelPath: InitializeOptions['modelPath'];
+    modelPath: ModelPathConfig;
     /**
      * Model quantization preference
      * - true: Prefer int8 quantized models (model.int8.onnx) - smaller, faster
@@ -16,24 +100,127 @@ export interface STTInitializeOptions {
     preferInt8?: boolean;
     /**
      * Explicit model type specification for STT models
-     * - 'transducer': Force detection as Zipformer/Transducer model
+     * - 'transducer': Force detection as Transducer model
+     * - 'zipformer_ctc' | 'ctc': Force detection as Zipformer CTC model
      * - 'paraformer': Force detection as Paraformer model
      * - 'nemo_ctc': Force detection as NeMo CTC model
      * - 'whisper': Force detection as Whisper model
      * - 'wenet_ctc': Force detection as WeNet CTC model
      * - 'sense_voice': Force detection as SenseVoice model
      * - 'funasr_nano': Force detection as FunASR Nano model
+     * - 'fire_red_asr': FireRed ASR (encoder/decoder)
+     * - 'moonshine': Moonshine (preprocess, encode, uncached_decode, cached_decode)
+     * - 'dolphin': Dolphin (single model)
+     * - 'canary': Canary (encoder/decoder)
+     * - 'omnilingual': Omnilingual CTC (single model)
+     * - 'medasr': MedASR CTC (single model)
+     * - 'telespeech_ctc': TeleSpeech CTC (single model)
      * - 'auto': Automatic detection based on files (default)
      */
-    modelType?: ModelType;
+    modelType?: STTModelType;
+    /**
+     * Enable debug logging in native layer and sherpa-onnx (config.model_config.debug).
+     * When true, wrapper and JNI emit verbose logs (config dumps, file checks, init/transcribe flow).
+     * Default: false.
+     */
+    debug?: boolean;
+    /**
+     * Path to hotwords file for keyword boosting (Kotlin OfflineRecognizerConfig.hotwordsFile).
+     */
+    hotwordsFile?: string;
+    /**
+     * Hotwords score/weight (Kotlin OfflineRecognizerConfig.hotwordsScore).
+     * Default in Kotlin: 1.5.
+     */
+    hotwordsScore?: number;
+    /**
+     * Modeling unit for hotwords tokenization (Kotlin OfflineModelConfig.modelingUnit).
+     * Only used when hotwords are set and model is transducer/nemo_transducer.
+     * Must match how the model was trained: 'bpe' (e.g. English zipformer), 'cjkchar' (e.g. Chinese conformer), 'cjkchar+bpe' (bilingual zh-en).
+     * See docs/stt.md "When to use which modelingUnit" and sherpa-onnx hotwords docs.
+     */
+    modelingUnit?: 'cjkchar' | 'bpe' | 'cjkchar+bpe';
+    /**
+     * Path to BPE vocabulary file for hotwords (Kotlin OfflineModelConfig.bpeVocab).
+     * Required when modelingUnit is 'bpe' or 'cjkchar+bpe'. Sentencepiece .vocab export (bpe.vocab), not the hotwords file.
+     */
+    bpeVocab?: string;
+    /**
+     * Number of threads for inference (Kotlin OfflineModelConfig.numThreads).
+     * Default in Kotlin: 1.
+     */
+    numThreads?: number;
+    /**
+     * Provider string (e.g. "cpu"). Stored in config only; no special logic on change.
+     * Kotlin OfflineModelConfig.provider.
+     */
+    provider?: string;
+    /**
+     * Path to rule FSTs (Kotlin OfflineRecognizerConfig.ruleFsts).
+     */
+    ruleFsts?: string;
+    /**
+     * Path to rule FARs (Kotlin OfflineRecognizerConfig.ruleFars).
+     */
+    ruleFars?: string;
+    /**
+     * Dither for feature extraction (Kotlin FeatureConfig.dither). Default 0.
+     */
+    dither?: number;
+    /**
+     * Model-specific options. Only options for the loaded model type are applied.
+     * E.g. when modelType is 'whisper', only modelOptions.whisper is used.
+     */
+    modelOptions?: SttModelOptions;
 }
 /**
- * Transcription result
+ * Full recognition result from offline STT (maps to Kotlin OfflineRecognizerResult).
  */
-export interface TranscriptionResult {
-    /**
-     * Transcribed text
-     */
+export interface SttRecognitionResult {
+    /** Transcribed text. */
     text: string;
+    /** Token strings. */
+    tokens: string[];
+    /** Timestamps per token (model-dependent). */
+    timestamps: number[];
+    /** Detected or specified language (model-dependent). */
+    lang: string;
+    /** Emotion label (model-dependent, e.g. SenseVoice). */
+    emotion: string;
+    /** Event label (model-dependent). */
+    event: string;
+    /** Durations (valid for TDT models). */
+    durations: number[];
+}
+/**
+ * Instance-based STT engine returned by createSTT().
+ * Call destroy() when done to free native resources.
+ */
+export interface SttEngine {
+    readonly instanceId: string;
+    transcribeFile(filePath: string): Promise<SttRecognitionResult>;
+    transcribeSamples(samples: number[], sampleRate: number): Promise<SttRecognitionResult>;
+    setConfig(options: SttRuntimeConfig): Promise<void>;
+    destroy(): Promise<void>;
+}
+/**
+ * Runtime config for the offline recognizer (Kotlin OfflineRecognizerConfig).
+ * Only fields that can be updated via setConfig are included.
+ */
+export interface SttRuntimeConfig {
+    /** Decoding method (e.g. greedy_search). */
+    decodingMethod?: string;
+    /** Max active paths (beam search). */
+    maxActivePaths?: number;
+    /** Path to hotwords file. */
+    hotwordsFile?: string;
+    /** Hotwords score. */
+    hotwordsScore?: number;
+    /** Blank penalty. */
+    blankPenalty?: number;
+    /** Path to rule FSTs. */
+    ruleFsts?: string;
+    /** Path to rule FARs. */
+    ruleFars?: string;
 }
 //# sourceMappingURL=types.d.ts.map

package/lib/typescript/src/stt/types.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../../../src/stt/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,iBAAiB,EAAE,SAAS,EAAE,MAAM,UAAU,CAAC;AAE7D;;GAEG;AACH,MAAM,WAAW,oBAAoB;IACnC;;OAEG;IACH,SAAS,EAAE,iBAAiB,CAAC,WAAW,CAAC,CAAC;IAE1C;;;;;OAKG;IACH,UAAU,CAAC,EAAE,OAAO,CAAC;IAErB;;;;;;;;;;OAUG;IACH,SAAS,CAAC,EAAE,SAAS,CAAC;CACvB;AAED;;GAEG;AACH,MAAM,WAAW,mBAAmB;IAClC;;OAEG;IACH,IAAI,EAAE,MAAM,CAAC;CACd"}
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../../../src/stt/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,UAAU,CAAC;AAEhD;;;GAGG;AACH,MAAM,MAAM,YAAY,GACpB,YAAY,GACZ,iBAAiB,GACjB,YAAY,GACZ,UAAU,GACV,WAAW,GACX,aAAa,GACb,eAAe,GACf,KAAK,GACL,SAAS,GACT,aAAa,GACb,cAAc,GACd,WAAW,GACX,SAAS,GACT,QAAQ,GACR,aAAa,GACb,QAAQ,GACR,gBAAgB,GAChB,MAAM,CAAC;AAEX,kKAAkK;AAClK,eAAO,MAAM,wBAAwB,EAAE,SAAS,YAAY,EAGlD,CAAC;AAEX;;;GAGG;AACH,wBAAgB,mBAAmB,CAAC,SAAS,EAAE,YAAY,GAAG,MAAM,GAAG,OAAO,CAE7E;AAED,0FAA0F;AAC1F,eAAO,MAAM,eAAe,EAAE,SAAS,YAAY,EAmBzC,CAAC;AAEX,6JAA6J;AAC7J,MAAM,WAAW,aAAa;IAC5B,OAAO,EAAE,OAAO,CAAC;IACjB,cAAc,EAAE,KAAK,CAAC;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;IAC1D,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AAID,4EAA4E;AAC5E,MAAM,WAAW,sBAAsB;IACrC,qFAAqF;IACrF,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,oGAAoG;IACpG,IAAI,CAAC,EAAE,YAAY,GAAG,WAAW,CAAC;IAClC,sEAAsE;IACtE,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,4DAA4D;IAC5D,qBAAqB,CAAC,EAAE,OAAO,CAAC;IAChC,8DAA8D;IAC9D,uBAAuB,CAAC,EAAE,OAAO,CAAC;CACnC;AAED,mFAAmF;AACnF,MAAM,WAAW,yBAAyB;IACxC,qBAAqB;IACrB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,uEAAuE;IACvE,MAAM,CAAC,EAAE,OAAO,CAAC;CAClB;AAED,0EAA0E;AAC1E,MAAM,WAAW,qBAAqB;IACpC,2CAA2C;IAC3C,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,2CAA2C;IAC3C,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,sCAAsC;IACtC,MAAM,CAAC,EAAE,OAAO,CAAC;CAClB;AAED,oFAAoF;AACpF,MAAM,WAAW,yBAAyB;IACxC,6DAA6D;IAC7D,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,2CAA2C;IAC3C,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,oCAAoC;IACpC,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,kCAAkC;IAClC,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,2BAA2B;IAC3B,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,gCAAgC;IAChC,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,qBAAqB;IACrB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,iDAAiD;IACjD,GAAG,CAAC,EAAE,OAAO,CAAC;IACd,uBAAuB;IACvB,QAAQ,CAAC,EAAE,MAAM,CAAC;CACnB;AAED;;;GAGG;AACH,MAAM,WAAW,eAAe;IAC9B,OAAO,CAAC,EAAE,sBAAsB,CAAC;IACjC,UAAU,CAAC,EAAE,yBAAyB,CAAC;IACvC,MAAM,CAAC,EAAE,qBAAqB,CAAC;IAC/B,UAAU,CAAC,EAAE,yBAAyB,CAAC;CACxC;AAED;;GAEG;AACH,MAAM,WAAW,oBAAoB;IACnC;;OAEG;IACH,SAAS,EAAE,eAAe,CAAC;IAE3B;;;;;OAKG;IACH,UAAU,CAAC,EAAE,OAAO,CAAC;IAErB;;;;;;;;;;;;;;;;;;OAkBG;IACH,SAAS,CAAC,EAAE,YAAY,CAAC;IAEzB;;;;OAIG;IACH,KAAK,CAAC,EAAE,OAAO,CAAC;IAEhB;;OAEG;IACH,YAAY,CAAC,EAAE,MAAM,CAAC;IAEtB;;;OAGG;IACH,aAAa,CAAC,EAAE,MAAM,CAAC;IAEvB;;;;;OAKG;IACH,YAAY,CAAC,EAAE,SAAS,GAAG,KAAK,GAAG,aAAa,CAAC;IAEjD;;;OAGG;IACH,QAAQ,CAAC,EAAE,MAAM,CAAC;IAElB;;;OAGG;IACH,UAAU,CAAC,EAAE,MAAM,CAAC;IAEpB;;;OAGG;IACH,QAAQ,CAAC,EAAE,MAAM,CAAC;IAElB;;OAEG;IACH,QAAQ,CAAC,EAAE,MAAM,CAAC;IAElB;;OAEG;IACH,QAAQ,CAAC,EAAE,MAAM,CAAC;IAElB;;OAEG;IACH,MAAM,CAAC,EAAE,MAAM,CAAC;IAEhB;;;OAGG;IACH,YAAY,CAAC,EAAE,eAAe,CAAC;CAChC;AAED;;GAEG;AACH,MAAM,WAAW,oBAAoB;IACnC,wBAAwB;IACxB,IAAI,EAAE,MAAM,CAAC;IACb,qBAAqB;IACrB,MAAM,EAAE,MAAM,EAAE,CAAC;IACjB,8CAA8C;IAC9C,UAAU,EAAE,MAAM,EAAE,CAAC;IACrB,wDAAwD;IACxD,IAAI,EAAE,MAAM,CAAC;IACb,wDAAwD;IACxD,OAAO,EAAE,MAAM,CAAC;IAChB,qCAAqC;IACrC,KAAK,EAAE,MAAM,CAAC;IACd,wCAAwC;IACxC,SAAS,EAAE,MAAM,EAAE,CAAC;CACrB;AAED;;;GAGG;AACH,MAAM,WAAW,SAAS;IACxB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,cAAc,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC,oBAAoB,CAAC,CAAC;IAChE,iBAAiB,CACf,OAAO,EAAE,MAAM,EAAE,EACjB,UAAU,EAAE,MAAM,GACjB,OAAO,CAAC,oBAAoB,CAAC,CAAC;IACjC,SAAS,CAAC,OAAO,EAAE,gBAAgB,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IACpD,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;CAC1B;AAED;;;GAGG;AACH,MAAM,WAAW,gBAAgB;IAC/B,4CAA4C;IAC5C,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,sCAAsC;IACtC,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,6BAA6B;IAC7B,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,sBAAsB;IACtB,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,qBAAqB;IACrB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,yBAAyB;IACzB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,yBAAyB;IACzB,QAAQ,CAAC,EAAE,MAAM,CAAC;CACnB"}