react-native-sherpa-onnx 0.3.5 → 0.3.7

package/scripts/setup-ios-framework.sh

@@ -72,6 +72,12 @@ fi
 # Create frameworks directory if it doesn't exist
 mkdir -p "$FRAMEWORKS_DIR"
 
+# Backward compatibility: if slug-specific sherpa-onnx version file exists, remove legacy .framework-version
+# so only one source of truth remains and build numbers (e.g. 1.12.28-1) are not confused with legacy (1.12.28).
+if [ -f "$FRAMEWORKS_DIR/.framework-version-sherpa-onnx" ] && [ -f "$FRAMEWORKS_DIR/.framework-version" ]; then
+  rm -f "$FRAMEWORKS_DIR/.framework-version"
+fi
+
 # Framework slugs to manage (order: sherpa-onnx first, then libarchive, then ffmpeg)
 FRAMEWORK_SLUGS=(sherpa-onnx)
 
@@ -93,7 +99,7 @@ get_framework_config() {
   case "$slug" in
     sherpa-onnx)
       TAG_FILE="$PROJECT_ROOT/third_party/sherpa-onnx-prebuilt/IOS_RELEASE_TAG"
-      TAG_PREFIX="framework-v"
+      TAG_PREFIX="sherpa-onnx-ios-v"
       XCFRAMEWORK_NAME="sherpa_onnx.xcframework"
       ZIP_ASSET_NAME="sherpa_onnx.xcframework.zip"
       VERSION_FILE="$FRAMEWORKS_DIR/.framework-version-sherpa-onnx"
@@ -145,19 +151,19 @@ framework_valid() {
 
 # Helper: get installed framework version (from version file or xcframework VERSION.txt)
 # Usage: get_installed_version <slug>
-# Backward compat: for sherpa-onnx, also checks legacy .framework-version
+# Prefer slug-specific VERSION_FILE so build numbers (e.g. 1.12.28-1) match; fall back to legacy .framework-version for sherpa-onnx.
 get_installed_version() {
   local slug="$1"
   get_framework_config "$slug" || return 1
-  # Backward compat: sherpa-onnx used to use .framework-version
-  if [ "$slug" = "sherpa-onnx" ] && [ -f "$FRAMEWORKS_DIR/.framework-version" ]; then
-    cat "$FRAMEWORKS_DIR/.framework-version" 2>/dev/null | tr -d '\r\n'
-    return 0
-  fi
   if [ -f "$VERSION_FILE" ]; then
     cat "$VERSION_FILE" 2>/dev/null | tr -d '\r\n'
     return 0
   fi
+  # Backward compatibility only: sherpa-onnx used to use .framework-version; prefer slug-specific file above.
+  if [ "$slug" = "sherpa-onnx" ] && [ -f "$FRAMEWORKS_DIR/.framework-version" ]; then
+    cat "$FRAMEWORKS_DIR/.framework-version" 2>/dev/null | tr -d '\r\n'
+    return 0
+  fi
   if [ -f "$FRAMEWORKS_DIR/$XCFRAMEWORK_NAME/VERSION.txt" ]; then
     grep -Eo '([0-9]+\.)+[0-9]+([-a-zA-Z0-9.]*)?' "$FRAMEWORKS_DIR/$XCFRAMEWORK_NAME/VERSION.txt" | head -n1 | tr -d '\r\n'
     return 0
@@ -294,6 +300,7 @@ get_local_framework_version() {
     cat "$VERSION_FILE"
     return 0
   fi
+  # Backward compatibility only: sherpa-onnx used to use .framework-version.
   if [ "$slug" = "sherpa-onnx" ] && [ -f "$FRAMEWORKS_DIR/.framework-version" ]; then
     cat "$FRAMEWORKS_DIR/.framework-version"
     return 0
@@ -446,10 +453,6 @@ if [ "$INTERACTIVE" = true ]; then
     fi
   done
   echo "" >&2
-  echo "Next steps:" >&2
-  echo "  1. cd example" >&2
-  echo "  2. pod install" >&2
-  echo "  3. Open ios/SherpaOnnxExample.xcworkspace in Xcode" >&2
 fi
 
 # Fix CocoaPods header flattening: delete FFmpeg's time.h so it doesn't shadow system time.h

package/scripts/update_commercial_use.js

@@ -0,0 +1,73 @@
+const fs = require('fs');
+const path = require('path');
+
+const commercialMapping = {
+  'agpl-3.0': 'conditional',
+  'apache-2.0': 'yes',
+  'bsd-3-clause': 'yes',
+  'cc-by': 'yes',
+  'cc-by-3.0': 'yes',
+  'cc-by-4.0': 'yes',
+  'cc-by-nc-4.0': 'no',
+  'cc-by-nc-nd-4.0': 'no',
+  'cc-by-nc-sa-4.0': 'no',
+  'cc-by-sa': 'conditional',
+  'cc-by-sa-3.0-es': 'conditional',
+  'cc-by-sa-4.0': 'conditional',
+  'cc0': 'yes',
+  'gpl-2.0': 'conditional',
+  'mit': 'yes',
+  'ngc-terms-of-use': 'restricted',
+  'no-license': 'no',
+  'nvidia-open-model-license': 'yes',
+  'proprietary-restricted': 'no',
+  'public-domain': 'yes',
+  'research-only': 'no',
+  'unlicense': 'yes',
+};
+
+const files = [
+  'android/src/main/assets/model_licenses/asr-models-license-status.csv',
+  'android/src/main/assets/model_licenses/qnn-asr-models-license-status.csv',
+  'android/src/main/assets/model_licenses/tts-models-license-status.csv',
+];
+
+files.forEach((fileName) => {
+  const filePath = path.resolve(__dirname, '..', fileName);
+  if (!fs.existsSync(filePath)) {
+    console.error(`File not found: ${filePath}`);
+    return;
+  }
+
+  const content = fs.readFileSync(filePath, 'utf-8');
+  const lines = content.split('\n');
+  const newLines = [];
+
+  // Keep header
+  newLines.push(lines[0]);
+
+  for (let i = 1; i < lines.length; i++) {
+    const line = lines[i];
+    if (!line.trim()) {
+      if (i < lines.length - 1 || line === '') newLines.push(line);
+      continue;
+    }
+
+    const columns = line.split(',');
+    if (columns.length > 2) {
+      const licenseType = columns[1].trim();
+      const currentCommercialUse = columns[2].trim();
+
+      if (currentCommercialUse === 'unknown') {
+        const mappedValue = commercialMapping[licenseType];
+        if (mappedValue) {
+          columns[2] = mappedValue;
+        }
+      }
+    }
+    newLines.push(columns.join(','));
+  }
+
+  fs.writeFileSync(filePath, newLines.join('\n'), 'utf-8');
+  console.log(`Updated commercial_use in ${fileName}`);
+});

package/src/NativeSherpaOnnx.ts

@@ -29,7 +29,7 @@ export interface Spec extends TurboModule {
    * @param provider - Optional: provider string e.g. 'cpu' (stored in config only)
    * @param ruleFsts - Optional: path(s) to rule FSTs for ITN (comma-separated)
    * @param ruleFars - Optional: path(s) to rule FARs for ITN (comma-separated)
-   * @param dither - Optional: dither for feature extraction (default 0)
+   * @param dither - Optional: dither for feature extraction. **Android:** applied. **iOS:** ignored (native API does not expose it)
    * @param modelOptions - Optional: model-specific options (whisper, senseVoice, canary, funasrNano). Only the block for the loaded model type is applied.
    * @param modelingUnit - Optional: 'cjkchar' | 'bpe' | 'cjkchar+bpe' for hotwords tokenization (OfflineModelConfig.modelingUnit)
    * @param bpeVocab - Optional: path to BPE vocab file (OfflineModelConfig.bpeVocab), used when modelingUnit is bpe or cjkchar+bpe
@@ -73,6 +73,8 @@ export interface Spec extends TurboModule {
     modelType?: string
   ): Promise<{
     success: boolean;
+    /** Present when success is false (or native included a message). */
+    error?: string;
     /** True when detection failed because the model targets unsupported hardware (RK35xx, Ascend, CANN). Use to show a specific message or block init. */
     isHardwareSpecificUnsupported?: boolean;
     detectedModels: Array<{ type: string; modelDir: string }>;
@@ -127,7 +129,8 @@ export interface Spec extends TurboModule {
   /**
    * Initialize OnlineRecognizer for streaming STT (single options object to avoid iOS TurboModule marshalling crash with many args).
    * @param instanceId - Unique ID for this engine instance (from createStreamingSTT)
-   * @param options - All init options (modelDir, modelType, enableEndpoint, decodingMethod, maxActivePaths, and optional endpoint/rule params)
+   * @param options - All init options (modelDir, modelType, enableEndpoint, decodingMethod, maxActivePaths, and optional endpoint/rule params).
+   *   `options.dither`: **Android** only; **iOS** ignores it (native `FeatureConfig` has no dither field).
    */
   initializeOnlineSttWithOptions(
     instanceId: string,
@@ -143,6 +146,8 @@ export interface Spec extends TurboModule {
       provider?: string;
       ruleFsts?: string;
       ruleFars?: string;
+      /** Feature dither. Android: applied. iOS: ignored. */
+      dither?: number;
       blankPenalty?: number;
       debug?: boolean;
       rule1MustContainNonSilence?: boolean;
@@ -280,6 +285,8 @@ export interface Spec extends TurboModule {
     modelType?: string
   ): Promise<{
     success: boolean;
+    /** Present when success is false (or native included a message). */
+    error?: string;
     detectedModels: Array<{ type: string; modelDir: string }>;
     modelType?: string;
     /** Language ids from detected lexicon files (e.g. "default" for lexicon.txt, "us-en", "zh" from lexicon-us-en.txt, lexicon-zh.txt). Present for Kokoro/Kitten when multiple or single lexicon files are found; use for language selection UI. */
@@ -310,7 +317,8 @@ export interface Spec extends TurboModule {
    * Generate speech from text.
    * @param instanceId - Unique ID for this engine instance
    * @param text - Text to convert to speech
-   * @param options - Generation options (sid, speed, referenceAudio, referenceText, numSteps, silenceScale, extra)
+   * @param options - Generation options: `sid`, `speed`, `silenceScale`, `numSteps`, `extra`.
+   *   Voice cloning (iOS & Android): `referenceAudio` + `referenceSampleRate` for Zipvoice/Pocket only; Zipvoice also needs non-empty `referenceText`.
    * @returns Object with { samples: number[], sampleRate: number }
    */
   generateTts(
@@ -326,7 +334,7 @@ export interface Spec extends TurboModule {
    * Generate speech with subtitle/timestamp metadata.
    * @param instanceId - Unique ID for this engine instance
    * @param text - Text to convert to speech
-   * @param options - Generation options (sid, speed, referenceAudio, referenceText, numSteps, silenceScale, extra)
+   * @param options - Same as {@link generateTts} options (cloning: Zipvoice/Pocket; Zipvoice needs `referenceText`).
    * @returns Object with samples, sampleRate, subtitles, and estimated flag
    */
   generateTtsWithTimestamps(
@@ -347,7 +355,7 @@ export interface Spec extends TurboModule {
    * @param instanceId - Unique ID for this engine instance
    * @param requestId - Unique ID for this generation (included in chunk/end/error events for routing)
    * @param text - Text to convert to speech
-   * @param options - Generation options (sid, speed, referenceAudio, referenceText, numSteps, silenceScale, extra)
+   * @param options - Same shape as batch TTS; reference streaming is **Pocket-only** (Zipvoice cloning uses non-streaming generate).
    */
   generateTtsStream(
     instanceId: string,
@@ -530,6 +538,13 @@ export interface Spec extends TurboModule {
    */
   getAssetPackPath(packName: string): Promise<string | null>;
 
+  /**
+   * Read the contents of a text file from the bundled assets (Android) or main bundle (iOS).
+   * @param assetPath The relative path to the asset file (e.g., 'model_licenses/asr-models-license-status.csv')
+   * @returns Resolves with the string content of the file or rejects if the file cannot be read.
+   */
+  readAssetFileAsUtf8(assetPath: string): Promise<string>;
+
   // ==================== Helper - Extraction ====================
 
   /**
@@ -552,6 +567,67 @@ export interface Spec extends TurboModule {
    */
   cancelExtractTarBz2(): Promise<void>;
 
+  /**
+   * Extract a .tar.zst (or .zst) archive to a target folder.
+   * Returns { success, path } or { success, reason }.
+   */
+  extractTarZst(
+    sourcePath: string,
+    targetPath: string,
+    force: boolean
+  ): Promise<{
+    success: boolean;
+    path?: string;
+    sha256?: string;
+    reason?: string;
+  }>;
+
+  /**
+   * Cancel any in-progress tar.zst extraction.
+   */
+  cancelExtractTarZst(): Promise<void>;
+
+  /**
+   * Cancel extraction for a specific source archive path (per-operation cancel for parallel extractions).
+   */
+  cancelExtractBySourcePath(sourcePath: string): Promise<void>;
+
+  /**
+   * List asset paths of .tar.zst and .tar.bz2 archives in a PAD pack when stored as APK_ASSETS.
+   * Android only; returns [] when pack is not available or not APK_ASSETS. Used by getBundledArchives.
+   */
+  listBundledArchiveAssetPaths(packName: string): Promise<string[]>;
+
+  /**
+   * Extract a .tar.zst archive from Android assets (AssetManager) to a target folder. Android only.
+   * Streams from asset; no copy of the archive to disk. Used when PAD pack is APK_ASSETS.
+   */
+  extractTarZstFromAsset(
+    assetPath: string,
+    targetPath: string,
+    force: boolean
+  ): Promise<{
+    success: boolean;
+    path?: string;
+    sha256?: string;
+    reason?: string;
+  }>;
+
+  /**
+   * Extract a .tar.bz2 archive from Android assets (AssetManager) to a target folder. Android only.
+   * Streams from asset; no copy of the archive to disk. Used when PAD pack is APK_ASSETS.
+   */
+  extractTarBz2FromAsset(
+    assetPath: string,
+    targetPath: string,
+    force: boolean
+  ): Promise<{
+    success: boolean;
+    path?: string;
+    sha256?: string;
+    reason?: string;
+  }>;
+
   /**
    * Compute SHA-256 of a file and return the hex digest.
    */
@@ -578,6 +654,17 @@ export interface Spec extends TurboModule {
    */
   convertAudioToWav16k(inputPath: string, outputPath: string): Promise<void>;
 
+  /**
+   * Decode an audio file to mono float samples in [-1, 1] and the effective sample rate.
+   * Supports the same inputs as convertAudioToFormat (file paths and Android content:// URIs).
+   * On Android, non-WAV formats require FFmpeg prebuilts; WAV may use a fast path via WaveReader.
+   * @param targetSampleRateHz - If > 0, resample to this rate; if 0 or omitted, keep the decoded stream rate.
+   */
+  decodeAudioFileToFloatSamples(
+    inputPath: string,
+    targetSampleRateHz?: number
+  ): Promise<{ samples: number[]; sampleRate: number }>;
+
   // ==================== Execution Provider Methods ====================
 
   /**

package/src/audio/index.ts

@@ -113,3 +113,23 @@ export function convertAudioToWav16k(
 ): Promise<void> {
   return SherpaOnnx.convertAudioToWav16k(inputPath, outputPath);
 }
+
+export type DecodedAudioFloatSamples = {
+  samples: number[];
+  sampleRate: number;
+};
+
+/**
+ * Decode a supported audio file to mono float PCM in [-1, 1] plus sample rate.
+ * Same decode coverage as {@link convertAudioToFormat} (FFmpeg-backed on Android when not WAV).
+ * @param targetSampleRateHz - Resample to this rate when > 0; use native decoded rate when 0 or omitted.
+ */
+export function decodeAudioFileToFloatSamples(
+  inputPath: string,
+  targetSampleRateHz?: number
+): Promise<DecodedAudioFloatSamples> {
+  return SherpaOnnx.decodeAudioFileToFloatSamples(
+    inputPath,
+    targetSampleRateHz ?? 0
+  );
+}