react-native-davoice 1.0.11 → 1.0.13

package/speech/index.ts

@@ -34,6 +34,12 @@ function sleep(ms: number) {
   return new Promise<void>((r) => setTimeout(r, ms));
 }
 
+let speechOpSeq = 0;
+function nextSpeechOpId(prefix: string) {
+  speechOpSeq += 1;
+  return `${prefix}-${speechOpSeq}-${Date.now()}`;
+}
+
 // If you use typed-array -> base64, Buffer is convenient (works in RN)
 let toBase64: (u8: Uint8Array) => string;
 try {
@@ -184,6 +190,8 @@ class Speech {
   // top of file (new state)
   private lastLocale: string | null = null;
   private lastModel: string | null = null;
+  private lastOnboardingJsonPath: string | null = null;
+  private hasCompletedFullInit = false;
   private iosTtsOnly = false; // when true, use NativeTTS directly on iOS
 
 
@@ -386,11 +394,13 @@ class Speech {
 
     this.lastLocale = opts.locale;
     this.lastModel = modelPath;
+    this.lastOnboardingJsonPath = opts.onboardingJsonPath ?? null;
 
     if (Platform.OS === 'ios' && NativeSpeech?.initAll) {
       this.iosTtsOnly = false;                  // full unified mode
       this.teardownListeners();                 // re-wire listeners for unified
       const r = await NativeSpeech.initAll({ ...opts, model: modelPath });
+      this.hasCompletedFullInit = true;
       this.ensureListeners();
       return r;
     }
@@ -440,6 +450,7 @@ class Speech {
     const modelExt = this.resolveModelExt(opts.model);
     console.log('[MODELDBG] initAll.modelExt (resolved)=', modelExt);
     await NativeTTS.initTTS({ model: modelPath, modelExt });
+    this.hasCompletedFullInit = true;
   }
 
   async destroyAll() {
@@ -450,6 +461,7 @@ class Speech {
     // iOS unified
     if (Platform.OS === 'ios' && NativeSpeech?.destroyAll) {
       const r = await NativeSpeech.destroyAll();
+      this.hasCompletedFullInit = false;
       this.iosTtsOnly = false;
       this.lastLocale = this.lastLocale ?? null;
       this.teardownListeners();
@@ -463,10 +475,97 @@ class Speech {
         NativeSTT.destroySpeech(() => res());
       });
     } catch {}
+    this.hasCompletedFullInit = false;
     this.teardownListeners();
     return 'Destroyed';
   }
 
+  async initWithoutModel() {
+    if (!this.hasCompletedFullInit || !this.lastLocale) {
+      throw new Error('initWithoutModel() requires a successful initAll() first.');
+    }
+
+    if (Platform.OS === 'ios' && NativeSpeech?.initWithoutModel) {
+      this.iosTtsOnly = false;
+      this.teardownListeners();
+      const r = await NativeSpeech.initWithoutModel({
+        locale: this.lastLocale,
+        onboardingJsonPath: this.lastOnboardingJsonPath,
+      });
+      this.ensureListeners();
+      return r;
+    }
+
+    if (!NativeSTT) {
+      throw new Error('Missing native STT bridge.');
+    }
+
+    this.ensureListeners();
+    return new Promise<void>((resolve, reject) => {
+      try {
+        if (typeof NativeSTT.initWithoutModel === 'function') {
+          NativeSTT.initWithoutModel(
+            this.lastLocale,
+            { onboardingJsonPath: this.lastOnboardingJsonPath ?? null },
+            (err: string) => (err ? reject(new Error(err)) : resolve()),
+          );
+          return;
+        }
+
+        if (Platform.OS === 'android') {
+          NativeSTT.startSpeech(
+            this.lastLocale,
+            {
+              EXTRA_LANGUAGE_MODEL: 'LANGUAGE_MODEL_FREE_FORM',
+              EXTRA_MAX_RESULTS: 5,
+              EXTRA_PARTIAL_RESULTS: true,
+              REQUEST_PERMISSIONS_AUTO: true,
+              onboardingJsonPath: this.lastOnboardingJsonPath ?? null,
+            },
+            (err: string) => (err ? reject(new Error(err)) : resolve()),
+          );
+          return;
+        }
+
+        NativeSTT.startSpeech(this.lastLocale, (err: string) =>
+          err ? reject(new Error(err)) : resolve(),
+        );
+      } catch (e) {
+        reject(e as any);
+      }
+    });
+  }
+
+  async destroyWihtouModel() {
+    if (!this.hasCompletedFullInit) {
+      throw new Error('destroyWihtouModel() requires a successful initAll() first.');
+    }
+
+    if (Platform.OS === 'ios' && NativeSpeech?.destroyWihtouModel) {
+      return NativeSpeech.destroyWihtouModel();
+    }
+
+    return new Promise<void>((resolve) => {
+      try {
+        if (typeof NativeSTT?.destroyWihtouModel === 'function') {
+          NativeSTT.destroyWihtouModel(() => resolve());
+          return;
+        }
+        if (!NativeSTT?.destroySpeech) {
+          resolve();
+          return;
+        }
+        NativeSTT.destroySpeech(() => resolve());
+      } catch {
+        resolve();
+      }
+    });
+  }
+
+  async destroyWithoutModel() {
+    return this.destroyWihtouModel();
+  }
+
   // ---------- STT ----------
   async start(locale: string, options: Record<string, any> = {}) {
     this.ensureListeners();
@@ -524,9 +623,11 @@ class Speech {
     return this.start(locale);
   }
   
-  // XXX BUG THE NATIVE SIDE DOES NOT REALLY AWAITS
+  // TODO: CHECK THE NATIVE SIDE DOES NOT REALLY AWAITS
   async pauseSpeechRecognition(): Promise<void> {
     this.logCall('pauseSpeechRecognitionLite');
+    const opId = nextSpeechOpId('pause-stt');
+    const startedAt = Date.now();
 
     const mod: any = Platform.OS === 'ios' ? NativeSpeech : NativeSTT;
     const fn = mod?.pauseSpeechRecognitionLite;
@@ -537,7 +638,9 @@ class Speech {
     }
 
     if (Platform.OS === 'ios' && typeof mod?.pauseSpeechRecognitionLiteAsync === 'function') {
+      dbg('[pauseSpeechRecognitionLiteAsync] begin', { opId, timeoutMs: 1500 });
       const result = await mod.pauseSpeechRecognitionLiteAsync(1500);
+      dbg('[pauseSpeechRecognitionLiteAsync] resolved', { opId, elapsedMs: Date.now() - startedAt, result });
       if (result?.ok === false) dbgErr('pauseSpeechRecognitionLiteAsync failed', result?.reason);
       return;
     }
@@ -569,8 +672,14 @@ class Speech {
     });
   }
 
-  async unPauseSpeechRecognition(times: number, preFetchMs: number = 0): Promise<void> {
+  async unPauseSpeechRecognition(
+    times: number,
+    preFetchMs: number = 0,
+    timeoutMs: number = 2500,
+  ): Promise<void> {
     this.logCall('unPauseSpeechRecognitionLite', { times, preFetchMs });
+    const opId = nextSpeechOpId('unpause-stt');
+    const startedAt = Date.now();
 
     const mod: any = Platform.OS === 'ios' ? NativeSpeech : NativeSTT;
     const fn = mod?.unPauseSpeechRecognitionLite;
@@ -581,7 +690,15 @@ class Speech {
     }
 
     if (Platform.OS === 'ios' && typeof mod?.unPauseSpeechRecognitionLiteAsync === 'function') {
-      const result = await mod.unPauseSpeechRecognitionLiteAsync(times, preFetchMs, 2500);
+      dbg('[unPauseSpeechRecognitionLiteAsync] begin', { opId, times, preFetchMs, timeoutMs });
+      const result = await mod.unPauseSpeechRecognitionLiteAsync(times, preFetchMs, timeoutMs);
+      dbg('[unPauseSpeechRecognitionLiteAsync] resolved', {
+        opId,
+        times,
+        preFetchMs,
+        elapsedMs: Date.now() - startedAt,
+        result,
+      });
       if (result?.ok === false) dbgErr('unPauseSpeechRecognitionLiteAsync failed', result?.reason);
       return;
     }
@@ -790,6 +907,23 @@ class Speech {
     return results.every(Boolean);
   }
 
+  async setAECEnabled(enabled: boolean): Promise<void> {
+    const value = !!enabled;
+
+    if (Platform.OS === 'ios' && NativeSpeech?.setAECEnabled) {
+      await NativeSpeech.setAECEnabled(value);
+      return;
+    }
+
+    if (NativeSTT?.setAECEnabled) {
+      await NativeSTT.setAECEnabled(value);
+      return;
+    }
+    console.warn("AEC not supported on this platform");
+    return;
+    //throw new Error('setAECEnabled not available on this platform.');
+  }
+
   // ---------- TTS ----------
   async initTTS(modelOrConfig: ModelRef | { model: ModelRef }) {
     const cfg =