@estuary-ai/sdk 0.1.27 → 0.1.29

package/dist/index.js

@@ -5045,9 +5045,9 @@ var init_livekit_voice = __esm({
       _isActive = false;
       speakingStateCallback = null;
       audioLevelCallback = null;
-      // Audio analyser (via livekit-client's createAudioAnalyser)
-      calculateVolume = null;
-      analyserCleanup = null;
+      // Audio level polling (via LiveKit's server-pushed participant.audioLevel)
+      botParticipant = null;
+      smoothedAudioLevel = 0;
       audioLevelPollTimer = null;
       _isBotSpeaking = false;
       constructor(socketManager, logger) {
@@ -5105,15 +5105,17 @@ var init_livekit_voice = __esm({
             }
             audioElement.play().catch(() => {
             });
-            this.setupAnalyser(track);
+            this.botParticipant = participant;
             if (this._isBotSpeaking) {
-              setTimeout(() => this.startAudioLevelPolling(), 50);
+              this.startAudioLevelPolling();
             }
           }
         });
         this.room.on(RoomEvent.TrackUnsubscribed, (track) => {
           if (track.kind === Track.Kind.Audio) {
-            this.teardownAnalyser();
+            this.stopAudioLevelPolling();
+            this.botParticipant = null;
+            this.smoothedAudioLevel = 0;
             track.detach().forEach((el) => el.remove());
           }
         });
@@ -5121,7 +5123,9 @@ var init_livekit_voice = __esm({
           this.logger.debug("LiveKit room disconnected");
           this._isActive = false;
           this._isBotSpeaking = false;
-          this.teardownAnalyser();
+          this.stopAudioLevelPolling();
+          this.botParticipant = null;
+          this.smoothedAudioLevel = 0;
           this.speakingStateCallback?.(false);
         });
         try {
@@ -5177,7 +5181,9 @@ var init_livekit_voice = __esm({
         } catch {
         }
         this._isBotSpeaking = false;
-        this.teardownAnalyser();
+        this.stopAudioLevelPolling();
+        this.botParticipant = null;
+        this.smoothedAudioLevel = 0;
         this.speakingStateCallback?.(false);
         if (this.room) {
           for (const [, publication] of this.room.localParticipant.trackPublications) {
@@ -5202,7 +5208,9 @@ var init_livekit_voice = __esm({
         this.speakingStateCallback = null;
         this.audioLevelCallback = null;
         this._isBotSpeaking = false;
-        this.teardownAnalyser();
+        this.stopAudioLevelPolling();
+        this.botParticipant = null;
+        this.smoothedAudioLevel = 0;
         if (this.room) {
           this.room.disconnect();
           this.room = null;
@@ -5210,44 +5218,19 @@ var init_livekit_voice = __esm({
         this._isActive = false;
         this._isMuted = false;
       }
-      // ─── Audio Analyser (livekit-client built-in) ───────────────────
-      async setupAnalyser(track) {
-        this.teardownAnalyser();
-        try {
-          const { createAudioAnalyser } = await import('livekit-client');
-          const { analyser, calculateVolume, cleanup } = createAudioAnalyser(track, {
-            fftSize: 256,
-            smoothingTimeConstant: 0.3
-          });
-          if (analyser.context.state === "suspended") {
-            await analyser.context.resume();
-          }
-          this.calculateVolume = calculateVolume;
-          this.analyserCleanup = cleanup;
-          this.logger.debug("Audio analyser created for bot track");
-        } catch (err) {
-          this.logger.debug("Failed to create audio analyser:", err);
-        }
-      }
-      teardownAnalyser() {
-        this.stopAudioLevelPolling();
-        if (this.analyserCleanup) {
-          this.analyserCleanup().catch(() => {
-          });
-          this.analyserCleanup = null;
-        }
-        this.calculateVolume = null;
-      }
+      // ─── Audio Level Polling (participant.audioLevel) ───────────────
       startAudioLevelPolling() {
         if (this.audioLevelPollTimer !== null) return;
-        if (!this.calculateVolume) return;
+        if (!this.botParticipant) return;
+        const alpha = 0.35;
         this.audioLevelPollTimer = setInterval(() => {
-          if (!this.calculateVolume) {
+          if (!this.botParticipant) {
             this.stopAudioLevelPolling();
             return;
           }
-          const vol = this.calculateVolume();
-          this.audioLevelCallback?.(vol);
+          const raw = this.botParticipant.audioLevel ?? 0;
+          this.smoothedAudioLevel += alpha * (raw - this.smoothedAudioLevel);
+          this.audioLevelCallback?.(this.smoothedAudioLevel);
         }, 33);
       }
       stopAudioLevelPolling() {
@@ -5255,6 +5238,7 @@ var init_livekit_voice = __esm({
           clearInterval(this.audioLevelPollTimer);
           this.audioLevelPollTimer = null;
         }
+        this.smoothedAudioLevel = 0;
       }
       // ─── Private ────────────────────────────────────────────────────
       requestToken() {
@@ -9183,24 +9167,24 @@ async function isLiveKitAvailable() {
 async function createVoiceManager(transport, socketManager, sampleRate, logger) {
   if (transport === "websocket") {
     const { WebSocketVoiceManager: WebSocketVoiceManager2 } = await Promise.resolve().then(() => (init_websocket_voice(), websocket_voice_exports));
-    return new WebSocketVoiceManager2(socketManager, sampleRate, logger);
+    return { manager: new WebSocketVoiceManager2(socketManager, sampleRate, logger), resolvedTransport: "websocket" };
   }
   if (transport === "livekit") {
     if (await isLiveKitAvailable()) {
       const { LiveKitVoiceManager: LiveKitVoiceManager2 } = await Promise.resolve().then(() => (init_livekit_voice(), livekit_voice_exports));
-      return new LiveKitVoiceManager2(socketManager, logger);
+      return { manager: new LiveKitVoiceManager2(socketManager, logger), resolvedTransport: "livekit" };
     }
     logger.warn("livekit-client not installed, falling back to WebSocket voice");
     const { WebSocketVoiceManager: WebSocketVoiceManager2 } = await Promise.resolve().then(() => (init_websocket_voice(), websocket_voice_exports));
-    return new WebSocketVoiceManager2(socketManager, sampleRate, logger);
+    return { manager: new WebSocketVoiceManager2(socketManager, sampleRate, logger), resolvedTransport: "websocket" };
   }
   if (transport === "auto") {
     if (await isLiveKitAvailable()) {
       const { LiveKitVoiceManager: LiveKitVoiceManager2 } = await Promise.resolve().then(() => (init_livekit_voice(), livekit_voice_exports));
-      return new LiveKitVoiceManager2(socketManager, logger);
+      return { manager: new LiveKitVoiceManager2(socketManager, logger), resolvedTransport: "livekit" };
     }
     const { WebSocketVoiceManager: WebSocketVoiceManager2 } = await Promise.resolve().then(() => (init_websocket_voice(), websocket_voice_exports));
-    return new WebSocketVoiceManager2(socketManager, sampleRate, logger);
+    return { manager: new WebSocketVoiceManager2(socketManager, sampleRate, logger), resolvedTransport: "websocket" };
   }
   return null;
 }
@@ -9688,11 +9672,12 @@ var EstuaryClient = class extends TypedEventEmitter {
     }
     const transport = this.config.voiceTransport ?? "auto";
     const sampleRate = this.config.audioSampleRate ?? DEFAULT_SAMPLE_RATE;
-    this.voiceManager = await createVoiceManager(transport, this.socketManager, sampleRate, this.logger);
-    if (!this.voiceManager) {
+    const result = await createVoiceManager(transport, this.socketManager, sampleRate, this.logger);
+    if (!result) {
       throw new exports.EstuaryError("VOICE_NOT_SUPPORTED" /* VOICE_NOT_SUPPORTED */, "No voice transport available");
     }
-    if (!this.audioPlayer && typeof AudioContext !== "undefined") {
+    this.voiceManager = result.manager;
+    if (!this.audioPlayer && result.resolvedTransport === "websocket" && typeof AudioContext !== "undefined") {
       this.audioPlayer = new AudioPlayer(sampleRate, (event) => {
         if (event.type === "started") {
           this._hasAutoInterrupted = true;