@layercode/js-sdk 2.8.0 → 2.8.1

package/dist/layercode-js-sdk.esm.js

@@ -5945,6 +5945,7 @@ class LayercodeClient {
         this._emitAudioInput();
         this.AMPLITUDE_MONITORING_SAMPLE_RATE = 2;
         this._websocketUrl = DEFAULT_WS_URL;
+        this.audioOutputReady = null;
         this.wavRecorder = new WavRecorder({ sampleRate: 8000 }); // TODO should be set my fetched agent config
         this.wavPlayer = new WavStreamPlayer({
             finishedPlayingCallback: this._clientResponseAudioReplayFinished.bind(this),
@@ -6106,12 +6107,19 @@ class LayercodeClient {
         this.status = status;
         this.options.onStatusChange(status);
     }
+    async _waitForAudioOutputReady() {
+        if (!this.audioOutputReady) {
+            return;
+        }
+        await this.audioOutputReady;
+    }
     _setAgentSpeaking(isSpeaking) {
-        if (this.agentIsSpeaking === isSpeaking) {
+        const shouldReportSpeaking = this.audioOutput && isSpeaking;
+        if (this.agentIsSpeaking === shouldReportSpeaking) {
             return;
         }
-        this.agentIsSpeaking = isSpeaking;
-        this.options.onAgentSpeakingChange(isSpeaking);
+        this.agentIsSpeaking = shouldReportSpeaking;
+        this.options.onAgentSpeakingChange(shouldReportSpeaking);
     }
     _setUserSpeaking(isSpeaking) {
         const shouldReportSpeaking = this._shouldCaptureUserAudio() && isSpeaking;
@@ -6179,7 +6187,6 @@ class LayercodeClient {
                     if (message.role === 'assistant') {
                         // Start tracking new agent turn
                         console.debug('Agent turn started, will track new turn ID from audio/text');
-                        this._setAgentSpeaking(true);
                         this._setUserSpeaking(false);
                     }
                     else if (message.role === 'user' && !this.pushToTalkEnabled) {
@@ -6201,6 +6208,7 @@ class LayercodeClient {
                     break;
                 }
                 case 'response.audio':
+                    await this._waitForAudioOutputReady();
                     this._setAgentSpeaking(true);
                     const audioBuffer = base64ToArrayBuffer(message.content);
                     this.wavPlayer.add16BitPCM(audioBuffer, message.turn_id);
@@ -6427,6 +6435,7 @@ class LayercodeClient {
             }
             else {
                 this.wavPlayer.mute();
+                this._setAgentSpeaking(false);
             }
         }
     }
@@ -6485,7 +6494,9 @@ class LayercodeClient {
             this.setupVadConfig(config);
             // Bind the websocket message callbacks
             this.bindWebsocketMessageCallbacks(ws, config);
-            await this.setupAudioOutput();
+            const audioOutputReady = this.setupAudioOutput();
+            this.audioOutputReady = audioOutputReady;
+            await audioOutputReady;
         }
         catch (error) {
             console.error('Error connecting to Layercode agent:', error);