@layercode/js-sdk 1.0.13 → 1.0.14

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

@@ -3475,6 +3475,7 @@ class LayercodeClient {
             authorizeSessionEndpoint: options.authorizeSessionEndpoint,
             metadata: options.metadata || {},
             vadEnabled: options.vadEnabled || true,
+            vadResumeDelay: options.vadResumeDelay || 500,
             onConnect: options.onConnect || (() => { }),
             onDisconnect: options.onDisconnect || (() => { }),
             onError: options.onError || (() => { }),
@@ -3497,10 +3498,10 @@ class LayercodeClient {
                 model: 'v5',
                 // baseAssetPath: '/', // Use if bundling model locally
                 // onnxWASMBasePath: '/', // Use if bundling model locally
-                positiveSpeechThreshold: 0.4,
-                negativeSpeechThreshold: 0.4,
-                redemptionFrames: 8,
-                minSpeechFrames: 20,
+                positiveSpeechThreshold: 0.3,
+                negativeSpeechThreshold: 0.2,
+                redemptionFrames: 25, // Number of frames of silence before onVADMisfire or onSpeechEnd is called. Effectively a delay before restarting.
+                minSpeechFrames: 15,
                 preSpeechPadFrames: 0,
                 onSpeechStart: () => {
                     // Only pause agent audio if it's currently playing
@@ -3511,23 +3512,35 @@ class LayercodeClient {
                     }
                     else {
                         console.log('onSpeechStart: WavPlayer is not playing, VAD will not pause.');
-                        this.vadPausedPlayer = false;
                     }
                 },
                 onVADMisfire: () => {
-                    // If the speech detected was for less than minSpeechFrames, this is called instead of onSpeechEnd, and we should resume the assistant audio as it was a false interruption
+                    // If the speech detected was for less than minSpeechFrames, this is called instead of onSpeechEnd, and we should resume the assistant audio as it was a false interruption. We include a configurable delay so the assistant isn't too quick to start speaking again.
                     if (this.vadPausedPlayer) {
-                        console.log('onVADMisfire: VAD paused the player, attempting to resume.');
+                        console.log('onSpeechEnd: VAD paused the player, resuming');
                         this.wavPlayer.play();
                         this.vadPausedPlayer = false; // Reset flag
+                        // Option to extend delay in the case where the transcriber takes longer to detect a new turn
+                        // console.log('onVADMisfire: VAD paused the player, resuming in ' + this.options.vadResumeDelay + 'ms');
+                        // // Add configurable delay before resuming playback
+                        // setTimeout(() => {
+                        //   this.wavPlayer.play();
+                        //   this.vadPausedPlayer = false; // Reset flag
+                        // }, this.options.vadResumeDelay);
                     }
                     else {
                         console.log('onVADMisfire: VAD did not pause the player, no action taken to resume.');
                     }
                 },
-                onSpeechEnd: () => {
-                    // We don't take any action here, as the user speech is for more than minSpeechFrames and is very likely to result in a new turn start from the transcriber
-                },
+                // onSpeechEnd: () => {
+                //   if (this.vadPausedPlayer) {
+                //     console.log('onSpeechEnd: VAD paused the player, resuming');
+                //     this.wavPlayer.play();
+                //     this.vadPausedPlayer = false; // Reset flag
+                //   } else {
+                //     console.log('onSpeechEnd: VAD did not pause the player, not resuming.');
+                //   }
+                // },
             })
                 .then((vad) => {
                 this.vad = vad;