@livekit/agents 1.0.48 → 1.0.50

package/dist/voice/agent_activity.cjs

@@ -341,13 +341,27 @@ class AgentActivity {
   attachAudioInput(audioStream) {
     void this.audioStream.close();
     this.audioStream = new import_multi_input_stream.MultiInputStream();
+    const aecWarmupAudioFilter = new import_web.TransformStream({
+      transform: (frame, controller) => {
+        const shouldDiscardForAecWarmup = this.agentSession.agentState === "speaking" && this.agentSession._aecWarmupRemaining > 0;
+        if (!shouldDiscardForAecWarmup) {
+          controller.enqueue(frame);
+        }
+      }
+    });
     this.audioStreamId = this.audioStream.addInputStream(audioStream);
-    const [realtimeAudioStream, recognitionAudioStream] = this.audioStream.stream.tee();
-    if (this.realtimeSession) {
+    if (this.realtimeSession && this.audioRecognition) {
+      const [realtimeAudioStream, recognitionAudioStream] = this.audioStream.stream.pipeThrough(aecWarmupAudioFilter).tee();
       this.realtimeSession.setInputAudioStream(realtimeAudioStream);
-    }
-    if (this.audioRecognition) {
       this.audioRecognition.setInputAudioStream(recognitionAudioStream);
+    } else if (this.realtimeSession) {
+      this.realtimeSession.setInputAudioStream(
+        this.audioStream.stream.pipeThrough(aecWarmupAudioFilter)
+      );
+    } else if (this.audioRecognition) {
+      this.audioRecognition.setInputAudioStream(
+        this.audioStream.stream.pipeThrough(aecWarmupAudioFilter)
+      );
     }
   }
   detachAudioInput() {
@@ -543,6 +557,9 @@ class AgentActivity {
   }
   interruptByAudioActivity() {
     var _a, _b;
+    if (this.agentSession._aecWarmupRemaining > 0) {
+      return;
+    }
     if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.turnDetection) {
       return;
     }
@@ -719,6 +736,9 @@ class AgentActivity {
           throw new Error("Speech queue is empty");
         }
         const speechHandle = heapItem[2];
+        if (speechHandle.interrupted || speechHandle.done()) {
+          continue;
+        }
         this._currentSpeech = speechHandle;
         speechHandle._authorizeGeneration();
         await speechHandle._waitForGeneration();
@@ -861,7 +881,16 @@ ${instructions}`;
       speech.interrupt(force);
     }
     (_a = this.realtimeSession) == null ? void 0 : _a.interrupt();
-    if (currentSpeech === void 0) {
+    if (force) {
+      for (const task of this.speechTasks) {
+        task.cancel();
+      }
+      if (currentSpeech && !currentSpeech.done()) {
+        currentSpeech._markDone();
+      }
+      this.speechQueue.clear();
+      future.resolve();
+    } else if (currentSpeech === void 0) {
       future.resolve();
     } else {
       currentSpeech.addDoneCallback(() => {
@@ -1229,9 +1258,7 @@ ${instructions}`;
         audioOutput.clearBuffer();
       }
       replyAbortController.abort();
-      await Promise.allSettled(
-        tasks.map((task) => task.cancelAndWait(AgentActivity.REPLY_TASK_CANCEL_TIMEOUT))
-      );
+      await (0, import_utils.cancelAndWait)(tasks, AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
       let forwardedText = (textOut == null ? void 0 : textOut.text) || "";
       if (audioOutput) {
         const playbackEv = await audioOutput.waitForPlayout();
@@ -1861,6 +1888,10 @@ ${instructions}`;
     const unlock = await this.lock.lock();
     try {
       this.cancelPreemptiveGeneration();
+      await (0, import_utils.cancelAndWait)(Array.from(this.speechTasks), AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
+      if (this._currentSpeech && !this._currentSpeech.done()) {
+        this._currentSpeech._markDone();
+      }
       await this._closeSessionResources();
       if (this._mainTask) {
         await this._mainTask.cancelAndWait();