@livekit/agents 1.0.35 → 1.0.36

package/dist/voice/agent_activity.cjs

@@ -974,10 +974,12 @@ ${instructions}` : instructions,
       replyAbortController
     );
     tasks.push(llmTask);
-    const [ttsTextInput, llmOutput] = llmGenData.textStream.tee();
     let ttsTask = null;
     let ttsStream = null;
+    let llmOutput;
     if (audioOutput) {
+      const [ttsTextInput, textOutput] = llmGenData.textStream.tee();
+      llmOutput = textOutput;
       [ttsTask, ttsStream] = (0, import_generation.performTTSInference)(
         (...args) => this.agent.ttsNode(...args),
         ttsTextInput,
@@ -985,6 +987,8 @@ ${instructions}` : instructions,
         replyAbortController
       );
       tasks.push(ttsTask);
+    } else {
+      llmOutput = llmGenData.textStream;
     }
     await speechHandle.waitIfNotInterrupted([speechHandle._waitForScheduled()]);
     if (newMessage && speechHandle.scheduled) {
@@ -1031,9 +1035,15 @@ ${instructions}` : instructions,
     } else {
       textOut == null ? void 0 : textOut.firstTextFut.await.finally(onFirstFrame);
     }
-    const onToolExecutionStarted = (_) => {
+    const onToolExecutionStarted = (f) => {
+      speechHandle._itemAdded([f]);
+      this.agent._chatCtx.items.push(f);
+      this.agentSession._toolItemsAdded([f]);
     };
-    const onToolExecutionCompleted = (_) => {
+    const onToolExecutionCompleted = (out) => {
+      if (out.toolCallOutput) {
+        speechHandle._itemAdded([out.toolCallOutput]);
+      }
     };
     const [executeToolsTask, toolOutput] = (0, import_generation.performToolExecutions)({
       session: this.agentSession,
@@ -1093,6 +1103,7 @@ ${instructions}` : instructions,
         });
         chatCtx.insert(message);
         this.agent._chatCtx.insert(message);
+        speechHandle._itemAdded([message]);
         this.agentSession._conversationItemAdded(message);
       }
       if (this.agentSession.agentState === "speaking") {
@@ -1116,6 +1127,7 @@ ${instructions}` : instructions,
       });
       chatCtx.insert(message);
       this.agent._chatCtx.insert(message);
+      speechHandle._itemAdded([message]);
       this.agentSession._conversationItemAdded(message);
       this.logger.info(
         { speech_id: speechHandle.id, message: textOut.text },
@@ -1184,24 +1196,12 @@ ${instructions}` : instructions,
     ];
     if (shouldGenerateToolReply) {
       chatCtx.insert(toolMessages);
-      const handle = import_speech_handle.SpeechHandle.create({
-        allowInterruptions: speechHandle.allowInterruptions,
-        stepIndex: speechHandle._stepIndex + 1,
-        parent: speechHandle
-      });
-      this.agentSession.emit(
-        import_events.AgentSessionEventTypes.SpeechCreated,
-        (0, import_events.createSpeechCreatedEvent)({
-          userInitiated: false,
-          source: "tool_response",
-          speechHandle: handle
-        })
-      );
+      speechHandle._numSteps += 1;
       const respondToolChoice = draining || modelSettings.toolChoice === "none" ? "none" : "auto";
       const toolResponseTask = this.createSpeechTask({
         task: import_utils.Task.from(
           () => this.pipelineReplyTask(
-            handle,
+            speechHandle,
             chatCtx,
             toolCtx,
             { toolChoice: respondToolChoice },
@@ -1211,11 +1211,11 @@ ${instructions}` : instructions,
             toolMessages
           )
         ),
-        ownedSpeechHandle: handle,
+        ownedSpeechHandle: speechHandle,
         name: "AgentActivity.pipelineReply"
       });
       toolResponseTask.finally(() => this.onPipelineReplyDone());
-      this.scheduleSpeech(handle, import_speech_handle.SpeechHandle.SPEECH_PRIORITY_NORMAL, true);
+      this.scheduleSpeech(speechHandle, import_speech_handle.SpeechHandle.SPEECH_PRIORITY_NORMAL, true);
     } else if (functionToolsExecutedEvent.functionCallOutputs.length > 0) {
       for (const msg of toolMessages) {
         msg.createdAt = replyStartedAt;