@livekit/agents 1.0.22 → 1.0.24

package/dist/voice/agent_activity.js

@@ -1,4 +1,5 @@
 import { Mutex } from "@livekit/mutex";
+import { ROOT_CONTEXT, trace } from "@opentelemetry/api";
 import { Heap } from "heap-js";
 import { AsyncLocalStorage } from "node:async_hooks";
 import { ReadableStream } from "node:stream/web";
@@ -11,6 +12,7 @@ import { isSameToolChoice, isSameToolContext } from "../llm/tool_context.js";
 import { log } from "../log.js";
 import { DeferredReadableStream } from "../stream/deferred_stream.js";
 import { STT } from "../stt/stt.js";
+import { traceTypes, tracer } from "../telemetry/index.js";
 import { splitWords } from "../tokenize/basic/word.js";
 import { TTS } from "../tts/tts.js";
 import { Future, Task, cancelAndWait, waitFor } from "../utils.js";
@@ -124,6 +126,11 @@ class AgentActivity {
   async start() {
     const unlock = await this.lock.lock();
     try {
+      const startSpan = tracer.startSpan({
+        name: "start_agent_activity",
+        attributes: { [traceTypes.ATTR_AGENT_LABEL]: this.agent.id },
+        context: ROOT_CONTEXT
+      });
       this.agent._agentActivity = this;
       if (this.llm instanceof RealtimeModel) {
         this.realtimeSession = this.llm.session();
@@ -191,15 +198,22 @@ class AgentActivity {
         turnDetector: typeof this.turnDetection === "string" ? void 0 : this.turnDetection,
         turnDetectionMode: this.turnDetectionMode,
         minEndpointingDelay: this.agentSession.options.minEndpointingDelay,
-        maxEndpointingDelay: this.agentSession.options.maxEndpointingDelay
+        maxEndpointingDelay: this.agentSession.options.maxEndpointingDelay,
+        rootSpanContext: this.agentSession.rootSpanContext
       });
       this.audioRecognition.start();
       this.started = true;
       this._mainTask = Task.from(({ signal }) => this.mainTask(signal));
+      const onEnterTask = tracer.startActiveSpan(async () => this.agent.onEnter(), {
+        name: "on_enter",
+        context: trace.setSpan(ROOT_CONTEXT, startSpan),
+        attributes: { [traceTypes.ATTR_AGENT_LABEL]: this.agent.id }
+      });
       this.createSpeechTask({
-        task: Task.from(() => this.agent.onEnter()),
+        task: Task.from(() => onEnterTask),
         name: "AgentActivity_onEnter"
       });
+      startSpan.end();
     } finally {
       unlock();
     }
@@ -900,9 +914,25 @@ ${instructions}` : instructions,
       this.agentSession._updateAgentState("listening");
     }
   }
-  // TODO(brian): PR3 - Wrap entire pipelineReplyTask() method with tracer.startActiveSpan('agent_turn')
-  async pipelineReplyTask(speechHandle, chatCtx, toolCtx, modelSettings, replyAbortController, instructions, newMessage, toolsMessages) {
+  _pipelineReplyTaskImpl = async ({
+    speechHandle,
+    chatCtx,
+    toolCtx,
+    modelSettings,
+    replyAbortController,
+    instructions,
+    newMessage,
+    toolsMessages,
+    span
+  }) => {
     var _a, _b, _c;
+    span.setAttribute(traceTypes.ATTR_SPEECH_ID, speechHandle.id);
+    if (instructions) {
+      span.setAttribute(traceTypes.ATTR_INSTRUCTIONS, instructions);
+    }
+    if (newMessage) {
+      span.setAttribute(traceTypes.ATTR_USER_INPUT, newMessage.textContent || "");
+    }
     speechHandleStorage.enterWith(speechHandle);
     const audioOutput = this.agentSession.output.audioEnabled ? this.agentSession.output.audio : null;
     const transcriptionOutput = this.agentSession.output.transcriptionEnabled ? this.agentSession.output.transcription : null;
@@ -1011,6 +1041,7 @@ ${instructions}` : instructions,
         msg.createdAt = replyStartedAt;
       }
       this.agent._chatCtx.insert(toolsMessages);
+      this.agentSession._toolItemsAdded(toolsMessages);
     }
     if (speechHandle.interrupted) {
       this.logger.debug(
@@ -1175,10 +1206,50 @@ ${instructions}` : instructions,
         msg.createdAt = replyStartedAt;
       }
       this.agent._chatCtx.insert(toolMessages);
+      this.agentSession._toolItemsAdded(toolMessages);
     }
-  }
+  };
+  pipelineReplyTask = async (speechHandle, chatCtx, toolCtx, modelSettings, replyAbortController, instructions, newMessage, toolsMessages) => tracer.startActiveSpan(
+    async (span) => this._pipelineReplyTaskImpl({
+      speechHandle,
+      chatCtx,
+      toolCtx,
+      modelSettings,
+      replyAbortController,
+      instructions,
+      newMessage,
+      toolsMessages,
+      span
+    }),
+    {
+      name: "agent_turn",
+      context: this.agentSession.rootSpanContext
+    }
+  );
   async realtimeGenerationTask(speechHandle, ev, modelSettings, replyAbortController) {
+    return tracer.startActiveSpan(
+      async (span) => this._realtimeGenerationTaskImpl({
+        speechHandle,
+        ev,
+        modelSettings,
+        replyAbortController,
+        span
+      }),
+      {
+        name: "agent_turn",
+        context: this.agentSession.rootSpanContext
+      }
+    );
+  }
+  async _realtimeGenerationTaskImpl({
+    speechHandle,
+    ev,
+    modelSettings,
+    replyAbortController,
+    span
+  }) {
     var _a, _b, _c;
+    span.setAttribute(traceTypes.ATTR_SPEECH_ID, speechHandle.id);
     speechHandleStorage.enterWith(speechHandle);
     if (!this.realtimeSession) {
       throw new Error("realtime session is not initialized");
@@ -1320,6 +1391,8 @@ ${instructions}` : instructions,
     );
     const onToolExecutionStarted = (f) => {
       speechHandle._itemAdded([f]);
+      this.agent._chatCtx.items.push(f);
+      this.agentSession._toolItemsAdded([f]);
     };
     const onToolExecutionCompleted = (out) => {
       if (out.toolCallOutput) {
@@ -1469,6 +1542,9 @@ ${instructions}` : instructions,
       }
       const chatCtx = this.realtimeSession.chatCtx.copy();
       chatCtx.items.push(...functionToolsExecutedEvent.functionCallOutputs);
+      this.agentSession._toolItemsAdded(
+        functionToolsExecutedEvent.functionCallOutputs
+      );
       try {
         await this.realtimeSession.updateChatCtx(chatCtx);
       } catch (error) {
@@ -1557,15 +1633,25 @@ ${instructions}` : instructions,
     speechHandle._markScheduled();
     this.wakeupMainTask();
   }
-  // TODO(brian): PR3 - Wrap entire drain() method with tracer.startActiveSpan('drain_agent_activity', { attributes: { 'lk.agent_label': this.agent.label } })
   async drain() {
+    return tracer.startActiveSpan(async (span) => this._drainImpl(span), {
+      name: "drain_agent_activity",
+      context: ROOT_CONTEXT
+    });
+  }
+  async _drainImpl(span) {
     var _a;
+    span.setAttribute(traceTypes.ATTR_AGENT_LABEL, this.agent.id);
     const unlock = await this.lock.lock();
     try {
       if (this._draining) return;
       this.cancelPreemptiveGeneration();
+      const onExitTask = tracer.startActiveSpan(async () => this.agent.onExit(), {
+        name: "on_exit",
+        attributes: { [traceTypes.ATTR_AGENT_LABEL]: this.agent.id }
+      });
       this.createSpeechTask({
-        task: Task.from(() => this.agent.onExit()),
+        task: Task.from(() => onExitTask),
         name: "AgentActivity_onExit"
       });
       this.wakeupMainTask();