@livekit/agents 1.1.0-dev.0 → 1.2.0

package/dist/voice/audio_recognition_span.test.js

@@ -1,12 +1,21 @@
 import { ParticipantKind } from "@livekit/rtc-node";
-import { InMemorySpanExporter, SimpleSpanProcessor } from "@opentelemetry/sdk-trace-base";
+import { ROOT_CONTEXT, context as otelContext, trace } from "@opentelemetry/api";
+import {
+  InMemorySpanExporter,
+  SimpleSpanProcessor
+} from "@opentelemetry/sdk-trace-base";
 import { NodeTracerProvider } from "@opentelemetry/sdk-trace-node";
+import { ReadableStream } from "node:stream/web";
 import { describe, expect, it, vi } from "vitest";
+import { ChatContext } from "../llm/chat_context.js";
 import { initializeLogger } from "../log.js";
 import { SpeechEventType } from "../stt/stt.js";
-import { setTracerProvider } from "../telemetry/index.js";
+import { setTracerProvider, tracer } from "../telemetry/index.js";
 import { VAD, VADEventType } from "../vad.js";
-import { AudioRecognition } from "./audio_recognition.js";
+import { AgentSession } from "./agent_session.js";
+import {
+  AudioRecognition
+} from "./audio_recognition.js";
 function setupInMemoryTracing() {
   const exporter = new InMemorySpanExporter();
   const provider = new NodeTracerProvider();
@@ -18,6 +27,20 @@ function setupInMemoryTracing() {
 function spanByName(spans, name) {
   return spans.find((s) => s.name === name);
 }
+function createFakeSession(rootSpanContext = ROOT_CONTEXT) {
+  return {
+    _agentState: "listening",
+    _roomIO: {
+      linkedParticipant: { sid: "p3", identity: "charlie", kind: ParticipantKind.AGENT }
+    },
+    _setUserAwayTimer: vi.fn(),
+    _cancelUserAwayTimer: vi.fn(),
+    _userSpeakingSpan: void 0,
+    _userState: "listening",
+    emit: vi.fn(),
+    rootSpanContext
+  };
+}
 class FakeVADStream extends Object {
   // We intentionally avoid extending the real VADStream (it is not exported as a value in JS output
   // in some bundling contexts). Instead we emulate the async iterator shape used by AudioRecognition.
@@ -56,6 +79,8 @@ class FakeVAD extends VAD {
   }
 }
 const alwaysTrueTurnDetector = {
+  model: "test-turn-detector",
+  provider: "test-provider",
   supportsLanguage: async () => true,
   unlikelyThreshold: async () => void 0,
   predictEndOfTurn: async () => 1
@@ -65,22 +90,14 @@ describe("AudioRecognition user_turn span parity", () => {
   it("creates user_turn and parents eou_detection under it (stt mode)", async () => {
     const { exporter } = setupInMemoryTracing();
     const hooks = {
+      onInterruption: vi.fn(),
       onStartOfSpeech: vi.fn(),
       onVADInferenceDone: vi.fn(),
       onEndOfSpeech: vi.fn(),
       onInterimTranscript: vi.fn(),
       onFinalTranscript: vi.fn(),
       onPreemptiveGeneration: vi.fn(),
-      retrieveChatCtx: () => ({
-        copy() {
-          return this;
-        },
-        addMessage() {
-        },
-        toJSON() {
-          return { items: [] };
-        }
-      }),
+      retrieveChatCtx: () => ChatContext.empty(),
       onEndOfTurn: vi.fn(async () => true)
     };
     const sttEvents = [
@@ -125,6 +142,9 @@ describe("AudioRecognition user_turn span parity", () => {
     const eou = spanByName(spans, "eou_detection");
     expect(userTurn, "user_turn span missing").toBeTruthy();
     expect(eou, "eou_detection span missing").toBeTruthy();
+    if (!userTurn || !eou) {
+      throw new Error("expected user_turn and eou_detection spans");
+    }
     expect(eou.parentSpanId).toBe(userTurn.spanContext().spanId);
     expect(userTurn.attributes["lk.participant_id"]).toBe("p1");
     expect(userTurn.attributes["lk.participant_identity"]).toBe("bob");
@@ -137,22 +157,14 @@ describe("AudioRecognition user_turn span parity", () => {
   it("creates user_turn from VAD startTime (vad mode) and keeps same parenting", async () => {
     const { exporter } = setupInMemoryTracing();
     const hooks = {
+      onInterruption: vi.fn(),
       onStartOfSpeech: vi.fn(),
       onVADInferenceDone: vi.fn(),
       onEndOfSpeech: vi.fn(),
       onInterimTranscript: vi.fn(),
       onFinalTranscript: vi.fn(),
       onPreemptiveGeneration: vi.fn(),
-      retrieveChatCtx: () => ({
-        copy() {
-          return this;
-        },
-        addMessage() {
-        },
-        toJSON() {
-          return { items: [] };
-        }
-      }),
+      retrieveChatCtx: () => ChatContext.empty(),
       onEndOfTurn: vi.fn(async () => true)
     };
     const now = Date.now();
@@ -224,9 +236,64 @@ describe("AudioRecognition user_turn span parity", () => {
     const eou = spanByName(spans, "eou_detection");
     expect(userTurn).toBeTruthy();
     expect(eou).toBeTruthy();
+    if (!userTurn || !eou) {
+      throw new Error("expected user_turn and eou_detection spans");
+    }
     expect(eou.parentSpanId).toBe(userTurn.spanContext().spanId);
     expect(hooks.onStartOfSpeech).toHaveBeenCalled();
     expect(hooks.onEndOfSpeech).toHaveBeenCalled();
   });
+  it("parents user_speaking under user_turn when an explicit speech context is provided", () => {
+    const { exporter } = setupInMemoryTracing();
+    const sessionSpan = tracer.startSpan({ name: "agent_session", context: ROOT_CONTEXT });
+    const sessionContext = trace.setSpan(ROOT_CONTEXT, sessionSpan);
+    const fakeSession = createFakeSession(sessionContext);
+    const userTurn = tracer.startSpan({ name: "user_turn", context: sessionContext });
+    const userTurnContext = trace.setSpan(sessionContext, userTurn);
+    const speakingStartedAt = Date.now() - 100;
+    const speakingEndedAt = Date.now();
+    otelContext.with(userTurnContext, () => {
+      AgentSession.prototype._updateUserState.call(fakeSession, "speaking", {
+        lastSpeakingTime: speakingStartedAt,
+        otelContext: otelContext.active()
+      });
+      AgentSession.prototype._updateUserState.call(fakeSession, "listening", {
+        lastSpeakingTime: speakingEndedAt,
+        otelContext: otelContext.active()
+      });
+    });
+    userTurn.end();
+    sessionSpan.end();
+    const spans = exporter.getFinishedSpans();
+    const userSpeaking = spanByName(spans, "user_speaking");
+    const exportedUserTurn = spanByName(spans, "user_turn");
+    expect(userSpeaking).toBeTruthy();
+    expect(exportedUserTurn).toBeTruthy();
+    if (!userSpeaking || !exportedUserTurn) {
+      throw new Error("expected user_speaking and user_turn spans");
+    }
+    expect(userSpeaking.parentSpanId).toBe(exportedUserTurn.spanContext().spanId);
+    expect(userSpeaking.attributes["lk.participant_id"]).toBe("p3");
+  });
+  it("keeps user_speaking attached to the session root without an explicit speech context", () => {
+    const { exporter } = setupInMemoryTracing();
+    const sessionSpan = tracer.startSpan({ name: "agent_session", context: ROOT_CONTEXT });
+    const sessionContext = trace.setSpan(ROOT_CONTEXT, sessionSpan);
+    const fakeSession = createFakeSession(sessionContext);
+    AgentSession.prototype._updateUserState.call(fakeSession, "speaking", {
+      lastSpeakingTime: Date.now() - 100
+    });
+    AgentSession.prototype._updateUserState.call(fakeSession, "listening", {
+      lastSpeakingTime: Date.now()
+    });
+    sessionSpan.end();
+    const spans = exporter.getFinishedSpans();
+    const userSpeaking = spanByName(spans, "user_speaking");
+    expect(userSpeaking).toBeTruthy();
+    if (!userSpeaking) {
+      throw new Error("expected user_speaking span");
+    }
+    expect(userSpeaking.parentSpanId).toBe(sessionSpan.spanContext().spanId);
+  });
 });
 //# sourceMappingURL=audio_recognition_span.test.js.map

package/dist/voice/audio_recognition_span.test.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/audio_recognition_span.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { ParticipantKind } from '@livekit/rtc-node';\nimport { InMemorySpanExporter, SimpleSpanProcessor } from '@opentelemetry/sdk-trace-base';\nimport { NodeTracerProvider } from '@opentelemetry/sdk-trace-node';\nimport { describe, expect, it, vi } from 'vitest';\nimport { initializeLogger } from '../log.js';\nimport { type SpeechEvent, SpeechEventType } from '../stt/stt.js';\nimport { setTracerProvider } from '../telemetry/index.js';\nimport { VAD, type VADEvent, VADEventType, type VADStream } from '../vad.js';\nimport { AudioRecognition, type _TurnDetector } from './audio_recognition.js';\n\nfunction setupInMemoryTracing() {\n  const exporter = new InMemorySpanExporter();\n  const provider = new NodeTracerProvider();\n  provider.addSpanProcessor(new SimpleSpanProcessor(exporter));\n  provider.register();\n  setTracerProvider(provider);\n  return { exporter };\n}\n\nfunction spanByName(spans: any[], name: string) {\n  return spans.find((s) => s.name === name);\n}\n\nclass FakeVADStream extends (Object as unknown as { new (): VADStream }) {\n  // We intentionally avoid extending the real VADStream (it is not exported as a value in JS output\n  // in some bundling contexts). Instead we emulate the async iterator shape used by AudioRecognition.\n  private events: VADEvent[];\n  private idx = 0;\n  constructor(events: VADEvent[]) {\n    super();\n    this.events = events;\n  }\n  updateInputStream() {}\n  detachInputStream() {}\n  close() {}\n  [Symbol.asyncIterator]() {\n    return this;\n  }\n  async next(): Promise<IteratorResult<VADEvent>> {\n    if (this.idx >= this.events.length) {\n      return { done: true, value: undefined };\n    }\n    const value = this.events[this.idx++]!;\n    return { done: false, value };\n  }\n}\n\nclass FakeVAD extends VAD {\n  label = 'fake-vad';\n  private events: VADEvent[];\n  constructor(events: VADEvent[]) {\n    super({ updateInterval: 1 });\n    this.events = events;\n  }\n  stream(): any {\n    return new FakeVADStream(this.events);\n  }\n}\n\nconst alwaysTrueTurnDetector: _TurnDetector = {\n  supportsLanguage: async () => true,\n  unlikelyThreshold: async () => undefined,\n  predictEndOfTurn: async () => 1.0,\n};\n\ndescribe('AudioRecognition user_turn span parity', () => {\n  initializeLogger({ pretty: false, level: 'silent' });\n\n  it('creates user_turn and parents eou_detection under it (stt mode)', async () => {\n    const { exporter } = setupInMemoryTracing();\n\n    const hooks = {\n      onStartOfSpeech: vi.fn(),\n      onVADInferenceDone: vi.fn(),\n      onEndOfSpeech: vi.fn(),\n      onInterimTranscript: vi.fn(),\n      onFinalTranscript: vi.fn(),\n      onPreemptiveGeneration: vi.fn(),\n      retrieveChatCtx: () =>\n        ({\n          copy() {\n            return this;\n          },\n          addMessage() {},\n          toJSON() {\n            return { items: [] };\n          },\n        }) as any,\n      onEndOfTurn: vi.fn(async () => true),\n    };\n\n    const sttEvents: SpeechEvent[] = [\n      { type: SpeechEventType.START_OF_SPEECH },\n      {\n        type: SpeechEventType.FINAL_TRANSCRIPT,\n        alternatives: [\n          {\n            language: 'en',\n            text: 'hello',\n            startTime: 0,\n            endTime: 0,\n            confidence: 0.9,\n          },\n        ],\n      },\n      { type: SpeechEventType.END_OF_SPEECH },\n    ];\n\n    const sttNode = async () =>\n      new ReadableStream<SpeechEvent>({\n        start(controller) {\n          for (const ev of sttEvents) controller.enqueue(ev);\n          controller.close();\n        },\n      });\n\n    const ar = new AudioRecognition({\n      recognitionHooks: hooks as any,\n      stt: sttNode as any,\n      vad: undefined,\n      turnDetector: alwaysTrueTurnDetector,\n      turnDetectionMode: 'stt',\n      minEndpointingDelay: 0,\n      maxEndpointingDelay: 0,\n      sttModel: 'deepgram-nova2',\n      sttProvider: 'deepgram',\n      getLinkedParticipant: () => ({ sid: 'p1', identity: 'bob', kind: ParticipantKind.AGENT }),\n    });\n\n    await ar.start();\n    // allow background task to drain\n    await new Promise((r) => setTimeout(r, 20));\n    await ar.close();\n\n    const spans = exporter.getFinishedSpans();\n    const userTurn = spanByName(spans, 'user_turn');\n    const eou = spanByName(spans, 'eou_detection');\n    expect(userTurn, 'user_turn span missing').toBeTruthy();\n    expect(eou, 'eou_detection span missing').toBeTruthy();\n\n    expect(eou.parentSpanId).toBe(userTurn.spanContext().spanId);\n\n    // creation-time attributes\n    expect(userTurn.attributes['lk.participant_id']).toBe('p1');\n    expect(userTurn.attributes['lk.participant_identity']).toBe('bob');\n    expect(userTurn.attributes['lk.participant_kind']).toBe('AGENT');\n    expect(userTurn.attributes['gen_ai.request.model']).toBe('deepgram-nova2');\n    expect(userTurn.attributes['gen_ai.provider.name']).toBe('deepgram');\n\n    // end-of-turn attributes\n    expect(userTurn.attributes['lk.user_transcript']).toContain('hello');\n    expect(userTurn.attributes['lk.transcript_confidence']).toBeGreaterThan(0);\n  });\n\n  it('creates user_turn from VAD startTime (vad mode) and keeps same parenting', async () => {\n    const { exporter } = setupInMemoryTracing();\n\n    const hooks = {\n      onStartOfSpeech: vi.fn(),\n      onVADInferenceDone: vi.fn(),\n      onEndOfSpeech: vi.fn(),\n      onInterimTranscript: vi.fn(),\n      onFinalTranscript: vi.fn(),\n      onPreemptiveGeneration: vi.fn(),\n      retrieveChatCtx: () =>\n        ({\n          copy() {\n            return this;\n          },\n          addMessage() {},\n          toJSON() {\n            return { items: [] };\n          },\n        }) as any,\n      onEndOfTurn: vi.fn(async () => true),\n    };\n\n    const now = Date.now();\n    const vadEvents: VADEvent[] = [\n      {\n        type: VADEventType.START_OF_SPEECH,\n        samplesIndex: 0,\n        timestamp: now,\n        speechDuration: 100,\n        silenceDuration: 0,\n        frames: [],\n        probability: 0,\n        inferenceDuration: 0,\n        speaking: true,\n        rawAccumulatedSilence: 0,\n        rawAccumulatedSpeech: 0,\n      },\n      {\n        type: VADEventType.END_OF_SPEECH,\n        samplesIndex: 0,\n        timestamp: now + 200,\n        speechDuration: 100,\n        silenceDuration: 100,\n        frames: [],\n        probability: 0,\n        inferenceDuration: 0,\n        speaking: false,\n        rawAccumulatedSilence: 0,\n        rawAccumulatedSpeech: 0,\n      },\n    ];\n\n    const sttEvents: SpeechEvent[] = [\n      {\n        type: SpeechEventType.FINAL_TRANSCRIPT,\n        alternatives: [\n          {\n            language: 'en',\n            text: 'test',\n            startTime: 0,\n            endTime: 0,\n            confidence: 0.8,\n          },\n        ],\n      },\n    ];\n\n    const sttNode = async () =>\n      new ReadableStream<SpeechEvent>({\n        start(controller) {\n          for (const ev of sttEvents) controller.enqueue(ev);\n          controller.close();\n        },\n      });\n\n    const ar = new AudioRecognition({\n      recognitionHooks: hooks as any,\n      stt: sttNode as any,\n      vad: new FakeVAD(vadEvents) as any,\n      turnDetector: alwaysTrueTurnDetector,\n      turnDetectionMode: 'vad',\n      minEndpointingDelay: 0,\n      maxEndpointingDelay: 0,\n      sttModel: 'stt-model',\n      sttProvider: 'stt-provider',\n      getLinkedParticipant: () => ({ sid: 'p2', identity: 'alice', kind: ParticipantKind.AGENT }),\n    });\n\n    await ar.start();\n    await new Promise((r) => setTimeout(r, 20));\n    await ar.close();\n\n    const spans = exporter.getFinishedSpans();\n    const userTurn = spanByName(spans, 'user_turn');\n    const eou = spanByName(spans, 'eou_detection');\n    expect(userTurn).toBeTruthy();\n    expect(eou).toBeTruthy();\n    expect(eou.parentSpanId).toBe(userTurn.spanContext().spanId);\n\n    expect(hooks.onStartOfSpeech).toHaveBeenCalled();\n    expect(hooks.onEndOfSpeech).toHaveBeenCalled();\n  });\n});\n"],"mappings":"AAGA,SAAS,uBAAuB;AAChC,SAAS,sBAAsB,2BAA2B;AAC1D,SAAS,0BAA0B;AACnC,SAAS,UAAU,QAAQ,IAAI,UAAU;AACzC,SAAS,wBAAwB;AACjC,SAA2B,uBAAuB;AAClD,SAAS,yBAAyB;AAClC,SAAS,KAAoB,oBAAoC;AACjE,SAAS,wBAA4C;AAErD,SAAS,uBAAuB;AAC9B,QAAM,WAAW,IAAI,qBAAqB;AAC1C,QAAM,WAAW,IAAI,mBAAmB;AACxC,WAAS,iBAAiB,IAAI,oBAAoB,QAAQ,CAAC;AAC3D,WAAS,SAAS;AAClB,oBAAkB,QAAQ;AAC1B,SAAO,EAAE,SAAS;AACpB;AAEA,SAAS,WAAW,OAAc,MAAc;AAC9C,SAAO,MAAM,KAAK,CAAC,MAAM,EAAE,SAAS,IAAI;AAC1C;AAEA,MAAM,sBAAuB,OAA4C;AAAA;AAAA;AAAA,EAG/D;AAAA,EACA,MAAM;AAAA,EACd,YAAY,QAAoB;AAC9B,UAAM;AACN,SAAK,SAAS;AAAA,EAChB;AAAA,EACA,oBAAoB;AAAA,EAAC;AAAA,EACrB,oBAAoB;AAAA,EAAC;AAAA,EACrB,QAAQ;AAAA,EAAC;AAAA,EACT,CAAC,OAAO,aAAa,IAAI;AACvB,WAAO;AAAA,EACT;AAAA,EACA,MAAM,OAA0C;AAC9C,QAAI,KAAK,OAAO,KAAK,OAAO,QAAQ;AAClC,aAAO,EAAE,MAAM,MAAM,OAAO,OAAU;AAAA,IACxC;AACA,UAAM,QAAQ,KAAK,OAAO,KAAK,KAAK;AACpC,WAAO,EAAE,MAAM,OAAO,MAAM;AAAA,EAC9B;AACF;AAEA,MAAM,gBAAgB,IAAI;AAAA,EACxB,QAAQ;AAAA,EACA;AAAA,EACR,YAAY,QAAoB;AAC9B,UAAM,EAAE,gBAAgB,EAAE,CAAC;AAC3B,SAAK,SAAS;AAAA,EAChB;AAAA,EACA,SAAc;AACZ,WAAO,IAAI,cAAc,KAAK,MAAM;AAAA,EACtC;AACF;AAEA,MAAM,yBAAwC;AAAA,EAC5C,kBAAkB,YAAY;AAAA,EAC9B,mBAAmB,YAAY;AAAA,EAC/B,kBAAkB,YAAY;AAChC;AAEA,SAAS,0CAA0C,MAAM;AACvD,mBAAiB,EAAE,QAAQ,OAAO,OAAO,SAAS,CAAC;AAEnD,KAAG,mEAAmE,YAAY;AAChF,UAAM,EAAE,SAAS,IAAI,qBAAqB;AAE1C,UAAM,QAAQ;AAAA,MACZ,iBAAiB,GAAG,GAAG;AAAA,MACvB,oBAAoB,GAAG,GAAG;AAAA,MAC1B,eAAe,GAAG,GAAG;AAAA,MACrB,qBAAqB,GAAG,GAAG;AAAA,MAC3B,mBAAmB,GAAG,GAAG;AAAA,MACzB,wBAAwB,GAAG,GAAG;AAAA,MAC9B,iBAAiB,OACd;AAAA,QACC,OAAO;AACL,iBAAO;AAAA,QACT;AAAA,QACA,aAAa;AAAA,QAAC;AAAA,QACd,SAAS;AACP,iBAAO,EAAE,OAAO,CAAC,EAAE;AAAA,QACrB;AAAA,MACF;AAAA,MACF,aAAa,GAAG,GAAG,YAAY,IAAI;AAAA,IACrC;AAEA,UAAM,YAA2B;AAAA,MAC/B,EAAE,MAAM,gBAAgB,gBAAgB;AAAA,MACxC;AAAA,QACE,MAAM,gBAAgB;AAAA,QACtB,cAAc;AAAA,UACZ;AAAA,YACE,UAAU;AAAA,YACV,MAAM;AAAA,YACN,WAAW;AAAA,YACX,SAAS;AAAA,YACT,YAAY;AAAA,UACd;AAAA,QACF;AAAA,MACF;AAAA,MACA,EAAE,MAAM,gBAAgB,cAAc;AAAA,IACxC;AAEA,UAAM,UAAU,YACd,IAAI,eAA4B;AAAA,MAC9B,MAAM,YAAY;AAChB,mBAAW,MAAM,UAAW,YAAW,QAAQ,EAAE;AACjD,mBAAW,MAAM;AAAA,MACnB;AAAA,IACF,CAAC;AAEH,UAAM,KAAK,IAAI,iBAAiB;AAAA,MAC9B,kBAAkB;AAAA,MAClB,KAAK;AAAA,MACL,KAAK;AAAA,MACL,cAAc;AAAA,MACd,mBAAmB;AAAA,MACnB,qBAAqB;AAAA,MACrB,qBAAqB;AAAA,MACrB,UAAU;AAAA,MACV,aAAa;AAAA,MACb,sBAAsB,OAAO,EAAE,KAAK,MAAM,UAAU,OAAO,MAAM,gBAAgB,MAAM;AAAA,IACzF,CAAC;AAED,UAAM,GAAG,MAAM;AAEf,UAAM,IAAI,QAAQ,CAAC,MAAM,WAAW,GAAG,EAAE,CAAC;AAC1C,UAAM,GAAG,MAAM;AAEf,UAAM,QAAQ,SAAS,iBAAiB;AACxC,UAAM,WAAW,WAAW,OAAO,WAAW;AAC9C,UAAM,MAAM,WAAW,OAAO,eAAe;AAC7C,WAAO,UAAU,wBAAwB,EAAE,WAAW;AACtD,WAAO,KAAK,4BAA4B,EAAE,WAAW;AAErD,WAAO,IAAI,YAAY,EAAE,KAAK,SAAS,YAAY,EAAE,MAAM;AAG3D,WAAO,SAAS,WAAW,mBAAmB,CAAC,EAAE,KAAK,IAAI;AAC1D,WAAO,SAAS,WAAW,yBAAyB,CAAC,EAAE,KAAK,KAAK;AACjE,WAAO,SAAS,WAAW,qBAAqB,CAAC,EAAE,KAAK,OAAO;AAC/D,WAAO,SAAS,WAAW,sBAAsB,CAAC,EAAE,KAAK,gBAAgB;AACzE,WAAO,SAAS,WAAW,sBAAsB,CAAC,EAAE,KAAK,UAAU;AAGnE,WAAO,SAAS,WAAW,oBAAoB,CAAC,EAAE,UAAU,OAAO;AACnE,WAAO,SAAS,WAAW,0BAA0B,CAAC,EAAE,gBAAgB,CAAC;AAAA,EAC3E,CAAC;AAED,KAAG,4EAA4E,YAAY;AACzF,UAAM,EAAE,SAAS,IAAI,qBAAqB;AAE1C,UAAM,QAAQ;AAAA,MACZ,iBAAiB,GAAG,GAAG;AAAA,MACvB,oBAAoB,GAAG,GAAG;AAAA,MAC1B,eAAe,GAAG,GAAG;AAAA,MACrB,qBAAqB,GAAG,GAAG;AAAA,MAC3B,mBAAmB,GAAG,GAAG;AAAA,MACzB,wBAAwB,GAAG,GAAG;AAAA,MAC9B,iBAAiB,OACd;AAAA,QACC,OAAO;AACL,iBAAO;AAAA,QACT;AAAA,QACA,aAAa;AAAA,QAAC;AAAA,QACd,SAAS;AACP,iBAAO,EAAE,OAAO,CAAC,EAAE;AAAA,QACrB;AAAA,MACF;AAAA,MACF,aAAa,GAAG,GAAG,YAAY,IAAI;AAAA,IACrC;AAEA,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,YAAwB;AAAA,MAC5B;AAAA,QACE,MAAM,aAAa;AAAA,QACnB,cAAc;AAAA,QACd,WAAW;AAAA,QACX,gBAAgB;AAAA,QAChB,iBAAiB;AAAA,QACjB,QAAQ,CAAC;AAAA,QACT,aAAa;AAAA,QACb,mBAAmB;AAAA,QACnB,UAAU;AAAA,QACV,uBAAuB;AAAA,QACvB,sBAAsB;AAAA,MACxB;AAAA,MACA;AAAA,QACE,MAAM,aAAa;AAAA,QACnB,cAAc;AAAA,QACd,WAAW,MAAM;AAAA,QACjB,gBAAgB;AAAA,QAChB,iBAAiB;AAAA,QACjB,QAAQ,CAAC;AAAA,QACT,aAAa;AAAA,QACb,mBAAmB;AAAA,QACnB,UAAU;AAAA,QACV,uBAAuB;AAAA,QACvB,sBAAsB;AAAA,MACxB;AAAA,IACF;AAEA,UAAM,YAA2B;AAAA,MAC/B;AAAA,QACE,MAAM,gBAAgB;AAAA,QACtB,cAAc;AAAA,UACZ;AAAA,YACE,UAAU;AAAA,YACV,MAAM;AAAA,YACN,WAAW;AAAA,YACX,SAAS;AAAA,YACT,YAAY;AAAA,UACd;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,UAAU,YACd,IAAI,eAA4B;AAAA,MAC9B,MAAM,YAAY;AAChB,mBAAW,MAAM,UAAW,YAAW,QAAQ,EAAE;AACjD,mBAAW,MAAM;AAAA,MACnB;AAAA,IACF,CAAC;AAEH,UAAM,KAAK,IAAI,iBAAiB;AAAA,MAC9B,kBAAkB;AAAA,MAClB,KAAK;AAAA,MACL,KAAK,IAAI,QAAQ,SAAS;AAAA,MAC1B,cAAc;AAAA,MACd,mBAAmB;AAAA,MACnB,qBAAqB;AAAA,MACrB,qBAAqB;AAAA,MACrB,UAAU;AAAA,MACV,aAAa;AAAA,MACb,sBAAsB,OAAO,EAAE,KAAK,MAAM,UAAU,SAAS,MAAM,gBAAgB,MAAM;AAAA,IAC3F,CAAC;AAED,UAAM,GAAG,MAAM;AACf,UAAM,IAAI,QAAQ,CAAC,MAAM,WAAW,GAAG,EAAE,CAAC;AAC1C,UAAM,GAAG,MAAM;AAEf,UAAM,QAAQ,SAAS,iBAAiB;AACxC,UAAM,WAAW,WAAW,OAAO,WAAW;AAC9C,UAAM,MAAM,WAAW,OAAO,eAAe;AAC7C,WAAO,QAAQ,EAAE,WAAW;AAC5B,WAAO,GAAG,EAAE,WAAW;AACvB,WAAO,IAAI,YAAY,EAAE,KAAK,SAAS,YAAY,EAAE,MAAM;AAE3D,WAAO,MAAM,eAAe,EAAE,iBAAiB;AAC/C,WAAO,MAAM,aAAa,EAAE,iBAAiB;AAAA,EAC/C,CAAC;AACH,CAAC;","names":[]}
+{"version":3,"sources":["../../src/voice/audio_recognition_span.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { ParticipantKind } from '@livekit/rtc-node';\nimport { ROOT_CONTEXT, context as otelContext, trace } from '@opentelemetry/api';\nimport {\n  InMemorySpanExporter,\n  type ReadableSpan,\n  SimpleSpanProcessor,\n} from '@opentelemetry/sdk-trace-base';\nimport { NodeTracerProvider } from '@opentelemetry/sdk-trace-node';\nimport { ReadableStream } from 'node:stream/web';\nimport { describe, expect, it, vi } from 'vitest';\nimport { ChatContext } from '../llm/chat_context.js';\nimport { initializeLogger } from '../log.js';\nimport { type SpeechEvent, SpeechEventType } from '../stt/stt.js';\nimport { setTracerProvider, tracer } from '../telemetry/index.js';\nimport { VAD, type VADEvent, VADEventType, type VADStream } from '../vad.js';\nimport { AgentSession } from './agent_session.js';\nimport {\n  AudioRecognition,\n  type RecognitionHooks,\n  type _TurnDetector,\n} from './audio_recognition.js';\nimport type { STTNode } from './io.js';\n\nfunction setupInMemoryTracing() {\n  const exporter = new InMemorySpanExporter();\n  const provider = new NodeTracerProvider();\n  provider.addSpanProcessor(new SimpleSpanProcessor(exporter));\n  provider.register();\n  setTracerProvider(provider);\n  return { exporter };\n}\n\nfunction spanByName(spans: ReadableSpan[], name: string) {\n  return spans.find((s) => s.name === name);\n}\n\nfunction createFakeSession(rootSpanContext = ROOT_CONTEXT): AgentSession {\n  return {\n    _agentState: 'listening',\n    _roomIO: {\n      linkedParticipant: { sid: 'p3', identity: 'charlie', kind: ParticipantKind.AGENT },\n    },\n    _setUserAwayTimer: vi.fn(),\n    _cancelUserAwayTimer: vi.fn(),\n    _userSpeakingSpan: undefined,\n    _userState: 'listening',\n    emit: vi.fn(),\n    rootSpanContext,\n  } as unknown as AgentSession;\n}\n\nclass FakeVADStream extends (Object as unknown as { new (): VADStream }) {\n  // We intentionally avoid extending the real VADStream (it is not exported as a value in JS output\n  // in some bundling contexts). Instead we emulate the async iterator shape used by AudioRecognition.\n  private events: VADEvent[];\n  private idx = 0;\n  constructor(events: VADEvent[]) {\n    super();\n    this.events = events;\n  }\n  updateInputStream() {}\n  detachInputStream() {}\n  close() {}\n  [Symbol.asyncIterator]() {\n    return this;\n  }\n  async next(): Promise<IteratorResult<VADEvent>> {\n    if (this.idx >= this.events.length) {\n      return { done: true, value: undefined };\n    }\n    const value = this.events[this.idx++]!;\n    return { done: false, value };\n  }\n}\n\nclass FakeVAD extends VAD {\n  label = 'fake-vad';\n  private events: VADEvent[];\n  constructor(events: VADEvent[]) {\n    super({ updateInterval: 1 });\n    this.events = events;\n  }\n  stream(): any {\n    return new FakeVADStream(this.events);\n  }\n}\n\nconst alwaysTrueTurnDetector: _TurnDetector = {\n  model: 'test-turn-detector',\n  provider: 'test-provider',\n  supportsLanguage: async () => true,\n  unlikelyThreshold: async () => undefined,\n  predictEndOfTurn: async () => 1.0,\n};\n\ndescribe('AudioRecognition user_turn span parity', () => {\n  initializeLogger({ pretty: false, level: 'silent' });\n\n  it('creates user_turn and parents eou_detection under it (stt mode)', async () => {\n    const { exporter } = setupInMemoryTracing();\n\n    const hooks: RecognitionHooks = {\n      onInterruption: vi.fn(),\n      onStartOfSpeech: vi.fn(),\n      onVADInferenceDone: vi.fn(),\n      onEndOfSpeech: vi.fn(),\n      onInterimTranscript: vi.fn(),\n      onFinalTranscript: vi.fn(),\n      onPreemptiveGeneration: vi.fn(),\n      retrieveChatCtx: () => ChatContext.empty(),\n      onEndOfTurn: vi.fn(async () => true),\n    };\n\n    const sttEvents: SpeechEvent[] = [\n      { type: SpeechEventType.START_OF_SPEECH },\n      {\n        type: SpeechEventType.FINAL_TRANSCRIPT,\n        alternatives: [\n          {\n            language: 'en',\n            text: 'hello',\n            startTime: 0,\n            endTime: 0,\n            confidence: 0.9,\n          },\n        ],\n      },\n      { type: SpeechEventType.END_OF_SPEECH },\n    ];\n\n    const sttNode: STTNode = async () =>\n      new ReadableStream<SpeechEvent | string>({\n        start(controller) {\n          for (const ev of sttEvents) controller.enqueue(ev);\n          controller.close();\n        },\n      });\n\n    const ar = new AudioRecognition({\n      recognitionHooks: hooks,\n      stt: sttNode,\n      vad: undefined,\n      turnDetector: alwaysTrueTurnDetector,\n      turnDetectionMode: 'stt',\n      minEndpointingDelay: 0,\n      maxEndpointingDelay: 0,\n      sttModel: 'deepgram-nova2',\n      sttProvider: 'deepgram',\n      getLinkedParticipant: () => ({ sid: 'p1', identity: 'bob', kind: ParticipantKind.AGENT }),\n    });\n\n    await ar.start();\n    // allow background task to drain\n    await new Promise((r) => setTimeout(r, 20));\n    await ar.close();\n\n    const spans = exporter.getFinishedSpans();\n    const userTurn = spanByName(spans, 'user_turn');\n    const eou = spanByName(spans, 'eou_detection');\n    expect(userTurn, 'user_turn span missing').toBeTruthy();\n    expect(eou, 'eou_detection span missing').toBeTruthy();\n    if (!userTurn || !eou) {\n      throw new Error('expected user_turn and eou_detection spans');\n    }\n\n    expect(eou.parentSpanId).toBe(userTurn.spanContext().spanId);\n\n    // creation-time attributes\n    expect(userTurn.attributes['lk.participant_id']).toBe('p1');\n    expect(userTurn.attributes['lk.participant_identity']).toBe('bob');\n    expect(userTurn.attributes['lk.participant_kind']).toBe('AGENT');\n    expect(userTurn.attributes['gen_ai.request.model']).toBe('deepgram-nova2');\n    expect(userTurn.attributes['gen_ai.provider.name']).toBe('deepgram');\n\n    // end-of-turn attributes\n    expect(userTurn.attributes['lk.user_transcript']).toContain('hello');\n    expect(userTurn.attributes['lk.transcript_confidence']).toBeGreaterThan(0);\n  });\n\n  it('creates user_turn from VAD startTime (vad mode) and keeps same parenting', async () => {\n    const { exporter } = setupInMemoryTracing();\n\n    const hooks: RecognitionHooks = {\n      onInterruption: vi.fn(),\n      onStartOfSpeech: vi.fn(),\n      onVADInferenceDone: vi.fn(),\n      onEndOfSpeech: vi.fn(),\n      onInterimTranscript: vi.fn(),\n      onFinalTranscript: vi.fn(),\n      onPreemptiveGeneration: vi.fn(),\n      retrieveChatCtx: () => ChatContext.empty(),\n      onEndOfTurn: vi.fn(async () => true),\n    };\n\n    const now = Date.now();\n    const vadEvents: VADEvent[] = [\n      {\n        type: VADEventType.START_OF_SPEECH,\n        samplesIndex: 0,\n        timestamp: now,\n        speechDuration: 100,\n        silenceDuration: 0,\n        frames: [],\n        probability: 0,\n        inferenceDuration: 0,\n        speaking: true,\n        rawAccumulatedSilence: 0,\n        rawAccumulatedSpeech: 0,\n      },\n      {\n        type: VADEventType.END_OF_SPEECH,\n        samplesIndex: 0,\n        timestamp: now + 200,\n        speechDuration: 100,\n        silenceDuration: 100,\n        frames: [],\n        probability: 0,\n        inferenceDuration: 0,\n        speaking: false,\n        rawAccumulatedSilence: 0,\n        rawAccumulatedSpeech: 0,\n      },\n    ];\n\n    const sttEvents: SpeechEvent[] = [\n      {\n        type: SpeechEventType.FINAL_TRANSCRIPT,\n        alternatives: [\n          {\n            language: 'en',\n            text: 'test',\n            startTime: 0,\n            endTime: 0,\n            confidence: 0.8,\n          },\n        ],\n      },\n    ];\n\n    const sttNode: STTNode = async () =>\n      new ReadableStream<SpeechEvent | string>({\n        start(controller) {\n          for (const ev of sttEvents) controller.enqueue(ev);\n          controller.close();\n        },\n      });\n\n    const ar = new AudioRecognition({\n      recognitionHooks: hooks,\n      stt: sttNode,\n      vad: new FakeVAD(vadEvents),\n      turnDetector: alwaysTrueTurnDetector,\n      turnDetectionMode: 'vad',\n      minEndpointingDelay: 0,\n      maxEndpointingDelay: 0,\n      sttModel: 'stt-model',\n      sttProvider: 'stt-provider',\n      getLinkedParticipant: () => ({ sid: 'p2', identity: 'alice', kind: ParticipantKind.AGENT }),\n    });\n\n    await ar.start();\n    await new Promise((r) => setTimeout(r, 20));\n    await ar.close();\n\n    const spans = exporter.getFinishedSpans();\n    const userTurn = spanByName(spans, 'user_turn');\n    const eou = spanByName(spans, 'eou_detection');\n    expect(userTurn).toBeTruthy();\n    expect(eou).toBeTruthy();\n    if (!userTurn || !eou) {\n      throw new Error('expected user_turn and eou_detection spans');\n    }\n    expect(eou.parentSpanId).toBe(userTurn.spanContext().spanId);\n\n    expect(hooks.onStartOfSpeech).toHaveBeenCalled();\n    expect(hooks.onEndOfSpeech).toHaveBeenCalled();\n  });\n\n  it('parents user_speaking under user_turn when an explicit speech context is provided', () => {\n    const { exporter } = setupInMemoryTracing();\n    const sessionSpan = tracer.startSpan({ name: 'agent_session', context: ROOT_CONTEXT });\n    const sessionContext = trace.setSpan(ROOT_CONTEXT, sessionSpan);\n    const fakeSession = createFakeSession(sessionContext);\n    const userTurn = tracer.startSpan({ name: 'user_turn', context: sessionContext });\n    const userTurnContext = trace.setSpan(sessionContext, userTurn);\n    const speakingStartedAt = Date.now() - 100;\n    const speakingEndedAt = Date.now();\n\n    otelContext.with(userTurnContext, () => {\n      AgentSession.prototype._updateUserState.call(fakeSession, 'speaking', {\n        lastSpeakingTime: speakingStartedAt,\n        otelContext: otelContext.active(),\n      });\n      AgentSession.prototype._updateUserState.call(fakeSession, 'listening', {\n        lastSpeakingTime: speakingEndedAt,\n        otelContext: otelContext.active(),\n      });\n    });\n\n    userTurn.end();\n    sessionSpan.end();\n\n    const spans = exporter.getFinishedSpans();\n    const userSpeaking = spanByName(spans, 'user_speaking');\n    const exportedUserTurn = spanByName(spans, 'user_turn');\n    expect(userSpeaking).toBeTruthy();\n    expect(exportedUserTurn).toBeTruthy();\n    if (!userSpeaking || !exportedUserTurn) {\n      throw new Error('expected user_speaking and user_turn spans');\n    }\n    expect(userSpeaking.parentSpanId).toBe(exportedUserTurn.spanContext().spanId);\n    expect(userSpeaking.attributes['lk.participant_id']).toBe('p3');\n  });\n\n  it('keeps user_speaking attached to the session root without an explicit speech context', () => {\n    const { exporter } = setupInMemoryTracing();\n    const sessionSpan = tracer.startSpan({ name: 'agent_session', context: ROOT_CONTEXT });\n    const sessionContext = trace.setSpan(ROOT_CONTEXT, sessionSpan);\n    const fakeSession = createFakeSession(sessionContext);\n\n    AgentSession.prototype._updateUserState.call(fakeSession, 'speaking', {\n      lastSpeakingTime: Date.now() - 100,\n    });\n    AgentSession.prototype._updateUserState.call(fakeSession, 'listening', {\n      lastSpeakingTime: Date.now(),\n    });\n\n    sessionSpan.end();\n\n    const spans = exporter.getFinishedSpans();\n    const userSpeaking = spanByName(spans, 'user_speaking');\n    expect(userSpeaking).toBeTruthy();\n    if (!userSpeaking) {\n      throw new Error('expected user_speaking span');\n    }\n    expect(userSpeaking.parentSpanId).toBe(sessionSpan.spanContext().spanId);\n  });\n});\n"],"mappings":"AAGA,SAAS,uBAAuB;AAChC,SAAS,cAAc,WAAW,aAAa,aAAa;AAC5D;AAAA,EACE;AAAA,EAEA;AAAA,OACK;AACP,SAAS,0BAA0B;AACnC,SAAS,sBAAsB;AAC/B,SAAS,UAAU,QAAQ,IAAI,UAAU;AACzC,SAAS,mBAAmB;AAC5B,SAAS,wBAAwB;AACjC,SAA2B,uBAAuB;AAClD,SAAS,mBAAmB,cAAc;AAC1C,SAAS,KAAoB,oBAAoC;AACjE,SAAS,oBAAoB;AAC7B;AAAA,EACE;AAAA,OAGK;AAGP,SAAS,uBAAuB;AAC9B,QAAM,WAAW,IAAI,qBAAqB;AAC1C,QAAM,WAAW,IAAI,mBAAmB;AACxC,WAAS,iBAAiB,IAAI,oBAAoB,QAAQ,CAAC;AAC3D,WAAS,SAAS;AAClB,oBAAkB,QAAQ;AAC1B,SAAO,EAAE,SAAS;AACpB;AAEA,SAAS,WAAW,OAAuB,MAAc;AACvD,SAAO,MAAM,KAAK,CAAC,MAAM,EAAE,SAAS,IAAI;AAC1C;AAEA,SAAS,kBAAkB,kBAAkB,cAA4B;AACvE,SAAO;AAAA,IACL,aAAa;AAAA,IACb,SAAS;AAAA,MACP,mBAAmB,EAAE,KAAK,MAAM,UAAU,WAAW,MAAM,gBAAgB,MAAM;AAAA,IACnF;AAAA,IACA,mBAAmB,GAAG,GAAG;AAAA,IACzB,sBAAsB,GAAG,GAAG;AAAA,IAC5B,mBAAmB;AAAA,IACnB,YAAY;AAAA,IACZ,MAAM,GAAG,GAAG;AAAA,IACZ;AAAA,EACF;AACF;AAEA,MAAM,sBAAuB,OAA4C;AAAA;AAAA;AAAA,EAG/D;AAAA,EACA,MAAM;AAAA,EACd,YAAY,QAAoB;AAC9B,UAAM;AACN,SAAK,SAAS;AAAA,EAChB;AAAA,EACA,oBAAoB;AAAA,EAAC;AAAA,EACrB,oBAAoB;AAAA,EAAC;AAAA,EACrB,QAAQ;AAAA,EAAC;AAAA,EACT,CAAC,OAAO,aAAa,IAAI;AACvB,WAAO;AAAA,EACT;AAAA,EACA,MAAM,OAA0C;AAC9C,QAAI,KAAK,OAAO,KAAK,OAAO,QAAQ;AAClC,aAAO,EAAE,MAAM,MAAM,OAAO,OAAU;AAAA,IACxC;AACA,UAAM,QAAQ,KAAK,OAAO,KAAK,KAAK;AACpC,WAAO,EAAE,MAAM,OAAO,MAAM;AAAA,EAC9B;AACF;AAEA,MAAM,gBAAgB,IAAI;AAAA,EACxB,QAAQ;AAAA,EACA;AAAA,EACR,YAAY,QAAoB;AAC9B,UAAM,EAAE,gBAAgB,EAAE,CAAC;AAC3B,SAAK,SAAS;AAAA,EAChB;AAAA,EACA,SAAc;AACZ,WAAO,IAAI,cAAc,KAAK,MAAM;AAAA,EACtC;AACF;AAEA,MAAM,yBAAwC;AAAA,EAC5C,OAAO;AAAA,EACP,UAAU;AAAA,EACV,kBAAkB,YAAY;AAAA,EAC9B,mBAAmB,YAAY;AAAA,EAC/B,kBAAkB,YAAY;AAChC;AAEA,SAAS,0CAA0C,MAAM;AACvD,mBAAiB,EAAE,QAAQ,OAAO,OAAO,SAAS,CAAC;AAEnD,KAAG,mEAAmE,YAAY;AAChF,UAAM,EAAE,SAAS,IAAI,qBAAqB;AAE1C,UAAM,QAA0B;AAAA,MAC9B,gBAAgB,GAAG,GAAG;AAAA,MACtB,iBAAiB,GAAG,GAAG;AAAA,MACvB,oBAAoB,GAAG,GAAG;AAAA,MAC1B,eAAe,GAAG,GAAG;AAAA,MACrB,qBAAqB,GAAG,GAAG;AAAA,MAC3B,mBAAmB,GAAG,GAAG;AAAA,MACzB,wBAAwB,GAAG,GAAG;AAAA,MAC9B,iBAAiB,MAAM,YAAY,MAAM;AAAA,MACzC,aAAa,GAAG,GAAG,YAAY,IAAI;AAAA,IACrC;AAEA,UAAM,YAA2B;AAAA,MAC/B,EAAE,MAAM,gBAAgB,gBAAgB;AAAA,MACxC;AAAA,QACE,MAAM,gBAAgB;AAAA,QACtB,cAAc;AAAA,UACZ;AAAA,YACE,UAAU;AAAA,YACV,MAAM;AAAA,YACN,WAAW;AAAA,YACX,SAAS;AAAA,YACT,YAAY;AAAA,UACd;AAAA,QACF;AAAA,MACF;AAAA,MACA,EAAE,MAAM,gBAAgB,cAAc;AAAA,IACxC;AAEA,UAAM,UAAmB,YACvB,IAAI,eAAqC;AAAA,MACvC,MAAM,YAAY;AAChB,mBAAW,MAAM,UAAW,YAAW,QAAQ,EAAE;AACjD,mBAAW,MAAM;AAAA,MACnB;AAAA,IACF,CAAC;AAEH,UAAM,KAAK,IAAI,iBAAiB;AAAA,MAC9B,kBAAkB;AAAA,MAClB,KAAK;AAAA,MACL,KAAK;AAAA,MACL,cAAc;AAAA,MACd,mBAAmB;AAAA,MACnB,qBAAqB;AAAA,MACrB,qBAAqB;AAAA,MACrB,UAAU;AAAA,MACV,aAAa;AAAA,MACb,sBAAsB,OAAO,EAAE,KAAK,MAAM,UAAU,OAAO,MAAM,gBAAgB,MAAM;AAAA,IACzF,CAAC;AAED,UAAM,GAAG,MAAM;AAEf,UAAM,IAAI,QAAQ,CAAC,MAAM,WAAW,GAAG,EAAE,CAAC;AAC1C,UAAM,GAAG,MAAM;AAEf,UAAM,QAAQ,SAAS,iBAAiB;AACxC,UAAM,WAAW,WAAW,OAAO,WAAW;AAC9C,UAAM,MAAM,WAAW,OAAO,eAAe;AAC7C,WAAO,UAAU,wBAAwB,EAAE,WAAW;AACtD,WAAO,KAAK,4BAA4B,EAAE,WAAW;AACrD,QAAI,CAAC,YAAY,CAAC,KAAK;AACrB,YAAM,IAAI,MAAM,4CAA4C;AAAA,IAC9D;AAEA,WAAO,IAAI,YAAY,EAAE,KAAK,SAAS,YAAY,EAAE,MAAM;AAG3D,WAAO,SAAS,WAAW,mBAAmB,CAAC,EAAE,KAAK,IAAI;AAC1D,WAAO,SAAS,WAAW,yBAAyB,CAAC,EAAE,KAAK,KAAK;AACjE,WAAO,SAAS,WAAW,qBAAqB,CAAC,EAAE,KAAK,OAAO;AAC/D,WAAO,SAAS,WAAW,sBAAsB,CAAC,EAAE,KAAK,gBAAgB;AACzE,WAAO,SAAS,WAAW,sBAAsB,CAAC,EAAE,KAAK,UAAU;AAGnE,WAAO,SAAS,WAAW,oBAAoB,CAAC,EAAE,UAAU,OAAO;AACnE,WAAO,SAAS,WAAW,0BAA0B,CAAC,EAAE,gBAAgB,CAAC;AAAA,EAC3E,CAAC;AAED,KAAG,4EAA4E,YAAY;AACzF,UAAM,EAAE,SAAS,IAAI,qBAAqB;AAE1C,UAAM,QAA0B;AAAA,MAC9B,gBAAgB,GAAG,GAAG;AAAA,MACtB,iBAAiB,GAAG,GAAG;AAAA,MACvB,oBAAoB,GAAG,GAAG;AAAA,MAC1B,eAAe,GAAG,GAAG;AAAA,MACrB,qBAAqB,GAAG,GAAG;AAAA,MAC3B,mBAAmB,GAAG,GAAG;AAAA,MACzB,wBAAwB,GAAG,GAAG;AAAA,MAC9B,iBAAiB,MAAM,YAAY,MAAM;AAAA,MACzC,aAAa,GAAG,GAAG,YAAY,IAAI;AAAA,IACrC;AAEA,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,YAAwB;AAAA,MAC5B;AAAA,QACE,MAAM,aAAa;AAAA,QACnB,cAAc;AAAA,QACd,WAAW;AAAA,QACX,gBAAgB;AAAA,QAChB,iBAAiB;AAAA,QACjB,QAAQ,CAAC;AAAA,QACT,aAAa;AAAA,QACb,mBAAmB;AAAA,QACnB,UAAU;AAAA,QACV,uBAAuB;AAAA,QACvB,sBAAsB;AAAA,MACxB;AAAA,MACA;AAAA,QACE,MAAM,aAAa;AAAA,QACnB,cAAc;AAAA,QACd,WAAW,MAAM;AAAA,QACjB,gBAAgB;AAAA,QAChB,iBAAiB;AAAA,QACjB,QAAQ,CAAC;AAAA,QACT,aAAa;AAAA,QACb,mBAAmB;AAAA,QACnB,UAAU;AAAA,QACV,uBAAuB;AAAA,QACvB,sBAAsB;AAAA,MACxB;AAAA,IACF;AAEA,UAAM,YAA2B;AAAA,MAC/B;AAAA,QACE,MAAM,gBAAgB;AAAA,QACtB,cAAc;AAAA,UACZ;AAAA,YACE,UAAU;AAAA,YACV,MAAM;AAAA,YACN,WAAW;AAAA,YACX,SAAS;AAAA,YACT,YAAY;AAAA,UACd;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,UAAmB,YACvB,IAAI,eAAqC;AAAA,MACvC,MAAM,YAAY;AAChB,mBAAW,MAAM,UAAW,YAAW,QAAQ,EAAE;AACjD,mBAAW,MAAM;AAAA,MACnB;AAAA,IACF,CAAC;AAEH,UAAM,KAAK,IAAI,iBAAiB;AAAA,MAC9B,kBAAkB;AAAA,MAClB,KAAK;AAAA,MACL,KAAK,IAAI,QAAQ,SAAS;AAAA,MAC1B,cAAc;AAAA,MACd,mBAAmB;AAAA,MACnB,qBAAqB;AAAA,MACrB,qBAAqB;AAAA,MACrB,UAAU;AAAA,MACV,aAAa;AAAA,MACb,sBAAsB,OAAO,EAAE,KAAK,MAAM,UAAU,SAAS,MAAM,gBAAgB,MAAM;AAAA,IAC3F,CAAC;AAED,UAAM,GAAG,MAAM;AACf,UAAM,IAAI,QAAQ,CAAC,MAAM,WAAW,GAAG,EAAE,CAAC;AAC1C,UAAM,GAAG,MAAM;AAEf,UAAM,QAAQ,SAAS,iBAAiB;AACxC,UAAM,WAAW,WAAW,OAAO,WAAW;AAC9C,UAAM,MAAM,WAAW,OAAO,eAAe;AAC7C,WAAO,QAAQ,EAAE,WAAW;AAC5B,WAAO,GAAG,EAAE,WAAW;AACvB,QAAI,CAAC,YAAY,CAAC,KAAK;AACrB,YAAM,IAAI,MAAM,4CAA4C;AAAA,IAC9D;AACA,WAAO,IAAI,YAAY,EAAE,KAAK,SAAS,YAAY,EAAE,MAAM;AAE3D,WAAO,MAAM,eAAe,EAAE,iBAAiB;AAC/C,WAAO,MAAM,aAAa,EAAE,iBAAiB;AAAA,EAC/C,CAAC;AAED,KAAG,qFAAqF,MAAM;AAC5F,UAAM,EAAE,SAAS,IAAI,qBAAqB;AAC1C,UAAM,cAAc,OAAO,UAAU,EAAE,MAAM,iBAAiB,SAAS,aAAa,CAAC;AACrF,UAAM,iBAAiB,MAAM,QAAQ,cAAc,WAAW;AAC9D,UAAM,cAAc,kBAAkB,cAAc;AACpD,UAAM,WAAW,OAAO,UAAU,EAAE,MAAM,aAAa,SAAS,eAAe,CAAC;AAChF,UAAM,kBAAkB,MAAM,QAAQ,gBAAgB,QAAQ;AAC9D,UAAM,oBAAoB,KAAK,IAAI,IAAI;AACvC,UAAM,kBAAkB,KAAK,IAAI;AAEjC,gBAAY,KAAK,iBAAiB,MAAM;AACtC,mBAAa,UAAU,iBAAiB,KAAK,aAAa,YAAY;AAAA,QACpE,kBAAkB;AAAA,QAClB,aAAa,YAAY,OAAO;AAAA,MAClC,CAAC;AACD,mBAAa,UAAU,iBAAiB,KAAK,aAAa,aAAa;AAAA,QACrE,kBAAkB;AAAA,QAClB,aAAa,YAAY,OAAO;AAAA,MAClC,CAAC;AAAA,IACH,CAAC;AAED,aAAS,IAAI;AACb,gBAAY,IAAI;AAEhB,UAAM,QAAQ,SAAS,iBAAiB;AACxC,UAAM,eAAe,WAAW,OAAO,eAAe;AACtD,UAAM,mBAAmB,WAAW,OAAO,WAAW;AACtD,WAAO,YAAY,EAAE,WAAW;AAChC,WAAO,gBAAgB,EAAE,WAAW;AACpC,QAAI,CAAC,gBAAgB,CAAC,kBAAkB;AACtC,YAAM,IAAI,MAAM,4CAA4C;AAAA,IAC9D;AACA,WAAO,aAAa,YAAY,EAAE,KAAK,iBAAiB,YAAY,EAAE,MAAM;AAC5E,WAAO,aAAa,WAAW,mBAAmB,CAAC,EAAE,KAAK,IAAI;AAAA,EAChE,CAAC;AAED,KAAG,uFAAuF,MAAM;AAC9F,UAAM,EAAE,SAAS,IAAI,qBAAqB;AAC1C,UAAM,cAAc,OAAO,UAAU,EAAE,MAAM,iBAAiB,SAAS,aAAa,CAAC;AACrF,UAAM,iBAAiB,MAAM,QAAQ,cAAc,WAAW;AAC9D,UAAM,cAAc,kBAAkB,cAAc;AAEpD,iBAAa,UAAU,iBAAiB,KAAK,aAAa,YAAY;AAAA,MACpE,kBAAkB,KAAK,IAAI,IAAI;AAAA,IACjC,CAAC;AACD,iBAAa,UAAU,iBAAiB,KAAK,aAAa,aAAa;AAAA,MACrE,kBAAkB,KAAK,IAAI;AAAA,IAC7B,CAAC;AAED,gBAAY,IAAI;AAEhB,UAAM,QAAQ,SAAS,iBAAiB;AACxC,UAAM,eAAe,WAAW,OAAO,eAAe;AACtD,WAAO,YAAY,EAAE,WAAW;AAChC,QAAI,CAAC,cAAc;AACjB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,WAAO,aAAa,YAAY,EAAE,KAAK,YAAY,YAAY,EAAE,MAAM;AAAA,EACzE,CAAC;AACH,CAAC;","names":[]}

package/dist/voice/events.cjs

@@ -40,7 +40,7 @@ var AgentSessionEventTypes = /* @__PURE__ */ ((AgentSessionEventTypes2) => {
   AgentSessionEventTypes2["FunctionToolsExecuted"] = "function_tools_executed";
   AgentSessionEventTypes2["MetricsCollected"] = "metrics_collected";
   AgentSessionEventTypes2["SpeechCreated"] = "speech_created";
-  AgentSessionEventTypes2["UserOverlappingSpeech"] = "user_overlapping_speech";
+  AgentSessionEventTypes2["OverlappingSpeech"] = "overlapping_speech";
   AgentSessionEventTypes2["Error"] = "error";
   AgentSessionEventTypes2["Close"] = "close";
   return AgentSessionEventTypes2;

package/dist/voice/events.cjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/events.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { InterruptionDetectionError } from '../inference/interruption/errors.js';\nimport type { OverlappingSpeechEvent } from '../inference/interruption/types.js';\nimport type {\n  ChatMessage,\n  FunctionCall,\n  FunctionCallOutput,\n  LLM,\n  RealtimeModel,\n  RealtimeModelError,\n} from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport type { AgentMetrics } from '../metrics/base.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport type { TTS } from '../tts/index.js';\nimport type { TTSError } from '../tts/tts.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport enum AgentSessionEventTypes {\n  UserInputTranscribed = 'user_input_transcribed',\n  AgentStateChanged = 'agent_state_changed',\n  UserStateChanged = 'user_state_changed',\n  ConversationItemAdded = 'conversation_item_added',\n  FunctionToolsExecuted = 'function_tools_executed',\n  MetricsCollected = 'metrics_collected',\n  SpeechCreated = 'speech_created',\n  UserOverlappingSpeech = 'user_overlapping_speech',\n  Error = 'error',\n  Close = 'close',\n}\n\nexport type UserState = 'speaking' | 'listening' | 'away';\nexport type AgentState = 'initializing' | 'idle' | 'listening' | 'thinking' | 'speaking';\n\nexport enum CloseReason {\n  ERROR = 'error',\n  JOB_SHUTDOWN = 'job_shutdown',\n  PARTICIPANT_DISCONNECTED = 'participant_disconnected',\n  USER_INITIATED = 'user_initiated',\n}\n\nexport type ShutdownReason = CloseReason | string;\n\nexport type SpeechSource = 'say' | 'generate_reply' | 'tool_response';\n\nexport type UserStateChangedEvent = {\n  type: 'user_state_changed';\n  oldState: UserState;\n  newState: UserState;\n  createdAt: number;\n};\n\nexport const createUserStateChangedEvent = (\n  oldState: UserState,\n  newState: UserState,\n  createdAt: number = Date.now(),\n): UserStateChangedEvent => ({\n  type: 'user_state_changed',\n  oldState,\n  newState,\n  createdAt,\n});\n\nexport type AgentStateChangedEvent = {\n  type: 'agent_state_changed';\n  oldState: AgentState;\n  newState: AgentState;\n  createdAt: number;\n};\n\nexport const createAgentStateChangedEvent = (\n  oldState: AgentState,\n  newState: AgentState,\n  createdAt: number = Date.now(),\n): AgentStateChangedEvent => ({\n  type: 'agent_state_changed',\n  oldState,\n  newState,\n  createdAt,\n});\n\nexport type UserInputTranscribedEvent = {\n  type: 'user_input_transcribed';\n  transcript: string;\n  isFinal: boolean;\n  // TODO(AJS-106): add multi participant support\n  /** Not supported yet. Always null by default. */\n  speakerId: string | null;\n  createdAt: number;\n  language: string | null;\n};\n\nexport const createUserInputTranscribedEvent = ({\n  transcript,\n  isFinal,\n  speakerId = null,\n  language = null,\n  createdAt = Date.now(),\n}: {\n  transcript: string;\n  isFinal: boolean;\n  speakerId?: string | null;\n  language?: string | null;\n  createdAt?: number;\n}): UserInputTranscribedEvent => ({\n  type: 'user_input_transcribed',\n  transcript,\n  isFinal,\n  speakerId,\n  language,\n  createdAt,\n});\n\nexport type MetricsCollectedEvent = {\n  type: 'metrics_collected';\n  metrics: AgentMetrics;\n  createdAt: number;\n};\n\nexport const createMetricsCollectedEvent = ({\n  metrics,\n  createdAt = Date.now(),\n}: {\n  metrics: AgentMetrics;\n  createdAt?: number;\n}): MetricsCollectedEvent => ({\n  type: 'metrics_collected',\n  metrics,\n  createdAt,\n});\n\nexport type ConversationItemAddedEvent = {\n  type: 'conversation_item_added';\n  item: ChatMessage;\n  createdAt: number;\n};\n\nexport const createConversationItemAddedEvent = (\n  item: ChatMessage,\n  createdAt: number = Date.now(),\n): ConversationItemAddedEvent => ({\n  type: 'conversation_item_added',\n  item,\n  createdAt,\n});\n\nexport type FunctionToolsExecutedEvent = {\n  type: 'function_tools_executed';\n  functionCalls: FunctionCall[];\n  functionCallOutputs: FunctionCallOutput[];\n  createdAt: number;\n};\n\nexport const createFunctionToolsExecutedEvent = ({\n  functionCalls,\n  functionCallOutputs,\n  createdAt = Date.now(),\n}: {\n  functionCalls: FunctionCall[];\n  functionCallOutputs: FunctionCallOutput[];\n  createdAt?: number;\n}): FunctionToolsExecutedEvent => {\n  return {\n    type: 'function_tools_executed',\n    functionCalls,\n    functionCallOutputs,\n    createdAt,\n  };\n};\n\nexport const zipFunctionCallsAndOutputs = (\n  event: FunctionToolsExecutedEvent,\n): Array<[FunctionCall, FunctionCallOutput]> => {\n  return event.functionCalls.map((call, index) => [call, event.functionCallOutputs[index]!]);\n};\n\nexport type SpeechCreatedEvent = {\n  type: 'speech_created';\n  /**\n   * True if the speech was created using public methods like `say` or `generate_reply`\n   */\n  userInitiated: boolean;\n  /**\n   * Source indicating how the speech handle was created\n   */\n  source: SpeechSource;\n  /**\n   * The speech handle that was created\n   */\n  // TODO(shubhra): we need to make sure this doesn't get serialized\n  speechHandle: SpeechHandle;\n  /**\n   * The timestamp when the speech handle was created\n   */\n  createdAt: number;\n};\n\nexport const createSpeechCreatedEvent = ({\n  userInitiated,\n  source,\n  speechHandle,\n  createdAt = Date.now(),\n}: {\n  userInitiated: boolean;\n  source: SpeechSource;\n  speechHandle: SpeechHandle;\n  createdAt?: number;\n}): SpeechCreatedEvent => ({\n  type: 'speech_created',\n  userInitiated,\n  source,\n  speechHandle,\n  createdAt,\n});\n\nexport type ErrorEvent = {\n  type: 'error';\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | unknown;\n  source: LLM | STT | TTS | RealtimeModel | unknown;\n  createdAt: number;\n};\n\nexport const createErrorEvent = (\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | unknown,\n  source: LLM | STT | TTS | RealtimeModel | unknown,\n  createdAt: number = Date.now(),\n): ErrorEvent => ({\n  type: 'error',\n  error,\n  source,\n  createdAt,\n});\n\nexport type CloseEvent = {\n  type: 'close';\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | null;\n  reason: ShutdownReason;\n  createdAt: number;\n};\n\nexport const createCloseEvent = (\n  reason: ShutdownReason,\n  error:\n    | RealtimeModelError\n    | STTError\n    | TTSError\n    | LLMError\n    | InterruptionDetectionError\n    | null = null,\n  createdAt: number = Date.now(),\n): CloseEvent => ({\n  type: 'close',\n  error,\n  reason,\n  createdAt,\n});\n\nexport type AgentEvent =\n  | UserInputTranscribedEvent\n  | UserStateChangedEvent\n  | AgentStateChangedEvent\n  | MetricsCollectedEvent\n  | ConversationItemAddedEvent\n  | FunctionToolsExecutedEvent\n  | SpeechCreatedEvent\n  | OverlappingSpeechEvent\n  | ErrorEvent\n  | CloseEvent;\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAqBO,IAAK,yBAAL,kBAAKA,4BAAL;AACL,EAAAA,wBAAA,0BAAuB;AACvB,EAAAA,wBAAA,uBAAoB;AACpB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,mBAAgB;AAChB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,WAAQ;AACR,EAAAA,wBAAA,WAAQ;AAVE,SAAAA;AAAA,GAAA;AAgBL,IAAK,cAAL,kBAAKC,iBAAL;AACL,EAAAA,aAAA,WAAQ;AACR,EAAAA,aAAA,kBAAe;AACf,EAAAA,aAAA,8BAA2B;AAC3B,EAAAA,aAAA,oBAAiB;AAJP,SAAAA;AAAA,GAAA;AAkBL,MAAM,8BAA8B,CACzC,UACA,UACA,YAAoB,KAAK,IAAI,OACF;AAAA,EAC3B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,+BAA+B,CAC1C,UACA,UACA,YAAoB,KAAK,IAAI,OACD;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AAaO,MAAM,kCAAkC,CAAC;AAAA,EAC9C;AAAA,EACA;AAAA,EACA,YAAY;AAAA,EACZ,WAAW;AAAA,EACX,YAAY,KAAK,IAAI;AACvB,OAMkC;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAQO,MAAM,8BAA8B,CAAC;AAAA,EAC1C;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAG8B;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AACF;AAQO,MAAM,mCAAmC,CAC9C,MACA,YAAoB,KAAK,IAAI,OACG;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AACF;AASO,MAAM,mCAAmC,CAAC;AAAA,EAC/C;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,MAIkC;AAChC,SAAO;AAAA,IACL,MAAM;AAAA,IACN;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEO,MAAM,6BAA6B,CACxC,UAC8C;AAC9C,SAAO,MAAM,cAAc,IAAI,CAAC,MAAM,UAAU,CAAC,MAAM,MAAM,oBAAoB,KAAK,CAAE,CAAC;AAC3F;AAuBO,MAAM,2BAA2B,CAAC;AAAA,EACvC;AAAA,EACA;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAK2B;AAAA,EACzB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,OACA,QACA,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,QACA,QAMW,MACX,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;","names":["AgentSessionEventTypes","CloseReason"]}
+{"version":3,"sources":["../../src/voice/events.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { InterruptionDetectionError } from '../inference/interruption/errors.js';\nimport type { OverlappingSpeechEvent } from '../inference/interruption/types.js';\nimport type { LanguageCode } from '../language.js';\nimport type {\n  ChatMessage,\n  FunctionCall,\n  FunctionCallOutput,\n  LLM,\n  RealtimeModel,\n  RealtimeModelError,\n} from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport type { AgentMetrics } from '../metrics/base.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport type { TTS } from '../tts/index.js';\nimport type { TTSError } from '../tts/tts.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport enum AgentSessionEventTypes {\n  UserInputTranscribed = 'user_input_transcribed',\n  AgentStateChanged = 'agent_state_changed',\n  UserStateChanged = 'user_state_changed',\n  ConversationItemAdded = 'conversation_item_added',\n  FunctionToolsExecuted = 'function_tools_executed',\n  MetricsCollected = 'metrics_collected',\n  SpeechCreated = 'speech_created',\n  OverlappingSpeech = 'overlapping_speech',\n  Error = 'error',\n  Close = 'close',\n}\n\nexport type UserState = 'speaking' | 'listening' | 'away';\nexport type AgentState = 'initializing' | 'idle' | 'listening' | 'thinking' | 'speaking';\n\nexport enum CloseReason {\n  ERROR = 'error',\n  JOB_SHUTDOWN = 'job_shutdown',\n  PARTICIPANT_DISCONNECTED = 'participant_disconnected',\n  USER_INITIATED = 'user_initiated',\n}\n\nexport type ShutdownReason = CloseReason | string;\n\nexport type SpeechSource = 'say' | 'generate_reply' | 'tool_response';\n\nexport type UserStateChangedEvent = {\n  type: 'user_state_changed';\n  oldState: UserState;\n  newState: UserState;\n  createdAt: number;\n};\n\nexport const createUserStateChangedEvent = (\n  oldState: UserState,\n  newState: UserState,\n  createdAt: number = Date.now(),\n): UserStateChangedEvent => ({\n  type: 'user_state_changed',\n  oldState,\n  newState,\n  createdAt,\n});\n\nexport type AgentStateChangedEvent = {\n  type: 'agent_state_changed';\n  oldState: AgentState;\n  newState: AgentState;\n  createdAt: number;\n};\n\nexport const createAgentStateChangedEvent = (\n  oldState: AgentState,\n  newState: AgentState,\n  createdAt: number = Date.now(),\n): AgentStateChangedEvent => ({\n  type: 'agent_state_changed',\n  oldState,\n  newState,\n  createdAt,\n});\n\nexport type UserInputTranscribedEvent = {\n  type: 'user_input_transcribed';\n  transcript: string;\n  isFinal: boolean;\n  // TODO(AJS-106): add multi participant support\n  /** Not supported yet. Always null by default. */\n  speakerId: string | null;\n  createdAt: number;\n  language: LanguageCode | null;\n};\n\nexport const createUserInputTranscribedEvent = ({\n  transcript,\n  isFinal,\n  speakerId = null,\n  language = null,\n  createdAt = Date.now(),\n}: {\n  transcript: string;\n  isFinal: boolean;\n  speakerId?: string | null;\n  language?: LanguageCode | null;\n  createdAt?: number;\n}): UserInputTranscribedEvent => ({\n  type: 'user_input_transcribed',\n  transcript,\n  isFinal,\n  speakerId,\n  language,\n  createdAt,\n});\n\nexport type MetricsCollectedEvent = {\n  type: 'metrics_collected';\n  metrics: AgentMetrics;\n  createdAt: number;\n};\n\nexport const createMetricsCollectedEvent = ({\n  metrics,\n  createdAt = Date.now(),\n}: {\n  metrics: AgentMetrics;\n  createdAt?: number;\n}): MetricsCollectedEvent => ({\n  type: 'metrics_collected',\n  metrics,\n  createdAt,\n});\n\nexport type ConversationItemAddedEvent = {\n  type: 'conversation_item_added';\n  item: ChatMessage;\n  createdAt: number;\n};\n\nexport const createConversationItemAddedEvent = (\n  item: ChatMessage,\n  createdAt: number = Date.now(),\n): ConversationItemAddedEvent => ({\n  type: 'conversation_item_added',\n  item,\n  createdAt,\n});\n\nexport type FunctionToolsExecutedEvent = {\n  type: 'function_tools_executed';\n  functionCalls: FunctionCall[];\n  functionCallOutputs: FunctionCallOutput[];\n  createdAt: number;\n};\n\nexport const createFunctionToolsExecutedEvent = ({\n  functionCalls,\n  functionCallOutputs,\n  createdAt = Date.now(),\n}: {\n  functionCalls: FunctionCall[];\n  functionCallOutputs: FunctionCallOutput[];\n  createdAt?: number;\n}): FunctionToolsExecutedEvent => {\n  return {\n    type: 'function_tools_executed',\n    functionCalls,\n    functionCallOutputs,\n    createdAt,\n  };\n};\n\nexport const zipFunctionCallsAndOutputs = (\n  event: FunctionToolsExecutedEvent,\n): Array<[FunctionCall, FunctionCallOutput]> => {\n  return event.functionCalls.map((call, index) => [call, event.functionCallOutputs[index]!]);\n};\n\nexport type SpeechCreatedEvent = {\n  type: 'speech_created';\n  /**\n   * True if the speech was created using public methods like `say` or `generate_reply`\n   */\n  userInitiated: boolean;\n  /**\n   * Source indicating how the speech handle was created\n   */\n  source: SpeechSource;\n  /**\n   * The speech handle that was created\n   */\n  // TODO(shubhra): we need to make sure this doesn't get serialized\n  speechHandle: SpeechHandle;\n  /**\n   * The timestamp when the speech handle was created\n   */\n  createdAt: number;\n};\n\nexport const createSpeechCreatedEvent = ({\n  userInitiated,\n  source,\n  speechHandle,\n  createdAt = Date.now(),\n}: {\n  userInitiated: boolean;\n  source: SpeechSource;\n  speechHandle: SpeechHandle;\n  createdAt?: number;\n}): SpeechCreatedEvent => ({\n  type: 'speech_created',\n  userInitiated,\n  source,\n  speechHandle,\n  createdAt,\n});\n\nexport type ErrorEvent = {\n  type: 'error';\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | unknown;\n  source: LLM | STT | TTS | RealtimeModel | unknown;\n  createdAt: number;\n};\n\nexport const createErrorEvent = (\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | unknown,\n  source: LLM | STT | TTS | RealtimeModel | unknown,\n  createdAt: number = Date.now(),\n): ErrorEvent => ({\n  type: 'error',\n  error,\n  source,\n  createdAt,\n});\n\nexport type CloseEvent = {\n  type: 'close';\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | null;\n  reason: ShutdownReason;\n  createdAt: number;\n};\n\nexport const createCloseEvent = (\n  reason: ShutdownReason,\n  error:\n    | RealtimeModelError\n    | STTError\n    | TTSError\n    | LLMError\n    | InterruptionDetectionError\n    | null = null,\n  createdAt: number = Date.now(),\n): CloseEvent => ({\n  type: 'close',\n  error,\n  reason,\n  createdAt,\n});\n\nexport type AgentEvent =\n  | UserInputTranscribedEvent\n  | UserStateChangedEvent\n  | AgentStateChangedEvent\n  | MetricsCollectedEvent\n  | ConversationItemAddedEvent\n  | FunctionToolsExecutedEvent\n  | SpeechCreatedEvent\n  | OverlappingSpeechEvent\n  | ErrorEvent\n  | CloseEvent;\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAsBO,IAAK,yBAAL,kBAAKA,4BAAL;AACL,EAAAA,wBAAA,0BAAuB;AACvB,EAAAA,wBAAA,uBAAoB;AACpB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,mBAAgB;AAChB,EAAAA,wBAAA,uBAAoB;AACpB,EAAAA,wBAAA,WAAQ;AACR,EAAAA,wBAAA,WAAQ;AAVE,SAAAA;AAAA,GAAA;AAgBL,IAAK,cAAL,kBAAKC,iBAAL;AACL,EAAAA,aAAA,WAAQ;AACR,EAAAA,aAAA,kBAAe;AACf,EAAAA,aAAA,8BAA2B;AAC3B,EAAAA,aAAA,oBAAiB;AAJP,SAAAA;AAAA,GAAA;AAkBL,MAAM,8BAA8B,CACzC,UACA,UACA,YAAoB,KAAK,IAAI,OACF;AAAA,EAC3B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,+BAA+B,CAC1C,UACA,UACA,YAAoB,KAAK,IAAI,OACD;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AAaO,MAAM,kCAAkC,CAAC;AAAA,EAC9C;AAAA,EACA;AAAA,EACA,YAAY;AAAA,EACZ,WAAW;AAAA,EACX,YAAY,KAAK,IAAI;AACvB,OAMkC;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAQO,MAAM,8BAA8B,CAAC;AAAA,EAC1C;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAG8B;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AACF;AAQO,MAAM,mCAAmC,CAC9C,MACA,YAAoB,KAAK,IAAI,OACG;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AACF;AASO,MAAM,mCAAmC,CAAC;AAAA,EAC/C;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,MAIkC;AAChC,SAAO;AAAA,IACL,MAAM;AAAA,IACN;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEO,MAAM,6BAA6B,CACxC,UAC8C;AAC9C,SAAO,MAAM,cAAc,IAAI,CAAC,MAAM,UAAU,CAAC,MAAM,MAAM,oBAAoB,KAAK,CAAE,CAAC;AAC3F;AAuBO,MAAM,2BAA2B,CAAC;AAAA,EACvC;AAAA,EACA;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAK2B;AAAA,EACzB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,OACA,QACA,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,QACA,QAMW,MACX,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;","names":["AgentSessionEventTypes","CloseReason"]}

package/dist/voice/events.d.cts

@@ -1,5 +1,6 @@
 import type { InterruptionDetectionError } from '../inference/interruption/errors.js';
 import type { OverlappingSpeechEvent } from '../inference/interruption/types.js';
+import type { LanguageCode } from '../language.js';
 import type { ChatMessage, FunctionCall, FunctionCallOutput, LLM, RealtimeModel, RealtimeModelError } from '../llm/index.js';
 import type { LLMError } from '../llm/llm.js';
 import type { AgentMetrics } from '../metrics/base.js';
@@ -16,7 +17,7 @@ export declare enum AgentSessionEventTypes {
     FunctionToolsExecuted = "function_tools_executed",
     MetricsCollected = "metrics_collected",
     SpeechCreated = "speech_created",
-    UserOverlappingSpeech = "user_overlapping_speech",
+    OverlappingSpeech = "overlapping_speech",
     Error = "error",
     Close = "close"
 }
@@ -51,13 +52,13 @@ export type UserInputTranscribedEvent = {
     /** Not supported yet. Always null by default. */
     speakerId: string | null;
     createdAt: number;
-    language: string | null;
+    language: LanguageCode | null;
 };
 export declare const createUserInputTranscribedEvent: ({ transcript, isFinal, speakerId, language, createdAt, }: {
     transcript: string;
     isFinal: boolean;
     speakerId?: string | null;
-    language?: string | null;
+    language?: LanguageCode | null;
     createdAt?: number;
 }) => UserInputTranscribedEvent;
 export type MetricsCollectedEvent = {

package/dist/voice/events.d.ts

@@ -1,5 +1,6 @@
 import type { InterruptionDetectionError } from '../inference/interruption/errors.js';
 import type { OverlappingSpeechEvent } from '../inference/interruption/types.js';
+import type { LanguageCode } from '../language.js';
 import type { ChatMessage, FunctionCall, FunctionCallOutput, LLM, RealtimeModel, RealtimeModelError } from '../llm/index.js';
 import type { LLMError } from '../llm/llm.js';
 import type { AgentMetrics } from '../metrics/base.js';
@@ -16,7 +17,7 @@ export declare enum AgentSessionEventTypes {
     FunctionToolsExecuted = "function_tools_executed",
     MetricsCollected = "metrics_collected",
     SpeechCreated = "speech_created",
-    UserOverlappingSpeech = "user_overlapping_speech",
+    OverlappingSpeech = "overlapping_speech",
     Error = "error",
     Close = "close"
 }
@@ -51,13 +52,13 @@ export type UserInputTranscribedEvent = {
     /** Not supported yet. Always null by default. */
     speakerId: string | null;
     createdAt: number;
-    language: string | null;
+    language: LanguageCode | null;
 };
 export declare const createUserInputTranscribedEvent: ({ transcript, isFinal, speakerId, language, createdAt, }: {
     transcript: string;
     isFinal: boolean;
     speakerId?: string | null;
-    language?: string | null;
+    language?: LanguageCode | null;
     createdAt?: number;
 }) => UserInputTranscribedEvent;
 export type MetricsCollectedEvent = {

package/dist/voice/events.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"events.d.ts","sourceRoot":"","sources":["../../src/voice/events.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,0BAA0B,EAAE,MAAM,qCAAqC,CAAC;AACtF,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,oCAAoC,CAAC;AACjF,OAAO,KAAK,EACV,WAAW,EACX,YAAY,EACZ,kBAAkB,EAClB,GAAG,EACH,aAAa,EACb,kBAAkB,EACnB,MAAM,iBAAiB,CAAC;AACzB,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AACvD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,oBAAY,sBAAsB;IAChC,oBAAoB,2BAA2B;IAC/C,iBAAiB,wBAAwB;IACzC,gBAAgB,uBAAuB;IACvC,qBAAqB,4BAA4B;IACjD,qBAAqB,4BAA4B;IACjD,gBAAgB,sBAAsB;IACtC,aAAa,mBAAmB;IAChC,qBAAqB,4BAA4B;IACjD,KAAK,UAAU;IACf,KAAK,UAAU;CAChB;AAED,MAAM,MAAM,SAAS,GAAG,UAAU,GAAG,WAAW,GAAG,MAAM,CAAC;AAC1D,MAAM,MAAM,UAAU,GAAG,cAAc,GAAG,MAAM,GAAG,WAAW,GAAG,UAAU,GAAG,UAAU,CAAC;AAEzF,oBAAY,WAAW;IACrB,KAAK,UAAU;IACf,YAAY,iBAAiB;IAC7B,wBAAwB,6BAA6B;IACrD,cAAc,mBAAmB;CAClC;AAED,MAAM,MAAM,cAAc,GAAG,WAAW,GAAG,MAAM,CAAC;AAElD,MAAM,MAAM,YAAY,GAAG,KAAK,GAAG,gBAAgB,GAAG,eAAe,CAAC;AAEtE,MAAM,MAAM,qBAAqB,GAAG;IAClC,IAAI,EAAE,oBAAoB,CAAC;IAC3B,QAAQ,EAAE,SAAS,CAAC;IACpB,QAAQ,EAAE,SAAS,CAAC;IACpB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,2BAA2B,aAC5B,SAAS,YACT,SAAS,cACR,MAAM,KAChB,qBAKD,CAAC;AAEH,MAAM,MAAM,sBAAsB,GAAG;IACnC,IAAI,EAAE,qBAAqB,CAAC;IAC5B,QAAQ,EAAE,UAAU,CAAC;IACrB,QAAQ,EAAE,UAAU,CAAC;IACrB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,4BAA4B,aAC7B,UAAU,YACV,UAAU,cACT,MAAM,KAChB,sBAKD,CAAC;AAEH,MAAM,MAAM,yBAAyB,GAAG;IACtC,IAAI,EAAE,wBAAwB,CAAC;IAC/B,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IAEjB,iDAAiD;IACjD,SAAS,EAAE,MAAM,GAAG,IAAI,CAAC;IACzB,SAAS,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAC;CACzB,CAAC;AAEF,eAAO,MAAM,+BAA+B,6DAMzC;IACD,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,SAAS,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;IAC1B,QAAQ,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;IACzB,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,KAAG,yBAOF,CAAC;AAEH,MAAM,MAAM,qBAAqB,GAAG;IAClC,IAAI,EAAE,mBAAmB,CAAC;IAC1B,OAAO,EAAE,YAAY,CAAC;IACtB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,2BAA2B,4BAGrC;IACD,OAAO,EAAE,YAAY,CAAC;IACtB,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,KAAG,qBAIF,CAAC;AAEH,MAAM,MAAM,0BAA0B,GAAG;IACvC,IAAI,EAAE,yBAAyB,CAAC;IAChC,IAAI,EAAE,WAAW,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,gCAAgC,SACrC,WAAW,cACN,MAAM,KAChB,0BAID,CAAC;AAEH,MAAM,MAAM,0BAA0B,GAAG;IACvC,IAAI,EAAE,yBAAyB,CAAC;IAChC,aAAa,EAAE,YAAY,EAAE,CAAC;IAC9B,mBAAmB,EAAE,kBAAkB,EAAE,CAAC;IAC1C,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,gCAAgC,uDAI1C;IACD,aAAa,EAAE,YAAY,EAAE,CAAC;IAC9B,mBAAmB,EAAE,kBAAkB,EAAE,CAAC;IAC1C,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,KAAG,0BAOH,CAAC;AAEF,eAAO,MAAM,0BAA0B,UAC9B,0BAA0B,KAChC,MAAM,CAAC,YAAY,EAAE,kBAAkB,CAAC,CAE1C,CAAC;AAEF,MAAM,MAAM,kBAAkB,GAAG;IAC/B,IAAI,EAAE,gBAAgB,CAAC;IACvB;;OAEG;IACH,aAAa,EAAE,OAAO,CAAC;IACvB;;OAEG;IACH,MAAM,EAAE,YAAY,CAAC;IACrB;;OAEG;IAEH,YAAY,EAAE,YAAY,CAAC;IAC3B;;OAEG;IACH,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,wBAAwB,wDAKlC;IACD,aAAa,EAAE,OAAO,CAAC;IACvB,MAAM,EAAE,YAAY,CAAC;IACrB,YAAY,EAAE,YAAY,CAAC;IAC3B,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,KAAG,kBAMF,CAAC;AAEH,MAAM,MAAM,UAAU,GAAG;IACvB,IAAI,EAAE,OAAO,CAAC;IACd,KAAK,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,0BAA0B,GAAG,OAAO,CAAC;IAClG,MAAM,EAAE,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,aAAa,GAAG,OAAO,CAAC;IAClD,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,gBAAgB,UACpB,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,0BAA0B,GAAG,OAAO,UACzF,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,aAAa,GAAG,OAAO,cACtC,MAAM,KAChB,UAKD,CAAC;AAEH,MAAM,MAAM,UAAU,GAAG;IACvB,IAAI,EAAE,OAAO,CAAC;IACd,KAAK,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,0BAA0B,GAAG,IAAI,CAAC;IAC/F,MAAM,EAAE,cAAc,CAAC;IACvB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,gBAAgB,WACnB,cAAc,UAElB,kBAAkB,GAClB,QAAQ,GACR,QAAQ,GACR,QAAQ,GACR,0BAA0B,GAC1B,IAAI,cACG,MAAM,KAChB,UAKD,CAAC;AAEH,MAAM,MAAM,UAAU,GAClB,yBAAyB,GACzB,qBAAqB,GACrB,sBAAsB,GACtB,qBAAqB,GACrB,0BAA0B,GAC1B,0BAA0B,GAC1B,kBAAkB,GAClB,sBAAsB,GACtB,UAAU,GACV,UAAU,CAAC"}
+{"version":3,"file":"events.d.ts","sourceRoot":"","sources":["../../src/voice/events.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,0BAA0B,EAAE,MAAM,qCAAqC,CAAC;AACtF,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,oCAAoC,CAAC;AACjF,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,gBAAgB,CAAC;AACnD,OAAO,KAAK,EACV,WAAW,EACX,YAAY,EACZ,kBAAkB,EAClB,GAAG,EACH,aAAa,EACb,kBAAkB,EACnB,MAAM,iBAAiB,CAAC;AACzB,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AACvD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,oBAAY,sBAAsB;IAChC,oBAAoB,2BAA2B;IAC/C,iBAAiB,wBAAwB;IACzC,gBAAgB,uBAAuB;IACvC,qBAAqB,4BAA4B;IACjD,qBAAqB,4BAA4B;IACjD,gBAAgB,sBAAsB;IACtC,aAAa,mBAAmB;IAChC,iBAAiB,uBAAuB;IACxC,KAAK,UAAU;IACf,KAAK,UAAU;CAChB;AAED,MAAM,MAAM,SAAS,GAAG,UAAU,GAAG,WAAW,GAAG,MAAM,CAAC;AAC1D,MAAM,MAAM,UAAU,GAAG,cAAc,GAAG,MAAM,GAAG,WAAW,GAAG,UAAU,GAAG,UAAU,CAAC;AAEzF,oBAAY,WAAW;IACrB,KAAK,UAAU;IACf,YAAY,iBAAiB;IAC7B,wBAAwB,6BAA6B;IACrD,cAAc,mBAAmB;CAClC;AAED,MAAM,MAAM,cAAc,GAAG,WAAW,GAAG,MAAM,CAAC;AAElD,MAAM,MAAM,YAAY,GAAG,KAAK,GAAG,gBAAgB,GAAG,eAAe,CAAC;AAEtE,MAAM,MAAM,qBAAqB,GAAG;IAClC,IAAI,EAAE,oBAAoB,CAAC;IAC3B,QAAQ,EAAE,SAAS,CAAC;IACpB,QAAQ,EAAE,SAAS,CAAC;IACpB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,2BAA2B,aAC5B,SAAS,YACT,SAAS,cACR,MAAM,KAChB,qBAKD,CAAC;AAEH,MAAM,MAAM,sBAAsB,GAAG;IACnC,IAAI,EAAE,qBAAqB,CAAC;IAC5B,QAAQ,EAAE,UAAU,CAAC;IACrB,QAAQ,EAAE,UAAU,CAAC;IACrB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,4BAA4B,aAC7B,UAAU,YACV,UAAU,cACT,MAAM,KAChB,sBAKD,CAAC;AAEH,MAAM,MAAM,yBAAyB,GAAG;IACtC,IAAI,EAAE,wBAAwB,CAAC;IAC/B,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IAEjB,iDAAiD;IACjD,SAAS,EAAE,MAAM,GAAG,IAAI,CAAC;IACzB,SAAS,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,YAAY,GAAG,IAAI,CAAC;CAC/B,CAAC;AAEF,eAAO,MAAM,+BAA+B,6DAMzC;IACD,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,SAAS,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;IAC1B,QAAQ,CAAC,EAAE,YAAY,GAAG,IAAI,CAAC;IAC/B,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,KAAG,yBAOF,CAAC;AAEH,MAAM,MAAM,qBAAqB,GAAG;IAClC,IAAI,EAAE,mBAAmB,CAAC;IAC1B,OAAO,EAAE,YAAY,CAAC;IACtB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,2BAA2B,4BAGrC;IACD,OAAO,EAAE,YAAY,CAAC;IACtB,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,KAAG,qBAIF,CAAC;AAEH,MAAM,MAAM,0BAA0B,GAAG;IACvC,IAAI,EAAE,yBAAyB,CAAC;IAChC,IAAI,EAAE,WAAW,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,gCAAgC,SACrC,WAAW,cACN,MAAM,KAChB,0BAID,CAAC;AAEH,MAAM,MAAM,0BAA0B,GAAG;IACvC,IAAI,EAAE,yBAAyB,CAAC;IAChC,aAAa,EAAE,YAAY,EAAE,CAAC;IAC9B,mBAAmB,EAAE,kBAAkB,EAAE,CAAC;IAC1C,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,gCAAgC,uDAI1C;IACD,aAAa,EAAE,YAAY,EAAE,CAAC;IAC9B,mBAAmB,EAAE,kBAAkB,EAAE,CAAC;IAC1C,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,KAAG,0BAOH,CAAC;AAEF,eAAO,MAAM,0BAA0B,UAC9B,0BAA0B,KAChC,MAAM,CAAC,YAAY,EAAE,kBAAkB,CAAC,CAE1C,CAAC;AAEF,MAAM,MAAM,kBAAkB,GAAG;IAC/B,IAAI,EAAE,gBAAgB,CAAC;IACvB;;OAEG;IACH,aAAa,EAAE,OAAO,CAAC;IACvB;;OAEG;IACH,MAAM,EAAE,YAAY,CAAC;IACrB;;OAEG;IAEH,YAAY,EAAE,YAAY,CAAC;IAC3B;;OAEG;IACH,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,wBAAwB,wDAKlC;IACD,aAAa,EAAE,OAAO,CAAC;IACvB,MAAM,EAAE,YAAY,CAAC;IACrB,YAAY,EAAE,YAAY,CAAC;IAC3B,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,KAAG,kBAMF,CAAC;AAEH,MAAM,MAAM,UAAU,GAAG;IACvB,IAAI,EAAE,OAAO,CAAC;IACd,KAAK,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,0BAA0B,GAAG,OAAO,CAAC;IAClG,MAAM,EAAE,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,aAAa,GAAG,OAAO,CAAC;IAClD,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,gBAAgB,UACpB,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,0BAA0B,GAAG,OAAO,UACzF,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,aAAa,GAAG,OAAO,cACtC,MAAM,KAChB,UAKD,CAAC;AAEH,MAAM,MAAM,UAAU,GAAG;IACvB,IAAI,EAAE,OAAO,CAAC;IACd,KAAK,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,0BAA0B,GAAG,IAAI,CAAC;IAC/F,MAAM,EAAE,cAAc,CAAC;IACvB,SAAS,EAAE,MAAM,CAAC;CACnB,CAAC;AAEF,eAAO,MAAM,gBAAgB,WACnB,cAAc,UAElB,kBAAkB,GAClB,QAAQ,GACR,QAAQ,GACR,QAAQ,GACR,0BAA0B,GAC1B,IAAI,cACG,MAAM,KAChB,UAKD,CAAC;AAEH,MAAM,MAAM,UAAU,GAClB,yBAAyB,GACzB,qBAAqB,GACrB,sBAAsB,GACtB,qBAAqB,GACrB,0BAA0B,GAC1B,0BAA0B,GAC1B,kBAAkB,GAClB,sBAAsB,GACtB,UAAU,GACV,UAAU,CAAC"}

package/dist/voice/events.js

@@ -6,7 +6,7 @@ var AgentSessionEventTypes = /* @__PURE__ */ ((AgentSessionEventTypes2) => {
   AgentSessionEventTypes2["FunctionToolsExecuted"] = "function_tools_executed";
   AgentSessionEventTypes2["MetricsCollected"] = "metrics_collected";
   AgentSessionEventTypes2["SpeechCreated"] = "speech_created";
-  AgentSessionEventTypes2["UserOverlappingSpeech"] = "user_overlapping_speech";
+  AgentSessionEventTypes2["OverlappingSpeech"] = "overlapping_speech";
   AgentSessionEventTypes2["Error"] = "error";
   AgentSessionEventTypes2["Close"] = "close";
   return AgentSessionEventTypes2;

package/dist/voice/events.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/events.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { InterruptionDetectionError } from '../inference/interruption/errors.js';\nimport type { OverlappingSpeechEvent } from '../inference/interruption/types.js';\nimport type {\n  ChatMessage,\n  FunctionCall,\n  FunctionCallOutput,\n  LLM,\n  RealtimeModel,\n  RealtimeModelError,\n} from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport type { AgentMetrics } from '../metrics/base.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport type { TTS } from '../tts/index.js';\nimport type { TTSError } from '../tts/tts.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport enum AgentSessionEventTypes {\n  UserInputTranscribed = 'user_input_transcribed',\n  AgentStateChanged = 'agent_state_changed',\n  UserStateChanged = 'user_state_changed',\n  ConversationItemAdded = 'conversation_item_added',\n  FunctionToolsExecuted = 'function_tools_executed',\n  MetricsCollected = 'metrics_collected',\n  SpeechCreated = 'speech_created',\n  UserOverlappingSpeech = 'user_overlapping_speech',\n  Error = 'error',\n  Close = 'close',\n}\n\nexport type UserState = 'speaking' | 'listening' | 'away';\nexport type AgentState = 'initializing' | 'idle' | 'listening' | 'thinking' | 'speaking';\n\nexport enum CloseReason {\n  ERROR = 'error',\n  JOB_SHUTDOWN = 'job_shutdown',\n  PARTICIPANT_DISCONNECTED = 'participant_disconnected',\n  USER_INITIATED = 'user_initiated',\n}\n\nexport type ShutdownReason = CloseReason | string;\n\nexport type SpeechSource = 'say' | 'generate_reply' | 'tool_response';\n\nexport type UserStateChangedEvent = {\n  type: 'user_state_changed';\n  oldState: UserState;\n  newState: UserState;\n  createdAt: number;\n};\n\nexport const createUserStateChangedEvent = (\n  oldState: UserState,\n  newState: UserState,\n  createdAt: number = Date.now(),\n): UserStateChangedEvent => ({\n  type: 'user_state_changed',\n  oldState,\n  newState,\n  createdAt,\n});\n\nexport type AgentStateChangedEvent = {\n  type: 'agent_state_changed';\n  oldState: AgentState;\n  newState: AgentState;\n  createdAt: number;\n};\n\nexport const createAgentStateChangedEvent = (\n  oldState: AgentState,\n  newState: AgentState,\n  createdAt: number = Date.now(),\n): AgentStateChangedEvent => ({\n  type: 'agent_state_changed',\n  oldState,\n  newState,\n  createdAt,\n});\n\nexport type UserInputTranscribedEvent = {\n  type: 'user_input_transcribed';\n  transcript: string;\n  isFinal: boolean;\n  // TODO(AJS-106): add multi participant support\n  /** Not supported yet. Always null by default. */\n  speakerId: string | null;\n  createdAt: number;\n  language: string | null;\n};\n\nexport const createUserInputTranscribedEvent = ({\n  transcript,\n  isFinal,\n  speakerId = null,\n  language = null,\n  createdAt = Date.now(),\n}: {\n  transcript: string;\n  isFinal: boolean;\n  speakerId?: string | null;\n  language?: string | null;\n  createdAt?: number;\n}): UserInputTranscribedEvent => ({\n  type: 'user_input_transcribed',\n  transcript,\n  isFinal,\n  speakerId,\n  language,\n  createdAt,\n});\n\nexport type MetricsCollectedEvent = {\n  type: 'metrics_collected';\n  metrics: AgentMetrics;\n  createdAt: number;\n};\n\nexport const createMetricsCollectedEvent = ({\n  metrics,\n  createdAt = Date.now(),\n}: {\n  metrics: AgentMetrics;\n  createdAt?: number;\n}): MetricsCollectedEvent => ({\n  type: 'metrics_collected',\n  metrics,\n  createdAt,\n});\n\nexport type ConversationItemAddedEvent = {\n  type: 'conversation_item_added';\n  item: ChatMessage;\n  createdAt: number;\n};\n\nexport const createConversationItemAddedEvent = (\n  item: ChatMessage,\n  createdAt: number = Date.now(),\n): ConversationItemAddedEvent => ({\n  type: 'conversation_item_added',\n  item,\n  createdAt,\n});\n\nexport type FunctionToolsExecutedEvent = {\n  type: 'function_tools_executed';\n  functionCalls: FunctionCall[];\n  functionCallOutputs: FunctionCallOutput[];\n  createdAt: number;\n};\n\nexport const createFunctionToolsExecutedEvent = ({\n  functionCalls,\n  functionCallOutputs,\n  createdAt = Date.now(),\n}: {\n  functionCalls: FunctionCall[];\n  functionCallOutputs: FunctionCallOutput[];\n  createdAt?: number;\n}): FunctionToolsExecutedEvent => {\n  return {\n    type: 'function_tools_executed',\n    functionCalls,\n    functionCallOutputs,\n    createdAt,\n  };\n};\n\nexport const zipFunctionCallsAndOutputs = (\n  event: FunctionToolsExecutedEvent,\n): Array<[FunctionCall, FunctionCallOutput]> => {\n  return event.functionCalls.map((call, index) => [call, event.functionCallOutputs[index]!]);\n};\n\nexport type SpeechCreatedEvent = {\n  type: 'speech_created';\n  /**\n   * True if the speech was created using public methods like `say` or `generate_reply`\n   */\n  userInitiated: boolean;\n  /**\n   * Source indicating how the speech handle was created\n   */\n  source: SpeechSource;\n  /**\n   * The speech handle that was created\n   */\n  // TODO(shubhra): we need to make sure this doesn't get serialized\n  speechHandle: SpeechHandle;\n  /**\n   * The timestamp when the speech handle was created\n   */\n  createdAt: number;\n};\n\nexport const createSpeechCreatedEvent = ({\n  userInitiated,\n  source,\n  speechHandle,\n  createdAt = Date.now(),\n}: {\n  userInitiated: boolean;\n  source: SpeechSource;\n  speechHandle: SpeechHandle;\n  createdAt?: number;\n}): SpeechCreatedEvent => ({\n  type: 'speech_created',\n  userInitiated,\n  source,\n  speechHandle,\n  createdAt,\n});\n\nexport type ErrorEvent = {\n  type: 'error';\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | unknown;\n  source: LLM | STT | TTS | RealtimeModel | unknown;\n  createdAt: number;\n};\n\nexport const createErrorEvent = (\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | unknown,\n  source: LLM | STT | TTS | RealtimeModel | unknown,\n  createdAt: number = Date.now(),\n): ErrorEvent => ({\n  type: 'error',\n  error,\n  source,\n  createdAt,\n});\n\nexport type CloseEvent = {\n  type: 'close';\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | null;\n  reason: ShutdownReason;\n  createdAt: number;\n};\n\nexport const createCloseEvent = (\n  reason: ShutdownReason,\n  error:\n    | RealtimeModelError\n    | STTError\n    | TTSError\n    | LLMError\n    | InterruptionDetectionError\n    | null = null,\n  createdAt: number = Date.now(),\n): CloseEvent => ({\n  type: 'close',\n  error,\n  reason,\n  createdAt,\n});\n\nexport type AgentEvent =\n  | UserInputTranscribedEvent\n  | UserStateChangedEvent\n  | AgentStateChangedEvent\n  | MetricsCollectedEvent\n  | ConversationItemAddedEvent\n  | FunctionToolsExecutedEvent\n  | SpeechCreatedEvent\n  | OverlappingSpeechEvent\n  | ErrorEvent\n  | CloseEvent;\n"],"mappings":"AAqBO,IAAK,yBAAL,kBAAKA,4BAAL;AACL,EAAAA,wBAAA,0BAAuB;AACvB,EAAAA,wBAAA,uBAAoB;AACpB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,mBAAgB;AAChB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,WAAQ;AACR,EAAAA,wBAAA,WAAQ;AAVE,SAAAA;AAAA,GAAA;AAgBL,IAAK,cAAL,kBAAKC,iBAAL;AACL,EAAAA,aAAA,WAAQ;AACR,EAAAA,aAAA,kBAAe;AACf,EAAAA,aAAA,8BAA2B;AAC3B,EAAAA,aAAA,oBAAiB;AAJP,SAAAA;AAAA,GAAA;AAkBL,MAAM,8BAA8B,CACzC,UACA,UACA,YAAoB,KAAK,IAAI,OACF;AAAA,EAC3B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,+BAA+B,CAC1C,UACA,UACA,YAAoB,KAAK,IAAI,OACD;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AAaO,MAAM,kCAAkC,CAAC;AAAA,EAC9C;AAAA,EACA;AAAA,EACA,YAAY;AAAA,EACZ,WAAW;AAAA,EACX,YAAY,KAAK,IAAI;AACvB,OAMkC;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAQO,MAAM,8BAA8B,CAAC;AAAA,EAC1C;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAG8B;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AACF;AAQO,MAAM,mCAAmC,CAC9C,MACA,YAAoB,KAAK,IAAI,OACG;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AACF;AASO,MAAM,mCAAmC,CAAC;AAAA,EAC/C;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,MAIkC;AAChC,SAAO;AAAA,IACL,MAAM;AAAA,IACN;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEO,MAAM,6BAA6B,CACxC,UAC8C;AAC9C,SAAO,MAAM,cAAc,IAAI,CAAC,MAAM,UAAU,CAAC,MAAM,MAAM,oBAAoB,KAAK,CAAE,CAAC;AAC3F;AAuBO,MAAM,2BAA2B,CAAC;AAAA,EACvC;AAAA,EACA;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAK2B;AAAA,EACzB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,OACA,QACA,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,QACA,QAMW,MACX,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;","names":["AgentSessionEventTypes","CloseReason"]}
+{"version":3,"sources":["../../src/voice/events.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { InterruptionDetectionError } from '../inference/interruption/errors.js';\nimport type { OverlappingSpeechEvent } from '../inference/interruption/types.js';\nimport type { LanguageCode } from '../language.js';\nimport type {\n  ChatMessage,\n  FunctionCall,\n  FunctionCallOutput,\n  LLM,\n  RealtimeModel,\n  RealtimeModelError,\n} from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport type { AgentMetrics } from '../metrics/base.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport type { TTS } from '../tts/index.js';\nimport type { TTSError } from '../tts/tts.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport enum AgentSessionEventTypes {\n  UserInputTranscribed = 'user_input_transcribed',\n  AgentStateChanged = 'agent_state_changed',\n  UserStateChanged = 'user_state_changed',\n  ConversationItemAdded = 'conversation_item_added',\n  FunctionToolsExecuted = 'function_tools_executed',\n  MetricsCollected = 'metrics_collected',\n  SpeechCreated = 'speech_created',\n  OverlappingSpeech = 'overlapping_speech',\n  Error = 'error',\n  Close = 'close',\n}\n\nexport type UserState = 'speaking' | 'listening' | 'away';\nexport type AgentState = 'initializing' | 'idle' | 'listening' | 'thinking' | 'speaking';\n\nexport enum CloseReason {\n  ERROR = 'error',\n  JOB_SHUTDOWN = 'job_shutdown',\n  PARTICIPANT_DISCONNECTED = 'participant_disconnected',\n  USER_INITIATED = 'user_initiated',\n}\n\nexport type ShutdownReason = CloseReason | string;\n\nexport type SpeechSource = 'say' | 'generate_reply' | 'tool_response';\n\nexport type UserStateChangedEvent = {\n  type: 'user_state_changed';\n  oldState: UserState;\n  newState: UserState;\n  createdAt: number;\n};\n\nexport const createUserStateChangedEvent = (\n  oldState: UserState,\n  newState: UserState,\n  createdAt: number = Date.now(),\n): UserStateChangedEvent => ({\n  type: 'user_state_changed',\n  oldState,\n  newState,\n  createdAt,\n});\n\nexport type AgentStateChangedEvent = {\n  type: 'agent_state_changed';\n  oldState: AgentState;\n  newState: AgentState;\n  createdAt: number;\n};\n\nexport const createAgentStateChangedEvent = (\n  oldState: AgentState,\n  newState: AgentState,\n  createdAt: number = Date.now(),\n): AgentStateChangedEvent => ({\n  type: 'agent_state_changed',\n  oldState,\n  newState,\n  createdAt,\n});\n\nexport type UserInputTranscribedEvent = {\n  type: 'user_input_transcribed';\n  transcript: string;\n  isFinal: boolean;\n  // TODO(AJS-106): add multi participant support\n  /** Not supported yet. Always null by default. */\n  speakerId: string | null;\n  createdAt: number;\n  language: LanguageCode | null;\n};\n\nexport const createUserInputTranscribedEvent = ({\n  transcript,\n  isFinal,\n  speakerId = null,\n  language = null,\n  createdAt = Date.now(),\n}: {\n  transcript: string;\n  isFinal: boolean;\n  speakerId?: string | null;\n  language?: LanguageCode | null;\n  createdAt?: number;\n}): UserInputTranscribedEvent => ({\n  type: 'user_input_transcribed',\n  transcript,\n  isFinal,\n  speakerId,\n  language,\n  createdAt,\n});\n\nexport type MetricsCollectedEvent = {\n  type: 'metrics_collected';\n  metrics: AgentMetrics;\n  createdAt: number;\n};\n\nexport const createMetricsCollectedEvent = ({\n  metrics,\n  createdAt = Date.now(),\n}: {\n  metrics: AgentMetrics;\n  createdAt?: number;\n}): MetricsCollectedEvent => ({\n  type: 'metrics_collected',\n  metrics,\n  createdAt,\n});\n\nexport type ConversationItemAddedEvent = {\n  type: 'conversation_item_added';\n  item: ChatMessage;\n  createdAt: number;\n};\n\nexport const createConversationItemAddedEvent = (\n  item: ChatMessage,\n  createdAt: number = Date.now(),\n): ConversationItemAddedEvent => ({\n  type: 'conversation_item_added',\n  item,\n  createdAt,\n});\n\nexport type FunctionToolsExecutedEvent = {\n  type: 'function_tools_executed';\n  functionCalls: FunctionCall[];\n  functionCallOutputs: FunctionCallOutput[];\n  createdAt: number;\n};\n\nexport const createFunctionToolsExecutedEvent = ({\n  functionCalls,\n  functionCallOutputs,\n  createdAt = Date.now(),\n}: {\n  functionCalls: FunctionCall[];\n  functionCallOutputs: FunctionCallOutput[];\n  createdAt?: number;\n}): FunctionToolsExecutedEvent => {\n  return {\n    type: 'function_tools_executed',\n    functionCalls,\n    functionCallOutputs,\n    createdAt,\n  };\n};\n\nexport const zipFunctionCallsAndOutputs = (\n  event: FunctionToolsExecutedEvent,\n): Array<[FunctionCall, FunctionCallOutput]> => {\n  return event.functionCalls.map((call, index) => [call, event.functionCallOutputs[index]!]);\n};\n\nexport type SpeechCreatedEvent = {\n  type: 'speech_created';\n  /**\n   * True if the speech was created using public methods like `say` or `generate_reply`\n   */\n  userInitiated: boolean;\n  /**\n   * Source indicating how the speech handle was created\n   */\n  source: SpeechSource;\n  /**\n   * The speech handle that was created\n   */\n  // TODO(shubhra): we need to make sure this doesn't get serialized\n  speechHandle: SpeechHandle;\n  /**\n   * The timestamp when the speech handle was created\n   */\n  createdAt: number;\n};\n\nexport const createSpeechCreatedEvent = ({\n  userInitiated,\n  source,\n  speechHandle,\n  createdAt = Date.now(),\n}: {\n  userInitiated: boolean;\n  source: SpeechSource;\n  speechHandle: SpeechHandle;\n  createdAt?: number;\n}): SpeechCreatedEvent => ({\n  type: 'speech_created',\n  userInitiated,\n  source,\n  speechHandle,\n  createdAt,\n});\n\nexport type ErrorEvent = {\n  type: 'error';\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | unknown;\n  source: LLM | STT | TTS | RealtimeModel | unknown;\n  createdAt: number;\n};\n\nexport const createErrorEvent = (\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | unknown,\n  source: LLM | STT | TTS | RealtimeModel | unknown,\n  createdAt: number = Date.now(),\n): ErrorEvent => ({\n  type: 'error',\n  error,\n  source,\n  createdAt,\n});\n\nexport type CloseEvent = {\n  type: 'close';\n  error: RealtimeModelError | STTError | TTSError | LLMError | InterruptionDetectionError | null;\n  reason: ShutdownReason;\n  createdAt: number;\n};\n\nexport const createCloseEvent = (\n  reason: ShutdownReason,\n  error:\n    | RealtimeModelError\n    | STTError\n    | TTSError\n    | LLMError\n    | InterruptionDetectionError\n    | null = null,\n  createdAt: number = Date.now(),\n): CloseEvent => ({\n  type: 'close',\n  error,\n  reason,\n  createdAt,\n});\n\nexport type AgentEvent =\n  | UserInputTranscribedEvent\n  | UserStateChangedEvent\n  | AgentStateChangedEvent\n  | MetricsCollectedEvent\n  | ConversationItemAddedEvent\n  | FunctionToolsExecutedEvent\n  | SpeechCreatedEvent\n  | OverlappingSpeechEvent\n  | ErrorEvent\n  | CloseEvent;\n"],"mappings":"AAsBO,IAAK,yBAAL,kBAAKA,4BAAL;AACL,EAAAA,wBAAA,0BAAuB;AACvB,EAAAA,wBAAA,uBAAoB;AACpB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,2BAAwB;AACxB,EAAAA,wBAAA,sBAAmB;AACnB,EAAAA,wBAAA,mBAAgB;AAChB,EAAAA,wBAAA,uBAAoB;AACpB,EAAAA,wBAAA,WAAQ;AACR,EAAAA,wBAAA,WAAQ;AAVE,SAAAA;AAAA,GAAA;AAgBL,IAAK,cAAL,kBAAKC,iBAAL;AACL,EAAAA,aAAA,WAAQ;AACR,EAAAA,aAAA,kBAAe;AACf,EAAAA,aAAA,8BAA2B;AAC3B,EAAAA,aAAA,oBAAiB;AAJP,SAAAA;AAAA,GAAA;AAkBL,MAAM,8BAA8B,CACzC,UACA,UACA,YAAoB,KAAK,IAAI,OACF;AAAA,EAC3B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,+BAA+B,CAC1C,UACA,UACA,YAAoB,KAAK,IAAI,OACD;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AAaO,MAAM,kCAAkC,CAAC;AAAA,EAC9C;AAAA,EACA;AAAA,EACA,YAAY;AAAA,EACZ,WAAW;AAAA,EACX,YAAY,KAAK,IAAI;AACvB,OAMkC;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAQO,MAAM,8BAA8B,CAAC;AAAA,EAC1C;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAG8B;AAAA,EAC5B,MAAM;AAAA,EACN;AAAA,EACA;AACF;AAQO,MAAM,mCAAmC,CAC9C,MACA,YAAoB,KAAK,IAAI,OACG;AAAA,EAChC,MAAM;AAAA,EACN;AAAA,EACA;AACF;AASO,MAAM,mCAAmC,CAAC;AAAA,EAC/C;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,MAIkC;AAChC,SAAO;AAAA,IACL,MAAM;AAAA,IACN;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEO,MAAM,6BAA6B,CACxC,UAC8C;AAC9C,SAAO,MAAM,cAAc,IAAI,CAAC,MAAM,UAAU,CAAC,MAAM,MAAM,oBAAoB,KAAK,CAAE,CAAC;AAC3F;AAuBO,MAAM,2BAA2B,CAAC;AAAA,EACvC;AAAA,EACA;AAAA,EACA;AAAA,EACA,YAAY,KAAK,IAAI;AACvB,OAK2B;AAAA,EACzB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,OACA,QACA,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;AASO,MAAM,mBAAmB,CAC9B,QACA,QAMW,MACX,YAAoB,KAAK,IAAI,OACb;AAAA,EAChB,MAAM;AAAA,EACN;AAAA,EACA;AAAA,EACA;AACF;","names":["AgentSessionEventTypes","CloseReason"]}

package/dist/voice/index.cjs

@@ -32,7 +32,11 @@ __export(voice_exports, {
   Agent: () => import_agent.Agent,
   AgentSession: () => import_agent_session.AgentSession,
   AgentTask: () => import_agent.AgentTask,
+  RemoteSession: () => import_remote_session.RemoteSession,
+  RoomSessionTransport: () => import_remote_session.RoomSessionTransport,
   RunContext: () => import_run_context.RunContext,
+  SessionHost: () => import_remote_session.SessionHost,
+  SessionTransport: () => import_remote_session.SessionTransport,
   StopResponse: () => import_agent.StopResponse,
   testing: () => testing
 });
@@ -41,7 +45,7 @@ var import_agent = require("./agent.cjs");
 var import_agent_session = require("./agent_session.cjs");
 __reExport(voice_exports, require("./avatar/index.cjs"), module.exports);
 __reExport(voice_exports, require("./background_audio.cjs"), module.exports);
-var import_client_events = require("./client_events.cjs");
+var import_remote_session = require("./remote_session.cjs");
 __reExport(voice_exports, require("./events.cjs"), module.exports);
 var import_io = require("./io.cjs");
 __reExport(voice_exports, require("./report.cjs"), module.exports);
@@ -53,7 +57,11 @@ var testing = __toESM(require("./testing/index.cjs"), 1);
   Agent,
   AgentSession,
   AgentTask,
+  RemoteSession,
+  RoomSessionTransport,
   RunContext,
+  SessionHost,
+  SessionTransport,
   StopResponse,
   testing,
   ...require("./avatar/index.cjs"),

package/dist/voice/index.cjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport { Agent, AgentTask, StopResponse, type AgentOptions, type ModelSettings } from './agent.js';\nexport { AgentSession, type AgentSessionOptions, type VoiceOptions } from './agent_session.js';\nexport * from './avatar/index.js';\nexport * from './background_audio.js';\nexport { type TextInputCallback, type TextInputEvent } from './client_events.js';\nexport * from './events.js';\nexport { type TimedString } from './io.js';\nexport * from './report.js';\nexport * from './room_io/index.js';\nexport { RunContext } from './run_context.js';\nexport * as testing from './testing/index.js';\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,mBAAsF;AACtF,2BAA0E;AAC1E,0BAAc,8BALd;AAMA,0BAAc,kCANd;AAOA,2BAA4D;AAC5D,0BAAc,wBARd;AASA,gBAAiC;AACjC,0BAAc,wBAVd;AAWA,0BAAc,+BAXd;AAYA,yBAA2B;AAC3B,cAAyB;","names":[]}
+{"version":3,"sources":["../../src/voice/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport { Agent, AgentTask, StopResponse, type AgentOptions, type ModelSettings } from './agent.js';\nexport { AgentSession, type AgentSessionOptions, type VoiceOptions } from './agent_session.js';\nexport * from './avatar/index.js';\nexport * from './background_audio.js';\nexport {\n  type TextInputCallback,\n  type TextInputEvent,\n  RemoteSession,\n  type RemoteSessionCallbacks,\n  type RemoteSessionEventTypes,\n  SessionHost,\n  SessionTransport,\n  RoomSessionTransport,\n} from './remote_session.js';\nexport * from './events.js';\nexport { type TimedString } from './io.js';\nexport * from './report.js';\nexport * from './room_io/index.js';\nexport { RunContext } from './run_context.js';\nexport * as testing from './testing/index.js';\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,mBAAsF;AACtF,2BAA0E;AAC1E,0BAAc,8BALd;AAMA,0BAAc,kCANd;AAOA,4BASO;AACP,0BAAc,wBAjBd;AAkBA,gBAAiC;AACjC,0BAAc,wBAnBd;AAoBA,0BAAc,+BApBd;AAqBA,yBAA2B;AAC3B,cAAyB;","names":[]}

package/dist/voice/index.d.cts

@@ -2,7 +2,7 @@ export { Agent, AgentTask, StopResponse, type AgentOptions, type ModelSettings }
 export { AgentSession, type AgentSessionOptions, type VoiceOptions } from './agent_session.js';
 export * from './avatar/index.js';
 export * from './background_audio.js';
-export { type TextInputCallback, type TextInputEvent } from './client_events.js';
+export { type TextInputCallback, type TextInputEvent, RemoteSession, type RemoteSessionCallbacks, type RemoteSessionEventTypes, SessionHost, SessionTransport, RoomSessionTransport, } from './remote_session.js';
 export * from './events.js';
 export { type TimedString } from './io.js';
 export * from './report.js';

package/dist/voice/index.d.ts

@@ -2,7 +2,7 @@ export { Agent, AgentTask, StopResponse, type AgentOptions, type ModelSettings }
 export { AgentSession, type AgentSessionOptions, type VoiceOptions } from './agent_session.js';
 export * from './avatar/index.js';
 export * from './background_audio.js';
-export { type TextInputCallback, type TextInputEvent } from './client_events.js';
+export { type TextInputCallback, type TextInputEvent, RemoteSession, type RemoteSessionCallbacks, type RemoteSessionEventTypes, SessionHost, SessionTransport, RoomSessionTransport, } from './remote_session.js';
 export * from './events.js';
 export { type TimedString } from './io.js';
 export * from './report.js';

package/dist/voice/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/voice/index.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,KAAK,EAAE,SAAS,EAAE,YAAY,EAAE,KAAK,YAAY,EAAE,KAAK,aAAa,EAAE,MAAM,YAAY,CAAC;AACnG,OAAO,EAAE,YAAY,EAAE,KAAK,mBAAmB,EAAE,KAAK,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAC/F,cAAc,mBAAmB,CAAC;AAClC,cAAc,uBAAuB,CAAC;AACtC,OAAO,EAAE,KAAK,iBAAiB,EAAE,KAAK,cAAc,EAAE,MAAM,oBAAoB,CAAC;AACjF,cAAc,aAAa,CAAC;AAC5B,OAAO,EAAE,KAAK,WAAW,EAAE,MAAM,SAAS,CAAC;AAC3C,cAAc,aAAa,CAAC;AAC5B,cAAc,oBAAoB,CAAC;AACnC,OAAO,EAAE,UAAU,EAAE,MAAM,kBAAkB,CAAC;AAC9C,OAAO,KAAK,OAAO,MAAM,oBAAoB,CAAC"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/voice/index.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,KAAK,EAAE,SAAS,EAAE,YAAY,EAAE,KAAK,YAAY,EAAE,KAAK,aAAa,EAAE,MAAM,YAAY,CAAC;AACnG,OAAO,EAAE,YAAY,EAAE,KAAK,mBAAmB,EAAE,KAAK,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAC/F,cAAc,mBAAmB,CAAC;AAClC,cAAc,uBAAuB,CAAC;AACtC,OAAO,EACL,KAAK,iBAAiB,EACtB,KAAK,cAAc,EACnB,aAAa,EACb,KAAK,sBAAsB,EAC3B,KAAK,uBAAuB,EAC5B,WAAW,EACX,gBAAgB,EAChB,oBAAoB,GACrB,MAAM,qBAAqB,CAAC;AAC7B,cAAc,aAAa,CAAC;AAC5B,OAAO,EAAE,KAAK,WAAW,EAAE,MAAM,SAAS,CAAC;AAC3C,cAAc,aAAa,CAAC;AAC5B,cAAc,oBAAoB,CAAC;AACnC,OAAO,EAAE,UAAU,EAAE,MAAM,kBAAkB,CAAC;AAC9C,OAAO,KAAK,OAAO,MAAM,oBAAoB,CAAC"}

package/dist/voice/index.js

@@ -2,7 +2,12 @@ import { Agent, AgentTask, StopResponse } from "./agent.js";
 import { AgentSession } from "./agent_session.js";
 export * from "./avatar/index.js";
 export * from "./background_audio.js";
-import {} from "./client_events.js";
+import {
+  RemoteSession,
+  SessionHost,
+  SessionTransport,
+  RoomSessionTransport
+} from "./remote_session.js";
 export * from "./events.js";
 import {} from "./io.js";
 export * from "./report.js";
@@ -13,7 +18,11 @@ export {
   Agent,
   AgentSession,
   AgentTask,
+  RemoteSession,
+  RoomSessionTransport,
   RunContext,
+  SessionHost,
+  SessionTransport,
   StopResponse,
   testing
 };

package/dist/voice/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport { Agent, AgentTask, StopResponse, type AgentOptions, type ModelSettings } from './agent.js';\nexport { AgentSession, type AgentSessionOptions, type VoiceOptions } from './agent_session.js';\nexport * from './avatar/index.js';\nexport * from './background_audio.js';\nexport { type TextInputCallback, type TextInputEvent } from './client_events.js';\nexport * from './events.js';\nexport { type TimedString } from './io.js';\nexport * from './report.js';\nexport * from './room_io/index.js';\nexport { RunContext } from './run_context.js';\nexport * as testing from './testing/index.js';\n"],"mappings":"AAGA,SAAS,OAAO,WAAW,oBAA2D;AACtF,SAAS,oBAAiE;AAC1E,cAAc;AACd,cAAc;AACd,eAA4D;AAC5D,cAAc;AACd,eAAiC;AACjC,cAAc;AACd,cAAc;AACd,SAAS,kBAAkB;AAC3B,YAAY,aAAa;","names":[]}
+{"version":3,"sources":["../../src/voice/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport { Agent, AgentTask, StopResponse, type AgentOptions, type ModelSettings } from './agent.js';\nexport { AgentSession, type AgentSessionOptions, type VoiceOptions } from './agent_session.js';\nexport * from './avatar/index.js';\nexport * from './background_audio.js';\nexport {\n  type TextInputCallback,\n  type TextInputEvent,\n  RemoteSession,\n  type RemoteSessionCallbacks,\n  type RemoteSessionEventTypes,\n  SessionHost,\n  SessionTransport,\n  RoomSessionTransport,\n} from './remote_session.js';\nexport * from './events.js';\nexport { type TimedString } from './io.js';\nexport * from './report.js';\nexport * from './room_io/index.js';\nexport { RunContext } from './run_context.js';\nexport * as testing from './testing/index.js';\n"],"mappings":"AAGA,SAAS,OAAO,WAAW,oBAA2D;AACtF,SAAS,oBAAiE;AAC1E,cAAc;AACd,cAAc;AACd;AAAA,EAGE;AAAA,EAGA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,cAAc;AACd,eAAiC;AACjC,cAAc;AACd,cAAc;AACd,SAAS,kBAAkB;AAC3B,YAAY,aAAa;","names":[]}