@livekit/agents 1.0.37 → 1.0.38

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (133) hide show
  1. package/dist/cli.cjs.map +1 -1
  2. package/dist/inference/api_protos.cjs +68 -0
  3. package/dist/inference/api_protos.cjs.map +1 -1
  4. package/dist/inference/api_protos.d.cts +345 -4
  5. package/dist/inference/api_protos.d.ts +345 -4
  6. package/dist/inference/api_protos.d.ts.map +1 -1
  7. package/dist/inference/api_protos.js +60 -0
  8. package/dist/inference/api_protos.js.map +1 -1
  9. package/dist/inference/stt.cjs +32 -21
  10. package/dist/inference/stt.cjs.map +1 -1
  11. package/dist/inference/stt.d.ts.map +1 -1
  12. package/dist/inference/stt.js +34 -21
  13. package/dist/inference/stt.js.map +1 -1
  14. package/dist/ipc/inference_proc_executor.cjs.map +1 -1
  15. package/dist/ipc/job_proc_executor.cjs.map +1 -1
  16. package/dist/stt/stt.cjs +10 -0
  17. package/dist/stt/stt.cjs.map +1 -1
  18. package/dist/stt/stt.d.cts +12 -0
  19. package/dist/stt/stt.d.ts +12 -0
  20. package/dist/stt/stt.d.ts.map +1 -1
  21. package/dist/stt/stt.js +10 -0
  22. package/dist/stt/stt.js.map +1 -1
  23. package/dist/telemetry/traces.cjs +4 -3
  24. package/dist/telemetry/traces.cjs.map +1 -1
  25. package/dist/telemetry/traces.d.cts +2 -0
  26. package/dist/telemetry/traces.d.ts +2 -0
  27. package/dist/telemetry/traces.d.ts.map +1 -1
  28. package/dist/telemetry/traces.js +4 -3
  29. package/dist/telemetry/traces.js.map +1 -1
  30. package/dist/utils.cjs +6 -0
  31. package/dist/utils.cjs.map +1 -1
  32. package/dist/utils.d.cts +2 -0
  33. package/dist/utils.d.ts +2 -0
  34. package/dist/utils.d.ts.map +1 -1
  35. package/dist/utils.js +6 -0
  36. package/dist/utils.js.map +1 -1
  37. package/dist/voice/agent.cjs +5 -0
  38. package/dist/voice/agent.cjs.map +1 -1
  39. package/dist/voice/agent.d.ts.map +1 -1
  40. package/dist/voice/agent.js +5 -0
  41. package/dist/voice/agent.js.map +1 -1
  42. package/dist/voice/agent_activity.cjs +49 -23
  43. package/dist/voice/agent_activity.cjs.map +1 -1
  44. package/dist/voice/agent_activity.d.cts +1 -1
  45. package/dist/voice/agent_activity.d.ts +1 -1
  46. package/dist/voice/agent_activity.d.ts.map +1 -1
  47. package/dist/voice/agent_activity.js +50 -24
  48. package/dist/voice/agent_activity.js.map +1 -1
  49. package/dist/voice/agent_session.cjs +7 -5
  50. package/dist/voice/agent_session.cjs.map +1 -1
  51. package/dist/voice/agent_session.d.cts +5 -2
  52. package/dist/voice/agent_session.d.ts +5 -2
  53. package/dist/voice/agent_session.d.ts.map +1 -1
  54. package/dist/voice/agent_session.js +7 -5
  55. package/dist/voice/agent_session.js.map +1 -1
  56. package/dist/voice/audio_recognition.cjs +3 -1
  57. package/dist/voice/audio_recognition.cjs.map +1 -1
  58. package/dist/voice/audio_recognition.d.ts.map +1 -1
  59. package/dist/voice/audio_recognition.js +3 -1
  60. package/dist/voice/audio_recognition.js.map +1 -1
  61. package/dist/voice/avatar/datastream_io.cjs +6 -0
  62. package/dist/voice/avatar/datastream_io.cjs.map +1 -1
  63. package/dist/voice/avatar/datastream_io.d.cts +1 -0
  64. package/dist/voice/avatar/datastream_io.d.ts +1 -0
  65. package/dist/voice/avatar/datastream_io.d.ts.map +1 -1
  66. package/dist/voice/avatar/datastream_io.js +6 -0
  67. package/dist/voice/avatar/datastream_io.js.map +1 -1
  68. package/dist/voice/background_audio.cjs.map +1 -1
  69. package/dist/voice/generation.cjs +14 -5
  70. package/dist/voice/generation.cjs.map +1 -1
  71. package/dist/voice/generation.d.cts +3 -2
  72. package/dist/voice/generation.d.ts +3 -2
  73. package/dist/voice/generation.d.ts.map +1 -1
  74. package/dist/voice/generation.js +14 -5
  75. package/dist/voice/generation.js.map +1 -1
  76. package/dist/voice/io.cjs +12 -0
  77. package/dist/voice/io.cjs.map +1 -1
  78. package/dist/voice/io.d.cts +19 -1
  79. package/dist/voice/io.d.ts +19 -1
  80. package/dist/voice/io.d.ts.map +1 -1
  81. package/dist/voice/io.js +12 -0
  82. package/dist/voice/io.js.map +1 -1
  83. package/dist/voice/recorder_io/recorder_io.cjs +91 -28
  84. package/dist/voice/recorder_io/recorder_io.cjs.map +1 -1
  85. package/dist/voice/recorder_io/recorder_io.d.cts +7 -1
  86. package/dist/voice/recorder_io/recorder_io.d.ts +7 -1
  87. package/dist/voice/recorder_io/recorder_io.d.ts.map +1 -1
  88. package/dist/voice/recorder_io/recorder_io.js +91 -28
  89. package/dist/voice/recorder_io/recorder_io.js.map +1 -1
  90. package/dist/voice/room_io/_input.cjs +40 -11
  91. package/dist/voice/room_io/_input.cjs.map +1 -1
  92. package/dist/voice/room_io/_input.d.cts +4 -1
  93. package/dist/voice/room_io/_input.d.ts +4 -1
  94. package/dist/voice/room_io/_input.d.ts.map +1 -1
  95. package/dist/voice/room_io/_input.js +31 -2
  96. package/dist/voice/room_io/_input.js.map +1 -1
  97. package/dist/voice/room_io/_output.cjs +6 -0
  98. package/dist/voice/room_io/_output.cjs.map +1 -1
  99. package/dist/voice/room_io/_output.d.cts +1 -0
  100. package/dist/voice/room_io/_output.d.ts +1 -0
  101. package/dist/voice/room_io/_output.d.ts.map +1 -1
  102. package/dist/voice/room_io/_output.js +6 -0
  103. package/dist/voice/room_io/_output.js.map +1 -1
  104. package/dist/voice/room_io/room_io.cjs.map +1 -1
  105. package/dist/voice/room_io/room_io.d.cts +2 -2
  106. package/dist/voice/room_io/room_io.d.ts +2 -2
  107. package/dist/voice/room_io/room_io.d.ts.map +1 -1
  108. package/dist/voice/room_io/room_io.js.map +1 -1
  109. package/dist/voice/speech_handle.cjs +2 -0
  110. package/dist/voice/speech_handle.cjs.map +1 -1
  111. package/dist/voice/speech_handle.d.cts +3 -0
  112. package/dist/voice/speech_handle.d.ts +3 -0
  113. package/dist/voice/speech_handle.d.ts.map +1 -1
  114. package/dist/voice/speech_handle.js +2 -0
  115. package/dist/voice/speech_handle.js.map +1 -1
  116. package/package.json +1 -1
  117. package/src/inference/api_protos.ts +83 -0
  118. package/src/inference/stt.ts +39 -22
  119. package/src/stt/stt.ts +21 -0
  120. package/src/telemetry/traces.ts +6 -2
  121. package/src/utils.ts +7 -0
  122. package/src/voice/agent.ts +9 -0
  123. package/src/voice/agent_activity.ts +72 -26
  124. package/src/voice/agent_session.ts +6 -5
  125. package/src/voice/audio_recognition.ts +2 -0
  126. package/src/voice/avatar/datastream_io.ts +8 -0
  127. package/src/voice/generation.ts +24 -12
  128. package/src/voice/io.ts +27 -5
  129. package/src/voice/recorder_io/recorder_io.ts +123 -31
  130. package/src/voice/room_io/_input.ts +32 -4
  131. package/src/voice/room_io/_output.ts +8 -0
  132. package/src/voice/room_io/room_io.ts +3 -1
  133. package/src/voice/speech_handle.ts +4 -0
@@ -1 +1 @@
1
- {"version":3,"sources":["../../../src/voice/room_io/room_io.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n ConnectionState,\n DisconnectReason,\n type NoiseCancellationOptions,\n type Participant,\n ParticipantKind,\n type RemoteParticipant,\n type Room,\n RoomEvent,\n type TextStreamInfo,\n type TextStreamReader,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport type { WritableStreamDefaultWriter } from 'node:stream/web';\nimport { ATTRIBUTE_PUBLISH_ON_BEHALF, TOPIC_CHAT } from '../../constants.js';\nimport { log } from '../../log.js';\nimport { IdentityTransform } from '../../stream/identity_transform.js';\nimport { Future, Task } from '../../utils.js';\nimport { type AgentSession } from '../agent_session.js';\nimport {\n AgentSessionEventTypes,\n type AgentStateChangedEvent,\n CloseReason,\n type UserInputTranscribedEvent,\n} from '../events.js';\nimport type { AudioOutput, TextOutput } from '../io.js';\nimport { TranscriptionSynchronizer } from '../transcription/synchronizer.js';\nimport { ParticipantAudioInputStream } from './_input.js';\nimport {\n ParalellTextOutput,\n ParticipantAudioOutput,\n ParticipantLegacyTranscriptionOutput,\n ParticipantTranscriptionOutput,\n} from './_output.js';\n\nexport interface TextInputEvent {\n text: string;\n info: TextStreamInfo;\n participant: RemoteParticipant;\n}\n\nexport type TextInputCallback = (sess: AgentSession, ev: TextInputEvent) => void | Promise<void>;\n\nconst DEFAULT_TEXT_INPUT_CALLBACK: TextInputCallback = (sess: AgentSession, ev: TextInputEvent) => {\n sess.interrupt();\n sess.generateReply({ userInput: ev.text });\n};\n\nconst DEFAULT_PARTICIPANT_KINDS: ParticipantKind[] = [\n ParticipantKind.CONNECTOR,\n ParticipantKind.SIP,\n ParticipantKind.STANDARD,\n];\n\nconst CLOSE_ON_DISCONNECT_REASONS: DisconnectReason[] = [\n DisconnectReason.CLIENT_INITIATED,\n DisconnectReason.ROOM_DELETED,\n DisconnectReason.USER_REJECTED,\n];\n\nexport interface RoomInputOptions {\n audioSampleRate: number;\n audioNumChannels: number;\n /** If not given, default to True. */\n textEnabled: boolean;\n /** If not given, default to True. */\n audioEnabled: boolean;\n /** If not given, default to False. */\n videoEnabled: boolean;\n /** The participant to link to. If not provided, link to the first participant.\n Can be overridden by the `participant` argument of RoomIO constructor or `set_participant`.\n */\n participantIdentity?: string;\n noiseCancellation?: NoiseCancellationOptions;\n textInputCallback?: TextInputCallback;\n /** Participant kinds accepted for auto subscription. If not provided,\n accept `DEFAULT_PARTICIPANT_KINDS`\n */\n participantKinds?: ParticipantKind[];\n /** Close the AgentSession if the linked participant disconnects with reasons in\n CLIENT_INITIATED, ROOM_DELETED, or USER_REJECTED.\n */\n closeOnDisconnect: boolean;\n}\n\nexport interface RoomOutputOptions {\n /** If not given, default to True. */\n transcriptionEnabled: boolean;\n /** If not given, default to True. */\n audioEnabled: boolean;\n audioSampleRate: number;\n audioNumChannels: number;\n /** False to disable transcription synchronization with audio output.\n Otherwise, transcription is emitted as quickly as available.\n */\n syncTranscription: boolean;\n /** The name of the audio track to publish. If not provided, default to \"roomio_audio\".\n */\n audioPublishOptions: TrackPublishOptions;\n}\n\nconst DEFAULT_ROOM_INPUT_OPTIONS: RoomInputOptions = {\n audioSampleRate: 24000,\n audioNumChannels: 1,\n textEnabled: true,\n audioEnabled: true,\n videoEnabled: false,\n textInputCallback: DEFAULT_TEXT_INPUT_CALLBACK,\n closeOnDisconnect: true,\n};\n\nconst DEFAULT_ROOM_OUTPUT_OPTIONS: RoomOutputOptions = {\n audioSampleRate: 24000,\n audioNumChannels: 1,\n transcriptionEnabled: true,\n audioEnabled: true,\n syncTranscription: true,\n audioPublishOptions: new TrackPublishOptions({ source: TrackSource.SOURCE_MICROPHONE }),\n};\n\nexport class RoomIO {\n private agentSession: AgentSession;\n private room: Room;\n private inputOptions: RoomInputOptions;\n private outputOptions: RoomOutputOptions;\n\n private audioInput?: ParticipantAudioInputStream;\n private participantAudioOutput?: ParticipantAudioOutput;\n private userTranscriptOutput?: ParalellTextOutput;\n private agentTranscriptOutput?: ParalellTextOutput;\n private transcriptionSynchronizer?: TranscriptionSynchronizer;\n private participantIdentity: string | null = null;\n\n private participantAvailableFuture: Future<RemoteParticipant> = new Future();\n private roomConnectedFuture: Future<void> = new Future();\n\n // Use stream API for transcript queue\n private userTranscriptStream = new IdentityTransform<UserInputTranscribedEvent>();\n private userTranscriptWriter: WritableStreamDefaultWriter<UserInputTranscribedEvent>;\n private forwardUserTranscriptTask?: Task<void>;\n private initTask?: Task<void>;\n\n private textStreamHandlerRegistered = false;\n\n private logger = log();\n\n constructor({\n agentSession,\n room,\n participant = null,\n inputOptions,\n outputOptions,\n }: {\n agentSession: AgentSession;\n room: Room;\n participant?: RemoteParticipant | string | null;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }) {\n this.agentSession = agentSession;\n this.room = room;\n this.inputOptions = { ...DEFAULT_ROOM_INPUT_OPTIONS, ...inputOptions };\n this.outputOptions = { ...DEFAULT_ROOM_OUTPUT_OPTIONS, ...outputOptions };\n\n this.userTranscriptWriter = this.userTranscriptStream.writable.getWriter();\n\n this.participantIdentity = participant\n ? typeof participant === 'string'\n ? participant\n : participant.identity\n : this.inputOptions.participantIdentity ?? null;\n }\n private async init(signal: AbortSignal): Promise<void> {\n await this.roomConnectedFuture.await;\n\n for (const participant of this.room.remoteParticipants.values()) {\n this.onParticipantConnected(participant);\n }\n if (signal.aborted) {\n return;\n }\n\n const participant = await this.participantAvailableFuture.await;\n this.setParticipant(participant.identity);\n\n // init agent outputs\n this.updateTranscriptionOutput({\n output: this.agentTranscriptOutput,\n participant: this.room.localParticipant?.identity ?? null,\n });\n\n await this.participantAudioOutput?.start(signal);\n }\n\n private onConnectionStateChanged = (state: ConnectionState) => {\n this.logger.debug({ state }, 'connection state changed');\n if (\n state === ConnectionState.CONN_CONNECTED &&\n this.room.isConnected &&\n !this.roomConnectedFuture.done\n ) {\n this.roomConnectedFuture.resolve();\n }\n };\n\n private onParticipantConnected = (participant: RemoteParticipant) => {\n if (this.participantAvailableFuture.done) {\n return;\n }\n\n if (this.participantIdentity) {\n if (participant.identity !== this.participantIdentity) {\n return;\n }\n } else if (\n // otherwise, skip participants that are marked as publishing for this agent\n participant.attributes?.[ATTRIBUTE_PUBLISH_ON_BEHALF] === this.room.localParticipant?.identity\n ) {\n return;\n }\n\n const acceptedKinds = this.inputOptions.participantKinds ?? DEFAULT_PARTICIPANT_KINDS;\n if (participant.info.kind !== undefined && !acceptedKinds.includes(participant.info.kind)) {\n return;\n }\n\n this.participantAvailableFuture.resolve(participant);\n };\n\n private onParticipantDisconnected = (participant: RemoteParticipant) => {\n if (participant.identity !== this.participantIdentity) {\n return;\n }\n this.participantAvailableFuture = new Future<RemoteParticipant>();\n if (\n this.inputOptions.closeOnDisconnect &&\n participant.disconnectReason &&\n CLOSE_ON_DISCONNECT_REASONS.includes(participant.disconnectReason)\n ) {\n this.logger.info(\n {\n participant: participant.identity,\n reason: DisconnectReason[participant.disconnectReason],\n },\n 'closing agent session due to participant disconnect ' +\n '(disable via `RoomInputOptions.closeOnDisconnect=False`)',\n );\n this.agentSession._closeSoon({\n reason: CloseReason.PARTICIPANT_DISCONNECTED,\n });\n }\n };\n\n private onUserInputTranscribed = (ev: UserInputTranscribedEvent) => {\n this.userTranscriptWriter.write(ev).catch((error) => {\n this.logger.error({ error }, 'Failed to write transcript event to stream');\n });\n };\n\n private onAgentStateChanged = async (ev: AgentStateChangedEvent) => {\n if (this.room.isConnected && this.room.localParticipant) {\n await this.room.localParticipant.setAttributes({\n [`lk.agent.state`]: ev.newState,\n });\n }\n };\n\n private onUserTextInput = (reader: TextStreamReader, participantInfo: { identity: string }) => {\n if (participantInfo.identity !== this.participantIdentity) {\n return;\n }\n\n const participant = this.room.remoteParticipants.get(participantInfo.identity);\n if (!participant) {\n this.logger.warn('participant not found, ignoring text input');\n return;\n }\n\n const readText = async () => {\n const text = await reader.readAll();\n\n const textInputResult = this.inputOptions.textInputCallback!(this.agentSession, {\n text,\n info: reader.info,\n participant,\n });\n\n // check if callback is a Promise\n if (textInputResult instanceof Promise) {\n await textInputResult;\n }\n };\n\n readText().catch((error) => {\n this.logger.error({ error }, 'Error reading text input');\n });\n };\n\n private async forwardUserTranscript(signal: AbortSignal): Promise<void> {\n const reader = this.userTranscriptStream.readable.getReader();\n try {\n while (!signal.aborted) {\n const { done, value } = await reader.read();\n if (done) break;\n\n const event = value;\n // IMPORTANT: need to await here to avoid race condition\n await this.userTranscriptOutput?.captureText(event.transcript);\n if (event.isFinal) {\n this.userTranscriptOutput?.flush();\n }\n }\n } catch (error) {\n this.logger.error({ error }, 'Error processing transcript stream');\n }\n }\n\n private createTranscriptionOutput(options: {\n isDeltaStream: boolean;\n participant: Participant | string | null;\n }) {\n return new ParalellTextOutput([\n new ParticipantLegacyTranscriptionOutput(\n this.room,\n options.isDeltaStream,\n options.participant,\n ),\n new ParticipantTranscriptionOutput(this.room, options.isDeltaStream, options.participant),\n ]);\n }\n\n private updateTranscriptionOutput({\n output,\n participant,\n }: {\n output?: ParalellTextOutput;\n participant: string | null;\n }) {\n if (!output) {\n return;\n }\n\n for (const sink of output._sinks) {\n if (\n sink instanceof ParticipantLegacyTranscriptionOutput ||\n sink instanceof ParticipantTranscriptionOutput\n ) {\n sink.setParticipant(participant);\n }\n }\n }\n\n get audioOutput(): AudioOutput | undefined {\n if (!this.transcriptionSynchronizer) {\n return this.participantAudioOutput;\n }\n\n return this.transcriptionSynchronizer.audioOutput;\n }\n\n get transcriptionOutput(): TextOutput | undefined {\n if (!this.transcriptionSynchronizer) {\n return this.agentTranscriptOutput;\n }\n\n return this.transcriptionSynchronizer.textOutput;\n }\n\n get isParticipantAvailable(): boolean {\n return this.participantAvailableFuture.done;\n }\n\n /** Switch to a different participant */\n setParticipant(participantIdentity: string | null) {\n this.logger.debug({ participantIdentity }, 'setting participant');\n if (participantIdentity === null) {\n this.unsetParticipant();\n return;\n }\n\n if (this.participantIdentity !== participantIdentity) {\n this.participantAvailableFuture = new Future<RemoteParticipant>();\n\n // check if new participant is already connected\n for (const participant of this.room.remoteParticipants.values()) {\n if (participant.identity === participantIdentity) {\n this.participantAvailableFuture.resolve(participant);\n break;\n }\n }\n }\n\n // update participant identity and handlers\n this.participantIdentity = participantIdentity;\n this.audioInput?.setParticipant(participantIdentity);\n this.updateTranscriptionOutput({\n output: this.userTranscriptOutput,\n participant: participantIdentity,\n });\n }\n\n unsetParticipant() {\n this.participantIdentity = null;\n this.participantAvailableFuture = new Future<RemoteParticipant>();\n this.audioInput?.setParticipant(null);\n this.updateTranscriptionOutput({\n output: this.userTranscriptOutput,\n participant: null,\n });\n }\n\n start() {\n if (this.inputOptions.textEnabled) {\n try {\n this.room.registerTextStreamHandler(TOPIC_CHAT, this.onUserTextInput);\n this.textStreamHandlerRegistered = true;\n } catch (error) {\n if (this.inputOptions.textEnabled) {\n this.logger.warn(`text stream handler for topic \"${TOPIC_CHAT}\" already set, ignoring`);\n }\n }\n }\n\n // -- create inputs --\n if (this.inputOptions.audioEnabled) {\n this.audioInput = new ParticipantAudioInputStream({\n room: this.room,\n sampleRate: this.inputOptions.audioSampleRate,\n numChannels: this.inputOptions.audioNumChannels,\n noiseCancellation: this.inputOptions.noiseCancellation,\n });\n }\n\n // -- create outputs --\n if (this.outputOptions.audioEnabled) {\n this.participantAudioOutput = new ParticipantAudioOutput(this.room, {\n sampleRate: this.outputOptions.audioSampleRate,\n numChannels: this.outputOptions.audioNumChannels,\n trackPublishOptions: this.outputOptions.audioPublishOptions,\n });\n }\n if (this.outputOptions.transcriptionEnabled) {\n this.userTranscriptOutput = this.createTranscriptionOutput({\n isDeltaStream: false,\n participant: this.participantIdentity,\n });\n // Start the transcript forwarding\n this.forwardUserTranscriptTask = Task.from((controller) =>\n this.forwardUserTranscript(controller.signal),\n );\n this.agentTranscriptOutput = this.createTranscriptionOutput({\n isDeltaStream: true,\n participant: null,\n });\n\n // use the RoomIO's audio output if available, otherwise use the agent's audio output\n // TODO(AJS-176): check for agent output\n const audioOutput = this.participantAudioOutput;\n if (this.outputOptions.syncTranscription && audioOutput) {\n this.transcriptionSynchronizer = new TranscriptionSynchronizer(\n audioOutput,\n this.agentTranscriptOutput,\n );\n }\n }\n\n // -- set the room event handlers --\n this.room.on(RoomEvent.ParticipantConnected, this.onParticipantConnected);\n this.room.on(RoomEvent.ConnectionStateChanged, this.onConnectionStateChanged);\n this.room.on(RoomEvent.ParticipantDisconnected, this.onParticipantDisconnected);\n if (this.room.isConnected) {\n this.onConnectionStateChanged(ConnectionState.CONN_CONNECTED);\n }\n\n this.initTask = Task.from((controller) => this.init(controller.signal));\n\n // attach the agent to the session\n if (this.audioInput) {\n this.agentSession.input.audio = this.audioInput;\n }\n if (this.audioOutput) {\n this.agentSession.output.audio = this.audioOutput;\n }\n if (this.transcriptionOutput) {\n this.agentSession.output.transcription = this.transcriptionOutput;\n }\n\n this.agentSession.on(AgentSessionEventTypes.AgentStateChanged, this.onAgentStateChanged);\n this.agentSession.on(AgentSessionEventTypes.UserInputTranscribed, this.onUserInputTranscribed);\n }\n\n async close() {\n this.room.off(RoomEvent.ParticipantConnected, this.onParticipantConnected);\n this.room.off(RoomEvent.ConnectionStateChanged, this.onConnectionStateChanged);\n this.room.off(RoomEvent.ParticipantDisconnected, this.onParticipantDisconnected);\n this.agentSession.off(AgentSessionEventTypes.UserInputTranscribed, this.onUserInputTranscribed);\n this.agentSession.off(AgentSessionEventTypes.AgentStateChanged, this.onAgentStateChanged);\n\n if (this.textStreamHandlerRegistered) {\n this.room.unregisterTextStreamHandler(TOPIC_CHAT);\n this.textStreamHandlerRegistered = false;\n }\n\n await this.initTask?.cancelAndWait();\n\n // Close stream FIRST so reader.read() in forwardUserTranscript can exit.\n // This is a workaround for a race condition in the stream API.\n this.userTranscriptWriter.close();\n await this.forwardUserTranscriptTask?.cancelAndWait();\n\n await this.audioInput?.close();\n await this.participantAudioOutput?.close();\n await this.transcriptionSynchronizer?.close();\n }\n}\n"],"mappings":"AAGA;AAAA,EACE;AAAA,EACA;AAAA,EAGA;AAAA,EAGA;AAAA,EAGA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,6BAA6B,kBAAkB;AACxD,SAAS,WAAW;AACpB,SAAS,yBAAyB;AAClC,SAAS,QAAQ,YAAY;AAC7B,eAAkC;AAClC;AAAA,EACE;AAAA,EAEA;AAAA,OAEK;AAEP,SAAS,iCAAiC;AAC1C,SAAS,mCAAmC;AAC5C;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAUP,MAAM,8BAAiD,CAAC,MAAoB,OAAuB;AACjG,OAAK,UAAU;AACf,OAAK,cAAc,EAAE,WAAW,GAAG,KAAK,CAAC;AAC3C;AAEA,MAAM,4BAA+C;AAAA,EACnD,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,gBAAgB;AAClB;AAEA,MAAM,8BAAkD;AAAA,EACtD,iBAAiB;AAAA,EACjB,iBAAiB;AAAA,EACjB,iBAAiB;AACnB;AA2CA,MAAM,6BAA+C;AAAA,EACnD,iBAAiB;AAAA,EACjB,kBAAkB;AAAA,EAClB,aAAa;AAAA,EACb,cAAc;AAAA,EACd,cAAc;AAAA,EACd,mBAAmB;AAAA,EACnB,mBAAmB;AACrB;AAEA,MAAM,8BAAiD;AAAA,EACrD,iBAAiB;AAAA,EACjB,kBAAkB;AAAA,EAClB,sBAAsB;AAAA,EACtB,cAAc;AAAA,EACd,mBAAmB;AAAA,EACnB,qBAAqB,IAAI,oBAAoB,EAAE,QAAQ,YAAY,kBAAkB,CAAC;AACxF;AAEO,MAAM,OAAO;AAAA,EACV;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,sBAAqC;AAAA,EAErC,6BAAwD,IAAI,OAAO;AAAA,EACnE,sBAAoC,IAAI,OAAO;AAAA;AAAA,EAG/C,uBAAuB,IAAI,kBAA6C;AAAA,EACxE;AAAA,EACA;AAAA,EACA;AAAA,EAEA,8BAA8B;AAAA,EAE9B,SAAS,IAAI;AAAA,EAErB,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA,cAAc;AAAA,IACd;AAAA,IACA;AAAA,EACF,GAMG;AACD,SAAK,eAAe;AACpB,SAAK,OAAO;AACZ,SAAK,eAAe,EAAE,GAAG,4BAA4B,GAAG,aAAa;AACrE,SAAK,gBAAgB,EAAE,GAAG,6BAA6B,GAAG,cAAc;AAExE,SAAK,uBAAuB,KAAK,qBAAqB,SAAS,UAAU;AAEzE,SAAK,sBAAsB,cACvB,OAAO,gBAAgB,WACrB,cACA,YAAY,WACd,KAAK,aAAa,uBAAuB;AAAA,EAC/C;AAAA,EACA,MAAc,KAAK,QAAoC;AAhLzD;AAiLI,UAAM,KAAK,oBAAoB;AAE/B,eAAWA,gBAAe,KAAK,KAAK,mBAAmB,OAAO,GAAG;AAC/D,WAAK,uBAAuBA,YAAW;AAAA,IACzC;AACA,QAAI,OAAO,SAAS;AAClB;AAAA,IACF;AAEA,UAAM,cAAc,MAAM,KAAK,2BAA2B;AAC1D,SAAK,eAAe,YAAY,QAAQ;AAGxC,SAAK,0BAA0B;AAAA,MAC7B,QAAQ,KAAK;AAAA,MACb,eAAa,UAAK,KAAK,qBAAV,mBAA4B,aAAY;AAAA,IACvD,CAAC;AAED,YAAM,UAAK,2BAAL,mBAA6B,MAAM;AAAA,EAC3C;AAAA,EAEQ,2BAA2B,CAAC,UAA2B;AAC7D,SAAK,OAAO,MAAM,EAAE,MAAM,GAAG,0BAA0B;AACvD,QACE,UAAU,gBAAgB,kBAC1B,KAAK,KAAK,eACV,CAAC,KAAK,oBAAoB,MAC1B;AACA,WAAK,oBAAoB,QAAQ;AAAA,IACnC;AAAA,EACF;AAAA,EAEQ,yBAAyB,CAAC,gBAAmC;AAjNvE;AAkNI,QAAI,KAAK,2BAA2B,MAAM;AACxC;AAAA,IACF;AAEA,QAAI,KAAK,qBAAqB;AAC5B,UAAI,YAAY,aAAa,KAAK,qBAAqB;AACrD;AAAA,MACF;AAAA,IACF;AAAA;AAAA,QAEE,iBAAY,eAAZ,mBAAyB,oCAAiC,UAAK,KAAK,qBAAV,mBAA4B;AAAA,MACtF;AACA;AAAA,IACF;AAEA,UAAM,gBAAgB,KAAK,aAAa,oBAAoB;AAC5D,QAAI,YAAY,KAAK,SAAS,UAAa,CAAC,cAAc,SAAS,YAAY,KAAK,IAAI,GAAG;AACzF;AAAA,IACF;AAEA,SAAK,2BAA2B,QAAQ,WAAW;AAAA,EACrD;AAAA,EAEQ,4BAA4B,CAAC,gBAAmC;AACtE,QAAI,YAAY,aAAa,KAAK,qBAAqB;AACrD;AAAA,IACF;AACA,SAAK,6BAA6B,IAAI,OAA0B;AAChE,QACE,KAAK,aAAa,qBAClB,YAAY,oBACZ,4BAA4B,SAAS,YAAY,gBAAgB,GACjE;AACA,WAAK,OAAO;AAAA,QACV;AAAA,UACE,aAAa,YAAY;AAAA,UACzB,QAAQ,iBAAiB,YAAY,gBAAgB;AAAA,QACvD;AAAA,QACA;AAAA,MAEF;AACA,WAAK,aAAa,WAAW;AAAA,QAC3B,QAAQ,YAAY;AAAA,MACtB,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEQ,yBAAyB,CAAC,OAAkC;AAClE,SAAK,qBAAqB,MAAM,EAAE,EAAE,MAAM,CAAC,UAAU;AACnD,WAAK,OAAO,MAAM,EAAE,MAAM,GAAG,4CAA4C;AAAA,IAC3E,CAAC;AAAA,EACH;AAAA,EAEQ,sBAAsB,OAAO,OAA+B;AAClE,QAAI,KAAK,KAAK,eAAe,KAAK,KAAK,kBAAkB;AACvD,YAAM,KAAK,KAAK,iBAAiB,cAAc;AAAA,QAC7C,CAAC,gBAAgB,GAAG,GAAG;AAAA,MACzB,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEQ,kBAAkB,CAAC,QAA0B,oBAA0C;AAC7F,QAAI,gBAAgB,aAAa,KAAK,qBAAqB;AACzD;AAAA,IACF;AAEA,UAAM,cAAc,KAAK,KAAK,mBAAmB,IAAI,gBAAgB,QAAQ;AAC7E,QAAI,CAAC,aAAa;AAChB,WAAK,OAAO,KAAK,4CAA4C;AAC7D;AAAA,IACF;AAEA,UAAM,WAAW,YAAY;AAC3B,YAAM,OAAO,MAAM,OAAO,QAAQ;AAElC,YAAM,kBAAkB,KAAK,aAAa,kBAAmB,KAAK,cAAc;AAAA,QAC9E;AAAA,QACA,MAAM,OAAO;AAAA,QACb;AAAA,MACF,CAAC;AAGD,UAAI,2BAA2B,SAAS;AACtC,cAAM;AAAA,MACR;AAAA,IACF;AAEA,aAAS,EAAE,MAAM,CAAC,UAAU;AAC1B,WAAK,OAAO,MAAM,EAAE,MAAM,GAAG,0BAA0B;AAAA,IACzD,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,sBAAsB,QAAoC;AA9S1E;AA+SI,UAAM,SAAS,KAAK,qBAAqB,SAAS,UAAU;AAC5D,QAAI;AACF,aAAO,CAAC,OAAO,SAAS;AACtB,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AAEV,cAAM,QAAQ;AAEd,gBAAM,UAAK,yBAAL,mBAA2B,YAAY,MAAM;AACnD,YAAI,MAAM,SAAS;AACjB,qBAAK,yBAAL,mBAA2B;AAAA,QAC7B;AAAA,MACF;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,EAAE,MAAM,GAAG,oCAAoC;AAAA,IACnE;AAAA,EACF;AAAA,EAEQ,0BAA0B,SAG/B;AACD,WAAO,IAAI,mBAAmB;AAAA,MAC5B,IAAI;AAAA,QACF,KAAK;AAAA,QACL,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA,IAAI,+BAA+B,KAAK,MAAM,QAAQ,eAAe,QAAQ,WAAW;AAAA,IAC1F,CAAC;AAAA,EACH;AAAA,EAEQ,0BAA0B;AAAA,IAChC;AAAA,IACA;AAAA,EACF,GAGG;AACD,QAAI,CAAC,QAAQ;AACX;AAAA,IACF;AAEA,eAAW,QAAQ,OAAO,QAAQ;AAChC,UACE,gBAAgB,wCAChB,gBAAgB,gCAChB;AACA,aAAK,eAAe,WAAW;AAAA,MACjC;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,cAAuC;AACzC,QAAI,CAAC,KAAK,2BAA2B;AACnC,aAAO,KAAK;AAAA,IACd;AAEA,WAAO,KAAK,0BAA0B;AAAA,EACxC;AAAA,EAEA,IAAI,sBAA8C;AAChD,QAAI,CAAC,KAAK,2BAA2B;AACnC,aAAO,KAAK;AAAA,IACd;AAEA,WAAO,KAAK,0BAA0B;AAAA,EACxC;AAAA,EAEA,IAAI,yBAAkC;AACpC,WAAO,KAAK,2BAA2B;AAAA,EACzC;AAAA;AAAA,EAGA,eAAe,qBAAoC;AAzXrD;AA0XI,SAAK,OAAO,MAAM,EAAE,oBAAoB,GAAG,qBAAqB;AAChE,QAAI,wBAAwB,MAAM;AAChC,WAAK,iBAAiB;AACtB;AAAA,IACF;AAEA,QAAI,KAAK,wBAAwB,qBAAqB;AACpD,WAAK,6BAA6B,IAAI,OAA0B;AAGhE,iBAAW,eAAe,KAAK,KAAK,mBAAmB,OAAO,GAAG;AAC/D,YAAI,YAAY,aAAa,qBAAqB;AAChD,eAAK,2BAA2B,QAAQ,WAAW;AACnD;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAGA,SAAK,sBAAsB;AAC3B,eAAK,eAAL,mBAAiB,eAAe;AAChC,SAAK,0BAA0B;AAAA,MAC7B,QAAQ,KAAK;AAAA,MACb,aAAa;AAAA,IACf,CAAC;AAAA,EACH;AAAA,EAEA,mBAAmB;AArZrB;AAsZI,SAAK,sBAAsB;AAC3B,SAAK,6BAA6B,IAAI,OAA0B;AAChE,eAAK,eAAL,mBAAiB,eAAe;AAChC,SAAK,0BAA0B;AAAA,MAC7B,QAAQ,KAAK;AAAA,MACb,aAAa;AAAA,IACf,CAAC;AAAA,EACH;AAAA,EAEA,QAAQ;AACN,QAAI,KAAK,aAAa,aAAa;AACjC,UAAI;AACF,aAAK,KAAK,0BAA0B,YAAY,KAAK,eAAe;AACpE,aAAK,8BAA8B;AAAA,MACrC,SAAS,OAAO;AACd,YAAI,KAAK,aAAa,aAAa;AACjC,eAAK,OAAO,KAAK,kCAAkC,UAAU,yBAAyB;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAGA,QAAI,KAAK,aAAa,cAAc;AAClC,WAAK,aAAa,IAAI,4BAA4B;AAAA,QAChD,MAAM,KAAK;AAAA,QACX,YAAY,KAAK,aAAa;AAAA,QAC9B,aAAa,KAAK,aAAa;AAAA,QAC/B,mBAAmB,KAAK,aAAa;AAAA,MACvC,CAAC;AAAA,IACH;AAGA,QAAI,KAAK,cAAc,cAAc;AACnC,WAAK,yBAAyB,IAAI,uBAAuB,KAAK,MAAM;AAAA,QAClE,YAAY,KAAK,cAAc;AAAA,QAC/B,aAAa,KAAK,cAAc;AAAA,QAChC,qBAAqB,KAAK,cAAc;AAAA,MAC1C,CAAC;AAAA,IACH;AACA,QAAI,KAAK,cAAc,sBAAsB;AAC3C,WAAK,uBAAuB,KAAK,0BAA0B;AAAA,QACzD,eAAe;AAAA,QACf,aAAa,KAAK;AAAA,MACpB,CAAC;AAED,WAAK,4BAA4B,KAAK;AAAA,QAAK,CAAC,eAC1C,KAAK,sBAAsB,WAAW,MAAM;AAAA,MAC9C;AACA,WAAK,wBAAwB,KAAK,0BAA0B;AAAA,QAC1D,eAAe;AAAA,QACf,aAAa;AAAA,MACf,CAAC;AAID,YAAM,cAAc,KAAK;AACzB,UAAI,KAAK,cAAc,qBAAqB,aAAa;AACvD,aAAK,4BAA4B,IAAI;AAAA,UACnC;AAAA,UACA,KAAK;AAAA,QACP;AAAA,MACF;AAAA,IACF;AAGA,SAAK,KAAK,GAAG,UAAU,sBAAsB,KAAK,sBAAsB;AACxE,SAAK,KAAK,GAAG,UAAU,wBAAwB,KAAK,wBAAwB;AAC5E,SAAK,KAAK,GAAG,UAAU,yBAAyB,KAAK,yBAAyB;AAC9E,QAAI,KAAK,KAAK,aAAa;AACzB,WAAK,yBAAyB,gBAAgB,cAAc;AAAA,IAC9D;AAEA,SAAK,WAAW,KAAK,KAAK,CAAC,eAAe,KAAK,KAAK,WAAW,MAAM,CAAC;AAGtE,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,MAAM,QAAQ,KAAK;AAAA,IACvC;AACA,QAAI,KAAK,aAAa;AACpB,WAAK,aAAa,OAAO,QAAQ,KAAK;AAAA,IACxC;AACA,QAAI,KAAK,qBAAqB;AAC5B,WAAK,aAAa,OAAO,gBAAgB,KAAK;AAAA,IAChD;AAEA,SAAK,aAAa,GAAG,uBAAuB,mBAAmB,KAAK,mBAAmB;AACvF,SAAK,aAAa,GAAG,uBAAuB,sBAAsB,KAAK,sBAAsB;AAAA,EAC/F;AAAA,EAEA,MAAM,QAAQ;AA/ehB;AAgfI,SAAK,KAAK,IAAI,UAAU,sBAAsB,KAAK,sBAAsB;AACzE,SAAK,KAAK,IAAI,UAAU,wBAAwB,KAAK,wBAAwB;AAC7E,SAAK,KAAK,IAAI,UAAU,yBAAyB,KAAK,yBAAyB;AAC/E,SAAK,aAAa,IAAI,uBAAuB,sBAAsB,KAAK,sBAAsB;AAC9F,SAAK,aAAa,IAAI,uBAAuB,mBAAmB,KAAK,mBAAmB;AAExF,QAAI,KAAK,6BAA6B;AACpC,WAAK,KAAK,4BAA4B,UAAU;AAChD,WAAK,8BAA8B;AAAA,IACrC;AAEA,YAAM,UAAK,aAAL,mBAAe;AAIrB,SAAK,qBAAqB,MAAM;AAChC,YAAM,UAAK,8BAAL,mBAAgC;AAEtC,YAAM,UAAK,eAAL,mBAAiB;AACvB,YAAM,UAAK,2BAAL,mBAA6B;AACnC,YAAM,UAAK,8BAAL,mBAAgC;AAAA,EACxC;AACF;","names":["participant"]}
1
+ {"version":3,"sources":["../../../src/voice/room_io/room_io.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type AudioFrame,\n ConnectionState,\n DisconnectReason,\n type FrameProcessor,\n type NoiseCancellationOptions,\n type Participant,\n ParticipantKind,\n type RemoteParticipant,\n type Room,\n RoomEvent,\n type TextStreamInfo,\n type TextStreamReader,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport type { WritableStreamDefaultWriter } from 'node:stream/web';\nimport { ATTRIBUTE_PUBLISH_ON_BEHALF, TOPIC_CHAT } from '../../constants.js';\nimport { log } from '../../log.js';\nimport { IdentityTransform } from '../../stream/identity_transform.js';\nimport { Future, Task } from '../../utils.js';\nimport { type AgentSession } from '../agent_session.js';\nimport {\n AgentSessionEventTypes,\n type AgentStateChangedEvent,\n CloseReason,\n type UserInputTranscribedEvent,\n} from '../events.js';\nimport type { AudioOutput, TextOutput } from '../io.js';\nimport { TranscriptionSynchronizer } from '../transcription/synchronizer.js';\nimport { ParticipantAudioInputStream } from './_input.js';\nimport {\n ParalellTextOutput,\n ParticipantAudioOutput,\n ParticipantLegacyTranscriptionOutput,\n ParticipantTranscriptionOutput,\n} from './_output.js';\n\nexport interface TextInputEvent {\n text: string;\n info: TextStreamInfo;\n participant: RemoteParticipant;\n}\n\nexport type TextInputCallback = (sess: AgentSession, ev: TextInputEvent) => void | Promise<void>;\n\nconst DEFAULT_TEXT_INPUT_CALLBACK: TextInputCallback = (sess: AgentSession, ev: TextInputEvent) => {\n sess.interrupt();\n sess.generateReply({ userInput: ev.text });\n};\n\nconst DEFAULT_PARTICIPANT_KINDS: ParticipantKind[] = [\n ParticipantKind.CONNECTOR,\n ParticipantKind.SIP,\n ParticipantKind.STANDARD,\n];\n\nconst CLOSE_ON_DISCONNECT_REASONS: DisconnectReason[] = [\n DisconnectReason.CLIENT_INITIATED,\n DisconnectReason.ROOM_DELETED,\n DisconnectReason.USER_REJECTED,\n];\n\nexport interface RoomInputOptions {\n audioSampleRate: number;\n audioNumChannels: number;\n /** If not given, default to True. */\n textEnabled: boolean;\n /** If not given, default to True. */\n audioEnabled: boolean;\n /** If not given, default to False. */\n videoEnabled: boolean;\n /** The participant to link to. If not provided, link to the first participant.\n Can be overridden by the `participant` argument of RoomIO constructor or `set_participant`.\n */\n participantIdentity?: string;\n noiseCancellation?: NoiseCancellationOptions | FrameProcessor<AudioFrame>;\n textInputCallback?: TextInputCallback;\n /** Participant kinds accepted for auto subscription. If not provided,\n accept `DEFAULT_PARTICIPANT_KINDS`\n */\n participantKinds?: ParticipantKind[];\n /** Close the AgentSession if the linked participant disconnects with reasons in\n CLIENT_INITIATED, ROOM_DELETED, or USER_REJECTED.\n */\n closeOnDisconnect: boolean;\n}\n\nexport interface RoomOutputOptions {\n /** If not given, default to True. */\n transcriptionEnabled: boolean;\n /** If not given, default to True. */\n audioEnabled: boolean;\n audioSampleRate: number;\n audioNumChannels: number;\n /** False to disable transcription synchronization with audio output.\n Otherwise, transcription is emitted as quickly as available.\n */\n syncTranscription: boolean;\n /** The name of the audio track to publish. If not provided, default to \"roomio_audio\".\n */\n audioPublishOptions: TrackPublishOptions;\n}\n\nconst DEFAULT_ROOM_INPUT_OPTIONS: RoomInputOptions = {\n audioSampleRate: 24000,\n audioNumChannels: 1,\n textEnabled: true,\n audioEnabled: true,\n videoEnabled: false,\n textInputCallback: DEFAULT_TEXT_INPUT_CALLBACK,\n closeOnDisconnect: true,\n};\n\nconst DEFAULT_ROOM_OUTPUT_OPTIONS: RoomOutputOptions = {\n audioSampleRate: 24000,\n audioNumChannels: 1,\n transcriptionEnabled: true,\n audioEnabled: true,\n syncTranscription: true,\n audioPublishOptions: new TrackPublishOptions({ source: TrackSource.SOURCE_MICROPHONE }),\n};\n\nexport class RoomIO {\n private agentSession: AgentSession;\n private room: Room;\n private inputOptions: RoomInputOptions;\n private outputOptions: RoomOutputOptions;\n\n private audioInput?: ParticipantAudioInputStream;\n private participantAudioOutput?: ParticipantAudioOutput;\n private userTranscriptOutput?: ParalellTextOutput;\n private agentTranscriptOutput?: ParalellTextOutput;\n private transcriptionSynchronizer?: TranscriptionSynchronizer;\n private participantIdentity: string | null = null;\n\n private participantAvailableFuture: Future<RemoteParticipant> = new Future();\n private roomConnectedFuture: Future<void> = new Future();\n\n // Use stream API for transcript queue\n private userTranscriptStream = new IdentityTransform<UserInputTranscribedEvent>();\n private userTranscriptWriter: WritableStreamDefaultWriter<UserInputTranscribedEvent>;\n private forwardUserTranscriptTask?: Task<void>;\n private initTask?: Task<void>;\n\n private textStreamHandlerRegistered = false;\n\n private logger = log();\n\n constructor({\n agentSession,\n room,\n participant = null,\n inputOptions,\n outputOptions,\n }: {\n agentSession: AgentSession;\n room: Room;\n participant?: RemoteParticipant | string | null;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }) {\n this.agentSession = agentSession;\n this.room = room;\n this.inputOptions = { ...DEFAULT_ROOM_INPUT_OPTIONS, ...inputOptions };\n this.outputOptions = { ...DEFAULT_ROOM_OUTPUT_OPTIONS, ...outputOptions };\n\n this.userTranscriptWriter = this.userTranscriptStream.writable.getWriter();\n\n this.participantIdentity = participant\n ? typeof participant === 'string'\n ? participant\n : participant.identity\n : this.inputOptions.participantIdentity ?? null;\n }\n private async init(signal: AbortSignal): Promise<void> {\n await this.roomConnectedFuture.await;\n\n for (const participant of this.room.remoteParticipants.values()) {\n this.onParticipantConnected(participant);\n }\n if (signal.aborted) {\n return;\n }\n\n const participant = await this.participantAvailableFuture.await;\n this.setParticipant(participant.identity);\n\n // init agent outputs\n this.updateTranscriptionOutput({\n output: this.agentTranscriptOutput,\n participant: this.room.localParticipant?.identity ?? null,\n });\n\n await this.participantAudioOutput?.start(signal);\n }\n\n private onConnectionStateChanged = (state: ConnectionState) => {\n this.logger.debug({ state }, 'connection state changed');\n if (\n state === ConnectionState.CONN_CONNECTED &&\n this.room.isConnected &&\n !this.roomConnectedFuture.done\n ) {\n this.roomConnectedFuture.resolve();\n }\n };\n\n private onParticipantConnected = (participant: RemoteParticipant) => {\n if (this.participantAvailableFuture.done) {\n return;\n }\n\n if (this.participantIdentity) {\n if (participant.identity !== this.participantIdentity) {\n return;\n }\n } else if (\n // otherwise, skip participants that are marked as publishing for this agent\n participant.attributes?.[ATTRIBUTE_PUBLISH_ON_BEHALF] === this.room.localParticipant?.identity\n ) {\n return;\n }\n\n const acceptedKinds = this.inputOptions.participantKinds ?? DEFAULT_PARTICIPANT_KINDS;\n if (participant.info.kind !== undefined && !acceptedKinds.includes(participant.info.kind)) {\n return;\n }\n\n this.participantAvailableFuture.resolve(participant);\n };\n\n private onParticipantDisconnected = (participant: RemoteParticipant) => {\n if (participant.identity !== this.participantIdentity) {\n return;\n }\n this.participantAvailableFuture = new Future<RemoteParticipant>();\n if (\n this.inputOptions.closeOnDisconnect &&\n participant.disconnectReason &&\n CLOSE_ON_DISCONNECT_REASONS.includes(participant.disconnectReason)\n ) {\n this.logger.info(\n {\n participant: participant.identity,\n reason: DisconnectReason[participant.disconnectReason],\n },\n 'closing agent session due to participant disconnect ' +\n '(disable via `RoomInputOptions.closeOnDisconnect=False`)',\n );\n this.agentSession._closeSoon({\n reason: CloseReason.PARTICIPANT_DISCONNECTED,\n });\n }\n };\n\n private onUserInputTranscribed = (ev: UserInputTranscribedEvent) => {\n this.userTranscriptWriter.write(ev).catch((error) => {\n this.logger.error({ error }, 'Failed to write transcript event to stream');\n });\n };\n\n private onAgentStateChanged = async (ev: AgentStateChangedEvent) => {\n if (this.room.isConnected && this.room.localParticipant) {\n await this.room.localParticipant.setAttributes({\n [`lk.agent.state`]: ev.newState,\n });\n }\n };\n\n private onUserTextInput = (reader: TextStreamReader, participantInfo: { identity: string }) => {\n if (participantInfo.identity !== this.participantIdentity) {\n return;\n }\n\n const participant = this.room.remoteParticipants.get(participantInfo.identity);\n if (!participant) {\n this.logger.warn('participant not found, ignoring text input');\n return;\n }\n\n const readText = async () => {\n const text = await reader.readAll();\n\n const textInputResult = this.inputOptions.textInputCallback!(this.agentSession, {\n text,\n info: reader.info,\n participant,\n });\n\n // check if callback is a Promise\n if (textInputResult instanceof Promise) {\n await textInputResult;\n }\n };\n\n readText().catch((error) => {\n this.logger.error({ error }, 'Error reading text input');\n });\n };\n\n private async forwardUserTranscript(signal: AbortSignal): Promise<void> {\n const reader = this.userTranscriptStream.readable.getReader();\n try {\n while (!signal.aborted) {\n const { done, value } = await reader.read();\n if (done) break;\n\n const event = value;\n // IMPORTANT: need to await here to avoid race condition\n await this.userTranscriptOutput?.captureText(event.transcript);\n if (event.isFinal) {\n this.userTranscriptOutput?.flush();\n }\n }\n } catch (error) {\n this.logger.error({ error }, 'Error processing transcript stream');\n }\n }\n\n private createTranscriptionOutput(options: {\n isDeltaStream: boolean;\n participant: Participant | string | null;\n }) {\n return new ParalellTextOutput([\n new ParticipantLegacyTranscriptionOutput(\n this.room,\n options.isDeltaStream,\n options.participant,\n ),\n new ParticipantTranscriptionOutput(this.room, options.isDeltaStream, options.participant),\n ]);\n }\n\n private updateTranscriptionOutput({\n output,\n participant,\n }: {\n output?: ParalellTextOutput;\n participant: string | null;\n }) {\n if (!output) {\n return;\n }\n\n for (const sink of output._sinks) {\n if (\n sink instanceof ParticipantLegacyTranscriptionOutput ||\n sink instanceof ParticipantTranscriptionOutput\n ) {\n sink.setParticipant(participant);\n }\n }\n }\n\n get audioOutput(): AudioOutput | undefined {\n if (!this.transcriptionSynchronizer) {\n return this.participantAudioOutput;\n }\n\n return this.transcriptionSynchronizer.audioOutput;\n }\n\n get transcriptionOutput(): TextOutput | undefined {\n if (!this.transcriptionSynchronizer) {\n return this.agentTranscriptOutput;\n }\n\n return this.transcriptionSynchronizer.textOutput;\n }\n\n get isParticipantAvailable(): boolean {\n return this.participantAvailableFuture.done;\n }\n\n /** Switch to a different participant */\n setParticipant(participantIdentity: string | null) {\n this.logger.debug({ participantIdentity }, 'setting participant');\n if (participantIdentity === null) {\n this.unsetParticipant();\n return;\n }\n\n if (this.participantIdentity !== participantIdentity) {\n this.participantAvailableFuture = new Future<RemoteParticipant>();\n\n // check if new participant is already connected\n for (const participant of this.room.remoteParticipants.values()) {\n if (participant.identity === participantIdentity) {\n this.participantAvailableFuture.resolve(participant);\n break;\n }\n }\n }\n\n // update participant identity and handlers\n this.participantIdentity = participantIdentity;\n this.audioInput?.setParticipant(participantIdentity);\n this.updateTranscriptionOutput({\n output: this.userTranscriptOutput,\n participant: participantIdentity,\n });\n }\n\n unsetParticipant() {\n this.participantIdentity = null;\n this.participantAvailableFuture = new Future<RemoteParticipant>();\n this.audioInput?.setParticipant(null);\n this.updateTranscriptionOutput({\n output: this.userTranscriptOutput,\n participant: null,\n });\n }\n\n start() {\n if (this.inputOptions.textEnabled) {\n try {\n this.room.registerTextStreamHandler(TOPIC_CHAT, this.onUserTextInput);\n this.textStreamHandlerRegistered = true;\n } catch (error) {\n if (this.inputOptions.textEnabled) {\n this.logger.warn(`text stream handler for topic \"${TOPIC_CHAT}\" already set, ignoring`);\n }\n }\n }\n\n // -- create inputs --\n if (this.inputOptions.audioEnabled) {\n this.audioInput = new ParticipantAudioInputStream({\n room: this.room,\n sampleRate: this.inputOptions.audioSampleRate,\n numChannels: this.inputOptions.audioNumChannels,\n noiseCancellation: this.inputOptions.noiseCancellation,\n });\n }\n\n // -- create outputs --\n if (this.outputOptions.audioEnabled) {\n this.participantAudioOutput = new ParticipantAudioOutput(this.room, {\n sampleRate: this.outputOptions.audioSampleRate,\n numChannels: this.outputOptions.audioNumChannels,\n trackPublishOptions: this.outputOptions.audioPublishOptions,\n });\n }\n if (this.outputOptions.transcriptionEnabled) {\n this.userTranscriptOutput = this.createTranscriptionOutput({\n isDeltaStream: false,\n participant: this.participantIdentity,\n });\n // Start the transcript forwarding\n this.forwardUserTranscriptTask = Task.from((controller) =>\n this.forwardUserTranscript(controller.signal),\n );\n this.agentTranscriptOutput = this.createTranscriptionOutput({\n isDeltaStream: true,\n participant: null,\n });\n\n // use the RoomIO's audio output if available, otherwise use the agent's audio output\n // TODO(AJS-176): check for agent output\n const audioOutput = this.participantAudioOutput;\n if (this.outputOptions.syncTranscription && audioOutput) {\n this.transcriptionSynchronizer = new TranscriptionSynchronizer(\n audioOutput,\n this.agentTranscriptOutput,\n );\n }\n }\n\n // -- set the room event handlers --\n this.room.on(RoomEvent.ParticipantConnected, this.onParticipantConnected);\n this.room.on(RoomEvent.ConnectionStateChanged, this.onConnectionStateChanged);\n this.room.on(RoomEvent.ParticipantDisconnected, this.onParticipantDisconnected);\n if (this.room.isConnected) {\n this.onConnectionStateChanged(ConnectionState.CONN_CONNECTED);\n }\n\n this.initTask = Task.from((controller) => this.init(controller.signal));\n\n // attach the agent to the session\n if (this.audioInput) {\n this.agentSession.input.audio = this.audioInput;\n }\n if (this.audioOutput) {\n this.agentSession.output.audio = this.audioOutput;\n }\n if (this.transcriptionOutput) {\n this.agentSession.output.transcription = this.transcriptionOutput;\n }\n\n this.agentSession.on(AgentSessionEventTypes.AgentStateChanged, this.onAgentStateChanged);\n this.agentSession.on(AgentSessionEventTypes.UserInputTranscribed, this.onUserInputTranscribed);\n }\n\n async close() {\n this.room.off(RoomEvent.ParticipantConnected, this.onParticipantConnected);\n this.room.off(RoomEvent.ConnectionStateChanged, this.onConnectionStateChanged);\n this.room.off(RoomEvent.ParticipantDisconnected, this.onParticipantDisconnected);\n this.agentSession.off(AgentSessionEventTypes.UserInputTranscribed, this.onUserInputTranscribed);\n this.agentSession.off(AgentSessionEventTypes.AgentStateChanged, this.onAgentStateChanged);\n\n if (this.textStreamHandlerRegistered) {\n this.room.unregisterTextStreamHandler(TOPIC_CHAT);\n this.textStreamHandlerRegistered = false;\n }\n\n await this.initTask?.cancelAndWait();\n\n // Close stream FIRST so reader.read() in forwardUserTranscript can exit.\n // This is a workaround for a race condition in the stream API.\n this.userTranscriptWriter.close();\n await this.forwardUserTranscriptTask?.cancelAndWait();\n\n await this.audioInput?.close();\n await this.participantAudioOutput?.close();\n await this.transcriptionSynchronizer?.close();\n }\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EAIA;AAAA,EAGA;AAAA,EAGA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,6BAA6B,kBAAkB;AACxD,SAAS,WAAW;AACpB,SAAS,yBAAyB;AAClC,SAAS,QAAQ,YAAY;AAC7B,eAAkC;AAClC;AAAA,EACE;AAAA,EAEA;AAAA,OAEK;AAEP,SAAS,iCAAiC;AAC1C,SAAS,mCAAmC;AAC5C;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAUP,MAAM,8BAAiD,CAAC,MAAoB,OAAuB;AACjG,OAAK,UAAU;AACf,OAAK,cAAc,EAAE,WAAW,GAAG,KAAK,CAAC;AAC3C;AAEA,MAAM,4BAA+C;AAAA,EACnD,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,gBAAgB;AAClB;AAEA,MAAM,8BAAkD;AAAA,EACtD,iBAAiB;AAAA,EACjB,iBAAiB;AAAA,EACjB,iBAAiB;AACnB;AA2CA,MAAM,6BAA+C;AAAA,EACnD,iBAAiB;AAAA,EACjB,kBAAkB;AAAA,EAClB,aAAa;AAAA,EACb,cAAc;AAAA,EACd,cAAc;AAAA,EACd,mBAAmB;AAAA,EACnB,mBAAmB;AACrB;AAEA,MAAM,8BAAiD;AAAA,EACrD,iBAAiB;AAAA,EACjB,kBAAkB;AAAA,EAClB,sBAAsB;AAAA,EACtB,cAAc;AAAA,EACd,mBAAmB;AAAA,EACnB,qBAAqB,IAAI,oBAAoB,EAAE,QAAQ,YAAY,kBAAkB,CAAC;AACxF;AAEO,MAAM,OAAO;AAAA,EACV;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,sBAAqC;AAAA,EAErC,6BAAwD,IAAI,OAAO;AAAA,EACnE,sBAAoC,IAAI,OAAO;AAAA;AAAA,EAG/C,uBAAuB,IAAI,kBAA6C;AAAA,EACxE;AAAA,EACA;AAAA,EACA;AAAA,EAEA,8BAA8B;AAAA,EAE9B,SAAS,IAAI;AAAA,EAErB,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA,cAAc;AAAA,IACd;AAAA,IACA;AAAA,EACF,GAMG;AACD,SAAK,eAAe;AACpB,SAAK,OAAO;AACZ,SAAK,eAAe,EAAE,GAAG,4BAA4B,GAAG,aAAa;AACrE,SAAK,gBAAgB,EAAE,GAAG,6BAA6B,GAAG,cAAc;AAExE,SAAK,uBAAuB,KAAK,qBAAqB,SAAS,UAAU;AAEzE,SAAK,sBAAsB,cACvB,OAAO,gBAAgB,WACrB,cACA,YAAY,WACd,KAAK,aAAa,uBAAuB;AAAA,EAC/C;AAAA,EACA,MAAc,KAAK,QAAoC;AAlLzD;AAmLI,UAAM,KAAK,oBAAoB;AAE/B,eAAWA,gBAAe,KAAK,KAAK,mBAAmB,OAAO,GAAG;AAC/D,WAAK,uBAAuBA,YAAW;AAAA,IACzC;AACA,QAAI,OAAO,SAAS;AAClB;AAAA,IACF;AAEA,UAAM,cAAc,MAAM,KAAK,2BAA2B;AAC1D,SAAK,eAAe,YAAY,QAAQ;AAGxC,SAAK,0BAA0B;AAAA,MAC7B,QAAQ,KAAK;AAAA,MACb,eAAa,UAAK,KAAK,qBAAV,mBAA4B,aAAY;AAAA,IACvD,CAAC;AAED,YAAM,UAAK,2BAAL,mBAA6B,MAAM;AAAA,EAC3C;AAAA,EAEQ,2BAA2B,CAAC,UAA2B;AAC7D,SAAK,OAAO,MAAM,EAAE,MAAM,GAAG,0BAA0B;AACvD,QACE,UAAU,gBAAgB,kBAC1B,KAAK,KAAK,eACV,CAAC,KAAK,oBAAoB,MAC1B;AACA,WAAK,oBAAoB,QAAQ;AAAA,IACnC;AAAA,EACF;AAAA,EAEQ,yBAAyB,CAAC,gBAAmC;AAnNvE;AAoNI,QAAI,KAAK,2BAA2B,MAAM;AACxC;AAAA,IACF;AAEA,QAAI,KAAK,qBAAqB;AAC5B,UAAI,YAAY,aAAa,KAAK,qBAAqB;AACrD;AAAA,MACF;AAAA,IACF;AAAA;AAAA,QAEE,iBAAY,eAAZ,mBAAyB,oCAAiC,UAAK,KAAK,qBAAV,mBAA4B;AAAA,MACtF;AACA;AAAA,IACF;AAEA,UAAM,gBAAgB,KAAK,aAAa,oBAAoB;AAC5D,QAAI,YAAY,KAAK,SAAS,UAAa,CAAC,cAAc,SAAS,YAAY,KAAK,IAAI,GAAG;AACzF;AAAA,IACF;AAEA,SAAK,2BAA2B,QAAQ,WAAW;AAAA,EACrD;AAAA,EAEQ,4BAA4B,CAAC,gBAAmC;AACtE,QAAI,YAAY,aAAa,KAAK,qBAAqB;AACrD;AAAA,IACF;AACA,SAAK,6BAA6B,IAAI,OAA0B;AAChE,QACE,KAAK,aAAa,qBAClB,YAAY,oBACZ,4BAA4B,SAAS,YAAY,gBAAgB,GACjE;AACA,WAAK,OAAO;AAAA,QACV;AAAA,UACE,aAAa,YAAY;AAAA,UACzB,QAAQ,iBAAiB,YAAY,gBAAgB;AAAA,QACvD;AAAA,QACA;AAAA,MAEF;AACA,WAAK,aAAa,WAAW;AAAA,QAC3B,QAAQ,YAAY;AAAA,MACtB,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEQ,yBAAyB,CAAC,OAAkC;AAClE,SAAK,qBAAqB,MAAM,EAAE,EAAE,MAAM,CAAC,UAAU;AACnD,WAAK,OAAO,MAAM,EAAE,MAAM,GAAG,4CAA4C;AAAA,IAC3E,CAAC;AAAA,EACH;AAAA,EAEQ,sBAAsB,OAAO,OAA+B;AAClE,QAAI,KAAK,KAAK,eAAe,KAAK,KAAK,kBAAkB;AACvD,YAAM,KAAK,KAAK,iBAAiB,cAAc;AAAA,QAC7C,CAAC,gBAAgB,GAAG,GAAG;AAAA,MACzB,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEQ,kBAAkB,CAAC,QAA0B,oBAA0C;AAC7F,QAAI,gBAAgB,aAAa,KAAK,qBAAqB;AACzD;AAAA,IACF;AAEA,UAAM,cAAc,KAAK,KAAK,mBAAmB,IAAI,gBAAgB,QAAQ;AAC7E,QAAI,CAAC,aAAa;AAChB,WAAK,OAAO,KAAK,4CAA4C;AAC7D;AAAA,IACF;AAEA,UAAM,WAAW,YAAY;AAC3B,YAAM,OAAO,MAAM,OAAO,QAAQ;AAElC,YAAM,kBAAkB,KAAK,aAAa,kBAAmB,KAAK,cAAc;AAAA,QAC9E;AAAA,QACA,MAAM,OAAO;AAAA,QACb;AAAA,MACF,CAAC;AAGD,UAAI,2BAA2B,SAAS;AACtC,cAAM;AAAA,MACR;AAAA,IACF;AAEA,aAAS,EAAE,MAAM,CAAC,UAAU;AAC1B,WAAK,OAAO,MAAM,EAAE,MAAM,GAAG,0BAA0B;AAAA,IACzD,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,sBAAsB,QAAoC;AAhT1E;AAiTI,UAAM,SAAS,KAAK,qBAAqB,SAAS,UAAU;AAC5D,QAAI;AACF,aAAO,CAAC,OAAO,SAAS;AACtB,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AAEV,cAAM,QAAQ;AAEd,gBAAM,UAAK,yBAAL,mBAA2B,YAAY,MAAM;AACnD,YAAI,MAAM,SAAS;AACjB,qBAAK,yBAAL,mBAA2B;AAAA,QAC7B;AAAA,MACF;AAAA,IACF,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,EAAE,MAAM,GAAG,oCAAoC;AAAA,IACnE;AAAA,EACF;AAAA,EAEQ,0BAA0B,SAG/B;AACD,WAAO,IAAI,mBAAmB;AAAA,MAC5B,IAAI;AAAA,QACF,KAAK;AAAA,QACL,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA,IAAI,+BAA+B,KAAK,MAAM,QAAQ,eAAe,QAAQ,WAAW;AAAA,IAC1F,CAAC;AAAA,EACH;AAAA,EAEQ,0BAA0B;AAAA,IAChC;AAAA,IACA;AAAA,EACF,GAGG;AACD,QAAI,CAAC,QAAQ;AACX;AAAA,IACF;AAEA,eAAW,QAAQ,OAAO,QAAQ;AAChC,UACE,gBAAgB,wCAChB,gBAAgB,gCAChB;AACA,aAAK,eAAe,WAAW;AAAA,MACjC;AAAA,IACF;AAAA,EACF;AAAA,EAEA,IAAI,cAAuC;AACzC,QAAI,CAAC,KAAK,2BAA2B;AACnC,aAAO,KAAK;AAAA,IACd;AAEA,WAAO,KAAK,0BAA0B;AAAA,EACxC;AAAA,EAEA,IAAI,sBAA8C;AAChD,QAAI,CAAC,KAAK,2BAA2B;AACnC,aAAO,KAAK;AAAA,IACd;AAEA,WAAO,KAAK,0BAA0B;AAAA,EACxC;AAAA,EAEA,IAAI,yBAAkC;AACpC,WAAO,KAAK,2BAA2B;AAAA,EACzC;AAAA;AAAA,EAGA,eAAe,qBAAoC;AA3XrD;AA4XI,SAAK,OAAO,MAAM,EAAE,oBAAoB,GAAG,qBAAqB;AAChE,QAAI,wBAAwB,MAAM;AAChC,WAAK,iBAAiB;AACtB;AAAA,IACF;AAEA,QAAI,KAAK,wBAAwB,qBAAqB;AACpD,WAAK,6BAA6B,IAAI,OAA0B;AAGhE,iBAAW,eAAe,KAAK,KAAK,mBAAmB,OAAO,GAAG;AAC/D,YAAI,YAAY,aAAa,qBAAqB;AAChD,eAAK,2BAA2B,QAAQ,WAAW;AACnD;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAGA,SAAK,sBAAsB;AAC3B,eAAK,eAAL,mBAAiB,eAAe;AAChC,SAAK,0BAA0B;AAAA,MAC7B,QAAQ,KAAK;AAAA,MACb,aAAa;AAAA,IACf,CAAC;AAAA,EACH;AAAA,EAEA,mBAAmB;AAvZrB;AAwZI,SAAK,sBAAsB;AAC3B,SAAK,6BAA6B,IAAI,OAA0B;AAChE,eAAK,eAAL,mBAAiB,eAAe;AAChC,SAAK,0BAA0B;AAAA,MAC7B,QAAQ,KAAK;AAAA,MACb,aAAa;AAAA,IACf,CAAC;AAAA,EACH;AAAA,EAEA,QAAQ;AACN,QAAI,KAAK,aAAa,aAAa;AACjC,UAAI;AACF,aAAK,KAAK,0BAA0B,YAAY,KAAK,eAAe;AACpE,aAAK,8BAA8B;AAAA,MACrC,SAAS,OAAO;AACd,YAAI,KAAK,aAAa,aAAa;AACjC,eAAK,OAAO,KAAK,kCAAkC,UAAU,yBAAyB;AAAA,QACxF;AAAA,MACF;AAAA,IACF;AAGA,QAAI,KAAK,aAAa,cAAc;AAClC,WAAK,aAAa,IAAI,4BAA4B;AAAA,QAChD,MAAM,KAAK;AAAA,QACX,YAAY,KAAK,aAAa;AAAA,QAC9B,aAAa,KAAK,aAAa;AAAA,QAC/B,mBAAmB,KAAK,aAAa;AAAA,MACvC,CAAC;AAAA,IACH;AAGA,QAAI,KAAK,cAAc,cAAc;AACnC,WAAK,yBAAyB,IAAI,uBAAuB,KAAK,MAAM;AAAA,QAClE,YAAY,KAAK,cAAc;AAAA,QAC/B,aAAa,KAAK,cAAc;AAAA,QAChC,qBAAqB,KAAK,cAAc;AAAA,MAC1C,CAAC;AAAA,IACH;AACA,QAAI,KAAK,cAAc,sBAAsB;AAC3C,WAAK,uBAAuB,KAAK,0BAA0B;AAAA,QACzD,eAAe;AAAA,QACf,aAAa,KAAK;AAAA,MACpB,CAAC;AAED,WAAK,4BAA4B,KAAK;AAAA,QAAK,CAAC,eAC1C,KAAK,sBAAsB,WAAW,MAAM;AAAA,MAC9C;AACA,WAAK,wBAAwB,KAAK,0BAA0B;AAAA,QAC1D,eAAe;AAAA,QACf,aAAa;AAAA,MACf,CAAC;AAID,YAAM,cAAc,KAAK;AACzB,UAAI,KAAK,cAAc,qBAAqB,aAAa;AACvD,aAAK,4BAA4B,IAAI;AAAA,UACnC;AAAA,UACA,KAAK;AAAA,QACP;AAAA,MACF;AAAA,IACF;AAGA,SAAK,KAAK,GAAG,UAAU,sBAAsB,KAAK,sBAAsB;AACxE,SAAK,KAAK,GAAG,UAAU,wBAAwB,KAAK,wBAAwB;AAC5E,SAAK,KAAK,GAAG,UAAU,yBAAyB,KAAK,yBAAyB;AAC9E,QAAI,KAAK,KAAK,aAAa;AACzB,WAAK,yBAAyB,gBAAgB,cAAc;AAAA,IAC9D;AAEA,SAAK,WAAW,KAAK,KAAK,CAAC,eAAe,KAAK,KAAK,WAAW,MAAM,CAAC;AAGtE,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,MAAM,QAAQ,KAAK;AAAA,IACvC;AACA,QAAI,KAAK,aAAa;AACpB,WAAK,aAAa,OAAO,QAAQ,KAAK;AAAA,IACxC;AACA,QAAI,KAAK,qBAAqB;AAC5B,WAAK,aAAa,OAAO,gBAAgB,KAAK;AAAA,IAChD;AAEA,SAAK,aAAa,GAAG,uBAAuB,mBAAmB,KAAK,mBAAmB;AACvF,SAAK,aAAa,GAAG,uBAAuB,sBAAsB,KAAK,sBAAsB;AAAA,EAC/F;AAAA,EAEA,MAAM,QAAQ;AAjfhB;AAkfI,SAAK,KAAK,IAAI,UAAU,sBAAsB,KAAK,sBAAsB;AACzE,SAAK,KAAK,IAAI,UAAU,wBAAwB,KAAK,wBAAwB;AAC7E,SAAK,KAAK,IAAI,UAAU,yBAAyB,KAAK,yBAAyB;AAC/E,SAAK,aAAa,IAAI,uBAAuB,sBAAsB,KAAK,sBAAsB;AAC9F,SAAK,aAAa,IAAI,uBAAuB,mBAAmB,KAAK,mBAAmB;AAExF,QAAI,KAAK,6BAA6B;AACpC,WAAK,KAAK,4BAA4B,UAAU;AAChD,WAAK,8BAA8B;AAAA,IACrC;AAEA,YAAM,UAAK,aAAL,mBAAe;AAIrB,SAAK,qBAAqB,MAAM;AAChC,YAAM,UAAK,8BAAL,mBAAgC;AAEtC,YAAM,UAAK,eAAL,mBAAiB;AACvB,YAAM,UAAK,2BAAL,mBAA6B;AACnC,YAAM,UAAK,8BAAL,mBAAgC;AAAA,EACxC;AACF;","names":["participant"]}
@@ -56,6 +56,8 @@ class SpeechHandle {
56
56
  _tasks = [];
57
57
  /** @internal */
58
58
  _numSteps = 1;
59
+ /** @internal - OpenTelemetry context for the agent turn span */
60
+ _agentTurnContext;
59
61
  itemAddedCallbacks = /* @__PURE__ */ new Set();
60
62
  doneCallbacks = /* @__PURE__ */ new Set();
61
63
  /** @internal Symbol marker for type identification */
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/voice/speech_handle.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { ChatItem } from '../llm/index.js';\nimport type { Task } from '../utils.js';\nimport { Event, Future, shortuuid } from '../utils.js';\nimport { asyncLocalStorage } from './agent.js';\n\n/** Symbol used to identify SpeechHandle instances */\nconst SPEECH_HANDLE_SYMBOL = Symbol.for('livekit.agents.SpeechHandle');\n\n/**\n * Type guard to check if a value is a SpeechHandle.\n */\nexport function isSpeechHandle(value: unknown): value is SpeechHandle {\n return (\n typeof value === 'object' &&\n value !== null &&\n SPEECH_HANDLE_SYMBOL in value &&\n (value as Record<symbol, boolean>)[SPEECH_HANDLE_SYMBOL] === true\n );\n}\n\nexport class SpeechHandle {\n /** Priority for messages that should be played after all other messages in the queue */\n static SPEECH_PRIORITY_LOW = 0;\n /** Every speech generates by the VoiceAgent defaults to this priority. */\n static SPEECH_PRIORITY_NORMAL = 5;\n /** Priority for important messages that should be played before others. */\n static SPEECH_PRIORITY_HIGH = 10;\n\n private interruptFut = new Future<void>();\n private authorizedEvent = new Event();\n private scheduledFut = new Future<void>();\n private doneFut = new Future<void>();\n private generations: Future<void>[] = [];\n private _chatItems: ChatItem[] = [];\n\n /** @internal */\n _tasks: Task<void>[] = [];\n\n /** @internal */\n _numSteps = 1;\n\n private itemAddedCallbacks: Set<(item: ChatItem) => void> = new Set();\n private doneCallbacks: Set<(sh: SpeechHandle) => void> = new Set();\n\n /** @internal Symbol marker for type identification */\n readonly [SPEECH_HANDLE_SYMBOL] = true;\n\n constructor(\n private _id: string,\n private _allowInterruptions: boolean,\n /** @internal */\n public _stepIndex: number,\n readonly parent?: SpeechHandle,\n ) {\n this.doneFut.await.finally(() => {\n for (const callback of this.doneCallbacks) {\n callback(this);\n }\n });\n }\n\n static create(options?: {\n allowInterruptions?: boolean;\n stepIndex?: number;\n parent?: SpeechHandle;\n }) {\n const { allowInterruptions = true, stepIndex = 0, parent } = options ?? {};\n\n return new SpeechHandle(shortuuid('speech_'), allowInterruptions, stepIndex, parent);\n }\n\n get interrupted(): boolean {\n return this.interruptFut.done;\n }\n\n get numSteps(): number {\n return this._numSteps;\n }\n\n get id(): string {\n return this._id;\n }\n\n get scheduled(): boolean {\n return this.scheduledFut.done;\n }\n\n get allowInterruptions(): boolean {\n return this._allowInterruptions;\n }\n\n /**\n * Allow or disallow interruptions on this SpeechHandle.\n *\n * When set to false, the SpeechHandle will no longer accept any incoming\n * interruption requests until re-enabled. If the handle is already\n * interrupted, clearing interruptions is not allowed.\n *\n * @param value - true to allow interruptions, false to disallow\n * @throws Error If attempting to disable interruptions when already interrupted\n */\n set allowInterruptions(value: boolean) {\n if (this.interrupted && !value) {\n throw new Error(\n 'Cannot set allow_interruptions to False, the SpeechHandle is already interrupted',\n );\n }\n this._allowInterruptions = value;\n }\n\n done(): boolean {\n return this.doneFut.done;\n }\n\n get chatItems(): ChatItem[] {\n return this._chatItems;\n }\n\n /**\n * Interrupt the current speech generation.\n *\n * @throws Error If this speech handle does not allow interruptions.\n *\n * @returns The same speech handle that was interrupted.\n */\n interrupt(force: boolean = false): SpeechHandle {\n if (!force && !this.allowInterruptions) {\n throw new Error('This generation handle does not allow interruptions');\n }\n\n this._cancel();\n return this;\n }\n\n /**\n * Waits for the entire assistant turn to complete playback.\n *\n * This method waits until the assistant has fully finished speaking,\n * including any finalization steps beyond initial response generation.\n * This is appropriate to call when you want to ensure the speech output\n * has entirely played out, including any tool calls and response follow-ups.\n */\n async waitForPlayout(): Promise<void> {\n const store = asyncLocalStorage.getStore();\n if (store && store?.functionCall) {\n throw new Error(\n `Cannot call 'SpeechHandle.waitForPlayout()' from inside the function tool '${store.functionCall.name}'. ` +\n 'This creates a circular wait: the speech handle is waiting for the function tool to complete, ' +\n 'while the function tool is simultaneously waiting for the speech handle.\\n' +\n \"To wait for the assistant's spoken response prior to running this tool, use RunContext.wait_for_playout() instead.\",\n );\n }\n await this.doneFut.await;\n }\n\n async waitIfNotInterrupted(aw: Promise<unknown>[]): Promise<void> {\n const allTasksPromise = Promise.all(aw);\n const fs: Promise<unknown>[] = [allTasksPromise, this.interruptFut.await];\n await Promise.race(fs);\n }\n\n addDoneCallback(callback: (sh: SpeechHandle) => void) {\n this.doneCallbacks.add(callback);\n }\n\n removeDoneCallback(callback: (sh: SpeechHandle) => void) {\n this.doneCallbacks.delete(callback);\n }\n\n /** @internal */\n _cancel(): SpeechHandle {\n if (this.done()) {\n return this;\n }\n\n if (!this.interruptFut.done) {\n this.interruptFut.resolve();\n }\n\n return this;\n }\n\n /** @internal */\n _authorizeGeneration(): void {\n const fut = new Future<void>();\n this.generations.push(fut);\n this.authorizedEvent.set();\n }\n\n /** @internal */\n _clearAuthorization(): void {\n this.authorizedEvent.clear();\n }\n\n /** @internal */\n async _waitForAuthorization(): Promise<void> {\n await this.authorizedEvent.wait();\n }\n\n /** @internal */\n async _waitForGeneration(stepIdx: number = -1): Promise<void> {\n if (this.generations.length === 0) {\n throw new Error('cannot use wait_for_generation: no active generation is running.');\n }\n\n const index = stepIdx === -1 ? this.generations.length - 1 : stepIdx;\n const generation = this.generations[index];\n if (!generation) {\n throw new Error(`Generation at index ${index} not found.`);\n }\n return generation.await;\n }\n\n /** @internal */\n async _waitForScheduled(): Promise<void> {\n return this.scheduledFut.await;\n }\n\n /** @internal */\n _markGenerationDone(): void {\n if (this.generations.length === 0) {\n throw new Error('cannot use mark_generation_done: no active generation is running.');\n }\n\n const lastGeneration = this.generations[this.generations.length - 1];\n if (lastGeneration && !lastGeneration.done) {\n lastGeneration.resolve();\n }\n }\n\n /** @internal */\n _markDone(): void {\n if (!this.doneFut.done) {\n this.doneFut.resolve();\n if (this.generations.length > 0) {\n this._markGenerationDone(); // preemptive generation could be cancelled before being scheduled\n }\n }\n }\n\n /** @internal */\n _markScheduled(): void {\n if (!this.scheduledFut.done) {\n this.scheduledFut.resolve();\n }\n }\n\n /** @internal */\n _addItemAddedCallback(callback: (item: ChatItem) => void): void {\n this.itemAddedCallbacks.add(callback);\n }\n\n /** @internal */\n _removeItemAddedCallback(callback: (item: ChatItem) => void): void {\n this.itemAddedCallbacks.delete(callback);\n }\n\n /** @internal */\n _itemAdded(items: ChatItem[]): void {\n for (const item of items) {\n for (const cb of this.itemAddedCallbacks) {\n cb(item);\n }\n this._chatItems.push(item);\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAKA,mBAAyC;AACzC,mBAAkC;AAGlC,MAAM,uBAAuB,OAAO,IAAI,6BAA6B;AAK9D,SAAS,eAAe,OAAuC;AACpE,SACE,OAAO,UAAU,YACjB,UAAU,QACV,wBAAwB,SACvB,MAAkC,oBAAoB,MAAM;AAEjE;AAEO,MAAM,aAAa;AAAA,EA2BxB,YACU,KACA,qBAED,YACE,QACT;AALQ;AACA;AAED;AACE;AAET,SAAK,QAAQ,MAAM,QAAQ,MAAM;AAC/B,iBAAW,YAAY,KAAK,eAAe;AACzC,iBAAS,IAAI;AAAA,MACf;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA,EArCA,OAAO,sBAAsB;AAAA;AAAA,EAE7B,OAAO,yBAAyB;AAAA;AAAA,EAEhC,OAAO,uBAAuB;AAAA,EAEtB,eAAe,IAAI,oBAAa;AAAA,EAChC,kBAAkB,IAAI,mBAAM;AAAA,EAC5B,eAAe,IAAI,oBAAa;AAAA,EAChC,UAAU,IAAI,oBAAa;AAAA,EAC3B,cAA8B,CAAC;AAAA,EAC/B,aAAyB,CAAC;AAAA;AAAA,EAGlC,SAAuB,CAAC;AAAA;AAAA,EAGxB,YAAY;AAAA,EAEJ,qBAAoD,oBAAI,IAAI;AAAA,EAC5D,gBAAiD,oBAAI,IAAI;AAAA;AAAA,EAGjE,CAAU,oBAAoB,IAAI;AAAA,EAgBlC,OAAO,OAAO,SAIX;AACD,UAAM,EAAE,qBAAqB,MAAM,YAAY,GAAG,OAAO,IAAI,WAAW,CAAC;AAEzE,WAAO,IAAI,iBAAa,wBAAU,SAAS,GAAG,oBAAoB,WAAW,MAAM;AAAA,EACrF;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAI,qBAA8B;AAChC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,IAAI,mBAAmB,OAAgB;AACrC,QAAI,KAAK,eAAe,CAAC,OAAO;AAC9B,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,SAAK,sBAAsB;AAAA,EAC7B;AAAA,EAEA,OAAgB;AACd,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA,EAEA,IAAI,YAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,UAAU,QAAiB,OAAqB;AAC9C,QAAI,CAAC,SAAS,CAAC,KAAK,oBAAoB;AACtC,YAAM,IAAI,MAAM,qDAAqD;AAAA,IACvE;AAEA,SAAK,QAAQ;AACb,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAM,iBAAgC;AACpC,UAAM,QAAQ,+BAAkB,SAAS;AACzC,QAAI,UAAS,+BAAO,eAAc;AAChC,YAAM,IAAI;AAAA,QACR,8EAA8E,MAAM,aAAa,IAAI;AAAA;AAAA,MAIvG;AAAA,IACF;AACA,UAAM,KAAK,QAAQ;AAAA,EACrB;AAAA,EAEA,MAAM,qBAAqB,IAAuC;AAChE,UAAM,kBAAkB,QAAQ,IAAI,EAAE;AACtC,UAAM,KAAyB,CAAC,iBAAiB,KAAK,aAAa,KAAK;AACxE,UAAM,QAAQ,KAAK,EAAE;AAAA,EACvB;AAAA,EAEA,gBAAgB,UAAsC;AACpD,SAAK,cAAc,IAAI,QAAQ;AAAA,EACjC;AAAA,EAEA,mBAAmB,UAAsC;AACvD,SAAK,cAAc,OAAO,QAAQ;AAAA,EACpC;AAAA;AAAA,EAGA,UAAwB;AACtB,QAAI,KAAK,KAAK,GAAG;AACf,aAAO;AAAA,IACT;AAEA,QAAI,CAAC,KAAK,aAAa,MAAM;AAC3B,WAAK,aAAa,QAAQ;AAAA,IAC5B;AAEA,WAAO;AAAA,EACT;AAAA;AAAA,EAGA,uBAA6B;AAC3B,UAAM,MAAM,IAAI,oBAAa;AAC7B,SAAK,YAAY,KAAK,GAAG;AACzB,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA;AAAA,EAGA,sBAA4B;AAC1B,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA;AAAA,EAGA,MAAM,wBAAuC;AAC3C,UAAM,KAAK,gBAAgB,KAAK;AAAA,EAClC;AAAA;AAAA,EAGA,MAAM,mBAAmB,UAAkB,IAAmB;AAC5D,QAAI,KAAK,YAAY,WAAW,GAAG;AACjC,YAAM,IAAI,MAAM,kEAAkE;AAAA,IACpF;AAEA,UAAM,QAAQ,YAAY,KAAK,KAAK,YAAY,SAAS,IAAI;AAC7D,UAAM,aAAa,KAAK,YAAY,KAAK;AACzC,QAAI,CAAC,YAAY;AACf,YAAM,IAAI,MAAM,uBAAuB,KAAK,aAAa;AAAA,IAC3D;AACA,WAAO,WAAW;AAAA,EACpB;AAAA;AAAA,EAGA,MAAM,oBAAmC;AACvC,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA;AAAA,EAGA,sBAA4B;AAC1B,QAAI,KAAK,YAAY,WAAW,GAAG;AACjC,YAAM,IAAI,MAAM,mEAAmE;AAAA,IACrF;AAEA,UAAM,iBAAiB,KAAK,YAAY,KAAK,YAAY,SAAS,CAAC;AACnE,QAAI,kBAAkB,CAAC,eAAe,MAAM;AAC1C,qBAAe,QAAQ;AAAA,IACzB;AAAA,EACF;AAAA;AAAA,EAGA,YAAkB;AAChB,QAAI,CAAC,KAAK,QAAQ,MAAM;AACtB,WAAK,QAAQ,QAAQ;AACrB,UAAI,KAAK,YAAY,SAAS,GAAG;AAC/B,aAAK,oBAAoB;AAAA,MAC3B;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,iBAAuB;AACrB,QAAI,CAAC,KAAK,aAAa,MAAM;AAC3B,WAAK,aAAa,QAAQ;AAAA,IAC5B;AAAA,EACF;AAAA;AAAA,EAGA,sBAAsB,UAA0C;AAC9D,SAAK,mBAAmB,IAAI,QAAQ;AAAA,EACtC;AAAA;AAAA,EAGA,yBAAyB,UAA0C;AACjE,SAAK,mBAAmB,OAAO,QAAQ;AAAA,EACzC;AAAA;AAAA,EAGA,WAAW,OAAyB;AAClC,eAAW,QAAQ,OAAO;AACxB,iBAAW,MAAM,KAAK,oBAAoB;AACxC,WAAG,IAAI;AAAA,MACT;AACA,WAAK,WAAW,KAAK,IAAI;AAAA,IAC3B;AAAA,EACF;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/voice/speech_handle.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { Context } from '@opentelemetry/api';\nimport type { ChatItem } from '../llm/index.js';\nimport type { Task } from '../utils.js';\nimport { Event, Future, shortuuid } from '../utils.js';\nimport { asyncLocalStorage } from './agent.js';\n\n/** Symbol used to identify SpeechHandle instances */\nconst SPEECH_HANDLE_SYMBOL = Symbol.for('livekit.agents.SpeechHandle');\n\n/**\n * Type guard to check if a value is a SpeechHandle.\n */\nexport function isSpeechHandle(value: unknown): value is SpeechHandle {\n return (\n typeof value === 'object' &&\n value !== null &&\n SPEECH_HANDLE_SYMBOL in value &&\n (value as Record<symbol, boolean>)[SPEECH_HANDLE_SYMBOL] === true\n );\n}\n\nexport class SpeechHandle {\n /** Priority for messages that should be played after all other messages in the queue */\n static SPEECH_PRIORITY_LOW = 0;\n /** Every speech generates by the VoiceAgent defaults to this priority. */\n static SPEECH_PRIORITY_NORMAL = 5;\n /** Priority for important messages that should be played before others. */\n static SPEECH_PRIORITY_HIGH = 10;\n\n private interruptFut = new Future<void>();\n private authorizedEvent = new Event();\n private scheduledFut = new Future<void>();\n private doneFut = new Future<void>();\n private generations: Future<void>[] = [];\n private _chatItems: ChatItem[] = [];\n\n /** @internal */\n _tasks: Task<void>[] = [];\n\n /** @internal */\n _numSteps = 1;\n\n /** @internal - OpenTelemetry context for the agent turn span */\n _agentTurnContext?: Context;\n\n private itemAddedCallbacks: Set<(item: ChatItem) => void> = new Set();\n private doneCallbacks: Set<(sh: SpeechHandle) => void> = new Set();\n\n /** @internal Symbol marker for type identification */\n readonly [SPEECH_HANDLE_SYMBOL] = true;\n\n constructor(\n private _id: string,\n private _allowInterruptions: boolean,\n /** @internal */\n public _stepIndex: number,\n readonly parent?: SpeechHandle,\n ) {\n this.doneFut.await.finally(() => {\n for (const callback of this.doneCallbacks) {\n callback(this);\n }\n });\n }\n\n static create(options?: {\n allowInterruptions?: boolean;\n stepIndex?: number;\n parent?: SpeechHandle;\n }) {\n const { allowInterruptions = true, stepIndex = 0, parent } = options ?? {};\n\n return new SpeechHandle(shortuuid('speech_'), allowInterruptions, stepIndex, parent);\n }\n\n get interrupted(): boolean {\n return this.interruptFut.done;\n }\n\n get numSteps(): number {\n return this._numSteps;\n }\n\n get id(): string {\n return this._id;\n }\n\n get scheduled(): boolean {\n return this.scheduledFut.done;\n }\n\n get allowInterruptions(): boolean {\n return this._allowInterruptions;\n }\n\n /**\n * Allow or disallow interruptions on this SpeechHandle.\n *\n * When set to false, the SpeechHandle will no longer accept any incoming\n * interruption requests until re-enabled. If the handle is already\n * interrupted, clearing interruptions is not allowed.\n *\n * @param value - true to allow interruptions, false to disallow\n * @throws Error If attempting to disable interruptions when already interrupted\n */\n set allowInterruptions(value: boolean) {\n if (this.interrupted && !value) {\n throw new Error(\n 'Cannot set allow_interruptions to False, the SpeechHandle is already interrupted',\n );\n }\n this._allowInterruptions = value;\n }\n\n done(): boolean {\n return this.doneFut.done;\n }\n\n get chatItems(): ChatItem[] {\n return this._chatItems;\n }\n\n /**\n * Interrupt the current speech generation.\n *\n * @throws Error If this speech handle does not allow interruptions.\n *\n * @returns The same speech handle that was interrupted.\n */\n interrupt(force: boolean = false): SpeechHandle {\n if (!force && !this.allowInterruptions) {\n throw new Error('This generation handle does not allow interruptions');\n }\n\n this._cancel();\n return this;\n }\n\n /**\n * Waits for the entire assistant turn to complete playback.\n *\n * This method waits until the assistant has fully finished speaking,\n * including any finalization steps beyond initial response generation.\n * This is appropriate to call when you want to ensure the speech output\n * has entirely played out, including any tool calls and response follow-ups.\n */\n async waitForPlayout(): Promise<void> {\n const store = asyncLocalStorage.getStore();\n if (store && store?.functionCall) {\n throw new Error(\n `Cannot call 'SpeechHandle.waitForPlayout()' from inside the function tool '${store.functionCall.name}'. ` +\n 'This creates a circular wait: the speech handle is waiting for the function tool to complete, ' +\n 'while the function tool is simultaneously waiting for the speech handle.\\n' +\n \"To wait for the assistant's spoken response prior to running this tool, use RunContext.wait_for_playout() instead.\",\n );\n }\n await this.doneFut.await;\n }\n\n async waitIfNotInterrupted(aw: Promise<unknown>[]): Promise<void> {\n const allTasksPromise = Promise.all(aw);\n const fs: Promise<unknown>[] = [allTasksPromise, this.interruptFut.await];\n await Promise.race(fs);\n }\n\n addDoneCallback(callback: (sh: SpeechHandle) => void) {\n this.doneCallbacks.add(callback);\n }\n\n removeDoneCallback(callback: (sh: SpeechHandle) => void) {\n this.doneCallbacks.delete(callback);\n }\n\n /** @internal */\n _cancel(): SpeechHandle {\n if (this.done()) {\n return this;\n }\n\n if (!this.interruptFut.done) {\n this.interruptFut.resolve();\n }\n\n return this;\n }\n\n /** @internal */\n _authorizeGeneration(): void {\n const fut = new Future<void>();\n this.generations.push(fut);\n this.authorizedEvent.set();\n }\n\n /** @internal */\n _clearAuthorization(): void {\n this.authorizedEvent.clear();\n }\n\n /** @internal */\n async _waitForAuthorization(): Promise<void> {\n await this.authorizedEvent.wait();\n }\n\n /** @internal */\n async _waitForGeneration(stepIdx: number = -1): Promise<void> {\n if (this.generations.length === 0) {\n throw new Error('cannot use wait_for_generation: no active generation is running.');\n }\n\n const index = stepIdx === -1 ? this.generations.length - 1 : stepIdx;\n const generation = this.generations[index];\n if (!generation) {\n throw new Error(`Generation at index ${index} not found.`);\n }\n return generation.await;\n }\n\n /** @internal */\n async _waitForScheduled(): Promise<void> {\n return this.scheduledFut.await;\n }\n\n /** @internal */\n _markGenerationDone(): void {\n if (this.generations.length === 0) {\n throw new Error('cannot use mark_generation_done: no active generation is running.');\n }\n\n const lastGeneration = this.generations[this.generations.length - 1];\n if (lastGeneration && !lastGeneration.done) {\n lastGeneration.resolve();\n }\n }\n\n /** @internal */\n _markDone(): void {\n if (!this.doneFut.done) {\n this.doneFut.resolve();\n if (this.generations.length > 0) {\n this._markGenerationDone(); // preemptive generation could be cancelled before being scheduled\n }\n }\n }\n\n /** @internal */\n _markScheduled(): void {\n if (!this.scheduledFut.done) {\n this.scheduledFut.resolve();\n }\n }\n\n /** @internal */\n _addItemAddedCallback(callback: (item: ChatItem) => void): void {\n this.itemAddedCallbacks.add(callback);\n }\n\n /** @internal */\n _removeItemAddedCallback(callback: (item: ChatItem) => void): void {\n this.itemAddedCallbacks.delete(callback);\n }\n\n /** @internal */\n _itemAdded(items: ChatItem[]): void {\n for (const item of items) {\n for (const cb of this.itemAddedCallbacks) {\n cb(item);\n }\n this._chatItems.push(item);\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAMA,mBAAyC;AACzC,mBAAkC;AAGlC,MAAM,uBAAuB,OAAO,IAAI,6BAA6B;AAK9D,SAAS,eAAe,OAAuC;AACpE,SACE,OAAO,UAAU,YACjB,UAAU,QACV,wBAAwB,SACvB,MAAkC,oBAAoB,MAAM;AAEjE;AAEO,MAAM,aAAa;AAAA,EA8BxB,YACU,KACA,qBAED,YACE,QACT;AALQ;AACA;AAED;AACE;AAET,SAAK,QAAQ,MAAM,QAAQ,MAAM;AAC/B,iBAAW,YAAY,KAAK,eAAe;AACzC,iBAAS,IAAI;AAAA,MACf;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA,EAxCA,OAAO,sBAAsB;AAAA;AAAA,EAE7B,OAAO,yBAAyB;AAAA;AAAA,EAEhC,OAAO,uBAAuB;AAAA,EAEtB,eAAe,IAAI,oBAAa;AAAA,EAChC,kBAAkB,IAAI,mBAAM;AAAA,EAC5B,eAAe,IAAI,oBAAa;AAAA,EAChC,UAAU,IAAI,oBAAa;AAAA,EAC3B,cAA8B,CAAC;AAAA,EAC/B,aAAyB,CAAC;AAAA;AAAA,EAGlC,SAAuB,CAAC;AAAA;AAAA,EAGxB,YAAY;AAAA;AAAA,EAGZ;AAAA,EAEQ,qBAAoD,oBAAI,IAAI;AAAA,EAC5D,gBAAiD,oBAAI,IAAI;AAAA;AAAA,EAGjE,CAAU,oBAAoB,IAAI;AAAA,EAgBlC,OAAO,OAAO,SAIX;AACD,UAAM,EAAE,qBAAqB,MAAM,YAAY,GAAG,OAAO,IAAI,WAAW,CAAC;AAEzE,WAAO,IAAI,iBAAa,wBAAU,SAAS,GAAG,oBAAoB,WAAW,MAAM;AAAA,EACrF;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAI,qBAA8B;AAChC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,IAAI,mBAAmB,OAAgB;AACrC,QAAI,KAAK,eAAe,CAAC,OAAO;AAC9B,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,SAAK,sBAAsB;AAAA,EAC7B;AAAA,EAEA,OAAgB;AACd,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA,EAEA,IAAI,YAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,UAAU,QAAiB,OAAqB;AAC9C,QAAI,CAAC,SAAS,CAAC,KAAK,oBAAoB;AACtC,YAAM,IAAI,MAAM,qDAAqD;AAAA,IACvE;AAEA,SAAK,QAAQ;AACb,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAM,iBAAgC;AACpC,UAAM,QAAQ,+BAAkB,SAAS;AACzC,QAAI,UAAS,+BAAO,eAAc;AAChC,YAAM,IAAI;AAAA,QACR,8EAA8E,MAAM,aAAa,IAAI;AAAA;AAAA,MAIvG;AAAA,IACF;AACA,UAAM,KAAK,QAAQ;AAAA,EACrB;AAAA,EAEA,MAAM,qBAAqB,IAAuC;AAChE,UAAM,kBAAkB,QAAQ,IAAI,EAAE;AACtC,UAAM,KAAyB,CAAC,iBAAiB,KAAK,aAAa,KAAK;AACxE,UAAM,QAAQ,KAAK,EAAE;AAAA,EACvB;AAAA,EAEA,gBAAgB,UAAsC;AACpD,SAAK,cAAc,IAAI,QAAQ;AAAA,EACjC;AAAA,EAEA,mBAAmB,UAAsC;AACvD,SAAK,cAAc,OAAO,QAAQ;AAAA,EACpC;AAAA;AAAA,EAGA,UAAwB;AACtB,QAAI,KAAK,KAAK,GAAG;AACf,aAAO;AAAA,IACT;AAEA,QAAI,CAAC,KAAK,aAAa,MAAM;AAC3B,WAAK,aAAa,QAAQ;AAAA,IAC5B;AAEA,WAAO;AAAA,EACT;AAAA;AAAA,EAGA,uBAA6B;AAC3B,UAAM,MAAM,IAAI,oBAAa;AAC7B,SAAK,YAAY,KAAK,GAAG;AACzB,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA;AAAA,EAGA,sBAA4B;AAC1B,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA;AAAA,EAGA,MAAM,wBAAuC;AAC3C,UAAM,KAAK,gBAAgB,KAAK;AAAA,EAClC;AAAA;AAAA,EAGA,MAAM,mBAAmB,UAAkB,IAAmB;AAC5D,QAAI,KAAK,YAAY,WAAW,GAAG;AACjC,YAAM,IAAI,MAAM,kEAAkE;AAAA,IACpF;AAEA,UAAM,QAAQ,YAAY,KAAK,KAAK,YAAY,SAAS,IAAI;AAC7D,UAAM,aAAa,KAAK,YAAY,KAAK;AACzC,QAAI,CAAC,YAAY;AACf,YAAM,IAAI,MAAM,uBAAuB,KAAK,aAAa;AAAA,IAC3D;AACA,WAAO,WAAW;AAAA,EACpB;AAAA;AAAA,EAGA,MAAM,oBAAmC;AACvC,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA;AAAA,EAGA,sBAA4B;AAC1B,QAAI,KAAK,YAAY,WAAW,GAAG;AACjC,YAAM,IAAI,MAAM,mEAAmE;AAAA,IACrF;AAEA,UAAM,iBAAiB,KAAK,YAAY,KAAK,YAAY,SAAS,CAAC;AACnE,QAAI,kBAAkB,CAAC,eAAe,MAAM;AAC1C,qBAAe,QAAQ;AAAA,IACzB;AAAA,EACF;AAAA;AAAA,EAGA,YAAkB;AAChB,QAAI,CAAC,KAAK,QAAQ,MAAM;AACtB,WAAK,QAAQ,QAAQ;AACrB,UAAI,KAAK,YAAY,SAAS,GAAG;AAC/B,aAAK,oBAAoB;AAAA,MAC3B;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,iBAAuB;AACrB,QAAI,CAAC,KAAK,aAAa,MAAM;AAC3B,WAAK,aAAa,QAAQ;AAAA,IAC5B;AAAA,EACF;AAAA;AAAA,EAGA,sBAAsB,UAA0C;AAC9D,SAAK,mBAAmB,IAAI,QAAQ;AAAA,EACtC;AAAA;AAAA,EAGA,yBAAyB,UAA0C;AACjE,SAAK,mBAAmB,OAAO,QAAQ;AAAA,EACzC;AAAA;AAAA,EAGA,WAAW,OAAyB;AAClC,eAAW,QAAQ,OAAO;AACxB,iBAAW,MAAM,KAAK,oBAAoB;AACxC,WAAG,IAAI;AAAA,MACT;AACA,WAAK,WAAW,KAAK,IAAI;AAAA,IAC3B;AAAA,EACF;AACF;","names":[]}
@@ -1,3 +1,4 @@
1
+ import type { Context } from '@opentelemetry/api';
1
2
  import type { ChatItem } from '../llm/index.js';
2
3
  import type { Task } from '../utils.js';
3
4
  /** Symbol used to identify SpeechHandle instances */
@@ -28,6 +29,8 @@ export declare class SpeechHandle {
28
29
  _tasks: Task<void>[];
29
30
  /** @internal */
30
31
  _numSteps: number;
32
+ /** @internal - OpenTelemetry context for the agent turn span */
33
+ _agentTurnContext?: Context;
31
34
  private itemAddedCallbacks;
32
35
  private doneCallbacks;
33
36
  /** @internal Symbol marker for type identification */
@@ -1,3 +1,4 @@
1
+ import type { Context } from '@opentelemetry/api';
1
2
  import type { ChatItem } from '../llm/index.js';
2
3
  import type { Task } from '../utils.js';
3
4
  /** Symbol used to identify SpeechHandle instances */
@@ -28,6 +29,8 @@ export declare class SpeechHandle {
28
29
  _tasks: Task<void>[];
29
30
  /** @internal */
30
31
  _numSteps: number;
32
+ /** @internal - OpenTelemetry context for the agent turn span */
33
+ _agentTurnContext?: Context;
31
34
  private itemAddedCallbacks;
32
35
  private doneCallbacks;
33
36
  /** @internal Symbol marker for type identification */
@@ -1 +1 @@
1
- {"version":3,"file":"speech_handle.d.ts","sourceRoot":"","sources":["../../src/voice/speech_handle.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,iBAAiB,CAAC;AAChD,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,aAAa,CAAC;AAIxC,qDAAqD;AACrD,QAAA,MAAM,oBAAoB,eAA4C,CAAC;AAEvE;;GAEG;AACH,wBAAgB,cAAc,CAAC,KAAK,EAAE,OAAO,GAAG,KAAK,IAAI,YAAY,CAOpE;AAED,qBAAa,YAAY;IA4BrB,OAAO,CAAC,GAAG;IACX,OAAO,CAAC,mBAAmB;IAC3B,gBAAgB;IACT,UAAU,EAAE,MAAM;IACzB,QAAQ,CAAC,MAAM,CAAC;IA/BlB,wFAAwF;IACxF,MAAM,CAAC,mBAAmB,SAAK;IAC/B,0EAA0E;IAC1E,MAAM,CAAC,sBAAsB,SAAK;IAClC,2EAA2E;IAC3E,MAAM,CAAC,oBAAoB,SAAM;IAEjC,OAAO,CAAC,YAAY,CAAsB;IAC1C,OAAO,CAAC,eAAe,CAAe;IACtC,OAAO,CAAC,YAAY,CAAsB;IAC1C,OAAO,CAAC,OAAO,CAAsB;IACrC,OAAO,CAAC,WAAW,CAAsB;IACzC,OAAO,CAAC,UAAU,CAAkB;IAEpC,gBAAgB;IAChB,MAAM,EAAE,IAAI,CAAC,IAAI,CAAC,EAAE,CAAM;IAE1B,gBAAgB;IAChB,SAAS,SAAK;IAEd,OAAO,CAAC,kBAAkB,CAA4C;IACtE,OAAO,CAAC,aAAa,CAA8C;IAEnE,sDAAsD;IACtD,QAAQ,CAAC,CAAC,oBAAoB,CAAC,QAAQ;gBAG7B,GAAG,EAAE,MAAM,EACX,mBAAmB,EAAE,OAAO;IACpC,gBAAgB;IACT,UAAU,EAAE,MAAM,EAChB,MAAM,CAAC,0BAAc;IAShC,MAAM,CAAC,MAAM,CAAC,OAAO,CAAC,EAAE;QACtB,kBAAkB,CAAC,EAAE,OAAO,CAAC;QAC7B,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,MAAM,CAAC,EAAE,YAAY,CAAC;KACvB;IAMD,IAAI,WAAW,IAAI,OAAO,CAEzB;IAED,IAAI,QAAQ,IAAI,MAAM,CAErB;IAED,IAAI,EAAE,IAAI,MAAM,CAEf;IAED,IAAI,SAAS,IAAI,OAAO,CAEvB;IAED,IAAI,kBAAkB,IAAI,OAAO,CAEhC;IAED;;;;;;;;;OASG;IACH,IAAI,kBAAkB,CAAC,KAAK,EAAE,OAAO,EAOpC;IAED,IAAI,IAAI,OAAO;IAIf,IAAI,SAAS,IAAI,QAAQ,EAAE,CAE1B;IAED;;;;;;OAMG;IACH,SAAS,CAAC,KAAK,GAAE,OAAe,GAAG,YAAY;IAS/C;;;;;;;OAOG;IACG,cAAc,IAAI,OAAO,CAAC,IAAI,CAAC;IAa/B,oBAAoB,CAAC,EAAE,EAAE,OAAO,CAAC,OAAO,CAAC,EAAE,GAAG,OAAO,CAAC,IAAI,CAAC;IAMjE,eAAe,CAAC,QAAQ,EAAE,CAAC,EAAE,EAAE,YAAY,KAAK,IAAI;IAIpD,kBAAkB,CAAC,QAAQ,EAAE,CAAC,EAAE,EAAE,YAAY,KAAK,IAAI;IAIvD,gBAAgB;IAChB,OAAO,IAAI,YAAY;IAYvB,gBAAgB;IAChB,oBAAoB,IAAI,IAAI;IAM5B,gBAAgB;IAChB,mBAAmB,IAAI,IAAI;IAI3B,gBAAgB;IACV,qBAAqB,IAAI,OAAO,CAAC,IAAI,CAAC;IAI5C,gBAAgB;IACV,kBAAkB,CAAC,OAAO,GAAE,MAAW,GAAG,OAAO,CAAC,IAAI,CAAC;IAa7D,gBAAgB;IACV,iBAAiB,IAAI,OAAO,CAAC,IAAI,CAAC;IAIxC,gBAAgB;IAChB,mBAAmB,IAAI,IAAI;IAW3B,gBAAgB;IAChB,SAAS,IAAI,IAAI;IASjB,gBAAgB;IAChB,cAAc,IAAI,IAAI;IAMtB,gBAAgB;IAChB,qBAAqB,CAAC,QAAQ,EAAE,CAAC,IAAI,EAAE,QAAQ,KAAK,IAAI,GAAG,IAAI;IAI/D,gBAAgB;IAChB,wBAAwB,CAAC,QAAQ,EAAE,CAAC,IAAI,EAAE,QAAQ,KAAK,IAAI,GAAG,IAAI;IAIlE,gBAAgB;IAChB,UAAU,CAAC,KAAK,EAAE,QAAQ,EAAE,GAAG,IAAI;CAQpC"}
1
+ {"version":3,"file":"speech_handle.d.ts","sourceRoot":"","sources":["../../src/voice/speech_handle.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,OAAO,EAAE,MAAM,oBAAoB,CAAC;AAClD,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,iBAAiB,CAAC;AAChD,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,aAAa,CAAC;AAIxC,qDAAqD;AACrD,QAAA,MAAM,oBAAoB,eAA4C,CAAC;AAEvE;;GAEG;AACH,wBAAgB,cAAc,CAAC,KAAK,EAAE,OAAO,GAAG,KAAK,IAAI,YAAY,CAOpE;AAED,qBAAa,YAAY;IA+BrB,OAAO,CAAC,GAAG;IACX,OAAO,CAAC,mBAAmB;IAC3B,gBAAgB;IACT,UAAU,EAAE,MAAM;IACzB,QAAQ,CAAC,MAAM,CAAC;IAlClB,wFAAwF;IACxF,MAAM,CAAC,mBAAmB,SAAK;IAC/B,0EAA0E;IAC1E,MAAM,CAAC,sBAAsB,SAAK;IAClC,2EAA2E;IAC3E,MAAM,CAAC,oBAAoB,SAAM;IAEjC,OAAO,CAAC,YAAY,CAAsB;IAC1C,OAAO,CAAC,eAAe,CAAe;IACtC,OAAO,CAAC,YAAY,CAAsB;IAC1C,OAAO,CAAC,OAAO,CAAsB;IACrC,OAAO,CAAC,WAAW,CAAsB;IACzC,OAAO,CAAC,UAAU,CAAkB;IAEpC,gBAAgB;IAChB,MAAM,EAAE,IAAI,CAAC,IAAI,CAAC,EAAE,CAAM;IAE1B,gBAAgB;IAChB,SAAS,SAAK;IAEd,gEAAgE;IAChE,iBAAiB,CAAC,EAAE,OAAO,CAAC;IAE5B,OAAO,CAAC,kBAAkB,CAA4C;IACtE,OAAO,CAAC,aAAa,CAA8C;IAEnE,sDAAsD;IACtD,QAAQ,CAAC,CAAC,oBAAoB,CAAC,QAAQ;gBAG7B,GAAG,EAAE,MAAM,EACX,mBAAmB,EAAE,OAAO;IACpC,gBAAgB;IACT,UAAU,EAAE,MAAM,EAChB,MAAM,CAAC,0BAAc;IAShC,MAAM,CAAC,MAAM,CAAC,OAAO,CAAC,EAAE;QACtB,kBAAkB,CAAC,EAAE,OAAO,CAAC;QAC7B,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,MAAM,CAAC,EAAE,YAAY,CAAC;KACvB;IAMD,IAAI,WAAW,IAAI,OAAO,CAEzB;IAED,IAAI,QAAQ,IAAI,MAAM,CAErB;IAED,IAAI,EAAE,IAAI,MAAM,CAEf;IAED,IAAI,SAAS,IAAI,OAAO,CAEvB;IAED,IAAI,kBAAkB,IAAI,OAAO,CAEhC;IAED;;;;;;;;;OASG;IACH,IAAI,kBAAkB,CAAC,KAAK,EAAE,OAAO,EAOpC;IAED,IAAI,IAAI,OAAO;IAIf,IAAI,SAAS,IAAI,QAAQ,EAAE,CAE1B;IAED;;;;;;OAMG;IACH,SAAS,CAAC,KAAK,GAAE,OAAe,GAAG,YAAY;IAS/C;;;;;;;OAOG;IACG,cAAc,IAAI,OAAO,CAAC,IAAI,CAAC;IAa/B,oBAAoB,CAAC,EAAE,EAAE,OAAO,CAAC,OAAO,CAAC,EAAE,GAAG,OAAO,CAAC,IAAI,CAAC;IAMjE,eAAe,CAAC,QAAQ,EAAE,CAAC,EAAE,EAAE,YAAY,KAAK,IAAI;IAIpD,kBAAkB,CAAC,QAAQ,EAAE,CAAC,EAAE,EAAE,YAAY,KAAK,IAAI;IAIvD,gBAAgB;IAChB,OAAO,IAAI,YAAY;IAYvB,gBAAgB;IAChB,oBAAoB,IAAI,IAAI;IAM5B,gBAAgB;IAChB,mBAAmB,IAAI,IAAI;IAI3B,gBAAgB;IACV,qBAAqB,IAAI,OAAO,CAAC,IAAI,CAAC;IAI5C,gBAAgB;IACV,kBAAkB,CAAC,OAAO,GAAE,MAAW,GAAG,OAAO,CAAC,IAAI,CAAC;IAa7D,gBAAgB;IACV,iBAAiB,IAAI,OAAO,CAAC,IAAI,CAAC;IAIxC,gBAAgB;IAChB,mBAAmB,IAAI,IAAI;IAW3B,gBAAgB;IAChB,SAAS,IAAI,IAAI;IASjB,gBAAgB;IAChB,cAAc,IAAI,IAAI;IAMtB,gBAAgB;IAChB,qBAAqB,CAAC,QAAQ,EAAE,CAAC,IAAI,EAAE,QAAQ,KAAK,IAAI,GAAG,IAAI;IAI/D,gBAAgB;IAChB,wBAAwB,CAAC,QAAQ,EAAE,CAAC,IAAI,EAAE,QAAQ,KAAK,IAAI,GAAG,IAAI;IAIlE,gBAAgB;IAChB,UAAU,CAAC,KAAK,EAAE,QAAQ,EAAE,GAAG,IAAI;CAQpC"}
@@ -32,6 +32,8 @@ class SpeechHandle {
32
32
  _tasks = [];
33
33
  /** @internal */
34
34
  _numSteps = 1;
35
+ /** @internal - OpenTelemetry context for the agent turn span */
36
+ _agentTurnContext;
35
37
  itemAddedCallbacks = /* @__PURE__ */ new Set();
36
38
  doneCallbacks = /* @__PURE__ */ new Set();
37
39
  /** @internal Symbol marker for type identification */
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/voice/speech_handle.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { ChatItem } from '../llm/index.js';\nimport type { Task } from '../utils.js';\nimport { Event, Future, shortuuid } from '../utils.js';\nimport { asyncLocalStorage } from './agent.js';\n\n/** Symbol used to identify SpeechHandle instances */\nconst SPEECH_HANDLE_SYMBOL = Symbol.for('livekit.agents.SpeechHandle');\n\n/**\n * Type guard to check if a value is a SpeechHandle.\n */\nexport function isSpeechHandle(value: unknown): value is SpeechHandle {\n return (\n typeof value === 'object' &&\n value !== null &&\n SPEECH_HANDLE_SYMBOL in value &&\n (value as Record<symbol, boolean>)[SPEECH_HANDLE_SYMBOL] === true\n );\n}\n\nexport class SpeechHandle {\n /** Priority for messages that should be played after all other messages in the queue */\n static SPEECH_PRIORITY_LOW = 0;\n /** Every speech generates by the VoiceAgent defaults to this priority. */\n static SPEECH_PRIORITY_NORMAL = 5;\n /** Priority for important messages that should be played before others. */\n static SPEECH_PRIORITY_HIGH = 10;\n\n private interruptFut = new Future<void>();\n private authorizedEvent = new Event();\n private scheduledFut = new Future<void>();\n private doneFut = new Future<void>();\n private generations: Future<void>[] = [];\n private _chatItems: ChatItem[] = [];\n\n /** @internal */\n _tasks: Task<void>[] = [];\n\n /** @internal */\n _numSteps = 1;\n\n private itemAddedCallbacks: Set<(item: ChatItem) => void> = new Set();\n private doneCallbacks: Set<(sh: SpeechHandle) => void> = new Set();\n\n /** @internal Symbol marker for type identification */\n readonly [SPEECH_HANDLE_SYMBOL] = true;\n\n constructor(\n private _id: string,\n private _allowInterruptions: boolean,\n /** @internal */\n public _stepIndex: number,\n readonly parent?: SpeechHandle,\n ) {\n this.doneFut.await.finally(() => {\n for (const callback of this.doneCallbacks) {\n callback(this);\n }\n });\n }\n\n static create(options?: {\n allowInterruptions?: boolean;\n stepIndex?: number;\n parent?: SpeechHandle;\n }) {\n const { allowInterruptions = true, stepIndex = 0, parent } = options ?? {};\n\n return new SpeechHandle(shortuuid('speech_'), allowInterruptions, stepIndex, parent);\n }\n\n get interrupted(): boolean {\n return this.interruptFut.done;\n }\n\n get numSteps(): number {\n return this._numSteps;\n }\n\n get id(): string {\n return this._id;\n }\n\n get scheduled(): boolean {\n return this.scheduledFut.done;\n }\n\n get allowInterruptions(): boolean {\n return this._allowInterruptions;\n }\n\n /**\n * Allow or disallow interruptions on this SpeechHandle.\n *\n * When set to false, the SpeechHandle will no longer accept any incoming\n * interruption requests until re-enabled. If the handle is already\n * interrupted, clearing interruptions is not allowed.\n *\n * @param value - true to allow interruptions, false to disallow\n * @throws Error If attempting to disable interruptions when already interrupted\n */\n set allowInterruptions(value: boolean) {\n if (this.interrupted && !value) {\n throw new Error(\n 'Cannot set allow_interruptions to False, the SpeechHandle is already interrupted',\n );\n }\n this._allowInterruptions = value;\n }\n\n done(): boolean {\n return this.doneFut.done;\n }\n\n get chatItems(): ChatItem[] {\n return this._chatItems;\n }\n\n /**\n * Interrupt the current speech generation.\n *\n * @throws Error If this speech handle does not allow interruptions.\n *\n * @returns The same speech handle that was interrupted.\n */\n interrupt(force: boolean = false): SpeechHandle {\n if (!force && !this.allowInterruptions) {\n throw new Error('This generation handle does not allow interruptions');\n }\n\n this._cancel();\n return this;\n }\n\n /**\n * Waits for the entire assistant turn to complete playback.\n *\n * This method waits until the assistant has fully finished speaking,\n * including any finalization steps beyond initial response generation.\n * This is appropriate to call when you want to ensure the speech output\n * has entirely played out, including any tool calls and response follow-ups.\n */\n async waitForPlayout(): Promise<void> {\n const store = asyncLocalStorage.getStore();\n if (store && store?.functionCall) {\n throw new Error(\n `Cannot call 'SpeechHandle.waitForPlayout()' from inside the function tool '${store.functionCall.name}'. ` +\n 'This creates a circular wait: the speech handle is waiting for the function tool to complete, ' +\n 'while the function tool is simultaneously waiting for the speech handle.\\n' +\n \"To wait for the assistant's spoken response prior to running this tool, use RunContext.wait_for_playout() instead.\",\n );\n }\n await this.doneFut.await;\n }\n\n async waitIfNotInterrupted(aw: Promise<unknown>[]): Promise<void> {\n const allTasksPromise = Promise.all(aw);\n const fs: Promise<unknown>[] = [allTasksPromise, this.interruptFut.await];\n await Promise.race(fs);\n }\n\n addDoneCallback(callback: (sh: SpeechHandle) => void) {\n this.doneCallbacks.add(callback);\n }\n\n removeDoneCallback(callback: (sh: SpeechHandle) => void) {\n this.doneCallbacks.delete(callback);\n }\n\n /** @internal */\n _cancel(): SpeechHandle {\n if (this.done()) {\n return this;\n }\n\n if (!this.interruptFut.done) {\n this.interruptFut.resolve();\n }\n\n return this;\n }\n\n /** @internal */\n _authorizeGeneration(): void {\n const fut = new Future<void>();\n this.generations.push(fut);\n this.authorizedEvent.set();\n }\n\n /** @internal */\n _clearAuthorization(): void {\n this.authorizedEvent.clear();\n }\n\n /** @internal */\n async _waitForAuthorization(): Promise<void> {\n await this.authorizedEvent.wait();\n }\n\n /** @internal */\n async _waitForGeneration(stepIdx: number = -1): Promise<void> {\n if (this.generations.length === 0) {\n throw new Error('cannot use wait_for_generation: no active generation is running.');\n }\n\n const index = stepIdx === -1 ? this.generations.length - 1 : stepIdx;\n const generation = this.generations[index];\n if (!generation) {\n throw new Error(`Generation at index ${index} not found.`);\n }\n return generation.await;\n }\n\n /** @internal */\n async _waitForScheduled(): Promise<void> {\n return this.scheduledFut.await;\n }\n\n /** @internal */\n _markGenerationDone(): void {\n if (this.generations.length === 0) {\n throw new Error('cannot use mark_generation_done: no active generation is running.');\n }\n\n const lastGeneration = this.generations[this.generations.length - 1];\n if (lastGeneration && !lastGeneration.done) {\n lastGeneration.resolve();\n }\n }\n\n /** @internal */\n _markDone(): void {\n if (!this.doneFut.done) {\n this.doneFut.resolve();\n if (this.generations.length > 0) {\n this._markGenerationDone(); // preemptive generation could be cancelled before being scheduled\n }\n }\n }\n\n /** @internal */\n _markScheduled(): void {\n if (!this.scheduledFut.done) {\n this.scheduledFut.resolve();\n }\n }\n\n /** @internal */\n _addItemAddedCallback(callback: (item: ChatItem) => void): void {\n this.itemAddedCallbacks.add(callback);\n }\n\n /** @internal */\n _removeItemAddedCallback(callback: (item: ChatItem) => void): void {\n this.itemAddedCallbacks.delete(callback);\n }\n\n /** @internal */\n _itemAdded(items: ChatItem[]): void {\n for (const item of items) {\n for (const cb of this.itemAddedCallbacks) {\n cb(item);\n }\n this._chatItems.push(item);\n }\n }\n}\n"],"mappings":"AAKA,SAAS,OAAO,QAAQ,iBAAiB;AACzC,SAAS,yBAAyB;AAGlC,MAAM,uBAAuB,OAAO,IAAI,6BAA6B;AAK9D,SAAS,eAAe,OAAuC;AACpE,SACE,OAAO,UAAU,YACjB,UAAU,QACV,wBAAwB,SACvB,MAAkC,oBAAoB,MAAM;AAEjE;AAEO,MAAM,aAAa;AAAA,EA2BxB,YACU,KACA,qBAED,YACE,QACT;AALQ;AACA;AAED;AACE;AAET,SAAK,QAAQ,MAAM,QAAQ,MAAM;AAC/B,iBAAW,YAAY,KAAK,eAAe;AACzC,iBAAS,IAAI;AAAA,MACf;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA,EArCA,OAAO,sBAAsB;AAAA;AAAA,EAE7B,OAAO,yBAAyB;AAAA;AAAA,EAEhC,OAAO,uBAAuB;AAAA,EAEtB,eAAe,IAAI,OAAa;AAAA,EAChC,kBAAkB,IAAI,MAAM;AAAA,EAC5B,eAAe,IAAI,OAAa;AAAA,EAChC,UAAU,IAAI,OAAa;AAAA,EAC3B,cAA8B,CAAC;AAAA,EAC/B,aAAyB,CAAC;AAAA;AAAA,EAGlC,SAAuB,CAAC;AAAA;AAAA,EAGxB,YAAY;AAAA,EAEJ,qBAAoD,oBAAI,IAAI;AAAA,EAC5D,gBAAiD,oBAAI,IAAI;AAAA;AAAA,EAGjE,CAAU,oBAAoB,IAAI;AAAA,EAgBlC,OAAO,OAAO,SAIX;AACD,UAAM,EAAE,qBAAqB,MAAM,YAAY,GAAG,OAAO,IAAI,WAAW,CAAC;AAEzE,WAAO,IAAI,aAAa,UAAU,SAAS,GAAG,oBAAoB,WAAW,MAAM;AAAA,EACrF;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAI,qBAA8B;AAChC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,IAAI,mBAAmB,OAAgB;AACrC,QAAI,KAAK,eAAe,CAAC,OAAO;AAC9B,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,SAAK,sBAAsB;AAAA,EAC7B;AAAA,EAEA,OAAgB;AACd,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA,EAEA,IAAI,YAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,UAAU,QAAiB,OAAqB;AAC9C,QAAI,CAAC,SAAS,CAAC,KAAK,oBAAoB;AACtC,YAAM,IAAI,MAAM,qDAAqD;AAAA,IACvE;AAEA,SAAK,QAAQ;AACb,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAM,iBAAgC;AACpC,UAAM,QAAQ,kBAAkB,SAAS;AACzC,QAAI,UAAS,+BAAO,eAAc;AAChC,YAAM,IAAI;AAAA,QACR,8EAA8E,MAAM,aAAa,IAAI;AAAA;AAAA,MAIvG;AAAA,IACF;AACA,UAAM,KAAK,QAAQ;AAAA,EACrB;AAAA,EAEA,MAAM,qBAAqB,IAAuC;AAChE,UAAM,kBAAkB,QAAQ,IAAI,EAAE;AACtC,UAAM,KAAyB,CAAC,iBAAiB,KAAK,aAAa,KAAK;AACxE,UAAM,QAAQ,KAAK,EAAE;AAAA,EACvB;AAAA,EAEA,gBAAgB,UAAsC;AACpD,SAAK,cAAc,IAAI,QAAQ;AAAA,EACjC;AAAA,EAEA,mBAAmB,UAAsC;AACvD,SAAK,cAAc,OAAO,QAAQ;AAAA,EACpC;AAAA;AAAA,EAGA,UAAwB;AACtB,QAAI,KAAK,KAAK,GAAG;AACf,aAAO;AAAA,IACT;AAEA,QAAI,CAAC,KAAK,aAAa,MAAM;AAC3B,WAAK,aAAa,QAAQ;AAAA,IAC5B;AAEA,WAAO;AAAA,EACT;AAAA;AAAA,EAGA,uBAA6B;AAC3B,UAAM,MAAM,IAAI,OAAa;AAC7B,SAAK,YAAY,KAAK,GAAG;AACzB,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA;AAAA,EAGA,sBAA4B;AAC1B,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA;AAAA,EAGA,MAAM,wBAAuC;AAC3C,UAAM,KAAK,gBAAgB,KAAK;AAAA,EAClC;AAAA;AAAA,EAGA,MAAM,mBAAmB,UAAkB,IAAmB;AAC5D,QAAI,KAAK,YAAY,WAAW,GAAG;AACjC,YAAM,IAAI,MAAM,kEAAkE;AAAA,IACpF;AAEA,UAAM,QAAQ,YAAY,KAAK,KAAK,YAAY,SAAS,IAAI;AAC7D,UAAM,aAAa,KAAK,YAAY,KAAK;AACzC,QAAI,CAAC,YAAY;AACf,YAAM,IAAI,MAAM,uBAAuB,KAAK,aAAa;AAAA,IAC3D;AACA,WAAO,WAAW;AAAA,EACpB;AAAA;AAAA,EAGA,MAAM,oBAAmC;AACvC,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA;AAAA,EAGA,sBAA4B;AAC1B,QAAI,KAAK,YAAY,WAAW,GAAG;AACjC,YAAM,IAAI,MAAM,mEAAmE;AAAA,IACrF;AAEA,UAAM,iBAAiB,KAAK,YAAY,KAAK,YAAY,SAAS,CAAC;AACnE,QAAI,kBAAkB,CAAC,eAAe,MAAM;AAC1C,qBAAe,QAAQ;AAAA,IACzB;AAAA,EACF;AAAA;AAAA,EAGA,YAAkB;AAChB,QAAI,CAAC,KAAK,QAAQ,MAAM;AACtB,WAAK,QAAQ,QAAQ;AACrB,UAAI,KAAK,YAAY,SAAS,GAAG;AAC/B,aAAK,oBAAoB;AAAA,MAC3B;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,iBAAuB;AACrB,QAAI,CAAC,KAAK,aAAa,MAAM;AAC3B,WAAK,aAAa,QAAQ;AAAA,IAC5B;AAAA,EACF;AAAA;AAAA,EAGA,sBAAsB,UAA0C;AAC9D,SAAK,mBAAmB,IAAI,QAAQ;AAAA,EACtC;AAAA;AAAA,EAGA,yBAAyB,UAA0C;AACjE,SAAK,mBAAmB,OAAO,QAAQ;AAAA,EACzC;AAAA;AAAA,EAGA,WAAW,OAAyB;AAClC,eAAW,QAAQ,OAAO;AACxB,iBAAW,MAAM,KAAK,oBAAoB;AACxC,WAAG,IAAI;AAAA,MACT;AACA,WAAK,WAAW,KAAK,IAAI;AAAA,IAC3B;AAAA,EACF;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/voice/speech_handle.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { Context } from '@opentelemetry/api';\nimport type { ChatItem } from '../llm/index.js';\nimport type { Task } from '../utils.js';\nimport { Event, Future, shortuuid } from '../utils.js';\nimport { asyncLocalStorage } from './agent.js';\n\n/** Symbol used to identify SpeechHandle instances */\nconst SPEECH_HANDLE_SYMBOL = Symbol.for('livekit.agents.SpeechHandle');\n\n/**\n * Type guard to check if a value is a SpeechHandle.\n */\nexport function isSpeechHandle(value: unknown): value is SpeechHandle {\n return (\n typeof value === 'object' &&\n value !== null &&\n SPEECH_HANDLE_SYMBOL in value &&\n (value as Record<symbol, boolean>)[SPEECH_HANDLE_SYMBOL] === true\n );\n}\n\nexport class SpeechHandle {\n /** Priority for messages that should be played after all other messages in the queue */\n static SPEECH_PRIORITY_LOW = 0;\n /** Every speech generates by the VoiceAgent defaults to this priority. */\n static SPEECH_PRIORITY_NORMAL = 5;\n /** Priority for important messages that should be played before others. */\n static SPEECH_PRIORITY_HIGH = 10;\n\n private interruptFut = new Future<void>();\n private authorizedEvent = new Event();\n private scheduledFut = new Future<void>();\n private doneFut = new Future<void>();\n private generations: Future<void>[] = [];\n private _chatItems: ChatItem[] = [];\n\n /** @internal */\n _tasks: Task<void>[] = [];\n\n /** @internal */\n _numSteps = 1;\n\n /** @internal - OpenTelemetry context for the agent turn span */\n _agentTurnContext?: Context;\n\n private itemAddedCallbacks: Set<(item: ChatItem) => void> = new Set();\n private doneCallbacks: Set<(sh: SpeechHandle) => void> = new Set();\n\n /** @internal Symbol marker for type identification */\n readonly [SPEECH_HANDLE_SYMBOL] = true;\n\n constructor(\n private _id: string,\n private _allowInterruptions: boolean,\n /** @internal */\n public _stepIndex: number,\n readonly parent?: SpeechHandle,\n ) {\n this.doneFut.await.finally(() => {\n for (const callback of this.doneCallbacks) {\n callback(this);\n }\n });\n }\n\n static create(options?: {\n allowInterruptions?: boolean;\n stepIndex?: number;\n parent?: SpeechHandle;\n }) {\n const { allowInterruptions = true, stepIndex = 0, parent } = options ?? {};\n\n return new SpeechHandle(shortuuid('speech_'), allowInterruptions, stepIndex, parent);\n }\n\n get interrupted(): boolean {\n return this.interruptFut.done;\n }\n\n get numSteps(): number {\n return this._numSteps;\n }\n\n get id(): string {\n return this._id;\n }\n\n get scheduled(): boolean {\n return this.scheduledFut.done;\n }\n\n get allowInterruptions(): boolean {\n return this._allowInterruptions;\n }\n\n /**\n * Allow or disallow interruptions on this SpeechHandle.\n *\n * When set to false, the SpeechHandle will no longer accept any incoming\n * interruption requests until re-enabled. If the handle is already\n * interrupted, clearing interruptions is not allowed.\n *\n * @param value - true to allow interruptions, false to disallow\n * @throws Error If attempting to disable interruptions when already interrupted\n */\n set allowInterruptions(value: boolean) {\n if (this.interrupted && !value) {\n throw new Error(\n 'Cannot set allow_interruptions to False, the SpeechHandle is already interrupted',\n );\n }\n this._allowInterruptions = value;\n }\n\n done(): boolean {\n return this.doneFut.done;\n }\n\n get chatItems(): ChatItem[] {\n return this._chatItems;\n }\n\n /**\n * Interrupt the current speech generation.\n *\n * @throws Error If this speech handle does not allow interruptions.\n *\n * @returns The same speech handle that was interrupted.\n */\n interrupt(force: boolean = false): SpeechHandle {\n if (!force && !this.allowInterruptions) {\n throw new Error('This generation handle does not allow interruptions');\n }\n\n this._cancel();\n return this;\n }\n\n /**\n * Waits for the entire assistant turn to complete playback.\n *\n * This method waits until the assistant has fully finished speaking,\n * including any finalization steps beyond initial response generation.\n * This is appropriate to call when you want to ensure the speech output\n * has entirely played out, including any tool calls and response follow-ups.\n */\n async waitForPlayout(): Promise<void> {\n const store = asyncLocalStorage.getStore();\n if (store && store?.functionCall) {\n throw new Error(\n `Cannot call 'SpeechHandle.waitForPlayout()' from inside the function tool '${store.functionCall.name}'. ` +\n 'This creates a circular wait: the speech handle is waiting for the function tool to complete, ' +\n 'while the function tool is simultaneously waiting for the speech handle.\\n' +\n \"To wait for the assistant's spoken response prior to running this tool, use RunContext.wait_for_playout() instead.\",\n );\n }\n await this.doneFut.await;\n }\n\n async waitIfNotInterrupted(aw: Promise<unknown>[]): Promise<void> {\n const allTasksPromise = Promise.all(aw);\n const fs: Promise<unknown>[] = [allTasksPromise, this.interruptFut.await];\n await Promise.race(fs);\n }\n\n addDoneCallback(callback: (sh: SpeechHandle) => void) {\n this.doneCallbacks.add(callback);\n }\n\n removeDoneCallback(callback: (sh: SpeechHandle) => void) {\n this.doneCallbacks.delete(callback);\n }\n\n /** @internal */\n _cancel(): SpeechHandle {\n if (this.done()) {\n return this;\n }\n\n if (!this.interruptFut.done) {\n this.interruptFut.resolve();\n }\n\n return this;\n }\n\n /** @internal */\n _authorizeGeneration(): void {\n const fut = new Future<void>();\n this.generations.push(fut);\n this.authorizedEvent.set();\n }\n\n /** @internal */\n _clearAuthorization(): void {\n this.authorizedEvent.clear();\n }\n\n /** @internal */\n async _waitForAuthorization(): Promise<void> {\n await this.authorizedEvent.wait();\n }\n\n /** @internal */\n async _waitForGeneration(stepIdx: number = -1): Promise<void> {\n if (this.generations.length === 0) {\n throw new Error('cannot use wait_for_generation: no active generation is running.');\n }\n\n const index = stepIdx === -1 ? this.generations.length - 1 : stepIdx;\n const generation = this.generations[index];\n if (!generation) {\n throw new Error(`Generation at index ${index} not found.`);\n }\n return generation.await;\n }\n\n /** @internal */\n async _waitForScheduled(): Promise<void> {\n return this.scheduledFut.await;\n }\n\n /** @internal */\n _markGenerationDone(): void {\n if (this.generations.length === 0) {\n throw new Error('cannot use mark_generation_done: no active generation is running.');\n }\n\n const lastGeneration = this.generations[this.generations.length - 1];\n if (lastGeneration && !lastGeneration.done) {\n lastGeneration.resolve();\n }\n }\n\n /** @internal */\n _markDone(): void {\n if (!this.doneFut.done) {\n this.doneFut.resolve();\n if (this.generations.length > 0) {\n this._markGenerationDone(); // preemptive generation could be cancelled before being scheduled\n }\n }\n }\n\n /** @internal */\n _markScheduled(): void {\n if (!this.scheduledFut.done) {\n this.scheduledFut.resolve();\n }\n }\n\n /** @internal */\n _addItemAddedCallback(callback: (item: ChatItem) => void): void {\n this.itemAddedCallbacks.add(callback);\n }\n\n /** @internal */\n _removeItemAddedCallback(callback: (item: ChatItem) => void): void {\n this.itemAddedCallbacks.delete(callback);\n }\n\n /** @internal */\n _itemAdded(items: ChatItem[]): void {\n for (const item of items) {\n for (const cb of this.itemAddedCallbacks) {\n cb(item);\n }\n this._chatItems.push(item);\n }\n }\n}\n"],"mappings":"AAMA,SAAS,OAAO,QAAQ,iBAAiB;AACzC,SAAS,yBAAyB;AAGlC,MAAM,uBAAuB,OAAO,IAAI,6BAA6B;AAK9D,SAAS,eAAe,OAAuC;AACpE,SACE,OAAO,UAAU,YACjB,UAAU,QACV,wBAAwB,SACvB,MAAkC,oBAAoB,MAAM;AAEjE;AAEO,MAAM,aAAa;AAAA,EA8BxB,YACU,KACA,qBAED,YACE,QACT;AALQ;AACA;AAED;AACE;AAET,SAAK,QAAQ,MAAM,QAAQ,MAAM;AAC/B,iBAAW,YAAY,KAAK,eAAe;AACzC,iBAAS,IAAI;AAAA,MACf;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA,EAxCA,OAAO,sBAAsB;AAAA;AAAA,EAE7B,OAAO,yBAAyB;AAAA;AAAA,EAEhC,OAAO,uBAAuB;AAAA,EAEtB,eAAe,IAAI,OAAa;AAAA,EAChC,kBAAkB,IAAI,MAAM;AAAA,EAC5B,eAAe,IAAI,OAAa;AAAA,EAChC,UAAU,IAAI,OAAa;AAAA,EAC3B,cAA8B,CAAC;AAAA,EAC/B,aAAyB,CAAC;AAAA;AAAA,EAGlC,SAAuB,CAAC;AAAA;AAAA,EAGxB,YAAY;AAAA;AAAA,EAGZ;AAAA,EAEQ,qBAAoD,oBAAI,IAAI;AAAA,EAC5D,gBAAiD,oBAAI,IAAI;AAAA;AAAA,EAGjE,CAAU,oBAAoB,IAAI;AAAA,EAgBlC,OAAO,OAAO,SAIX;AACD,UAAM,EAAE,qBAAqB,MAAM,YAAY,GAAG,OAAO,IAAI,WAAW,CAAC;AAEzE,WAAO,IAAI,aAAa,UAAU,SAAS,GAAG,oBAAoB,WAAW,MAAM;AAAA,EACrF;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAI,qBAA8B;AAChC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,IAAI,mBAAmB,OAAgB;AACrC,QAAI,KAAK,eAAe,CAAC,OAAO;AAC9B,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,SAAK,sBAAsB;AAAA,EAC7B;AAAA,EAEA,OAAgB;AACd,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA,EAEA,IAAI,YAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,UAAU,QAAiB,OAAqB;AAC9C,QAAI,CAAC,SAAS,CAAC,KAAK,oBAAoB;AACtC,YAAM,IAAI,MAAM,qDAAqD;AAAA,IACvE;AAEA,SAAK,QAAQ;AACb,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAM,iBAAgC;AACpC,UAAM,QAAQ,kBAAkB,SAAS;AACzC,QAAI,UAAS,+BAAO,eAAc;AAChC,YAAM,IAAI;AAAA,QACR,8EAA8E,MAAM,aAAa,IAAI;AAAA;AAAA,MAIvG;AAAA,IACF;AACA,UAAM,KAAK,QAAQ;AAAA,EACrB;AAAA,EAEA,MAAM,qBAAqB,IAAuC;AAChE,UAAM,kBAAkB,QAAQ,IAAI,EAAE;AACtC,UAAM,KAAyB,CAAC,iBAAiB,KAAK,aAAa,KAAK;AACxE,UAAM,QAAQ,KAAK,EAAE;AAAA,EACvB;AAAA,EAEA,gBAAgB,UAAsC;AACpD,SAAK,cAAc,IAAI,QAAQ;AAAA,EACjC;AAAA,EAEA,mBAAmB,UAAsC;AACvD,SAAK,cAAc,OAAO,QAAQ;AAAA,EACpC;AAAA;AAAA,EAGA,UAAwB;AACtB,QAAI,KAAK,KAAK,GAAG;AACf,aAAO;AAAA,IACT;AAEA,QAAI,CAAC,KAAK,aAAa,MAAM;AAC3B,WAAK,aAAa,QAAQ;AAAA,IAC5B;AAEA,WAAO;AAAA,EACT;AAAA;AAAA,EAGA,uBAA6B;AAC3B,UAAM,MAAM,IAAI,OAAa;AAC7B,SAAK,YAAY,KAAK,GAAG;AACzB,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA;AAAA,EAGA,sBAA4B;AAC1B,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA;AAAA,EAGA,MAAM,wBAAuC;AAC3C,UAAM,KAAK,gBAAgB,KAAK;AAAA,EAClC;AAAA;AAAA,EAGA,MAAM,mBAAmB,UAAkB,IAAmB;AAC5D,QAAI,KAAK,YAAY,WAAW,GAAG;AACjC,YAAM,IAAI,MAAM,kEAAkE;AAAA,IACpF;AAEA,UAAM,QAAQ,YAAY,KAAK,KAAK,YAAY,SAAS,IAAI;AAC7D,UAAM,aAAa,KAAK,YAAY,KAAK;AACzC,QAAI,CAAC,YAAY;AACf,YAAM,IAAI,MAAM,uBAAuB,KAAK,aAAa;AAAA,IAC3D;AACA,WAAO,WAAW;AAAA,EACpB;AAAA;AAAA,EAGA,MAAM,oBAAmC;AACvC,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA;AAAA,EAGA,sBAA4B;AAC1B,QAAI,KAAK,YAAY,WAAW,GAAG;AACjC,YAAM,IAAI,MAAM,mEAAmE;AAAA,IACrF;AAEA,UAAM,iBAAiB,KAAK,YAAY,KAAK,YAAY,SAAS,CAAC;AACnE,QAAI,kBAAkB,CAAC,eAAe,MAAM;AAC1C,qBAAe,QAAQ;AAAA,IACzB;AAAA,EACF;AAAA;AAAA,EAGA,YAAkB;AAChB,QAAI,CAAC,KAAK,QAAQ,MAAM;AACtB,WAAK,QAAQ,QAAQ;AACrB,UAAI,KAAK,YAAY,SAAS,GAAG;AAC/B,aAAK,oBAAoB;AAAA,MAC3B;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,iBAAuB;AACrB,QAAI,CAAC,KAAK,aAAa,MAAM;AAC3B,WAAK,aAAa,QAAQ;AAAA,IAC5B;AAAA,EACF;AAAA;AAAA,EAGA,sBAAsB,UAA0C;AAC9D,SAAK,mBAAmB,IAAI,QAAQ;AAAA,EACtC;AAAA;AAAA,EAGA,yBAAyB,UAA0C;AACjE,SAAK,mBAAmB,OAAO,QAAQ;AAAA,EACzC;AAAA;AAAA,EAGA,WAAW,OAAyB;AAClC,eAAW,QAAQ,OAAO;AACxB,iBAAW,MAAM,KAAK,oBAAoB;AACxC,WAAG,IAAI;AAAA,MACT;AACA,WAAK,WAAW,KAAK,IAAI;AAAA,IAC3B;AAAA,EACF;AACF;","names":[]}
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@livekit/agents",
3
- "version": "1.0.37",
3
+ "version": "1.0.38",
4
4
  "description": "LiveKit Agents - Node.js",
5
5
  "main": "dist/index.js",
6
6
  "require": "dist/index.cjs",
@@ -80,3 +80,86 @@ export type TtsSessionClosedEvent = z.infer<typeof ttsSessionClosedEventSchema>;
80
80
  export type TtsErrorEvent = z.infer<typeof ttsErrorEventSchema>;
81
81
  export type TtsClientEvent = z.infer<typeof ttsClientEventSchema>;
82
82
  export type TtsServerEvent = z.infer<typeof ttsServerEventSchema>;
83
+
84
+ // ============================================================================
85
+ // STT Schemas
86
+ // ============================================================================
87
+
88
+ // Word-level timing data
89
+ export const sttWordSchema = z.object({
90
+ word: z.string().optional().default(''),
91
+ start: z.number().optional().default(0),
92
+ end: z.number().optional().default(0),
93
+ confidence: z.number().optional().default(0.0),
94
+ extra: z.unknown().nullable().optional(),
95
+ });
96
+
97
+ // Interim transcript event
98
+ export const sttInterimTranscriptEventSchema = z.object({
99
+ type: z.literal('interim_transcript'),
100
+ session_id: z.string().optional(),
101
+ transcript: z.string().optional().default(''),
102
+ language: z.string().optional().default(''),
103
+ start: z.number().optional().default(0),
104
+ duration: z.number().optional().default(0),
105
+ confidence: z.number().optional().default(1.0),
106
+ words: z.array(sttWordSchema).optional().default([]),
107
+ extra: z.unknown().nullable().optional(),
108
+ });
109
+
110
+ // Final transcript event
111
+ export const sttFinalTranscriptEventSchema = z.object({
112
+ type: z.literal('final_transcript'),
113
+ session_id: z.string().optional(),
114
+ transcript: z.string().optional().default(''),
115
+ language: z.string().optional().default(''),
116
+ start: z.number().optional().default(0),
117
+ duration: z.number().optional().default(0),
118
+ confidence: z.number().optional().default(1.0),
119
+ words: z.array(sttWordSchema).optional().default([]),
120
+ extra: z.unknown().nullable().optional(),
121
+ });
122
+
123
+ // Session created event
124
+ export const sttSessionCreatedEventSchema = z.object({
125
+ type: z.literal('session.created'),
126
+ session_id: z.string().optional(),
127
+ });
128
+
129
+ // Session finalized event
130
+ export const sttSessionFinalizedEventSchema = z.object({
131
+ type: z.literal('session.finalized'),
132
+ });
133
+
134
+ // Session closed event
135
+ export const sttSessionClosedEventSchema = z.object({
136
+ type: z.literal('session.closed'),
137
+ });
138
+
139
+ // Error event
140
+ export const sttErrorEventSchema = z.object({
141
+ type: z.literal('error'),
142
+ message: z.string().optional(),
143
+ code: z.string().optional(),
144
+ });
145
+
146
+ // Discriminated union for all STT server events
147
+ export const sttServerEventSchema = z.discriminatedUnion('type', [
148
+ sttSessionCreatedEventSchema,
149
+ sttSessionFinalizedEventSchema,
150
+ sttSessionClosedEventSchema,
151
+ sttInterimTranscriptEventSchema,
152
+ sttFinalTranscriptEventSchema,
153
+ sttErrorEventSchema,
154
+ ]);
155
+
156
+ // Type exports for STT
157
+ export type SttWord = z.infer<typeof sttWordSchema>;
158
+ export type SttInterimTranscriptEvent = z.infer<typeof sttInterimTranscriptEventSchema>;
159
+ export type SttFinalTranscriptEvent = z.infer<typeof sttFinalTranscriptEventSchema>;
160
+ export type SttTranscriptEvent = SttInterimTranscriptEvent | SttFinalTranscriptEvent;
161
+ export type SttSessionCreatedEvent = z.infer<typeof sttSessionCreatedEventSchema>;
162
+ export type SttSessionFinalizedEvent = z.infer<typeof sttSessionFinalizedEventSchema>;
163
+ export type SttSessionClosedEvent = z.infer<typeof sttSessionClosedEventSchema>;
164
+ export type SttErrorEvent = z.infer<typeof sttErrorEventSchema>;
165
+ export type SttServerEvent = z.infer<typeof sttServerEventSchema>;
@@ -16,6 +16,12 @@ import {
16
16
  } from '../stt/index.js';
17
17
  import { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';
18
18
  import { type AudioBuffer, Event, Task, cancelAndWait, shortuuid, waitForAbort } from '../utils.js';
19
+ import type { TimedString } from '../voice/io.js';
20
+ import {
21
+ type SttServerEvent,
22
+ type SttTranscriptEvent,
23
+ sttServerEventSchema,
24
+ } from './api_protos.js';
19
25
  import { type AnyString, connectWs, createAccessToken } from './utils.js';
20
26
 
21
27
  export type DeepgramModels =
@@ -122,7 +128,7 @@ export class STT<TModel extends STTModels> extends BaseSTT {
122
128
  apiSecret?: string;
123
129
  modelOptions?: STTOptions<TModel>;
124
130
  }) {
125
- super({ streaming: true, interimResults: true });
131
+ super({ streaming: true, interimResults: true, alignedTranscript: 'word' });
126
132
 
127
133
  const {
128
134
  model,
@@ -271,7 +277,6 @@ export class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {
271
277
  let closing = false;
272
278
  let finalReceived = false;
273
279
 
274
- type SttServerEvent = Record<string, any>;
275
280
  const eventChannel = createStreamChannel<SttServerEvent>();
276
281
 
277
282
  const resourceCleanup = () => {
@@ -380,10 +385,19 @@ export class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {
380
385
  if (signal.aborted) return;
381
386
  if (result.done) return;
382
387
 
383
- const json = result.value;
384
- const type = json.type as string | undefined;
388
+ // Parse and validate with Zod schema
389
+ const parseResult = await sttServerEventSchema.safeParseAsync(result.value);
390
+ if (!parseResult.success) {
391
+ this.#logger.warn(
392
+ { error: parseResult.error, rawData: result.value },
393
+ 'Failed to parse STT server event',
394
+ );
395
+ continue;
396
+ }
397
+
398
+ const event: SttServerEvent = parseResult.data;
385
399
 
386
- switch (type) {
400
+ switch (event.type) {
387
401
  case 'session.created':
388
402
  case 'session.finalized':
389
403
  break;
@@ -392,21 +406,15 @@ export class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {
392
406
  resourceCleanup();
393
407
  break;
394
408
  case 'interim_transcript':
395
- this.processTranscript(json, false);
409
+ this.processTranscript(event, false);
396
410
  break;
397
411
  case 'final_transcript':
398
- this.processTranscript(json, true);
412
+ this.processTranscript(event, true);
399
413
  break;
400
414
  case 'error':
401
- this.#logger.error({ error: json }, 'Received error from LiveKit STT');
415
+ this.#logger.error({ error: event }, 'Received error from LiveKit STT');
402
416
  resourceCleanup();
403
- throw new APIError(`LiveKit STT returned error: ${JSON.stringify(json)}`);
404
- default:
405
- this.#logger.warn(
406
- { message: json },
407
- 'Received unexpected message from LiveKit STT',
408
- );
409
- break;
417
+ throw new APIError(`LiveKit STT returned error: ${JSON.stringify(event)}`);
410
418
  }
411
419
  }
412
420
  } finally {
@@ -457,13 +465,13 @@ export class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {
457
465
  }
458
466
  }
459
467
 
460
- private processTranscript(data: Record<string, any>, isFinal: boolean) {
468
+ private processTranscript(data: SttTranscriptEvent, isFinal: boolean) {
461
469
  // Check if queue is closed to avoid race condition during disconnect
462
470
  if (this.queue.closed) return;
463
471
 
464
- const requestId = data.request_id ?? this.requestId;
465
- const text = data.transcript ?? '';
466
- const language = data.language ?? this.opts.language ?? 'en';
472
+ const requestId = data.session_id || this.requestId;
473
+ const text = data.transcript;
474
+ const language = data.language || this.opts.language || 'en';
467
475
 
468
476
  if (!text && !isFinal) return;
469
477
 
@@ -476,10 +484,19 @@ export class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {
476
484
 
477
485
  const speechData: SpeechData = {
478
486
  language,
479
- startTime: data.start ?? 0,
480
- endTime: data.duration ?? 0,
481
- confidence: data.confidence ?? 1.0,
487
+ startTime: this.startTimeOffset + data.start,
488
+ endTime: this.startTimeOffset + data.start + data.duration,
489
+ confidence: data.confidence,
482
490
  text,
491
+ words: data.words.map(
492
+ (word): TimedString => ({
493
+ text: word.word,
494
+ startTime: word.start + this.startTimeOffset,
495
+ endTime: word.end + this.startTimeOffset,
496
+ startTimeOffset: this.startTimeOffset,
497
+ confidence: word.confidence,
498
+ }),
499
+ ),
483
500
  };
484
501
 
485
502
  if (isFinal) {
package/src/stt/stt.ts CHANGED
@@ -13,6 +13,7 @@ import { DeferredReadableStream } from '../stream/deferred_stream.js';
13
13
  import { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, intervalForRetry } from '../types.js';
14
14
  import type { AudioBuffer } from '../utils.js';
15
15
  import { AsyncIterableQueue, delay, startSoon, toError } from '../utils.js';
16
+ import type { TimedString } from '../voice/index.js';
16
17
 
17
18
  /** Indicates start/middle/end of speech */
18
19
  export enum SpeechEventType {
@@ -53,6 +54,7 @@ export interface SpeechData {
53
54
  startTime: number;
54
55
  endTime: number;
55
56
  confidence: number;
57
+ words?: TimedString[];
56
58
  }
57
59
 
58
60
  export interface RecognitionUsage {
@@ -76,6 +78,13 @@ export interface SpeechEvent {
76
78
  export interface STTCapabilities {
77
79
  streaming: boolean;
78
80
  interimResults: boolean;
81
+ /**
82
+ * Whether this STT supports aligned transcripts with word/chunk timestamps.
83
+ * - 'word': Provider returns word-level timestamps
84
+ * - 'chunk': Provider returns chunk-level timestamps (e.g., sentence/phrase boundaries)
85
+ * - false: Provider does not support aligned transcripts
86
+ */
87
+ alignedTranscript?: 'word' | 'chunk' | false;
79
88
  }
80
89
 
81
90
  export interface STTError {
@@ -176,6 +185,7 @@ export abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent>
176
185
  private deferredInputStream: DeferredReadableStream<AudioFrame>;
177
186
  private logger = log();
178
187
  private _connOptions: APIConnectOptions;
188
+ private _startTimeOffset: number = 0;
179
189
 
180
190
  protected abortController = new AbortController();
181
191
 
@@ -300,6 +310,17 @@ export abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent>
300
310
  return this.abortController.signal;
301
311
  }
302
312
 
313
+ get startTimeOffset(): number {
314
+ return this._startTimeOffset;
315
+ }
316
+
317
+ set startTimeOffset(value: number) {
318
+ if (value < 0) {
319
+ throw new Error('startTimeOffset must be non-negative');
320
+ }
321
+ this._startTimeOffset = value;
322
+ }
323
+
303
324
  updateInputStream(audioStream: ReadableStream<AudioFrame>) {
304
325
  this.deferredInputStream.setSource(audioStream);
305
326
  }
@@ -37,6 +37,8 @@ export interface StartSpanOptions {
37
37
  attributes?: Attributes;
38
38
  /** Whether to end the span when the function exits (default: true) */
39
39
  endOnExit?: boolean;
40
+ /** Optional start time for the span in milliseconds (Date.now() format) */
41
+ startTime?: number;
40
42
  }
41
43
 
42
44
  /**
@@ -79,10 +81,12 @@ class DynamicTracer {
79
81
  */
80
82
  startSpan(options: StartSpanOptions): Span {
81
83
  const ctx = options.context || otelContext.active();
84
+
82
85
  const span = this.tracer.startSpan(
83
86
  options.name,
84
87
  {
85
88
  attributes: options.attributes,
89
+ startTime: options.startTime,
86
90
  },
87
91
  ctx,
88
92
  );
@@ -101,7 +105,7 @@ class DynamicTracer {
101
105
  async startActiveSpan<T>(fn: (span: Span) => Promise<T>, options: StartSpanOptions): Promise<T> {
102
106
  const ctx = options.context || otelContext.active();
103
107
  const endOnExit = options.endOnExit === undefined ? true : options.endOnExit; // default true
104
- const opts: SpanOptions = { attributes: options.attributes };
108
+ const opts: SpanOptions = { attributes: options.attributes, startTime: options.startTime };
105
109
 
106
110
  // Directly return the tracer's startActiveSpan result - it handles async correctly
107
111
  return await this.tracer.startActiveSpan(options.name, opts, ctx, async (span) => {
@@ -125,7 +129,7 @@ class DynamicTracer {
125
129
  startActiveSpanSync<T>(fn: (span: Span) => T, options: StartSpanOptions): T {
126
130
  const ctx = options.context || otelContext.active();
127
131
  const endOnExit = options.endOnExit === undefined ? true : options.endOnExit; // default true
128
- const opts: SpanOptions = { attributes: options.attributes };
132
+ const opts: SpanOptions = { attributes: options.attributes, startTime: options.startTime };
129
133
 
130
134
  return this.tracer.startActiveSpan(options.name, opts, ctx, (span) => {
131
135
  try {