@livekit/agents 1.0.40 → 1.0.41
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.
- package/dist/cli.cjs +20 -18
- package/dist/cli.cjs.map +1 -1
- package/dist/cli.d.ts.map +1 -1
- package/dist/cli.js +20 -18
- package/dist/cli.js.map +1 -1
- package/dist/index.cjs +5 -0
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +1 -0
- package/dist/index.d.ts +1 -0
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +3 -0
- package/dist/index.js.map +1 -1
- package/dist/inference/stt.cjs +2 -1
- package/dist/inference/stt.cjs.map +1 -1
- package/dist/inference/stt.d.ts.map +1 -1
- package/dist/inference/stt.js +2 -1
- package/dist/inference/stt.js.map +1 -1
- package/dist/llm/realtime.cjs.map +1 -1
- package/dist/llm/realtime.d.cts +5 -1
- package/dist/llm/realtime.d.ts +5 -1
- package/dist/llm/realtime.d.ts.map +1 -1
- package/dist/llm/realtime.js.map +1 -1
- package/dist/tts/stream_adapter.cjs +15 -1
- package/dist/tts/stream_adapter.cjs.map +1 -1
- package/dist/tts/stream_adapter.d.ts.map +1 -1
- package/dist/tts/stream_adapter.js +15 -1
- package/dist/tts/stream_adapter.js.map +1 -1
- package/dist/tts/tts.cjs.map +1 -1
- package/dist/tts/tts.d.cts +9 -1
- package/dist/tts/tts.d.ts +9 -1
- package/dist/tts/tts.d.ts.map +1 -1
- package/dist/tts/tts.js.map +1 -1
- package/dist/types.cjs +3 -0
- package/dist/types.cjs.map +1 -1
- package/dist/types.d.cts +4 -0
- package/dist/types.d.ts +4 -0
- package/dist/types.d.ts.map +1 -1
- package/dist/types.js +2 -0
- package/dist/types.js.map +1 -1
- package/dist/voice/agent.cjs +11 -1
- package/dist/voice/agent.cjs.map +1 -1
- package/dist/voice/agent.d.cts +7 -3
- package/dist/voice/agent.d.ts +7 -3
- package/dist/voice/agent.d.ts.map +1 -1
- package/dist/voice/agent.js +11 -1
- package/dist/voice/agent.js.map +1 -1
- package/dist/voice/agent_activity.cjs +30 -14
- package/dist/voice/agent_activity.cjs.map +1 -1
- package/dist/voice/agent_activity.d.cts +1 -0
- package/dist/voice/agent_activity.d.ts +1 -0
- package/dist/voice/agent_activity.d.ts.map +1 -1
- package/dist/voice/agent_activity.js +30 -14
- package/dist/voice/agent_activity.js.map +1 -1
- package/dist/voice/agent_session.cjs +5 -1
- package/dist/voice/agent_session.cjs.map +1 -1
- package/dist/voice/agent_session.d.cts +2 -0
- package/dist/voice/agent_session.d.ts +2 -0
- package/dist/voice/agent_session.d.ts.map +1 -1
- package/dist/voice/agent_session.js +5 -1
- package/dist/voice/agent_session.js.map +1 -1
- package/dist/voice/background_audio.cjs +2 -1
- package/dist/voice/background_audio.cjs.map +1 -1
- package/dist/voice/background_audio.d.cts +4 -2
- package/dist/voice/background_audio.d.ts +4 -2
- package/dist/voice/background_audio.d.ts.map +1 -1
- package/dist/voice/background_audio.js +2 -1
- package/dist/voice/background_audio.js.map +1 -1
- package/dist/voice/generation.cjs +58 -5
- package/dist/voice/generation.cjs.map +1 -1
- package/dist/voice/generation.d.cts +17 -3
- package/dist/voice/generation.d.ts +17 -3
- package/dist/voice/generation.d.ts.map +1 -1
- package/dist/voice/generation.js +63 -6
- package/dist/voice/generation.js.map +1 -1
- package/dist/voice/index.cjs.map +1 -1
- package/dist/voice/index.d.cts +1 -1
- package/dist/voice/index.d.ts +1 -1
- package/dist/voice/index.d.ts.map +1 -1
- package/dist/voice/index.js.map +1 -1
- package/dist/voice/io.cjs +22 -2
- package/dist/voice/io.cjs.map +1 -1
- package/dist/voice/io.d.cts +21 -5
- package/dist/voice/io.d.ts +21 -5
- package/dist/voice/io.d.ts.map +1 -1
- package/dist/voice/io.js +18 -1
- package/dist/voice/io.js.map +1 -1
- package/dist/voice/room_io/_output.cjs +3 -2
- package/dist/voice/room_io/_output.cjs.map +1 -1
- package/dist/voice/room_io/_output.d.cts +3 -3
- package/dist/voice/room_io/_output.d.ts +3 -3
- package/dist/voice/room_io/_output.d.ts.map +1 -1
- package/dist/voice/room_io/_output.js +4 -3
- package/dist/voice/room_io/_output.js.map +1 -1
- package/dist/voice/transcription/synchronizer.cjs +137 -13
- package/dist/voice/transcription/synchronizer.cjs.map +1 -1
- package/dist/voice/transcription/synchronizer.d.cts +34 -4
- package/dist/voice/transcription/synchronizer.d.ts +34 -4
- package/dist/voice/transcription/synchronizer.d.ts.map +1 -1
- package/dist/voice/transcription/synchronizer.js +141 -14
- package/dist/voice/transcription/synchronizer.js.map +1 -1
- package/dist/voice/transcription/synchronizer.test.cjs +151 -0
- package/dist/voice/transcription/synchronizer.test.cjs.map +1 -0
- package/dist/voice/transcription/synchronizer.test.js +150 -0
- package/dist/voice/transcription/synchronizer.test.js.map +1 -0
- package/package.json +1 -1
- package/src/cli.ts +20 -18
- package/src/index.ts +1 -0
- package/src/inference/stt.ts +9 -8
- package/src/llm/realtime.ts +5 -1
- package/src/tts/stream_adapter.ts +23 -1
- package/src/tts/tts.ts +10 -1
- package/src/types.ts +5 -0
- package/src/voice/agent.ts +19 -4
- package/src/voice/agent_activity.ts +38 -13
- package/src/voice/agent_session.ts +6 -0
- package/src/voice/background_audio.ts +6 -3
- package/src/voice/generation.ts +115 -10
- package/src/voice/index.ts +1 -1
- package/src/voice/io.ts +40 -5
- package/src/voice/room_io/_output.ts +6 -5
- package/src/voice/transcription/synchronizer.test.ts +206 -0
- package/src/voice/transcription/synchronizer.ts +202 -17
package/dist/voice/agent.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/voice/agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { AsyncLocalStorage } from 'node:async_hooks';\nimport { ReadableStream } from 'node:stream/web';\nimport {\n LLM as InferenceLLM,\n STT as InferenceSTT,\n TTS as InferenceTTS,\n type LLMModels,\n type STTModelString,\n type TTSModelString,\n} from '../inference/index.js';\nimport { ReadonlyChatContext } from '../llm/chat_context.js';\nimport type { ChatMessage, FunctionCall, RealtimeModel } from '../llm/index.js';\nimport {\n type ChatChunk,\n ChatContext,\n LLM,\n type ToolChoice,\n type ToolContext,\n} from '../llm/index.js';\nimport type { STT, SpeechEvent } from '../stt/index.js';\nimport { StreamAdapter as STTStreamAdapter } from '../stt/index.js';\nimport { SentenceTokenizer as BasicSentenceTokenizer } from '../tokenize/basic/index.js';\nimport type { TTS } from '../tts/index.js';\nimport { SynthesizeStream, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport type { VAD } from '../vad.js';\nimport type { AgentActivity } from './agent_activity.js';\nimport type { AgentSession, TurnDetectionMode } from './agent_session.js';\n\nexport const asyncLocalStorage = new AsyncLocalStorage<{ functionCall?: FunctionCall }>();\nexport const STOP_RESPONSE_SYMBOL = Symbol('StopResponse');\n\nexport class StopResponse extends Error {\n constructor() {\n super();\n this.name = 'StopResponse';\n\n Object.defineProperty(this, STOP_RESPONSE_SYMBOL, {\n value: true,\n });\n }\n}\n\nexport function isStopResponse(value: unknown): value is StopResponse {\n return (\n value !== undefined &&\n value !== null &&\n typeof value === 'object' &&\n STOP_RESPONSE_SYMBOL in value\n );\n}\n\nexport interface ModelSettings {\n /** The tool choice to use when calling the LLM. */\n toolChoice?: ToolChoice;\n}\n\nexport interface AgentOptions<UserData> {\n id?: string;\n instructions: string;\n chatCtx?: ChatContext;\n tools?: ToolContext<UserData>;\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n allowInterruptions?: boolean;\n minConsecutiveSpeechDelay?: number;\n}\n\nexport class Agent<UserData = any> {\n private _id: string;\n private turnDetection?: TurnDetectionMode;\n private _stt?: STT;\n private _vad?: VAD;\n private _llm?: LLM | RealtimeModel;\n private _tts?: TTS;\n\n /** @internal */\n _agentActivity?: AgentActivity;\n\n /** @internal */\n _chatCtx: ChatContext;\n\n /** @internal */\n _instructions: string;\n\n /** @internal */\n _tools?: ToolContext<UserData>;\n\n constructor({\n id,\n instructions,\n chatCtx,\n tools,\n turnDetection,\n stt,\n vad,\n llm,\n tts,\n }: AgentOptions<UserData>) {\n if (id) {\n this._id = id;\n } else {\n // Convert class name to snake_case\n const className = this.constructor.name;\n if (className === 'Agent') {\n this._id = 'default_agent';\n } else {\n this._id = className\n .replace(/([A-Z])/g, '_$1')\n .toLowerCase()\n .replace(/^_/, '');\n }\n }\n\n this._instructions = instructions;\n this._tools = { ...tools };\n this._chatCtx = chatCtx\n ? chatCtx.copy({\n toolCtx: this._tools,\n })\n : ChatContext.empty();\n\n this.turnDetection = turnDetection;\n this._vad = vad;\n\n if (typeof stt === 'string') {\n this._stt = InferenceSTT.fromModelString(stt);\n } else {\n this._stt = stt;\n }\n\n if (typeof llm === 'string') {\n this._llm = InferenceLLM.fromModelString(llm);\n } else {\n this._llm = llm;\n }\n\n if (typeof tts === 'string') {\n this._tts = InferenceTTS.fromModelString(tts);\n } else {\n this._tts = tts;\n }\n\n this._agentActivity = undefined;\n }\n\n get vad(): VAD | undefined {\n return this._vad;\n }\n\n get stt(): STT | undefined {\n return this._stt;\n }\n\n get llm(): LLM | RealtimeModel | undefined {\n return this._llm;\n }\n\n get tts(): TTS | undefined {\n return this._tts;\n }\n\n get chatCtx(): ReadonlyChatContext {\n return new ReadonlyChatContext(this._chatCtx.items);\n }\n\n get id(): string {\n return this._id;\n }\n\n get instructions(): string {\n return this._instructions;\n }\n\n get toolCtx(): ToolContext<UserData> {\n return { ...this._tools };\n }\n\n get session(): AgentSession<UserData> {\n return this.getActivityOrThrow().agentSession as AgentSession<UserData>;\n }\n\n async onEnter(): Promise<void> {}\n\n async onExit(): Promise<void> {}\n\n async transcriptionNode(\n text: ReadableStream<string>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<string> | null> {\n return Agent.default.transcriptionNode(this, text, modelSettings);\n }\n\n async onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void> {}\n\n async sttNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n return Agent.default.sttNode(this, audio, modelSettings);\n }\n\n async llmNode(\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n return Agent.default.llmNode(this, chatCtx, toolCtx, modelSettings);\n }\n\n async ttsNode(\n text: ReadableStream<string>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.ttsNode(this, text, modelSettings);\n }\n\n async realtimeAudioOutputNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.realtimeAudioOutputNode(this, audio, modelSettings);\n }\n\n // realtime_audio_output_node\n\n getActivityOrThrow(): AgentActivity {\n if (!this._agentActivity) {\n throw new Error('Agent activity not found');\n }\n return this._agentActivity;\n }\n\n async updateChatCtx(chatCtx: ChatContext): Promise<void> {\n if (!this._agentActivity) {\n this._chatCtx = chatCtx.copy({ toolCtx: this.toolCtx });\n return;\n }\n\n this._agentActivity.updateChatCtx(chatCtx);\n }\n\n static default = {\n async sttNode(\n agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.stt) {\n throw new Error('sttNode called but no STT node is available');\n }\n\n let wrapped_stt = activity.stt;\n\n if (!wrapped_stt.capabilities.streaming) {\n const vad = agent.vad || activity.vad;\n if (!vad) {\n throw new Error(\n 'STT does not support streaming, add a VAD to the AgentTask/VoiceAgent to enable streaming',\n );\n }\n wrapped_stt = new STTStreamAdapter(wrapped_stt, vad);\n }\n\n const connOptions = activity.agentSession.connOptions.sttConnOptions;\n const stream = wrapped_stt.stream({ connOptions });\n\n // Set startTimeOffset to provide linear timestamps across reconnections\n const audioInputStartedAt =\n activity.agentSession._recorderIO?.recordingStartedAt ?? // Use recording start time if available\n activity.agentSession._startedAt ?? // Fallback to session start time\n Date.now(); // Fallback to current time\n\n stream.startTimeOffset = (Date.now() - audioInputStartedAt) / 1000;\n\n stream.updateInputStream(audio);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.detachInputStream();\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n } finally {\n // Always clean up the STT stream, whether it ends naturally or is cancelled\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async llmNode(\n agent: Agent,\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.llm) {\n throw new Error('llmNode called but no LLM node is available');\n }\n\n if (!(activity.llm instanceof LLM)) {\n throw new Error(\n 'llmNode should only be used with LLM (non-multimodal/realtime APIs) nodes',\n );\n }\n\n const { toolChoice } = modelSettings;\n const connOptions = activity.agentSession.connOptions.llmConnOptions;\n\n // parallelToolCalls is not passed here - it will use the value from LLM's modelOptions\n // This allows users to configure it via: new inference.LLM({ modelOptions: { parallel_tool_calls: false } })\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n connOptions,\n });\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async ttsNode(\n agent: Agent,\n text: ReadableStream<string>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.tts) {\n throw new Error('ttsNode called but no TTS node is available');\n }\n\n let wrapped_tts = activity.tts;\n\n if (!activity.tts.capabilities.streaming) {\n wrapped_tts = new TTSStreamAdapter(wrapped_tts, new BasicSentenceTokenizer());\n }\n\n const connOptions = activity.agentSession.connOptions.ttsConnOptions;\n const stream = wrapped_tts.stream({ connOptions });\n stream.updateInputStream(text);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n if (chunk === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n controller.enqueue(chunk.frame);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async transcriptionNode(\n agent: Agent,\n text: ReadableStream<string>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<string> | null> {\n return text;\n },\n\n async realtimeAudioOutputNode(\n _agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return audio;\n },\n };\n}\n"],"mappings":"AAIA,SAAS,yBAAyB;AAClC,SAAS,sBAAsB;AAC/B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,2BAA2B;AAEpC;AAAA,EAEE;AAAA,EACA;AAAA,OAGK;AAEP,SAAS,iBAAiB,wBAAwB;AAClD,SAAS,qBAAqB,8BAA8B;AAE5D,SAAS,kBAAkB,iBAAiB,wBAAwB;AAK7D,MAAM,oBAAoB,IAAI,kBAAmD;AACjF,MAAM,uBAAuB,OAAO,cAAc;AAElD,MAAM,qBAAqB,MAAM;AAAA,EACtC,cAAc;AACZ,UAAM;AACN,SAAK,OAAO;AAEZ,WAAO,eAAe,MAAM,sBAAsB;AAAA,MAChD,OAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AAEO,SAAS,eAAe,OAAuC;AACpE,SACE,UAAU,UACV,UAAU,QACV,OAAO,UAAU,YACjB,wBAAwB;AAE5B;AAqBO,MAAM,MAAsB;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAA2B;AACzB,QAAI,IAAI;AACN,WAAK,MAAM;AAAA,IACb,OAAO;AAEL,YAAM,YAAY,KAAK,YAAY;AACnC,UAAI,cAAc,SAAS;AACzB,aAAK,MAAM;AAAA,MACb,OAAO;AACL,aAAK,MAAM,UACR,QAAQ,YAAY,KAAK,EACzB,YAAY,EACZ,QAAQ,MAAM,EAAE;AAAA,MACrB;AAAA,IACF;AAEA,SAAK,gBAAgB;AACrB,SAAK,SAAS,EAAE,GAAG,MAAM;AACzB,SAAK,WAAW,UACZ,QAAQ,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,IAChB,CAAC,IACD,YAAY,MAAM;AAEtB,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAA+B;AACjC,WAAO,IAAI,oBAAoB,KAAK,SAAS,KAAK;AAAA,EACpD;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAiC;AACnC,WAAO,EAAE,GAAG,KAAK,OAAO;AAAA,EAC1B;AAAA,EAEA,IAAI,UAAkC;AACpC,WAAO,KAAK,mBAAmB,EAAE;AAAA,EACnC;AAAA,EAEA,MAAM,UAAyB;AAAA,EAAC;AAAA,EAEhC,MAAM,SAAwB;AAAA,EAAC;AAAA,EAE/B,MAAM,kBACJ,MACA,eACwC;AACxC,WAAO,MAAM,QAAQ,kBAAkB,MAAM,MAAM,aAAa;AAAA,EAClE;AAAA,EAEA,MAAM,oBAAoB,UAAuB,aAAyC;AAAA,EAAC;AAAA,EAE3F,MAAM,QACJ,OACA,eACsD;AACtD,WAAO,MAAM,QAAQ,QAAQ,MAAM,OAAO,aAAa;AAAA,EACzD;AAAA,EAEA,MAAM,QACJ,SACA,SACA,eACoD;AACpD,WAAO,MAAM,QAAQ,QAAQ,MAAM,SAAS,SAAS,aAAa;AAAA,EACpE;AAAA,EAEA,MAAM,QACJ,MACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,QAAQ,MAAM,MAAM,aAAa;AAAA,EACxD;AAAA,EAEA,MAAM,wBACJ,OACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,wBAAwB,MAAM,OAAO,aAAa;AAAA,EACzE;AAAA;AAAA,EAIA,qBAAoC;AAClC,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI,MAAM,0BAA0B;AAAA,IAC5C;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,cAAc,SAAqC;AACvD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,WAAW,QAAQ,KAAK,EAAE,SAAS,KAAK,QAAQ,CAAC;AACtD;AAAA,IACF;AAEA,SAAK,eAAe,cAAc,OAAO;AAAA,EAC3C;AAAA,EAEA,OAAO,UAAU;AAAA,IACf,MAAM,QACJ,OACA,OACA,gBACsD;AA7P5D;AA8PM,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,YAAY,aAAa,WAAW;AACvC,cAAM,MAAM,MAAM,OAAO,SAAS;AAClC,YAAI,CAAC,KAAK;AACR,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,iBAAiB,aAAa,GAAG;AAAA,MACrD;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AAGjD,YAAM,wBACJ,cAAS,aAAa,gBAAtB,mBAAmC;AAAA,MACnC,SAAS,aAAa;AAAA,MACtB,KAAK,IAAI;AAEX,aAAO,mBAAmB,KAAK,IAAI,IAAI,uBAAuB;AAE9D,aAAO,kBAAkB,KAAK;AAE9B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,kBAAkB;AACzB,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AAEA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,SACA,SACA,eACoD;AACpD,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,EAAE,SAAS,eAAe,MAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAEA,YAAM,EAAE,WAAW,IAAI;AACvB,YAAM,cAAc,SAAS,aAAa,YAAY;AAItD,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC;AAED,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,MACA,gBAC4C;AAC5C,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,SAAS,IAAI,aAAa,WAAW;AACxC,sBAAc,IAAI,iBAAiB,aAAa,IAAI,uBAAuB,CAAC;AAAA,MAC9E;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AACjD,aAAO,kBAAkB,IAAI;AAE7B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,kBAAI,UAAU,iBAAiB,eAAe;AAC5C;AAAA,cACF;AACA,yBAAW,QAAQ,MAAM,KAAK;AAAA,YAChC;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,kBACJ,OACA,MACA,gBACwC;AACxC,aAAO;AAAA,IACT;AAAA,IAEA,MAAM,wBACJ,QACA,OACA,gBAC4C;AAC5C,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../../src/voice/agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { AsyncLocalStorage } from 'node:async_hooks';\nimport { ReadableStream } from 'node:stream/web';\nimport {\n LLM as InferenceLLM,\n STT as InferenceSTT,\n TTS as InferenceTTS,\n type LLMModels,\n type STTModelString,\n type TTSModelString,\n} from '../inference/index.js';\nimport { ReadonlyChatContext } from '../llm/chat_context.js';\nimport type { ChatMessage, FunctionCall, RealtimeModel } from '../llm/index.js';\nimport {\n type ChatChunk,\n ChatContext,\n LLM,\n type ToolChoice,\n type ToolContext,\n} from '../llm/index.js';\nimport type { STT, SpeechEvent } from '../stt/index.js';\nimport { StreamAdapter as STTStreamAdapter } from '../stt/index.js';\nimport { SentenceTokenizer as BasicSentenceTokenizer } from '../tokenize/basic/index.js';\nimport type { TTS } from '../tts/index.js';\nimport { SynthesizeStream, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport { USERDATA_TIMED_TRANSCRIPT } from '../types.js';\nimport type { VAD } from '../vad.js';\nimport type { AgentActivity } from './agent_activity.js';\nimport type { AgentSession, TurnDetectionMode } from './agent_session.js';\nimport type { TimedString } from './io.js';\n\nexport const asyncLocalStorage = new AsyncLocalStorage<{ functionCall?: FunctionCall }>();\nexport const STOP_RESPONSE_SYMBOL = Symbol('StopResponse');\n\nexport class StopResponse extends Error {\n constructor() {\n super();\n this.name = 'StopResponse';\n\n Object.defineProperty(this, STOP_RESPONSE_SYMBOL, {\n value: true,\n });\n }\n}\n\nexport function isStopResponse(value: unknown): value is StopResponse {\n return (\n value !== undefined &&\n value !== null &&\n typeof value === 'object' &&\n STOP_RESPONSE_SYMBOL in value\n );\n}\n\nexport interface ModelSettings {\n /** The tool choice to use when calling the LLM. */\n toolChoice?: ToolChoice;\n}\n\nexport interface AgentOptions<UserData> {\n id?: string;\n instructions: string;\n chatCtx?: ChatContext;\n tools?: ToolContext<UserData>;\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n allowInterruptions?: boolean;\n minConsecutiveSpeechDelay?: number;\n useTtsAlignedTranscript?: boolean;\n}\n\nexport class Agent<UserData = any> {\n private _id: string;\n private turnDetection?: TurnDetectionMode;\n private _stt?: STT;\n private _vad?: VAD;\n private _llm?: LLM | RealtimeModel;\n private _tts?: TTS;\n private _useTtsAlignedTranscript?: boolean;\n\n /** @internal */\n _agentActivity?: AgentActivity;\n\n /** @internal */\n _chatCtx: ChatContext;\n\n /** @internal */\n _instructions: string;\n\n /** @internal */\n _tools?: ToolContext<UserData>;\n\n constructor({\n id,\n instructions,\n chatCtx,\n tools,\n turnDetection,\n stt,\n vad,\n llm,\n tts,\n useTtsAlignedTranscript,\n }: AgentOptions<UserData>) {\n if (id) {\n this._id = id;\n } else {\n // Convert class name to snake_case\n const className = this.constructor.name;\n if (className === 'Agent') {\n this._id = 'default_agent';\n } else {\n this._id = className\n .replace(/([A-Z])/g, '_$1')\n .toLowerCase()\n .replace(/^_/, '');\n }\n }\n\n this._instructions = instructions;\n this._tools = { ...tools };\n this._chatCtx = chatCtx\n ? chatCtx.copy({\n toolCtx: this._tools,\n })\n : ChatContext.empty();\n\n this.turnDetection = turnDetection;\n this._vad = vad;\n\n if (typeof stt === 'string') {\n this._stt = InferenceSTT.fromModelString(stt);\n } else {\n this._stt = stt;\n }\n\n if (typeof llm === 'string') {\n this._llm = InferenceLLM.fromModelString(llm);\n } else {\n this._llm = llm;\n }\n\n if (typeof tts === 'string') {\n this._tts = InferenceTTS.fromModelString(tts);\n } else {\n this._tts = tts;\n }\n\n this._useTtsAlignedTranscript = useTtsAlignedTranscript;\n\n this._agentActivity = undefined;\n }\n\n get vad(): VAD | undefined {\n return this._vad;\n }\n\n get stt(): STT | undefined {\n return this._stt;\n }\n\n get llm(): LLM | RealtimeModel | undefined {\n return this._llm;\n }\n\n get tts(): TTS | undefined {\n return this._tts;\n }\n\n get useTtsAlignedTranscript(): boolean | undefined {\n return this._useTtsAlignedTranscript;\n }\n\n get chatCtx(): ReadonlyChatContext {\n return new ReadonlyChatContext(this._chatCtx.items);\n }\n\n get id(): string {\n return this._id;\n }\n\n get instructions(): string {\n return this._instructions;\n }\n\n get toolCtx(): ToolContext<UserData> {\n return { ...this._tools };\n }\n\n get session(): AgentSession<UserData> {\n return this.getActivityOrThrow().agentSession as AgentSession<UserData>;\n }\n\n async onEnter(): Promise<void> {}\n\n async onExit(): Promise<void> {}\n\n async transcriptionNode(\n text: ReadableStream<string | TimedString>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<string | TimedString> | null> {\n return Agent.default.transcriptionNode(this, text, modelSettings);\n }\n\n async onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void> {}\n\n async sttNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n return Agent.default.sttNode(this, audio, modelSettings);\n }\n\n async llmNode(\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n return Agent.default.llmNode(this, chatCtx, toolCtx, modelSettings);\n }\n\n async ttsNode(\n text: ReadableStream<string>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.ttsNode(this, text, modelSettings);\n }\n\n async realtimeAudioOutputNode(\n audio: ReadableStream<AudioFrame>,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return Agent.default.realtimeAudioOutputNode(this, audio, modelSettings);\n }\n\n // realtime_audio_output_node\n\n getActivityOrThrow(): AgentActivity {\n if (!this._agentActivity) {\n throw new Error('Agent activity not found');\n }\n return this._agentActivity;\n }\n\n async updateChatCtx(chatCtx: ChatContext): Promise<void> {\n if (!this._agentActivity) {\n this._chatCtx = chatCtx.copy({ toolCtx: this.toolCtx });\n return;\n }\n\n this._agentActivity.updateChatCtx(chatCtx);\n }\n\n static default = {\n async sttNode(\n agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<SpeechEvent | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.stt) {\n throw new Error('sttNode called but no STT node is available');\n }\n\n let wrapped_stt = activity.stt;\n\n if (!wrapped_stt.capabilities.streaming) {\n const vad = agent.vad || activity.vad;\n if (!vad) {\n throw new Error(\n 'STT does not support streaming, add a VAD to the AgentTask/VoiceAgent to enable streaming',\n );\n }\n wrapped_stt = new STTStreamAdapter(wrapped_stt, vad);\n }\n\n const connOptions = activity.agentSession.connOptions.sttConnOptions;\n const stream = wrapped_stt.stream({ connOptions });\n\n // Set startTimeOffset to provide linear timestamps across reconnections\n const audioInputStartedAt =\n activity.agentSession._recorderIO?.recordingStartedAt ?? // Use recording start time if available\n activity.agentSession._startedAt ?? // Fallback to session start time\n Date.now(); // Fallback to current time\n\n stream.startTimeOffset = (Date.now() - audioInputStartedAt) / 1000;\n\n stream.updateInputStream(audio);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.detachInputStream();\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const event of stream) {\n controller.enqueue(event);\n }\n controller.close();\n } finally {\n // Always clean up the STT stream, whether it ends naturally or is cancelled\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async llmNode(\n agent: Agent,\n chatCtx: ChatContext,\n toolCtx: ToolContext,\n modelSettings: ModelSettings,\n ): Promise<ReadableStream<ChatChunk | string> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.llm) {\n throw new Error('llmNode called but no LLM node is available');\n }\n\n if (!(activity.llm instanceof LLM)) {\n throw new Error(\n 'llmNode should only be used with LLM (non-multimodal/realtime APIs) nodes',\n );\n }\n\n const { toolChoice } = modelSettings;\n const connOptions = activity.agentSession.connOptions.llmConnOptions;\n\n // parallelToolCalls is not passed here - it will use the value from LLM's modelOptions\n // This allows users to configure it via: new inference.LLM({ modelOptions: { parallel_tool_calls: false } })\n const stream = activity.llm.chat({\n chatCtx,\n toolCtx,\n toolChoice,\n connOptions,\n });\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n controller.enqueue(chunk);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async ttsNode(\n agent: Agent,\n text: ReadableStream<string>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n const activity = agent.getActivityOrThrow();\n if (!activity.tts) {\n throw new Error('ttsNode called but no TTS node is available');\n }\n\n let wrapped_tts = activity.tts;\n\n if (!activity.tts.capabilities.streaming) {\n wrapped_tts = new TTSStreamAdapter(wrapped_tts, new BasicSentenceTokenizer());\n }\n\n const connOptions = activity.agentSession.connOptions.ttsConnOptions;\n const stream = wrapped_tts.stream({ connOptions });\n stream.updateInputStream(text);\n\n let cleaned = false;\n const cleanup = () => {\n if (cleaned) return;\n cleaned = true;\n stream.close();\n };\n\n return new ReadableStream({\n async start(controller) {\n try {\n for await (const chunk of stream) {\n if (chunk === SynthesizeStream.END_OF_STREAM) {\n break;\n }\n // Attach timed transcripts to frame.userdata\n if (chunk.timedTranscripts && chunk.timedTranscripts.length > 0) {\n chunk.frame.userdata[USERDATA_TIMED_TRANSCRIPT] = chunk.timedTranscripts;\n }\n controller.enqueue(chunk.frame);\n }\n controller.close();\n } finally {\n cleanup();\n }\n },\n cancel() {\n cleanup();\n },\n });\n },\n\n async transcriptionNode(\n agent: Agent,\n text: ReadableStream<string | TimedString>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<string | TimedString> | null> {\n return text;\n },\n\n async realtimeAudioOutputNode(\n _agent: Agent,\n audio: ReadableStream<AudioFrame>,\n _modelSettings: ModelSettings,\n ): Promise<ReadableStream<AudioFrame> | null> {\n return audio;\n },\n };\n}\n"],"mappings":"AAIA,SAAS,yBAAyB;AAClC,SAAS,sBAAsB;AAC/B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,2BAA2B;AAEpC;AAAA,EAEE;AAAA,EACA;AAAA,OAGK;AAEP,SAAS,iBAAiB,wBAAwB;AAClD,SAAS,qBAAqB,8BAA8B;AAE5D,SAAS,kBAAkB,iBAAiB,wBAAwB;AACpE,SAAS,iCAAiC;AAMnC,MAAM,oBAAoB,IAAI,kBAAmD;AACjF,MAAM,uBAAuB,OAAO,cAAc;AAElD,MAAM,qBAAqB,MAAM;AAAA,EACtC,cAAc;AACZ,UAAM;AACN,SAAK,OAAO;AAEZ,WAAO,eAAe,MAAM,sBAAsB;AAAA,MAChD,OAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AAEO,SAAS,eAAe,OAAuC;AACpE,SACE,UAAU,UACV,UAAU,QACV,OAAO,UAAU,YACjB,wBAAwB;AAE5B;AAsBO,MAAM,MAAsB;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAA2B;AACzB,QAAI,IAAI;AACN,WAAK,MAAM;AAAA,IACb,OAAO;AAEL,YAAM,YAAY,KAAK,YAAY;AACnC,UAAI,cAAc,SAAS;AACzB,aAAK,MAAM;AAAA,MACb,OAAO;AACL,aAAK,MAAM,UACR,QAAQ,YAAY,KAAK,EACzB,YAAY,EACZ,QAAQ,MAAM,EAAE;AAAA,MACrB;AAAA,IACF;AAEA,SAAK,gBAAgB;AACrB,SAAK,SAAS,EAAE,GAAG,MAAM;AACzB,SAAK,WAAW,UACZ,QAAQ,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,IAChB,CAAC,IACD,YAAY,MAAM;AAEtB,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,SAAK,2BAA2B;AAEhC,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,0BAA+C;AACjD,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAA+B;AACjC,WAAO,IAAI,oBAAoB,KAAK,SAAS,KAAK;AAAA,EACpD;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAiC;AACnC,WAAO,EAAE,GAAG,KAAK,OAAO;AAAA,EAC1B;AAAA,EAEA,IAAI,UAAkC;AACpC,WAAO,KAAK,mBAAmB,EAAE;AAAA,EACnC;AAAA,EAEA,MAAM,UAAyB;AAAA,EAAC;AAAA,EAEhC,MAAM,SAAwB;AAAA,EAAC;AAAA,EAE/B,MAAM,kBACJ,MACA,eACsD;AACtD,WAAO,MAAM,QAAQ,kBAAkB,MAAM,MAAM,aAAa;AAAA,EAClE;AAAA,EAEA,MAAM,oBAAoB,UAAuB,aAAyC;AAAA,EAAC;AAAA,EAE3F,MAAM,QACJ,OACA,eACsD;AACtD,WAAO,MAAM,QAAQ,QAAQ,MAAM,OAAO,aAAa;AAAA,EACzD;AAAA,EAEA,MAAM,QACJ,SACA,SACA,eACoD;AACpD,WAAO,MAAM,QAAQ,QAAQ,MAAM,SAAS,SAAS,aAAa;AAAA,EACpE;AAAA,EAEA,MAAM,QACJ,MACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,QAAQ,MAAM,MAAM,aAAa;AAAA,EACxD;AAAA,EAEA,MAAM,wBACJ,OACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,wBAAwB,MAAM,OAAO,aAAa;AAAA,EACzE;AAAA;AAAA,EAIA,qBAAoC;AAClC,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI,MAAM,0BAA0B;AAAA,IAC5C;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,cAAc,SAAqC;AACvD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,WAAW,QAAQ,KAAK,EAAE,SAAS,KAAK,QAAQ,CAAC;AACtD;AAAA,IACF;AAEA,SAAK,eAAe,cAAc,OAAO;AAAA,EAC3C;AAAA,EAEA,OAAO,UAAU;AAAA,IACf,MAAM,QACJ,OACA,OACA,gBACsD;AAxQ5D;AAyQM,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,YAAY,aAAa,WAAW;AACvC,cAAM,MAAM,MAAM,OAAO,SAAS;AAClC,YAAI,CAAC,KAAK;AACR,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,iBAAiB,aAAa,GAAG;AAAA,MACrD;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AAGjD,YAAM,wBACJ,cAAS,aAAa,gBAAtB,mBAAmC;AAAA,MACnC,SAAS,aAAa;AAAA,MACtB,KAAK,IAAI;AAEX,aAAO,mBAAmB,KAAK,IAAI,IAAI,uBAAuB;AAE9D,aAAO,kBAAkB,KAAK;AAE9B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,kBAAkB;AACzB,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AAEA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,SACA,SACA,eACoD;AACpD,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,EAAE,SAAS,eAAe,MAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAEA,YAAM,EAAE,WAAW,IAAI;AACvB,YAAM,cAAc,SAAS,aAAa,YAAY;AAItD,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC;AAED,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,MACA,gBAC4C;AAC5C,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,SAAS,IAAI,aAAa,WAAW;AACxC,sBAAc,IAAI,iBAAiB,aAAa,IAAI,uBAAuB,CAAC;AAAA,MAC9E;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AACjD,aAAO,kBAAkB,IAAI;AAE7B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,kBAAI,UAAU,iBAAiB,eAAe;AAC5C;AAAA,cACF;AAEA,kBAAI,MAAM,oBAAoB,MAAM,iBAAiB,SAAS,GAAG;AAC/D,sBAAM,MAAM,SAAS,yBAAyB,IAAI,MAAM;AAAA,cAC1D;AACA,yBAAW,QAAQ,MAAM,KAAK;AAAA,YAChC;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,kBACJ,OACA,MACA,gBACsD;AACtD,aAAO;AAAA,IACT;AAAA,IAEA,MAAM,wBACJ,QACA,OACA,gBAC4C;AAC5C,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":[]}
|
|
@@ -251,6 +251,9 @@ class AgentActivity {
|
|
|
251
251
|
get allowInterruptions() {
|
|
252
252
|
return this.agentSession.options.allowInterruptions;
|
|
253
253
|
}
|
|
254
|
+
get useTtsAlignedTranscript() {
|
|
255
|
+
return this.agent.useTtsAlignedTranscript ?? this.agentSession.useTtsAlignedTranscript;
|
|
256
|
+
}
|
|
254
257
|
get turnDetection() {
|
|
255
258
|
return this.agentSession.turnDetection;
|
|
256
259
|
}
|
|
@@ -896,7 +899,7 @@ ${instructions}` : instructions,
|
|
|
896
899
|
} else {
|
|
897
900
|
let audioOut = null;
|
|
898
901
|
if (!audio) {
|
|
899
|
-
const [ttsTask,
|
|
902
|
+
const [ttsTask, ttsGenData] = (0, import_generation.performTTSInference)(
|
|
900
903
|
(...args) => this.agent.ttsNode(...args),
|
|
901
904
|
audioSource,
|
|
902
905
|
modelSettings,
|
|
@@ -904,7 +907,7 @@ ${instructions}` : instructions,
|
|
|
904
907
|
);
|
|
905
908
|
tasks.push(ttsTask);
|
|
906
909
|
const [forwardTask, _audioOut] = (0, import_generation.performAudioForwarding)(
|
|
907
|
-
|
|
910
|
+
ttsGenData.audioStream,
|
|
908
911
|
audioOutput,
|
|
909
912
|
replyAbortController
|
|
910
913
|
);
|
|
@@ -957,7 +960,7 @@ ${instructions}` : instructions,
|
|
|
957
960
|
toolsMessages,
|
|
958
961
|
span
|
|
959
962
|
}) => {
|
|
960
|
-
var _a, _b, _c;
|
|
963
|
+
var _a, _b, _c, _d;
|
|
961
964
|
speechHandle._agentTurnContext = import_api.context.active();
|
|
962
965
|
span.setAttribute(import_telemetry.traceTypes.ATTR_SPEECH_ID, speechHandle.id);
|
|
963
966
|
if (instructions) {
|
|
@@ -995,12 +998,12 @@ ${instructions}` : instructions,
|
|
|
995
998
|
);
|
|
996
999
|
tasks.push(llmTask);
|
|
997
1000
|
let ttsTask = null;
|
|
998
|
-
let
|
|
1001
|
+
let ttsGenData = null;
|
|
999
1002
|
let llmOutput;
|
|
1000
1003
|
if (audioOutput) {
|
|
1001
1004
|
const [ttsTextInput, textOutput] = llmGenData.textStream.tee();
|
|
1002
1005
|
llmOutput = textOutput;
|
|
1003
|
-
[ttsTask,
|
|
1006
|
+
[ttsTask, ttsGenData] = (0, import_generation.performTTSInference)(
|
|
1004
1007
|
(...args) => this.agent.ttsNode(...args),
|
|
1005
1008
|
ttsTextInput,
|
|
1006
1009
|
modelSettings,
|
|
@@ -1024,7 +1027,20 @@ ${instructions}` : instructions,
|
|
|
1024
1027
|
await speechHandle.waitIfNotInterrupted([speechHandle._waitForAuthorization()]);
|
|
1025
1028
|
speechHandle._clearAuthorization();
|
|
1026
1029
|
const replyStartedAt = Date.now();
|
|
1027
|
-
|
|
1030
|
+
let transcriptionInput = llmOutput;
|
|
1031
|
+
if (this.useTtsAlignedTranscript && ((_a = this.tts) == null ? void 0 : _a.capabilities.alignedTranscript) && ttsGenData) {
|
|
1032
|
+
const timedTextsStream = await Promise.race([
|
|
1033
|
+
ttsGenData.timedTextsFut.await,
|
|
1034
|
+
(ttsTask == null ? void 0 : ttsTask.result.catch(
|
|
1035
|
+
() => this.logger.warn("TTS task failed before resolving timedTextsFut")
|
|
1036
|
+
)) ?? Promise.resolve()
|
|
1037
|
+
]);
|
|
1038
|
+
if (timedTextsStream) {
|
|
1039
|
+
this.logger.debug("Using TTS aligned transcripts for transcription node input");
|
|
1040
|
+
transcriptionInput = timedTextsStream;
|
|
1041
|
+
}
|
|
1042
|
+
}
|
|
1043
|
+
const trNodeResult = await this.agent.transcriptionNode(transcriptionInput, modelSettings);
|
|
1028
1044
|
let textOut = null;
|
|
1029
1045
|
if (trNodeResult) {
|
|
1030
1046
|
const [textForwardTask, _textOut] = (0, import_generation.performTextForwarding)(
|
|
@@ -1043,9 +1059,9 @@ ${instructions}` : instructions,
|
|
|
1043
1059
|
};
|
|
1044
1060
|
let audioOut = null;
|
|
1045
1061
|
if (audioOutput) {
|
|
1046
|
-
if (
|
|
1062
|
+
if (ttsGenData) {
|
|
1047
1063
|
const [forwardTask, _audioOut] = (0, import_generation.performAudioForwarding)(
|
|
1048
|
-
|
|
1064
|
+
ttsGenData.audioStream,
|
|
1049
1065
|
audioOutput,
|
|
1050
1066
|
replyAbortController
|
|
1051
1067
|
);
|
|
@@ -1053,7 +1069,7 @@ ${instructions}` : instructions,
|
|
|
1053
1069
|
tasks.push(forwardTask);
|
|
1054
1070
|
audioOut.firstFrameFut.await.then((ts) => onFirstFrame(ts)).catch(() => this.logger.debug("firstFrameFut cancelled before first frame"));
|
|
1055
1071
|
} else {
|
|
1056
|
-
throw Error("
|
|
1072
|
+
throw Error("ttsGenData is null when audioOutput is enabled");
|
|
1057
1073
|
}
|
|
1058
1074
|
} else {
|
|
1059
1075
|
textOut == null ? void 0 : textOut.firstTextFut.await.then(() => onFirstFrame()).catch(() => this.logger.debug("firstTextFut cancelled before first frame"));
|
|
@@ -1201,10 +1217,10 @@ ${instructions}` : instructions,
|
|
|
1201
1217
|
this.logger.debug(
|
|
1202
1218
|
{
|
|
1203
1219
|
speechId: speechHandle.id,
|
|
1204
|
-
name: (
|
|
1220
|
+
name: (_b = sanitizedOut.toolCall) == null ? void 0 : _b.name,
|
|
1205
1221
|
args: sanitizedOut.toolCall.args,
|
|
1206
|
-
output: (
|
|
1207
|
-
isError: (
|
|
1222
|
+
output: (_c = sanitizedOut.toolCallOutput) == null ? void 0 : _c.output,
|
|
1223
|
+
isError: (_d = sanitizedOut.toolCallOutput) == null ? void 0 : _d.isError
|
|
1208
1224
|
},
|
|
1209
1225
|
"Tool call execution finished"
|
|
1210
1226
|
);
|
|
@@ -1371,14 +1387,14 @@ ${instructions}` : instructions,
|
|
|
1371
1387
|
if (audioOutput) {
|
|
1372
1388
|
let realtimeAudioResult = null;
|
|
1373
1389
|
if (ttsTextInput) {
|
|
1374
|
-
const [ttsTask,
|
|
1390
|
+
const [ttsTask, ttsGenData] = (0, import_generation.performTTSInference)(
|
|
1375
1391
|
(...args) => this.agent.ttsNode(...args),
|
|
1376
1392
|
ttsTextInput,
|
|
1377
1393
|
modelSettings,
|
|
1378
1394
|
abortController
|
|
1379
1395
|
);
|
|
1380
1396
|
tasks.push(ttsTask);
|
|
1381
|
-
realtimeAudioResult =
|
|
1397
|
+
realtimeAudioResult = ttsGenData.audioStream;
|
|
1382
1398
|
} else if (msgModalities && msgModalities.includes("audio")) {
|
|
1383
1399
|
realtimeAudioResult = await this.agent.realtimeAudioOutputNode(
|
|
1384
1400
|
msg.audioStream,
|