@livekit/agents 0.6.0 → 0.6.2
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/job.cjs.map +1 -1
- package/dist/job.js.map +1 -1
- package/dist/multimodal/multimodal_agent.cjs +27 -3
- package/dist/multimodal/multimodal_agent.cjs.map +1 -1
- package/dist/multimodal/multimodal_agent.d.ts +3 -1
- package/dist/multimodal/multimodal_agent.d.ts.map +1 -1
- package/dist/multimodal/multimodal_agent.js +27 -3
- package/dist/multimodal/multimodal_agent.js.map +1 -1
- package/dist/pipeline/pipeline_agent.cjs +19 -7
- package/dist/pipeline/pipeline_agent.cjs.map +1 -1
- package/dist/pipeline/pipeline_agent.js +19 -7
- package/dist/pipeline/pipeline_agent.js.map +1 -1
- package/dist/utils.cjs +11 -4
- package/dist/utils.cjs.map +1 -1
- package/dist/utils.d.ts.map +1 -1
- package/dist/utils.js +11 -4
- package/dist/utils.js.map +1 -1
- package/dist/worker.cjs +5 -2
- package/dist/worker.cjs.map +1 -1
- package/dist/worker.d.ts.map +1 -1
- package/dist/worker.js +5 -2
- package/dist/worker.js.map +1 -1
- package/package.json +3 -3
- package/src/job.ts +3 -3
- package/src/multimodal/multimodal_agent.ts +45 -6
- package/src/pipeline/pipeline_agent.ts +19 -13
- package/src/utils.ts +12 -4
- package/src/worker.ts +7 -3
package/dist/job.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/job.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type * as proto from '@livekit/protocol';\nimport type {\n E2EEOptions,\n LocalParticipant,\n RemoteParticipant,\n Room,\n RtcConfiguration,\n} from '@livekit/rtc-node';\nimport { ParticipantKind, RoomEvent, TrackKind } from '@livekit/rtc-node';\nimport type { Logger } from 'pino';\nimport { log } from './log.js';\n\n/** Which tracks, if any, should the agent automatically subscribe to? */\nexport enum AutoSubscribe {\n SUBSCRIBE_ALL,\n SUBSCRIBE_NONE,\n VIDEO_ONLY,\n AUDIO_ONLY,\n}\n\nexport type JobAcceptArguments = {\n name: string;\n identity: string;\n metadata: string;\n attributes?: { [key: string]: string };\n};\n\nexport type RunningJobInfo = {\n acceptArguments: JobAcceptArguments;\n job: proto.Job;\n url: string;\n token: string;\n};\n\n/** Attempted to add a function callback, but the function already exists. */\nexport class FunctionExistsError extends Error {\n constructor(msg?: string) {\n super(msg);\n Object.setPrototypeOf(this, new.target.prototype);\n }\n}\n\n/** The job and environment context as seen by the agent, accessible by the entrypoint function. */\nexport class JobContext {\n #proc: JobProcess;\n #info: RunningJobInfo;\n #room: Room;\n #onConnect: () => void;\n #onShutdown: (s: string) => void;\n /** @internal */\n shutdownCallbacks: (() => Promise<void>)[] = [];\n #participantEntrypoints: ((job: JobContext, p: RemoteParticipant) => Promise<void>)[] = [];\n #participantTasks: {\n [id: string]: {\n callback: (job: JobContext, p: RemoteParticipant) => Promise<void>;\n result: Promise<void>;\n };\n } = {};\n #logger: Logger;\n\n constructor(\n proc: JobProcess,\n info: RunningJobInfo,\n room: Room,\n onConnect: () => void,\n onShutdown: (s: string) => void,\n ) {\n this.#proc = proc;\n this.#info = info;\n this.#room = room;\n this.#onConnect = onConnect;\n this.#onShutdown = onShutdown;\n this.onParticipantConnected = this.onParticipantConnected.bind(this);\n this.#room.on(RoomEvent.ParticipantConnected, this.onParticipantConnected);\n this.#logger = log().child({ info: this.#info });\n }\n\n get proc(): JobProcess {\n return this.#proc;\n }\n\n get job(): proto.Job {\n return this.#info.job;\n }\n\n /** @returns The room the agent was called into */\n get room(): Room {\n return this.#room;\n }\n\n /** @returns The agent's participant if connected to the room, otherwise `undefined` */\n get agent(): LocalParticipant | undefined {\n return this.#room.localParticipant;\n }\n\n /** Adds a promise to be awaited when {@link JobContext.shutdown | shutdown} is called. */\n addShutdownCallback(callback: () => Promise<void>) {\n this.shutdownCallbacks.push(callback);\n }\n\n async waitForParticipant(identity?: string): Promise<RemoteParticipant> {\n if (!this.#room.isConnected) {\n throw new Error('room is not connected');\n }\n\n for (const p of this.#room.remoteParticipants.values()) {\n if ((!identity || p.identity === identity) && p.info.kind != ParticipantKind.AGENT) {\n return p;\n }\n }\n\n return new Promise((resolve, reject) => {\n const onParticipantConnected = (participant: RemoteParticipant) => {\n if (\n (!identity || participant.identity === identity) &&\n participant.info.kind != ParticipantKind.AGENT\n ) {\n clearHandlers();\n resolve(participant);\n }\n };\n const onDisconnected = () => {\n clearHandlers();\n reject(new Error('Room disconnected while waiting for participant'));\n };\n\n const clearHandlers = () => {\n this.#room.off(RoomEvent.ParticipantConnected, onParticipantConnected);\n this.#room.off(RoomEvent.Disconnected, onDisconnected);\n };\n\n this.#room.on(RoomEvent.ParticipantConnected, onParticipantConnected);\n this.#room.on(RoomEvent.Disconnected, onDisconnected);\n });\n }\n\n /**\n * Connects the agent to the room.\n *\n * @remarks\n * It is recommended to run this command as early in the function as possible, as executing it\n * later may cause noticeable delay between user and agent joins.\n *\n * @see {@link https://github.com/livekit/node-sdks/tree/main/packages/livekit-rtc#readme |\n * @livekit/rtc-node} for more information about the parameters.\n */\n async connect(\n e2ee?: E2EEOptions,\n autoSubscribe: AutoSubscribe = AutoSubscribe.SUBSCRIBE_ALL,\n rtcConfig?: RtcConfiguration,\n ) {\n const opts = {\n e2ee,\n autoSubscribe: autoSubscribe == AutoSubscribe.SUBSCRIBE_ALL,\n rtcConfig,\n dynacast: false,\n };\n\n await this.#room.connect(this.#info.url, this.#info.token, opts);\n this.#onConnect();\n\n this.#room.remoteParticipants.forEach(this.onParticipantConnected);\n\n if ([AutoSubscribe.AUDIO_ONLY, AutoSubscribe.VIDEO_ONLY].includes(autoSubscribe)) {\n this.#room.remoteParticipants.forEach((p) => {\n p.trackPublications.forEach((pub) => {\n if (\n (autoSubscribe === AutoSubscribe.AUDIO_ONLY && pub.kind === TrackKind.KIND_AUDIO) ||\n (autoSubscribe === AutoSubscribe.VIDEO_ONLY && pub.kind === TrackKind.KIND_VIDEO)\n ) {\n pub.setSubscribed(true);\n }\n });\n });\n }\n }\n\n /**\n * Gracefully shuts down the job, and runs all shutdown promises.\n *\n * @param reason - Optional reason for shutdown\n */\n shutdown(reason = '') {\n this.#onShutdown(reason);\n }\n\n /** @internal */\n onParticipantConnected(p: RemoteParticipant) {\n for (const callback of this.#participantEntrypoints) {\n if (this.#participantTasks[p.identity]?.callback == callback) {\n this.#logger.warn(\n 'a participant has joined before a prior prticipant task matching the same identity has finished:',\n p.identity,\n );\n }\n const result = callback(this, p);\n result.finally(() => delete this.#participantTasks[p.identity]);\n this.#participantTasks[p.identity] = { callback, result };\n }\n }\n\n /**\n * Adds a promise to be awaited whenever a new participant joins the room.\n *\n * @throws {@link FunctionExistsError} if an entrypoint already exists\n */\n addParticipantEntrypoint(callback: (job: JobContext, p: RemoteParticipant) => Promise<void>) {\n if (this.#participantEntrypoints.includes(callback)) {\n throw new FunctionExistsError('entrypoints cannot be added more than once');\n }\n\n this.#participantEntrypoints.push(callback);\n }\n}\n\nexport class JobProcess {\n #pid = process.pid;\n userData: { [id: string]: unknown } = {};\n\n get pid(): number {\n return this.#pid;\n }\n}\n\n/**\n * A request sent by the server to spawn a new agent job.\n *\n * @remarks\n * For most applications, this is best left to the default, which simply accepts the job and\n * handles the logic inside the entrypoint function. This class is useful for vetting which\n * requests should fill idle processes and which should be outright rejected.\n */\nexport class JobRequest {\n #job: proto.Job;\n #onReject: () => Promise<void>;\n #onAccept: (args: JobAcceptArguments) => Promise<void>;\n\n /** @internal */\n constructor(\n job: proto.Job,\n onReject: () => Promise<void>,\n onAccept: (args: JobAcceptArguments) => Promise<void>,\n ) {\n this.#job = job;\n this.#onReject = onReject;\n this.#onAccept = onAccept;\n }\n\n /** @returns The ID of the job, set by the LiveKit server */\n get id(): string {\n return this.#job.id;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get job(): proto.Job {\n return this.#job;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get room(): proto.Room | undefined {\n return this.#job.room;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get publisher(): proto.ParticipantInfo | undefined {\n return this.#job.participant;\n }\n\n /** @returns The agent's name, as set in {@link WorkerOptions} */\n get agentName(): string {\n return this.#job.agentName;\n }\n\n /** Rejects the job. */\n async reject() {\n await this.#onReject();\n }\n\n /** Accepts the job, launching it on an idle child process. */\n async accept(name = '', identity = '', metadata = '', attributes?: { [key: string]: string }) {\n if (identity === '') identity = 'agent-' + this.id;\n\n this.#onAccept({ name, identity, metadata, attributes });\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWA,sBAAsD;AAEtD,iBAAoB;AAGb,IAAK,gBAAL,kBAAKA,mBAAL;AACL,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AAJU,SAAAA;AAAA,GAAA;AAsBL,MAAM,4BAA4B,MAAM;AAAA,EAC7C,YAAY,KAAc;AACxB,UAAM,GAAG;AACT,WAAO,eAAe,MAAM,WAAW,SAAS;AAAA,EAClD;AACF;AAGO,MAAM,WAAW;AAAA,EACtB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAEA,oBAA6C,CAAC;AAAA,EAC9C,0BAAwF,CAAC;AAAA,EACzF,oBAKI,CAAC;AAAA,EACL;AAAA,EAEA,YACE,MACA,MACA,MACA,WACA,YACA;AACA,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,aAAa;AAClB,SAAK,cAAc;AACnB,SAAK,yBAAyB,KAAK,uBAAuB,KAAK,IAAI;AACnE,SAAK,MAAM,GAAG,0BAAU,sBAAsB,KAAK,sBAAsB;AACzE,SAAK,cAAU,gBAAI,EAAE,MAAM,EAAE,MAAM,KAAK,MAAM,CAAC;AAAA,EACjD;AAAA,EAEA,IAAI,OAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAiB;AACnB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,OAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,QAAsC;AACxC,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,oBAAoB,UAA+B;AACjD,SAAK,kBAAkB,KAAK,QAAQ;AAAA,EACtC;AAAA,EAEA,MAAM,mBAAmB,UAA+C;AACtE,QAAI,CAAC,KAAK,MAAM,aAAa;AAC3B,YAAM,IAAI,MAAM,uBAAuB;AAAA,IACzC;AAEA,eAAW,KAAK,KAAK,MAAM,mBAAmB,OAAO,GAAG;AACtD,WAAK,CAAC,YAAY,EAAE,aAAa,aAAa,EAAE,KAAK,QAAQ,gCAAgB,OAAO;AAClF,eAAO;AAAA,MACT;AAAA,IACF;AAEA,WAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,YAAM,yBAAyB,CAAC,gBAAmC;AACjE,aACG,CAAC,YAAY,YAAY,aAAa,aACvC,YAAY,KAAK,QAAQ,gCAAgB,OACzC;AACA,wBAAc;AACd,kBAAQ,WAAW;AAAA,QACrB;AAAA,MACF;AACA,YAAM,iBAAiB,MAAM;AAC3B,sBAAc;AACd,eAAO,IAAI,MAAM,iDAAiD,CAAC;AAAA,MACrE;AAEA,YAAM,gBAAgB,MAAM;AAC1B,aAAK,MAAM,IAAI,0BAAU,sBAAsB,sBAAsB;AACrE,aAAK,MAAM,IAAI,0BAAU,cAAc,cAAc;AAAA,MACvD;AAEA,WAAK,MAAM,GAAG,0BAAU,sBAAsB,sBAAsB;AACpE,WAAK,MAAM,GAAG,0BAAU,cAAc,cAAc;AAAA,IACtD,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAM,QACJ,MACA,gBAA+B,uBAC/B,WACA;AACA,UAAM,OAAO;AAAA,MACX;AAAA,MACA,eAAe,iBAAiB;AAAA,MAChC;AAAA,MACA,UAAU;AAAA,IACZ;AAEA,UAAM,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,KAAK,MAAM,OAAO,IAAI;AAC/D,SAAK,WAAW;AAEhB,SAAK,MAAM,mBAAmB,QAAQ,KAAK,sBAAsB;AAEjE,QAAI,CAAC,oBAA0B,kBAAwB,EAAE,SAAS,aAAa,GAAG;AAChF,WAAK,MAAM,mBAAmB,QAAQ,CAAC,MAAM;AAC3C,UAAE,kBAAkB,QAAQ,CAAC,QAAQ;AACnC,cACG,kBAAkB,sBAA4B,IAAI,SAAS,0BAAU,cACrE,kBAAkB,sBAA4B,IAAI,SAAS,0BAAU,YACtE;AACA,gBAAI,cAAc,IAAI;AAAA,UACxB;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,SAAS,SAAS,IAAI;AACpB,SAAK,YAAY,MAAM;AAAA,EACzB;AAAA;AAAA,EAGA,uBAAuB,GAAsB;AA9L/C;AA+LI,eAAW,YAAY,KAAK,yBAAyB;AACnD,YAAI,UAAK,kBAAkB,EAAE,QAAQ,MAAjC,mBAAoC,aAAY,UAAU;AAC5D,aAAK,QAAQ;AAAA,UACX;AAAA,UACA,EAAE;AAAA,QACJ;AAAA,MACF;AACA,YAAM,SAAS,SAAS,MAAM,CAAC;AAC/B,aAAO,QAAQ,MAAM,OAAO,KAAK,kBAAkB,EAAE,QAAQ,CAAC;AAC9D,WAAK,kBAAkB,EAAE,QAAQ,IAAI,EAAE,UAAU,OAAO;AAAA,IAC1D;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,yBAAyB,UAAoE;AAC3F,QAAI,KAAK,wBAAwB,SAAS,QAAQ,GAAG;AACnD,YAAM,IAAI,oBAAoB,4CAA4C;AAAA,IAC5E;AAEA,SAAK,wBAAwB,KAAK,QAAQ;AAAA,EAC5C;AACF;AAEO,MAAM,WAAW;AAAA,EACtB,OAAO,QAAQ;AAAA,EACf,WAAsC,CAAC;AAAA,EAEvC,IAAI,MAAc;AAChB,WAAO,KAAK;AAAA,EACd;AACF;AAUO,MAAM,WAAW;AAAA,EACtB;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGA,YACE,KACA,UACA,UACA;AACA,SAAK,OAAO;AACZ,SAAK,YAAY;AACjB,SAAK,YAAY;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,KAAa;AACf,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,MAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,OAA+B;AACjC,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,YAA+C;AACjD,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,YAAoB;AACtB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,MAAM,SAAS;AACb,UAAM,KAAK,UAAU;AAAA,EACvB;AAAA;AAAA,EAGA,MAAM,OAAO,OAAO,IAAI,WAAW,IAAI,WAAW,IAAI,YAAwC;AAC5F,QAAI,aAAa,GAAI,YAAW,WAAW,KAAK;AAEhD,SAAK,UAAU,EAAE,MAAM,UAAU,UAAU,WAAW,CAAC;AAAA,EACzD;AACF;","names":["AutoSubscribe"]}
|
|
1
|
+
{"version":3,"sources":["../src/job.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type * as proto from '@livekit/protocol';\nimport type {\n E2EEOptions,\n LocalParticipant,\n RemoteParticipant,\n Room,\n RtcConfiguration,\n} from '@livekit/rtc-node';\nimport { ParticipantKind, RoomEvent, TrackKind } from '@livekit/rtc-node';\nimport type { Logger } from 'pino';\nimport { log } from './log.js';\n\n/** Which tracks, if any, should the agent automatically subscribe to? */\nexport enum AutoSubscribe {\n SUBSCRIBE_ALL,\n SUBSCRIBE_NONE,\n VIDEO_ONLY,\n AUDIO_ONLY,\n}\n\nexport type JobAcceptArguments = {\n name: string;\n identity: string;\n metadata: string;\n attributes?: { [key: string]: string };\n};\n\nexport type RunningJobInfo = {\n acceptArguments: JobAcceptArguments;\n job: proto.Job;\n url: string;\n token: string;\n};\n\n/** Attempted to add a function callback, but the function already exists. */\nexport class FunctionExistsError extends Error {\n constructor(msg?: string) {\n super(msg);\n Object.setPrototypeOf(this, new.target.prototype);\n }\n}\n\n/** The job and environment context as seen by the agent, accessible by the entrypoint function. */\nexport class JobContext {\n #proc: JobProcess;\n #info: RunningJobInfo;\n #room: Room;\n #onConnect: () => void;\n #onShutdown: (s: string) => void;\n /** @internal */\n shutdownCallbacks: (() => Promise<void>)[] = [];\n #participantEntrypoints: ((job: JobContext, p: RemoteParticipant) => Promise<void>)[] = [];\n #participantTasks: {\n [id: string]: {\n callback: (job: JobContext, p: RemoteParticipant) => Promise<void>;\n result: Promise<void>;\n };\n } = {};\n #logger: Logger;\n\n constructor(\n proc: JobProcess,\n info: RunningJobInfo,\n room: Room,\n onConnect: () => void,\n onShutdown: (s: string) => void,\n ) {\n this.#proc = proc;\n this.#info = info;\n this.#room = room;\n this.#onConnect = onConnect;\n this.#onShutdown = onShutdown;\n this.onParticipantConnected = this.onParticipantConnected.bind(this);\n this.#room.on(RoomEvent.ParticipantConnected, this.onParticipantConnected);\n this.#logger = log().child({ info: this.#info });\n }\n\n get proc(): JobProcess {\n return this.#proc;\n }\n\n get job(): proto.Job {\n return this.#info.job;\n }\n\n /** @returns The room the agent was called into */\n get room(): Room {\n return this.#room;\n }\n\n /** @returns The agent's participant if connected to the room, otherwise `undefined` */\n get agent(): LocalParticipant | undefined {\n return this.#room.localParticipant;\n }\n\n /** Adds a promise to be awaited when {@link JobContext.shutdown | shutdown} is called. */\n addShutdownCallback(callback: () => Promise<void>) {\n this.shutdownCallbacks.push(callback);\n }\n\n async waitForParticipant(identity?: string): Promise<RemoteParticipant> {\n if (!this.#room.isConnected) {\n throw new Error('room is not connected');\n }\n\n for (const p of this.#room.remoteParticipants.values()) {\n if ((!identity || p.identity === identity) && p.info.kind != ParticipantKind.AGENT) {\n return p;\n }\n }\n\n return new Promise((resolve, reject) => {\n const onParticipantConnected = (participant: RemoteParticipant) => {\n if (\n (!identity || participant.identity === identity) &&\n participant.info.kind != ParticipantKind.AGENT\n ) {\n clearHandlers();\n resolve(participant);\n }\n };\n const onDisconnected = () => {\n clearHandlers();\n reject(new Error('Room disconnected while waiting for participant'));\n };\n\n const clearHandlers = () => {\n this.#room.off(RoomEvent.ParticipantConnected, onParticipantConnected);\n this.#room.off(RoomEvent.Disconnected, onDisconnected);\n };\n\n this.#room.on(RoomEvent.ParticipantConnected, onParticipantConnected);\n this.#room.on(RoomEvent.Disconnected, onDisconnected);\n });\n }\n\n /**\n * Connects the agent to the room.\n *\n * @remarks\n * It is recommended to run this command as early in the function as possible, as executing it\n * later may cause noticeable delay between user and agent joins.\n *\n * @see {@link https://github.com/livekit/node-sdks/tree/main/packages/livekit-rtc#readme |\n * @livekit/rtc-node} for more information about the parameters.\n */\n async connect(\n e2ee?: E2EEOptions,\n autoSubscribe: AutoSubscribe = AutoSubscribe.SUBSCRIBE_ALL,\n rtcConfig?: RtcConfiguration,\n ) {\n const opts = {\n e2ee,\n autoSubscribe: autoSubscribe == AutoSubscribe.SUBSCRIBE_ALL,\n rtcConfig,\n dynacast: false,\n };\n\n await this.#room.connect(this.#info.url, this.#info.token, opts);\n this.#onConnect();\n\n this.#room.remoteParticipants.forEach(this.onParticipantConnected);\n\n if ([AutoSubscribe.AUDIO_ONLY, AutoSubscribe.VIDEO_ONLY].includes(autoSubscribe)) {\n this.#room.remoteParticipants.forEach((p) => {\n p.trackPublications.forEach((pub) => {\n if (\n (autoSubscribe === AutoSubscribe.AUDIO_ONLY && pub.kind === TrackKind.KIND_AUDIO) ||\n (autoSubscribe === AutoSubscribe.VIDEO_ONLY && pub.kind === TrackKind.KIND_VIDEO)\n ) {\n pub.setSubscribed(true);\n }\n });\n });\n }\n }\n\n /**\n * Gracefully shuts down the job, and runs all shutdown promises.\n *\n * @param reason - Optional reason for shutdown\n */\n shutdown(reason = '') {\n this.#onShutdown(reason);\n }\n\n /** @internal */\n onParticipantConnected(p: RemoteParticipant) {\n for (const callback of this.#participantEntrypoints) {\n if (this.#participantTasks[p.identity!]?.callback == callback) {\n this.#logger.warn(\n 'a participant has joined before a prior prticipant task matching the same identity has finished:',\n p.identity,\n );\n }\n const result = callback(this, p);\n result.finally(() => delete this.#participantTasks[p.identity!]);\n this.#participantTasks[p.identity!] = { callback, result };\n }\n }\n\n /**\n * Adds a promise to be awaited whenever a new participant joins the room.\n *\n * @throws {@link FunctionExistsError} if an entrypoint already exists\n */\n addParticipantEntrypoint(callback: (job: JobContext, p: RemoteParticipant) => Promise<void>) {\n if (this.#participantEntrypoints.includes(callback)) {\n throw new FunctionExistsError('entrypoints cannot be added more than once');\n }\n\n this.#participantEntrypoints.push(callback);\n }\n}\n\nexport class JobProcess {\n #pid = process.pid;\n userData: { [id: string]: unknown } = {};\n\n get pid(): number {\n return this.#pid;\n }\n}\n\n/**\n * A request sent by the server to spawn a new agent job.\n *\n * @remarks\n * For most applications, this is best left to the default, which simply accepts the job and\n * handles the logic inside the entrypoint function. This class is useful for vetting which\n * requests should fill idle processes and which should be outright rejected.\n */\nexport class JobRequest {\n #job: proto.Job;\n #onReject: () => Promise<void>;\n #onAccept: (args: JobAcceptArguments) => Promise<void>;\n\n /** @internal */\n constructor(\n job: proto.Job,\n onReject: () => Promise<void>,\n onAccept: (args: JobAcceptArguments) => Promise<void>,\n ) {\n this.#job = job;\n this.#onReject = onReject;\n this.#onAccept = onAccept;\n }\n\n /** @returns The ID of the job, set by the LiveKit server */\n get id(): string {\n return this.#job.id;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get job(): proto.Job {\n return this.#job;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get room(): proto.Room | undefined {\n return this.#job.room;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get publisher(): proto.ParticipantInfo | undefined {\n return this.#job.participant;\n }\n\n /** @returns The agent's name, as set in {@link WorkerOptions} */\n get agentName(): string {\n return this.#job.agentName;\n }\n\n /** Rejects the job. */\n async reject() {\n await this.#onReject();\n }\n\n /** Accepts the job, launching it on an idle child process. */\n async accept(name = '', identity = '', metadata = '', attributes?: { [key: string]: string }) {\n if (identity === '') identity = 'agent-' + this.id;\n\n this.#onAccept({ name, identity, metadata, attributes });\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWA,sBAAsD;AAEtD,iBAAoB;AAGb,IAAK,gBAAL,kBAAKA,mBAAL;AACL,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AAJU,SAAAA;AAAA,GAAA;AAsBL,MAAM,4BAA4B,MAAM;AAAA,EAC7C,YAAY,KAAc;AACxB,UAAM,GAAG;AACT,WAAO,eAAe,MAAM,WAAW,SAAS;AAAA,EAClD;AACF;AAGO,MAAM,WAAW;AAAA,EACtB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAEA,oBAA6C,CAAC;AAAA,EAC9C,0BAAwF,CAAC;AAAA,EACzF,oBAKI,CAAC;AAAA,EACL;AAAA,EAEA,YACE,MACA,MACA,MACA,WACA,YACA;AACA,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,aAAa;AAClB,SAAK,cAAc;AACnB,SAAK,yBAAyB,KAAK,uBAAuB,KAAK,IAAI;AACnE,SAAK,MAAM,GAAG,0BAAU,sBAAsB,KAAK,sBAAsB;AACzE,SAAK,cAAU,gBAAI,EAAE,MAAM,EAAE,MAAM,KAAK,MAAM,CAAC;AAAA,EACjD;AAAA,EAEA,IAAI,OAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAiB;AACnB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,OAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,QAAsC;AACxC,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,oBAAoB,UAA+B;AACjD,SAAK,kBAAkB,KAAK,QAAQ;AAAA,EACtC;AAAA,EAEA,MAAM,mBAAmB,UAA+C;AACtE,QAAI,CAAC,KAAK,MAAM,aAAa;AAC3B,YAAM,IAAI,MAAM,uBAAuB;AAAA,IACzC;AAEA,eAAW,KAAK,KAAK,MAAM,mBAAmB,OAAO,GAAG;AACtD,WAAK,CAAC,YAAY,EAAE,aAAa,aAAa,EAAE,KAAK,QAAQ,gCAAgB,OAAO;AAClF,eAAO;AAAA,MACT;AAAA,IACF;AAEA,WAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,YAAM,yBAAyB,CAAC,gBAAmC;AACjE,aACG,CAAC,YAAY,YAAY,aAAa,aACvC,YAAY,KAAK,QAAQ,gCAAgB,OACzC;AACA,wBAAc;AACd,kBAAQ,WAAW;AAAA,QACrB;AAAA,MACF;AACA,YAAM,iBAAiB,MAAM;AAC3B,sBAAc;AACd,eAAO,IAAI,MAAM,iDAAiD,CAAC;AAAA,MACrE;AAEA,YAAM,gBAAgB,MAAM;AAC1B,aAAK,MAAM,IAAI,0BAAU,sBAAsB,sBAAsB;AACrE,aAAK,MAAM,IAAI,0BAAU,cAAc,cAAc;AAAA,MACvD;AAEA,WAAK,MAAM,GAAG,0BAAU,sBAAsB,sBAAsB;AACpE,WAAK,MAAM,GAAG,0BAAU,cAAc,cAAc;AAAA,IACtD,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAM,QACJ,MACA,gBAA+B,uBAC/B,WACA;AACA,UAAM,OAAO;AAAA,MACX;AAAA,MACA,eAAe,iBAAiB;AAAA,MAChC;AAAA,MACA,UAAU;AAAA,IACZ;AAEA,UAAM,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,KAAK,MAAM,OAAO,IAAI;AAC/D,SAAK,WAAW;AAEhB,SAAK,MAAM,mBAAmB,QAAQ,KAAK,sBAAsB;AAEjE,QAAI,CAAC,oBAA0B,kBAAwB,EAAE,SAAS,aAAa,GAAG;AAChF,WAAK,MAAM,mBAAmB,QAAQ,CAAC,MAAM;AAC3C,UAAE,kBAAkB,QAAQ,CAAC,QAAQ;AACnC,cACG,kBAAkB,sBAA4B,IAAI,SAAS,0BAAU,cACrE,kBAAkB,sBAA4B,IAAI,SAAS,0BAAU,YACtE;AACA,gBAAI,cAAc,IAAI;AAAA,UACxB;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,SAAS,SAAS,IAAI;AACpB,SAAK,YAAY,MAAM;AAAA,EACzB;AAAA;AAAA,EAGA,uBAAuB,GAAsB;AA9L/C;AA+LI,eAAW,YAAY,KAAK,yBAAyB;AACnD,YAAI,UAAK,kBAAkB,EAAE,QAAS,MAAlC,mBAAqC,aAAY,UAAU;AAC7D,aAAK,QAAQ;AAAA,UACX;AAAA,UACA,EAAE;AAAA,QACJ;AAAA,MACF;AACA,YAAM,SAAS,SAAS,MAAM,CAAC;AAC/B,aAAO,QAAQ,MAAM,OAAO,KAAK,kBAAkB,EAAE,QAAS,CAAC;AAC/D,WAAK,kBAAkB,EAAE,QAAS,IAAI,EAAE,UAAU,OAAO;AAAA,IAC3D;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,yBAAyB,UAAoE;AAC3F,QAAI,KAAK,wBAAwB,SAAS,QAAQ,GAAG;AACnD,YAAM,IAAI,oBAAoB,4CAA4C;AAAA,IAC5E;AAEA,SAAK,wBAAwB,KAAK,QAAQ;AAAA,EAC5C;AACF;AAEO,MAAM,WAAW;AAAA,EACtB,OAAO,QAAQ;AAAA,EACf,WAAsC,CAAC;AAAA,EAEvC,IAAI,MAAc;AAChB,WAAO,KAAK;AAAA,EACd;AACF;AAUO,MAAM,WAAW;AAAA,EACtB;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGA,YACE,KACA,UACA,UACA;AACA,SAAK,OAAO;AACZ,SAAK,YAAY;AACjB,SAAK,YAAY;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,KAAa;AACf,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,MAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,OAA+B;AACjC,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,YAA+C;AACjD,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,YAAoB;AACtB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,MAAM,SAAS;AACb,UAAM,KAAK,UAAU;AAAA,EACvB;AAAA;AAAA,EAGA,MAAM,OAAO,OAAO,IAAI,WAAW,IAAI,WAAW,IAAI,YAAwC;AAC5F,QAAI,aAAa,GAAI,YAAW,WAAW,KAAK;AAEhD,SAAK,UAAU,EAAE,MAAM,UAAU,UAAU,WAAW,CAAC;AAAA,EACzD;AACF;","names":["AutoSubscribe"]}
|
package/dist/job.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/job.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type * as proto from '@livekit/protocol';\nimport type {\n E2EEOptions,\n LocalParticipant,\n RemoteParticipant,\n Room,\n RtcConfiguration,\n} from '@livekit/rtc-node';\nimport { ParticipantKind, RoomEvent, TrackKind } from '@livekit/rtc-node';\nimport type { Logger } from 'pino';\nimport { log } from './log.js';\n\n/** Which tracks, if any, should the agent automatically subscribe to? */\nexport enum AutoSubscribe {\n SUBSCRIBE_ALL,\n SUBSCRIBE_NONE,\n VIDEO_ONLY,\n AUDIO_ONLY,\n}\n\nexport type JobAcceptArguments = {\n name: string;\n identity: string;\n metadata: string;\n attributes?: { [key: string]: string };\n};\n\nexport type RunningJobInfo = {\n acceptArguments: JobAcceptArguments;\n job: proto.Job;\n url: string;\n token: string;\n};\n\n/** Attempted to add a function callback, but the function already exists. */\nexport class FunctionExistsError extends Error {\n constructor(msg?: string) {\n super(msg);\n Object.setPrototypeOf(this, new.target.prototype);\n }\n}\n\n/** The job and environment context as seen by the agent, accessible by the entrypoint function. */\nexport class JobContext {\n #proc: JobProcess;\n #info: RunningJobInfo;\n #room: Room;\n #onConnect: () => void;\n #onShutdown: (s: string) => void;\n /** @internal */\n shutdownCallbacks: (() => Promise<void>)[] = [];\n #participantEntrypoints: ((job: JobContext, p: RemoteParticipant) => Promise<void>)[] = [];\n #participantTasks: {\n [id: string]: {\n callback: (job: JobContext, p: RemoteParticipant) => Promise<void>;\n result: Promise<void>;\n };\n } = {};\n #logger: Logger;\n\n constructor(\n proc: JobProcess,\n info: RunningJobInfo,\n room: Room,\n onConnect: () => void,\n onShutdown: (s: string) => void,\n ) {\n this.#proc = proc;\n this.#info = info;\n this.#room = room;\n this.#onConnect = onConnect;\n this.#onShutdown = onShutdown;\n this.onParticipantConnected = this.onParticipantConnected.bind(this);\n this.#room.on(RoomEvent.ParticipantConnected, this.onParticipantConnected);\n this.#logger = log().child({ info: this.#info });\n }\n\n get proc(): JobProcess {\n return this.#proc;\n }\n\n get job(): proto.Job {\n return this.#info.job;\n }\n\n /** @returns The room the agent was called into */\n get room(): Room {\n return this.#room;\n }\n\n /** @returns The agent's participant if connected to the room, otherwise `undefined` */\n get agent(): LocalParticipant | undefined {\n return this.#room.localParticipant;\n }\n\n /** Adds a promise to be awaited when {@link JobContext.shutdown | shutdown} is called. */\n addShutdownCallback(callback: () => Promise<void>) {\n this.shutdownCallbacks.push(callback);\n }\n\n async waitForParticipant(identity?: string): Promise<RemoteParticipant> {\n if (!this.#room.isConnected) {\n throw new Error('room is not connected');\n }\n\n for (const p of this.#room.remoteParticipants.values()) {\n if ((!identity || p.identity === identity) && p.info.kind != ParticipantKind.AGENT) {\n return p;\n }\n }\n\n return new Promise((resolve, reject) => {\n const onParticipantConnected = (participant: RemoteParticipant) => {\n if (\n (!identity || participant.identity === identity) &&\n participant.info.kind != ParticipantKind.AGENT\n ) {\n clearHandlers();\n resolve(participant);\n }\n };\n const onDisconnected = () => {\n clearHandlers();\n reject(new Error('Room disconnected while waiting for participant'));\n };\n\n const clearHandlers = () => {\n this.#room.off(RoomEvent.ParticipantConnected, onParticipantConnected);\n this.#room.off(RoomEvent.Disconnected, onDisconnected);\n };\n\n this.#room.on(RoomEvent.ParticipantConnected, onParticipantConnected);\n this.#room.on(RoomEvent.Disconnected, onDisconnected);\n });\n }\n\n /**\n * Connects the agent to the room.\n *\n * @remarks\n * It is recommended to run this command as early in the function as possible, as executing it\n * later may cause noticeable delay between user and agent joins.\n *\n * @see {@link https://github.com/livekit/node-sdks/tree/main/packages/livekit-rtc#readme |\n * @livekit/rtc-node} for more information about the parameters.\n */\n async connect(\n e2ee?: E2EEOptions,\n autoSubscribe: AutoSubscribe = AutoSubscribe.SUBSCRIBE_ALL,\n rtcConfig?: RtcConfiguration,\n ) {\n const opts = {\n e2ee,\n autoSubscribe: autoSubscribe == AutoSubscribe.SUBSCRIBE_ALL,\n rtcConfig,\n dynacast: false,\n };\n\n await this.#room.connect(this.#info.url, this.#info.token, opts);\n this.#onConnect();\n\n this.#room.remoteParticipants.forEach(this.onParticipantConnected);\n\n if ([AutoSubscribe.AUDIO_ONLY, AutoSubscribe.VIDEO_ONLY].includes(autoSubscribe)) {\n this.#room.remoteParticipants.forEach((p) => {\n p.trackPublications.forEach((pub) => {\n if (\n (autoSubscribe === AutoSubscribe.AUDIO_ONLY && pub.kind === TrackKind.KIND_AUDIO) ||\n (autoSubscribe === AutoSubscribe.VIDEO_ONLY && pub.kind === TrackKind.KIND_VIDEO)\n ) {\n pub.setSubscribed(true);\n }\n });\n });\n }\n }\n\n /**\n * Gracefully shuts down the job, and runs all shutdown promises.\n *\n * @param reason - Optional reason for shutdown\n */\n shutdown(reason = '') {\n this.#onShutdown(reason);\n }\n\n /** @internal */\n onParticipantConnected(p: RemoteParticipant) {\n for (const callback of this.#participantEntrypoints) {\n if (this.#participantTasks[p.identity]?.callback == callback) {\n this.#logger.warn(\n 'a participant has joined before a prior prticipant task matching the same identity has finished:',\n p.identity,\n );\n }\n const result = callback(this, p);\n result.finally(() => delete this.#participantTasks[p.identity]);\n this.#participantTasks[p.identity] = { callback, result };\n }\n }\n\n /**\n * Adds a promise to be awaited whenever a new participant joins the room.\n *\n * @throws {@link FunctionExistsError} if an entrypoint already exists\n */\n addParticipantEntrypoint(callback: (job: JobContext, p: RemoteParticipant) => Promise<void>) {\n if (this.#participantEntrypoints.includes(callback)) {\n throw new FunctionExistsError('entrypoints cannot be added more than once');\n }\n\n this.#participantEntrypoints.push(callback);\n }\n}\n\nexport class JobProcess {\n #pid = process.pid;\n userData: { [id: string]: unknown } = {};\n\n get pid(): number {\n return this.#pid;\n }\n}\n\n/**\n * A request sent by the server to spawn a new agent job.\n *\n * @remarks\n * For most applications, this is best left to the default, which simply accepts the job and\n * handles the logic inside the entrypoint function. This class is useful for vetting which\n * requests should fill idle processes and which should be outright rejected.\n */\nexport class JobRequest {\n #job: proto.Job;\n #onReject: () => Promise<void>;\n #onAccept: (args: JobAcceptArguments) => Promise<void>;\n\n /** @internal */\n constructor(\n job: proto.Job,\n onReject: () => Promise<void>,\n onAccept: (args: JobAcceptArguments) => Promise<void>,\n ) {\n this.#job = job;\n this.#onReject = onReject;\n this.#onAccept = onAccept;\n }\n\n /** @returns The ID of the job, set by the LiveKit server */\n get id(): string {\n return this.#job.id;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get job(): proto.Job {\n return this.#job;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get room(): proto.Room | undefined {\n return this.#job.room;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get publisher(): proto.ParticipantInfo | undefined {\n return this.#job.participant;\n }\n\n /** @returns The agent's name, as set in {@link WorkerOptions} */\n get agentName(): string {\n return this.#job.agentName;\n }\n\n /** Rejects the job. */\n async reject() {\n await this.#onReject();\n }\n\n /** Accepts the job, launching it on an idle child process. */\n async accept(name = '', identity = '', metadata = '', attributes?: { [key: string]: string }) {\n if (identity === '') identity = 'agent-' + this.id;\n\n this.#onAccept({ name, identity, metadata, attributes });\n }\n}\n"],"mappings":"AAWA,SAAS,iBAAiB,WAAW,iBAAiB;AAEtD,SAAS,WAAW;AAGb,IAAK,gBAAL,kBAAKA,mBAAL;AACL,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AAJU,SAAAA;AAAA,GAAA;AAsBL,MAAM,4BAA4B,MAAM;AAAA,EAC7C,YAAY,KAAc;AACxB,UAAM,GAAG;AACT,WAAO,eAAe,MAAM,WAAW,SAAS;AAAA,EAClD;AACF;AAGO,MAAM,WAAW;AAAA,EACtB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAEA,oBAA6C,CAAC;AAAA,EAC9C,0BAAwF,CAAC;AAAA,EACzF,oBAKI,CAAC;AAAA,EACL;AAAA,EAEA,YACE,MACA,MACA,MACA,WACA,YACA;AACA,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,aAAa;AAClB,SAAK,cAAc;AACnB,SAAK,yBAAyB,KAAK,uBAAuB,KAAK,IAAI;AACnE,SAAK,MAAM,GAAG,UAAU,sBAAsB,KAAK,sBAAsB;AACzE,SAAK,UAAU,IAAI,EAAE,MAAM,EAAE,MAAM,KAAK,MAAM,CAAC;AAAA,EACjD;AAAA,EAEA,IAAI,OAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAiB;AACnB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,OAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,QAAsC;AACxC,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,oBAAoB,UAA+B;AACjD,SAAK,kBAAkB,KAAK,QAAQ;AAAA,EACtC;AAAA,EAEA,MAAM,mBAAmB,UAA+C;AACtE,QAAI,CAAC,KAAK,MAAM,aAAa;AAC3B,YAAM,IAAI,MAAM,uBAAuB;AAAA,IACzC;AAEA,eAAW,KAAK,KAAK,MAAM,mBAAmB,OAAO,GAAG;AACtD,WAAK,CAAC,YAAY,EAAE,aAAa,aAAa,EAAE,KAAK,QAAQ,gBAAgB,OAAO;AAClF,eAAO;AAAA,MACT;AAAA,IACF;AAEA,WAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,YAAM,yBAAyB,CAAC,gBAAmC;AACjE,aACG,CAAC,YAAY,YAAY,aAAa,aACvC,YAAY,KAAK,QAAQ,gBAAgB,OACzC;AACA,wBAAc;AACd,kBAAQ,WAAW;AAAA,QACrB;AAAA,MACF;AACA,YAAM,iBAAiB,MAAM;AAC3B,sBAAc;AACd,eAAO,IAAI,MAAM,iDAAiD,CAAC;AAAA,MACrE;AAEA,YAAM,gBAAgB,MAAM;AAC1B,aAAK,MAAM,IAAI,UAAU,sBAAsB,sBAAsB;AACrE,aAAK,MAAM,IAAI,UAAU,cAAc,cAAc;AAAA,MACvD;AAEA,WAAK,MAAM,GAAG,UAAU,sBAAsB,sBAAsB;AACpE,WAAK,MAAM,GAAG,UAAU,cAAc,cAAc;AAAA,IACtD,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAM,QACJ,MACA,gBAA+B,uBAC/B,WACA;AACA,UAAM,OAAO;AAAA,MACX;AAAA,MACA,eAAe,iBAAiB;AAAA,MAChC;AAAA,MACA,UAAU;AAAA,IACZ;AAEA,UAAM,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,KAAK,MAAM,OAAO,IAAI;AAC/D,SAAK,WAAW;AAEhB,SAAK,MAAM,mBAAmB,QAAQ,KAAK,sBAAsB;AAEjE,QAAI,CAAC,oBAA0B,kBAAwB,EAAE,SAAS,aAAa,GAAG;AAChF,WAAK,MAAM,mBAAmB,QAAQ,CAAC,MAAM;AAC3C,UAAE,kBAAkB,QAAQ,CAAC,QAAQ;AACnC,cACG,kBAAkB,sBAA4B,IAAI,SAAS,UAAU,cACrE,kBAAkB,sBAA4B,IAAI,SAAS,UAAU,YACtE;AACA,gBAAI,cAAc,IAAI;AAAA,UACxB;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,SAAS,SAAS,IAAI;AACpB,SAAK,YAAY,MAAM;AAAA,EACzB;AAAA;AAAA,EAGA,uBAAuB,GAAsB;AA9L/C;AA+LI,eAAW,YAAY,KAAK,yBAAyB;AACnD,YAAI,UAAK,kBAAkB,EAAE,QAAQ,MAAjC,mBAAoC,aAAY,UAAU;AAC5D,aAAK,QAAQ;AAAA,UACX;AAAA,UACA,EAAE;AAAA,QACJ;AAAA,MACF;AACA,YAAM,SAAS,SAAS,MAAM,CAAC;AAC/B,aAAO,QAAQ,MAAM,OAAO,KAAK,kBAAkB,EAAE,QAAQ,CAAC;AAC9D,WAAK,kBAAkB,EAAE,QAAQ,IAAI,EAAE,UAAU,OAAO;AAAA,IAC1D;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,yBAAyB,UAAoE;AAC3F,QAAI,KAAK,wBAAwB,SAAS,QAAQ,GAAG;AACnD,YAAM,IAAI,oBAAoB,4CAA4C;AAAA,IAC5E;AAEA,SAAK,wBAAwB,KAAK,QAAQ;AAAA,EAC5C;AACF;AAEO,MAAM,WAAW;AAAA,EACtB,OAAO,QAAQ;AAAA,EACf,WAAsC,CAAC;AAAA,EAEvC,IAAI,MAAc;AAChB,WAAO,KAAK;AAAA,EACd;AACF;AAUO,MAAM,WAAW;AAAA,EACtB;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGA,YACE,KACA,UACA,UACA;AACA,SAAK,OAAO;AACZ,SAAK,YAAY;AACjB,SAAK,YAAY;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,KAAa;AACf,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,MAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,OAA+B;AACjC,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,YAA+C;AACjD,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,YAAoB;AACtB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,MAAM,SAAS;AACb,UAAM,KAAK,UAAU;AAAA,EACvB;AAAA;AAAA,EAGA,MAAM,OAAO,OAAO,IAAI,WAAW,IAAI,WAAW,IAAI,YAAwC;AAC5F,QAAI,aAAa,GAAI,YAAW,WAAW,KAAK;AAEhD,SAAK,UAAU,EAAE,MAAM,UAAU,UAAU,WAAW,CAAC;AAAA,EACzD;AACF;","names":["AutoSubscribe"]}
|
|
1
|
+
{"version":3,"sources":["../src/job.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type * as proto from '@livekit/protocol';\nimport type {\n E2EEOptions,\n LocalParticipant,\n RemoteParticipant,\n Room,\n RtcConfiguration,\n} from '@livekit/rtc-node';\nimport { ParticipantKind, RoomEvent, TrackKind } from '@livekit/rtc-node';\nimport type { Logger } from 'pino';\nimport { log } from './log.js';\n\n/** Which tracks, if any, should the agent automatically subscribe to? */\nexport enum AutoSubscribe {\n SUBSCRIBE_ALL,\n SUBSCRIBE_NONE,\n VIDEO_ONLY,\n AUDIO_ONLY,\n}\n\nexport type JobAcceptArguments = {\n name: string;\n identity: string;\n metadata: string;\n attributes?: { [key: string]: string };\n};\n\nexport type RunningJobInfo = {\n acceptArguments: JobAcceptArguments;\n job: proto.Job;\n url: string;\n token: string;\n};\n\n/** Attempted to add a function callback, but the function already exists. */\nexport class FunctionExistsError extends Error {\n constructor(msg?: string) {\n super(msg);\n Object.setPrototypeOf(this, new.target.prototype);\n }\n}\n\n/** The job and environment context as seen by the agent, accessible by the entrypoint function. */\nexport class JobContext {\n #proc: JobProcess;\n #info: RunningJobInfo;\n #room: Room;\n #onConnect: () => void;\n #onShutdown: (s: string) => void;\n /** @internal */\n shutdownCallbacks: (() => Promise<void>)[] = [];\n #participantEntrypoints: ((job: JobContext, p: RemoteParticipant) => Promise<void>)[] = [];\n #participantTasks: {\n [id: string]: {\n callback: (job: JobContext, p: RemoteParticipant) => Promise<void>;\n result: Promise<void>;\n };\n } = {};\n #logger: Logger;\n\n constructor(\n proc: JobProcess,\n info: RunningJobInfo,\n room: Room,\n onConnect: () => void,\n onShutdown: (s: string) => void,\n ) {\n this.#proc = proc;\n this.#info = info;\n this.#room = room;\n this.#onConnect = onConnect;\n this.#onShutdown = onShutdown;\n this.onParticipantConnected = this.onParticipantConnected.bind(this);\n this.#room.on(RoomEvent.ParticipantConnected, this.onParticipantConnected);\n this.#logger = log().child({ info: this.#info });\n }\n\n get proc(): JobProcess {\n return this.#proc;\n }\n\n get job(): proto.Job {\n return this.#info.job;\n }\n\n /** @returns The room the agent was called into */\n get room(): Room {\n return this.#room;\n }\n\n /** @returns The agent's participant if connected to the room, otherwise `undefined` */\n get agent(): LocalParticipant | undefined {\n return this.#room.localParticipant;\n }\n\n /** Adds a promise to be awaited when {@link JobContext.shutdown | shutdown} is called. */\n addShutdownCallback(callback: () => Promise<void>) {\n this.shutdownCallbacks.push(callback);\n }\n\n async waitForParticipant(identity?: string): Promise<RemoteParticipant> {\n if (!this.#room.isConnected) {\n throw new Error('room is not connected');\n }\n\n for (const p of this.#room.remoteParticipants.values()) {\n if ((!identity || p.identity === identity) && p.info.kind != ParticipantKind.AGENT) {\n return p;\n }\n }\n\n return new Promise((resolve, reject) => {\n const onParticipantConnected = (participant: RemoteParticipant) => {\n if (\n (!identity || participant.identity === identity) &&\n participant.info.kind != ParticipantKind.AGENT\n ) {\n clearHandlers();\n resolve(participant);\n }\n };\n const onDisconnected = () => {\n clearHandlers();\n reject(new Error('Room disconnected while waiting for participant'));\n };\n\n const clearHandlers = () => {\n this.#room.off(RoomEvent.ParticipantConnected, onParticipantConnected);\n this.#room.off(RoomEvent.Disconnected, onDisconnected);\n };\n\n this.#room.on(RoomEvent.ParticipantConnected, onParticipantConnected);\n this.#room.on(RoomEvent.Disconnected, onDisconnected);\n });\n }\n\n /**\n * Connects the agent to the room.\n *\n * @remarks\n * It is recommended to run this command as early in the function as possible, as executing it\n * later may cause noticeable delay between user and agent joins.\n *\n * @see {@link https://github.com/livekit/node-sdks/tree/main/packages/livekit-rtc#readme |\n * @livekit/rtc-node} for more information about the parameters.\n */\n async connect(\n e2ee?: E2EEOptions,\n autoSubscribe: AutoSubscribe = AutoSubscribe.SUBSCRIBE_ALL,\n rtcConfig?: RtcConfiguration,\n ) {\n const opts = {\n e2ee,\n autoSubscribe: autoSubscribe == AutoSubscribe.SUBSCRIBE_ALL,\n rtcConfig,\n dynacast: false,\n };\n\n await this.#room.connect(this.#info.url, this.#info.token, opts);\n this.#onConnect();\n\n this.#room.remoteParticipants.forEach(this.onParticipantConnected);\n\n if ([AutoSubscribe.AUDIO_ONLY, AutoSubscribe.VIDEO_ONLY].includes(autoSubscribe)) {\n this.#room.remoteParticipants.forEach((p) => {\n p.trackPublications.forEach((pub) => {\n if (\n (autoSubscribe === AutoSubscribe.AUDIO_ONLY && pub.kind === TrackKind.KIND_AUDIO) ||\n (autoSubscribe === AutoSubscribe.VIDEO_ONLY && pub.kind === TrackKind.KIND_VIDEO)\n ) {\n pub.setSubscribed(true);\n }\n });\n });\n }\n }\n\n /**\n * Gracefully shuts down the job, and runs all shutdown promises.\n *\n * @param reason - Optional reason for shutdown\n */\n shutdown(reason = '') {\n this.#onShutdown(reason);\n }\n\n /** @internal */\n onParticipantConnected(p: RemoteParticipant) {\n for (const callback of this.#participantEntrypoints) {\n if (this.#participantTasks[p.identity!]?.callback == callback) {\n this.#logger.warn(\n 'a participant has joined before a prior prticipant task matching the same identity has finished:',\n p.identity,\n );\n }\n const result = callback(this, p);\n result.finally(() => delete this.#participantTasks[p.identity!]);\n this.#participantTasks[p.identity!] = { callback, result };\n }\n }\n\n /**\n * Adds a promise to be awaited whenever a new participant joins the room.\n *\n * @throws {@link FunctionExistsError} if an entrypoint already exists\n */\n addParticipantEntrypoint(callback: (job: JobContext, p: RemoteParticipant) => Promise<void>) {\n if (this.#participantEntrypoints.includes(callback)) {\n throw new FunctionExistsError('entrypoints cannot be added more than once');\n }\n\n this.#participantEntrypoints.push(callback);\n }\n}\n\nexport class JobProcess {\n #pid = process.pid;\n userData: { [id: string]: unknown } = {};\n\n get pid(): number {\n return this.#pid;\n }\n}\n\n/**\n * A request sent by the server to spawn a new agent job.\n *\n * @remarks\n * For most applications, this is best left to the default, which simply accepts the job and\n * handles the logic inside the entrypoint function. This class is useful for vetting which\n * requests should fill idle processes and which should be outright rejected.\n */\nexport class JobRequest {\n #job: proto.Job;\n #onReject: () => Promise<void>;\n #onAccept: (args: JobAcceptArguments) => Promise<void>;\n\n /** @internal */\n constructor(\n job: proto.Job,\n onReject: () => Promise<void>,\n onAccept: (args: JobAcceptArguments) => Promise<void>,\n ) {\n this.#job = job;\n this.#onReject = onReject;\n this.#onAccept = onAccept;\n }\n\n /** @returns The ID of the job, set by the LiveKit server */\n get id(): string {\n return this.#job.id;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get job(): proto.Job {\n return this.#job;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get room(): proto.Room | undefined {\n return this.#job.room;\n }\n\n /** @see {@link https://www.npmjs.com/package/@livekit/protocol | @livekit/protocol} */\n get publisher(): proto.ParticipantInfo | undefined {\n return this.#job.participant;\n }\n\n /** @returns The agent's name, as set in {@link WorkerOptions} */\n get agentName(): string {\n return this.#job.agentName;\n }\n\n /** Rejects the job. */\n async reject() {\n await this.#onReject();\n }\n\n /** Accepts the job, launching it on an idle child process. */\n async accept(name = '', identity = '', metadata = '', attributes?: { [key: string]: string }) {\n if (identity === '') identity = 'agent-' + this.id;\n\n this.#onAccept({ name, identity, metadata, attributes });\n }\n}\n"],"mappings":"AAWA,SAAS,iBAAiB,WAAW,iBAAiB;AAEtD,SAAS,WAAW;AAGb,IAAK,gBAAL,kBAAKA,mBAAL;AACL,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AACA,EAAAA,8BAAA;AAJU,SAAAA;AAAA,GAAA;AAsBL,MAAM,4BAA4B,MAAM;AAAA,EAC7C,YAAY,KAAc;AACxB,UAAM,GAAG;AACT,WAAO,eAAe,MAAM,WAAW,SAAS;AAAA,EAClD;AACF;AAGO,MAAM,WAAW;AAAA,EACtB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAEA,oBAA6C,CAAC;AAAA,EAC9C,0BAAwF,CAAC;AAAA,EACzF,oBAKI,CAAC;AAAA,EACL;AAAA,EAEA,YACE,MACA,MACA,MACA,WACA,YACA;AACA,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,aAAa;AAClB,SAAK,cAAc;AACnB,SAAK,yBAAyB,KAAK,uBAAuB,KAAK,IAAI;AACnE,SAAK,MAAM,GAAG,UAAU,sBAAsB,KAAK,sBAAsB;AACzE,SAAK,UAAU,IAAI,EAAE,MAAM,EAAE,MAAM,KAAK,MAAM,CAAC;AAAA,EACjD;AAAA,EAEA,IAAI,OAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAiB;AACnB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,OAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,QAAsC;AACxC,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,oBAAoB,UAA+B;AACjD,SAAK,kBAAkB,KAAK,QAAQ;AAAA,EACtC;AAAA,EAEA,MAAM,mBAAmB,UAA+C;AACtE,QAAI,CAAC,KAAK,MAAM,aAAa;AAC3B,YAAM,IAAI,MAAM,uBAAuB;AAAA,IACzC;AAEA,eAAW,KAAK,KAAK,MAAM,mBAAmB,OAAO,GAAG;AACtD,WAAK,CAAC,YAAY,EAAE,aAAa,aAAa,EAAE,KAAK,QAAQ,gBAAgB,OAAO;AAClF,eAAO;AAAA,MACT;AAAA,IACF;AAEA,WAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,YAAM,yBAAyB,CAAC,gBAAmC;AACjE,aACG,CAAC,YAAY,YAAY,aAAa,aACvC,YAAY,KAAK,QAAQ,gBAAgB,OACzC;AACA,wBAAc;AACd,kBAAQ,WAAW;AAAA,QACrB;AAAA,MACF;AACA,YAAM,iBAAiB,MAAM;AAC3B,sBAAc;AACd,eAAO,IAAI,MAAM,iDAAiD,CAAC;AAAA,MACrE;AAEA,YAAM,gBAAgB,MAAM;AAC1B,aAAK,MAAM,IAAI,UAAU,sBAAsB,sBAAsB;AACrE,aAAK,MAAM,IAAI,UAAU,cAAc,cAAc;AAAA,MACvD;AAEA,WAAK,MAAM,GAAG,UAAU,sBAAsB,sBAAsB;AACpE,WAAK,MAAM,GAAG,UAAU,cAAc,cAAc;AAAA,IACtD,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAM,QACJ,MACA,gBAA+B,uBAC/B,WACA;AACA,UAAM,OAAO;AAAA,MACX;AAAA,MACA,eAAe,iBAAiB;AAAA,MAChC;AAAA,MACA,UAAU;AAAA,IACZ;AAEA,UAAM,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,KAAK,MAAM,OAAO,IAAI;AAC/D,SAAK,WAAW;AAEhB,SAAK,MAAM,mBAAmB,QAAQ,KAAK,sBAAsB;AAEjE,QAAI,CAAC,oBAA0B,kBAAwB,EAAE,SAAS,aAAa,GAAG;AAChF,WAAK,MAAM,mBAAmB,QAAQ,CAAC,MAAM;AAC3C,UAAE,kBAAkB,QAAQ,CAAC,QAAQ;AACnC,cACG,kBAAkB,sBAA4B,IAAI,SAAS,UAAU,cACrE,kBAAkB,sBAA4B,IAAI,SAAS,UAAU,YACtE;AACA,gBAAI,cAAc,IAAI;AAAA,UACxB;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,SAAS,SAAS,IAAI;AACpB,SAAK,YAAY,MAAM;AAAA,EACzB;AAAA;AAAA,EAGA,uBAAuB,GAAsB;AA9L/C;AA+LI,eAAW,YAAY,KAAK,yBAAyB;AACnD,YAAI,UAAK,kBAAkB,EAAE,QAAS,MAAlC,mBAAqC,aAAY,UAAU;AAC7D,aAAK,QAAQ;AAAA,UACX;AAAA,UACA,EAAE;AAAA,QACJ;AAAA,MACF;AACA,YAAM,SAAS,SAAS,MAAM,CAAC;AAC/B,aAAO,QAAQ,MAAM,OAAO,KAAK,kBAAkB,EAAE,QAAS,CAAC;AAC/D,WAAK,kBAAkB,EAAE,QAAS,IAAI,EAAE,UAAU,OAAO;AAAA,IAC3D;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,yBAAyB,UAAoE;AAC3F,QAAI,KAAK,wBAAwB,SAAS,QAAQ,GAAG;AACnD,YAAM,IAAI,oBAAoB,4CAA4C;AAAA,IAC5E;AAEA,SAAK,wBAAwB,KAAK,QAAQ;AAAA,EAC5C;AACF;AAEO,MAAM,WAAW;AAAA,EACtB,OAAO,QAAQ;AAAA,EACf,WAAsC,CAAC;AAAA,EAEvC,IAAI,MAAc;AAChB,WAAO,KAAK;AAAA,EACd;AACF;AAUO,MAAM,WAAW;AAAA,EACtB;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGA,YACE,KACA,UACA,UACA;AACA,SAAK,OAAO;AACZ,SAAK,YAAY;AACjB,SAAK,YAAY;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,KAAa;AACf,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,MAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,OAA+B;AACjC,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,YAA+C;AACjD,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,IAAI,YAAoB;AACtB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA;AAAA,EAGA,MAAM,SAAS;AACb,UAAM,KAAK,UAAU;AAAA,EACvB;AAAA;AAAA,EAGA,MAAM,OAAO,OAAO,IAAI,WAAW,IAAI,WAAW,IAAI,YAAwC;AAC5F,QAAI,aAAa,GAAI,YAAW,WAAW,KAAK;AAEhD,SAAK,UAAU,EAAE,MAAM,UAAU,UAAU,WAAW,CAAC;AAAA,EACzD;AACF;","names":["AutoSubscribe"]}
|
|
@@ -53,15 +53,19 @@ class MultimodalAgent extends import_node_events.EventEmitter {
|
|
|
53
53
|
linkedParticipant = null;
|
|
54
54
|
subscribedTrack = null;
|
|
55
55
|
readMicroTask = null;
|
|
56
|
+
#textResponseRetries = 0;
|
|
57
|
+
#maxTextResponseRetries;
|
|
56
58
|
constructor({
|
|
57
59
|
model,
|
|
58
60
|
chatCtx,
|
|
59
|
-
fncCtx
|
|
61
|
+
fncCtx,
|
|
62
|
+
maxTextResponseRetries = 5
|
|
60
63
|
}) {
|
|
61
64
|
super();
|
|
62
65
|
this.model = model;
|
|
63
66
|
this.#chatCtx = chatCtx;
|
|
64
67
|
this.#fncCtx = fncCtx;
|
|
68
|
+
this.#maxTextResponseRetries = maxTextResponseRetries;
|
|
65
69
|
}
|
|
66
70
|
#participant = null;
|
|
67
71
|
#agentPublication = null;
|
|
@@ -190,6 +194,7 @@ class MultimodalAgent extends import_node_events.EventEmitter {
|
|
|
190
194
|
this.#started = true;
|
|
191
195
|
this.#session.on("response_content_added", (message) => {
|
|
192
196
|
var _a2;
|
|
197
|
+
if (message.contentType === "text") return;
|
|
193
198
|
const trFwd = new import_transcription.BasicTranscriptionForwarder(
|
|
194
199
|
this.room,
|
|
195
200
|
this.room.localParticipant.identity,
|
|
@@ -205,6 +210,26 @@ class MultimodalAgent extends import_node_events.EventEmitter {
|
|
|
205
210
|
);
|
|
206
211
|
this.#playingHandle = handle;
|
|
207
212
|
});
|
|
213
|
+
this.#session.on("response_content_done", (message) => {
|
|
214
|
+
if (message.contentType === "text") {
|
|
215
|
+
if (this.#textResponseRetries >= this.#maxTextResponseRetries) {
|
|
216
|
+
throw new Error(
|
|
217
|
+
`The OpenAI Realtime API returned a text response after ${this.#maxTextResponseRetries} retries. Please try to reduce the number of text system or assistant messages in the chat context.`
|
|
218
|
+
);
|
|
219
|
+
}
|
|
220
|
+
this.#textResponseRetries++;
|
|
221
|
+
this.#logger.child({
|
|
222
|
+
itemId: message.itemId,
|
|
223
|
+
text: message.text,
|
|
224
|
+
retries: this.#textResponseRetries
|
|
225
|
+
}).warn(
|
|
226
|
+
"The OpenAI Realtime API returned a text response instead of audio. Attempting to recover to audio mode..."
|
|
227
|
+
);
|
|
228
|
+
this.#session.recoverFromTextResponse(message.itemId);
|
|
229
|
+
} else {
|
|
230
|
+
this.#textResponseRetries = 0;
|
|
231
|
+
}
|
|
232
|
+
});
|
|
208
233
|
this.#session.on("input_speech_committed", (ev) => {
|
|
209
234
|
var _a2, _b;
|
|
210
235
|
const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
|
|
@@ -334,9 +359,8 @@ class MultimodalAgent extends import_node_events.EventEmitter {
|
|
|
334
359
|
});
|
|
335
360
|
}
|
|
336
361
|
#getLocalTrackSid() {
|
|
337
|
-
var _a;
|
|
338
362
|
if (!this.#localTrackSid && this.room && this.room.localParticipant) {
|
|
339
|
-
this.#localTrackSid = (0, import_utils.findMicroTrackId)(this.room,
|
|
363
|
+
this.#localTrackSid = (0, import_utils.findMicroTrackId)(this.room, this.room.localParticipant.identity);
|
|
340
364
|
}
|
|
341
365
|
return this.#localTrackSid;
|
|
342
366
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { MultimodalLLMMetrics } from '../metrics/base.js';\nimport { BasicTranscriptionForwarder } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: Promise<void> | null = null;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.transcriptionFwd.text;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n const trFwd = new BasicTranscriptionForwarder(\n this.room!,\n this.room!.localParticipant!.identity,\n this.#getLocalTrackSid()!,\n message.responseId,\n );\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n trFwd,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', (ev: any) => {\n this.emit('user_started_speaking');\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n this.#session.on('metrics_collected', (metrics: MultimodalLLMMetrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n this.readMicroTask = new Promise<void>((resolve, reject) => {\n readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels))\n .then(resolve)\n .catch(reject);\n });\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant?.identity);\n }\n return this.#localTrackSid;\n }\n\n #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): void {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes[AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWA,sBAOO;AACP,yBAA6B;AAC7B,mBAAgC;AAChC,UAAqB;AACrB,iBAAoB;AAEpB,2BAA4C;AAC5C,mBAAiC;AACjC,2BAAiD;AAM1C,MAAe,wBAAwB,gCAAa;AAM3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,gCAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAsC;AAAA,EAEtC,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAIG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AAAA,EACjB;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,cAAU,gBAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EAExC,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AAzIlD;AA0IM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,0BAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAQ;AAAA,MAC5C,CAAC;AACD,WAAK;AAAA,QACH,0BAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,4BAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,0BAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,4BAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,iBAAiB;AAChD,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gCAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oCAAoB;AACxC,cAAQ,SAAS,4BAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAQ;AAAA,QAC5C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAQ;AAC1C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AA5OnE,YAAAC;AA8OQ,cAAM,QAAQ,IAAI;AAAA,UAChB,KAAK;AAAA,UACL,KAAK,KAAM,iBAAkB;AAAA,UAC7B,KAAK,kBAAkB;AAAA,UACvB,QAAQ;AAAA,QACV;AAEA,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,CAAC,OAAY;AAhQ9D,YAAAA,KAAA;AAkQQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,CAAC,OAAY;AA5Q5E,YAAAA,KAAA;AA8QQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,eAAe,MAAM,GAAG,MAAM;AAAA,QAC1F,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AA9R5D,YAAAA,KAAA;AA+RQ,aAAK,KAAK,uBAAuB;AACjC,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,WAAK,SAAS,GAAG,qBAAqB,CAAC,YAAkC;AACvE,aAAK,KAAK,qBAAqB,OAAO;AAAA,MACxC,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AArYJ;AAsYI,QACE,YAAY,WAAW,4BAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,SAAK,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC1D,0BAAoB,IAAI,4BAAY,OAAO,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW,CAAC,EACtF,KAAK,OAAO,EACZ,MAAM,MAAM;AAAA,IACjB,CAAC;AAAA,EACH;AAAA,EAEA,oBAAmC;AAnarC;AAoaI,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,qBAAiB,+BAAiB,KAAK,OAAM,UAAK,KAAK,qBAAV,mBAA4B,QAAQ;AAAA,IACxF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,sBACE,qBACA,UACA,MACA,SACA,IACM;AAhbV;AAibI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AAtd/B;AAudI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAW,qBAAqB;AAChF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
|
|
1
|
+
{"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { MultimodalLLMMetrics } from '../metrics/base.js';\nimport { BasicTranscriptionForwarder } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n abstract recoverFromTextResponse(itemId: string): void;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: Promise<void> | null = null;\n\n #textResponseRetries = 0;\n #maxTextResponseRetries: number;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n maxTextResponseRetries = 5,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n maxTextResponseRetries?: number;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n this.#maxTextResponseRetries = maxTextResponseRetries;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity!);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.transcriptionFwd.text;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity!);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity!);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') return;\n\n const trFwd = new BasicTranscriptionForwarder(\n this.room!,\n this.room!.localParticipant!.identity!,\n this.#getLocalTrackSid()!,\n message.responseId,\n );\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n trFwd,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_done', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') {\n if (this.#textResponseRetries >= this.#maxTextResponseRetries) {\n throw new Error(\n 'The OpenAI Realtime API returned a text response ' +\n `after ${this.#maxTextResponseRetries} retries. ` +\n 'Please try to reduce the number of text system or ' +\n 'assistant messages in the chat context.',\n );\n }\n\n this.#textResponseRetries++;\n this.#logger\n .child({\n itemId: message.itemId,\n text: message.text,\n retries: this.#textResponseRetries,\n })\n .warn(\n 'The OpenAI Realtime API returned a text response instead of audio. ' +\n 'Attempting to recover to audio mode...',\n );\n this.#session!.recoverFromTextResponse(message.itemId);\n } else {\n this.#textResponseRetries = 0;\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', (ev: any) => {\n this.emit('user_started_speaking');\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n this.#session.on('metrics_collected', (metrics: MultimodalLLMMetrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n this.readMicroTask = new Promise<void>((resolve, reject) => {\n readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels))\n .then(resolve)\n .catch(reject);\n });\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant!.identity!);\n }\n return this.#localTrackSid;\n }\n\n #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): void {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes![AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWA,sBAOO;AACP,yBAA6B;AAC7B,mBAAgC;AAChC,UAAqB;AACrB,iBAAoB;AAEpB,2BAA4C;AAC5C,mBAAiC;AACjC,2BAAiD;AAM1C,MAAe,wBAAwB,gCAAa;AAO3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,gCAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAsC;AAAA,EAEtC,uBAAuB;AAAA,EACvB;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA,yBAAyB;AAAA,EAC3B,GAKG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AACf,SAAK,0BAA0B;AAAA,EACjC;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,cAAU,gBAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EAExC,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AAhJlD;AAiJM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,0BAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAS;AAAA,MAC7C,CAAC;AACD,WAAK;AAAA,QACH,0BAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,4BAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,0BAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,4BAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,iBAAiB;AAChD,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gCAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oCAAoB;AACxC,cAAQ,SAAS,4BAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAS;AAAA,QAC7C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAS;AAC3C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AAnPnE,YAAAC;AAqPQ,YAAI,QAAQ,gBAAgB,OAAQ;AAEpC,cAAM,QAAQ,IAAI;AAAA,UAChB,KAAK;AAAA,UACL,KAAK,KAAM,iBAAkB;AAAA,UAC7B,KAAK,kBAAkB;AAAA,UACvB,QAAQ;AAAA,QACV;AAEA,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,YAAiB;AAE1D,YAAI,QAAQ,gBAAgB,QAAQ;AAClC,cAAI,KAAK,wBAAwB,KAAK,yBAAyB;AAC7D,kBAAM,IAAI;AAAA,cACR,0DACW,KAAK,uBAAuB;AAAA,YAGzC;AAAA,UACF;AAEA,eAAK;AACL,eAAK,QACF,MAAM;AAAA,YACL,QAAQ,QAAQ;AAAA,YAChB,MAAM,QAAQ;AAAA,YACd,SAAS,KAAK;AAAA,UAChB,CAAC,EACA;AAAA,YACC;AAAA,UAEF;AACF,eAAK,SAAU,wBAAwB,QAAQ,MAAM;AAAA,QACvD,OAAO;AACL,eAAK,uBAAuB;AAAA,QAC9B;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,CAAC,OAAY;AAvS9D,YAAAA,KAAA;AAySQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,CAAC,OAAY;AAnT5E,YAAAA,KAAA;AAqTQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,eAAe,MAAM,GAAG,MAAM;AAAA,QAC1F,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AArU5D,YAAAA,KAAA;AAsUQ,aAAK,KAAK,uBAAuB;AACjC,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,WAAK,SAAS,GAAG,qBAAqB,CAAC,YAAkC;AACvE,aAAK,KAAK,qBAAqB,OAAO;AAAA,MACxC,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AA5aJ;AA6aI,QACE,YAAY,WAAW,4BAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,SAAK,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC1D,0BAAoB,IAAI,4BAAY,OAAO,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW,CAAC,EACtF,KAAK,OAAO,EACZ,MAAM,MAAM;AAAA,IACjB,CAAC;AAAA,EACH;AAAA,EAEA,oBAAmC;AACjC,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,qBAAiB,+BAAiB,KAAK,MAAM,KAAK,KAAK,iBAAkB,QAAS;AAAA,IACzF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,sBACE,qBACA,UACA,MACA,SACA,IACM;AAvdV;AAwdI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AA7f/B;AA8fI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAY,qBAAqB;AACjF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
|
|
@@ -10,6 +10,7 @@ export declare abstract class RealtimeSession extends EventEmitter {
|
|
|
10
10
|
abstract conversation: any;
|
|
11
11
|
abstract inputAudioBuffer: any;
|
|
12
12
|
abstract fncCtx: llm.FunctionContext | undefined;
|
|
13
|
+
abstract recoverFromTextResponse(itemId: string): void;
|
|
13
14
|
}
|
|
14
15
|
/**
|
|
15
16
|
* @internal
|
|
@@ -33,10 +34,11 @@ export declare class MultimodalAgent extends EventEmitter {
|
|
|
33
34
|
linkedParticipant: RemoteParticipant | null;
|
|
34
35
|
subscribedTrack: RemoteAudioTrack | null;
|
|
35
36
|
readMicroTask: Promise<void> | null;
|
|
36
|
-
constructor({ model, chatCtx, fncCtx, }: {
|
|
37
|
+
constructor({ model, chatCtx, fncCtx, maxTextResponseRetries, }: {
|
|
37
38
|
model: RealtimeModel;
|
|
38
39
|
chatCtx?: llm.ChatContext;
|
|
39
40
|
fncCtx?: llm.FunctionContext;
|
|
41
|
+
maxTextResponseRetries?: number;
|
|
40
42
|
});
|
|
41
43
|
get fncCtx(): llm.FunctionContext | undefined;
|
|
42
44
|
set fncCtx(ctx: llm.FunctionContext | undefined);
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"multimodal_agent.d.ts","sourceRoot":"","sources":["../../src/multimodal/multimodal_agent.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAEV,gBAAgB,EAChB,iBAAiB,EAGjB,IAAI,EACL,MAAM,mBAAmB,CAAC;AAS3B,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,KAAK,GAAG,MAAM,iBAAiB,CAAC;AAOvC;;;GAGG;AACH,8BAAsB,eAAgB,SAAQ,YAAY;IAExD,QAAQ,CAAC,YAAY,EAAE,GAAG,CAAC;IAE3B,QAAQ,CAAC,gBAAgB,EAAE,GAAG,CAAC;IAC/B,QAAQ,CAAC,MAAM,EAAE,GAAG,CAAC,eAAe,GAAG,SAAS,CAAC;
|
|
1
|
+
{"version":3,"file":"multimodal_agent.d.ts","sourceRoot":"","sources":["../../src/multimodal/multimodal_agent.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAEV,gBAAgB,EAChB,iBAAiB,EAGjB,IAAI,EACL,MAAM,mBAAmB,CAAC;AAS3B,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,KAAK,GAAG,MAAM,iBAAiB,CAAC;AAOvC;;;GAGG;AACH,8BAAsB,eAAgB,SAAQ,YAAY;IAExD,QAAQ,CAAC,YAAY,EAAE,GAAG,CAAC;IAE3B,QAAQ,CAAC,gBAAgB,EAAE,GAAG,CAAC;IAC/B,QAAQ,CAAC,MAAM,EAAE,GAAG,CAAC,eAAe,GAAG,SAAS,CAAC;IACjD,QAAQ,CAAC,uBAAuB,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;CACvD;AAED;;;GAGG;AACH,8BAAsB,aAAa;IAEjC,QAAQ,CAAC,OAAO,CAAC,OAAO,EAAE,GAAG,GAAG,eAAe;IAC/C,QAAQ,CAAC,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAC/B,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;CAC/B;AAED,MAAM,MAAM,UAAU,GAAG,cAAc,GAAG,UAAU,GAAG,WAAW,GAAG,UAAU,CAAC;AAChF,eAAO,MAAM,qBAAqB,mBAAmB,CAAC;AAEtD,YAAY;AACZ,qBAAa,eAAgB,SAAQ,YAAY;;IAC/C,KAAK,EAAE,aAAa,CAAC;IACrB,IAAI,EAAE,IAAI,GAAG,IAAI,CAAQ;IACzB,iBAAiB,EAAE,iBAAiB,GAAG,IAAI,CAAQ;IACnD,eAAe,EAAE,gBAAgB,GAAG,IAAI,CAAQ;IAChD,aAAa,EAAE,OAAO,CAAC,IAAI,CAAC,GAAG,IAAI,CAAQ;gBAK/B,EACV,KAAK,EACL,OAAO,EACP,MAAM,EACN,sBAA0B,GAC3B,EAAE;QACD,KAAK,EAAE,aAAa,CAAC;QACrB,OAAO,CAAC,EAAE,GAAG,CAAC,WAAW,CAAC;QAC1B,MAAM,CAAC,EAAE,GAAG,CAAC,eAAe,CAAC;QAC7B,sBAAsB,CAAC,EAAE,MAAM,CAAC;KACjC;IAuBD,IAAI,MAAM,IAAI,GAAG,CAAC,eAAe,GAAG,SAAS,CAE5C;IAED,IAAI,MAAM,CAAC,GAAG,EAAE,GAAG,CAAC,eAAe,GAAG,SAAS,EAK9C;IA6BD,KAAK,CACH,IAAI,EAAE,IAAI,EACV,WAAW,GAAE,iBAAiB,GAAG,MAAM,GAAG,IAAW,GACpD,OAAO,CAAC,eAAe,CAAC;CAyX5B"}
|
|
@@ -24,15 +24,19 @@ class MultimodalAgent extends EventEmitter {
|
|
|
24
24
|
linkedParticipant = null;
|
|
25
25
|
subscribedTrack = null;
|
|
26
26
|
readMicroTask = null;
|
|
27
|
+
#textResponseRetries = 0;
|
|
28
|
+
#maxTextResponseRetries;
|
|
27
29
|
constructor({
|
|
28
30
|
model,
|
|
29
31
|
chatCtx,
|
|
30
|
-
fncCtx
|
|
32
|
+
fncCtx,
|
|
33
|
+
maxTextResponseRetries = 5
|
|
31
34
|
}) {
|
|
32
35
|
super();
|
|
33
36
|
this.model = model;
|
|
34
37
|
this.#chatCtx = chatCtx;
|
|
35
38
|
this.#fncCtx = fncCtx;
|
|
39
|
+
this.#maxTextResponseRetries = maxTextResponseRetries;
|
|
36
40
|
}
|
|
37
41
|
#participant = null;
|
|
38
42
|
#agentPublication = null;
|
|
@@ -161,6 +165,7 @@ class MultimodalAgent extends EventEmitter {
|
|
|
161
165
|
this.#started = true;
|
|
162
166
|
this.#session.on("response_content_added", (message) => {
|
|
163
167
|
var _a2;
|
|
168
|
+
if (message.contentType === "text") return;
|
|
164
169
|
const trFwd = new BasicTranscriptionForwarder(
|
|
165
170
|
this.room,
|
|
166
171
|
this.room.localParticipant.identity,
|
|
@@ -176,6 +181,26 @@ class MultimodalAgent extends EventEmitter {
|
|
|
176
181
|
);
|
|
177
182
|
this.#playingHandle = handle;
|
|
178
183
|
});
|
|
184
|
+
this.#session.on("response_content_done", (message) => {
|
|
185
|
+
if (message.contentType === "text") {
|
|
186
|
+
if (this.#textResponseRetries >= this.#maxTextResponseRetries) {
|
|
187
|
+
throw new Error(
|
|
188
|
+
`The OpenAI Realtime API returned a text response after ${this.#maxTextResponseRetries} retries. Please try to reduce the number of text system or assistant messages in the chat context.`
|
|
189
|
+
);
|
|
190
|
+
}
|
|
191
|
+
this.#textResponseRetries++;
|
|
192
|
+
this.#logger.child({
|
|
193
|
+
itemId: message.itemId,
|
|
194
|
+
text: message.text,
|
|
195
|
+
retries: this.#textResponseRetries
|
|
196
|
+
}).warn(
|
|
197
|
+
"The OpenAI Realtime API returned a text response instead of audio. Attempting to recover to audio mode..."
|
|
198
|
+
);
|
|
199
|
+
this.#session.recoverFromTextResponse(message.itemId);
|
|
200
|
+
} else {
|
|
201
|
+
this.#textResponseRetries = 0;
|
|
202
|
+
}
|
|
203
|
+
});
|
|
179
204
|
this.#session.on("input_speech_committed", (ev) => {
|
|
180
205
|
var _a2, _b;
|
|
181
206
|
const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
|
|
@@ -305,9 +330,8 @@ class MultimodalAgent extends EventEmitter {
|
|
|
305
330
|
});
|
|
306
331
|
}
|
|
307
332
|
#getLocalTrackSid() {
|
|
308
|
-
var _a;
|
|
309
333
|
if (!this.#localTrackSid && this.room && this.room.localParticipant) {
|
|
310
|
-
this.#localTrackSid = findMicroTrackId(this.room,
|
|
334
|
+
this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant.identity);
|
|
311
335
|
}
|
|
312
336
|
return this.#localTrackSid;
|
|
313
337
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { MultimodalLLMMetrics } from '../metrics/base.js';\nimport { BasicTranscriptionForwarder } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: Promise<void> | null = null;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.transcriptionFwd.text;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n const trFwd = new BasicTranscriptionForwarder(\n this.room!,\n this.room!.localParticipant!.identity,\n this.#getLocalTrackSid()!,\n message.responseId,\n );\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n trFwd,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', (ev: any) => {\n this.emit('user_started_speaking');\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n this.#session.on('metrics_collected', (metrics: MultimodalLLMMetrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n this.readMicroTask = new Promise<void>((resolve, reject) => {\n readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels))\n .then(resolve)\n .catch(reject);\n });\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant?.identity);\n }\n return this.#localTrackSid;\n }\n\n #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): void {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes[AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":"AAWA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,oBAAoB;AAC7B,SAAS,uBAAuB;AAChC,YAAY,SAAS;AACrB,SAAS,WAAW;AAEpB,SAAS,mCAAmC;AAC5C,SAAS,wBAAwB;AACjC,SAAS,oBAAwC;AAM1C,MAAe,wBAAwB,aAAa;AAM3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,aAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAsC;AAAA,EAEtC,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAIG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AAAA,EACjB;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,UAAU,IAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EAExC,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AAzIlD;AA0IM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,UAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAQ;AAAA,MAC5C,CAAC;AACD,WAAK;AAAA,QACH,UAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,YAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,UAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,YAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,iBAAiB;AAChD,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gBAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oBAAoB;AACxC,cAAQ,SAAS,YAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAQ;AAAA,QAC5C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAQ;AAC1C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AA5OnE,YAAAC;AA8OQ,cAAM,QAAQ,IAAI;AAAA,UAChB,KAAK;AAAA,UACL,KAAK,KAAM,iBAAkB;AAAA,UAC7B,KAAK,kBAAkB;AAAA,UACvB,QAAQ;AAAA,QACV;AAEA,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,CAAC,OAAY;AAhQ9D,YAAAA,KAAA;AAkQQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,CAAC,OAAY;AA5Q5E,YAAAA,KAAA;AA8QQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,eAAe,MAAM,GAAG,MAAM;AAAA,QAC1F,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AA9R5D,YAAAA,KAAA;AA+RQ,aAAK,KAAK,uBAAuB;AACjC,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,WAAK,SAAS,GAAG,qBAAqB,CAAC,YAAkC;AACvE,aAAK,KAAK,qBAAqB,OAAO;AAAA,MACxC,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AArYJ;AAsYI,QACE,YAAY,WAAW,YAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,SAAK,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC1D,0BAAoB,IAAI,YAAY,OAAO,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW,CAAC,EACtF,KAAK,OAAO,EACZ,MAAM,MAAM;AAAA,IACjB,CAAC;AAAA,EACH;AAAA,EAEA,oBAAmC;AAnarC;AAoaI,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,iBAAiB,iBAAiB,KAAK,OAAM,UAAK,KAAK,qBAAV,mBAA4B,QAAQ;AAAA,IACxF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,sBACE,qBACA,UACA,MACA,SACA,IACM;AAhbV;AAibI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AAtd/B;AAudI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAW,qBAAqB;AAChF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
|
|
1
|
+
{"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { MultimodalLLMMetrics } from '../metrics/base.js';\nimport { BasicTranscriptionForwarder } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n abstract recoverFromTextResponse(itemId: string): void;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: Promise<void> | null = null;\n\n #textResponseRetries = 0;\n #maxTextResponseRetries: number;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n maxTextResponseRetries = 5,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n maxTextResponseRetries?: number;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n this.#maxTextResponseRetries = maxTextResponseRetries;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity!);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.transcriptionFwd.text;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity!);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity!);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') return;\n\n const trFwd = new BasicTranscriptionForwarder(\n this.room!,\n this.room!.localParticipant!.identity!,\n this.#getLocalTrackSid()!,\n message.responseId,\n );\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n trFwd,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_done', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') {\n if (this.#textResponseRetries >= this.#maxTextResponseRetries) {\n throw new Error(\n 'The OpenAI Realtime API returned a text response ' +\n `after ${this.#maxTextResponseRetries} retries. ` +\n 'Please try to reduce the number of text system or ' +\n 'assistant messages in the chat context.',\n );\n }\n\n this.#textResponseRetries++;\n this.#logger\n .child({\n itemId: message.itemId,\n text: message.text,\n retries: this.#textResponseRetries,\n })\n .warn(\n 'The OpenAI Realtime API returned a text response instead of audio. ' +\n 'Attempting to recover to audio mode...',\n );\n this.#session!.recoverFromTextResponse(message.itemId);\n } else {\n this.#textResponseRetries = 0;\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', (ev: any) => {\n this.emit('user_started_speaking');\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n this.#session.on('metrics_collected', (metrics: MultimodalLLMMetrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n this.readMicroTask = new Promise<void>((resolve, reject) => {\n readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels))\n .then(resolve)\n .catch(reject);\n });\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant!.identity!);\n }\n return this.#localTrackSid;\n }\n\n #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): void {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes![AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":"AAWA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,oBAAoB;AAC7B,SAAS,uBAAuB;AAChC,YAAY,SAAS;AACrB,SAAS,WAAW;AAEpB,SAAS,mCAAmC;AAC5C,SAAS,wBAAwB;AACjC,SAAS,oBAAwC;AAM1C,MAAe,wBAAwB,aAAa;AAO3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,aAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAsC;AAAA,EAEtC,uBAAuB;AAAA,EACvB;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA,yBAAyB;AAAA,EAC3B,GAKG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AACf,SAAK,0BAA0B;AAAA,EACjC;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,UAAU,IAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EAExC,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AAhJlD;AAiJM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,UAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAS;AAAA,MAC7C,CAAC;AACD,WAAK;AAAA,QACH,UAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,YAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,UAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,YAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,iBAAiB;AAChD,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gBAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oBAAoB;AACxC,cAAQ,SAAS,YAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAS;AAAA,QAC7C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAS;AAC3C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AAnPnE,YAAAC;AAqPQ,YAAI,QAAQ,gBAAgB,OAAQ;AAEpC,cAAM,QAAQ,IAAI;AAAA,UAChB,KAAK;AAAA,UACL,KAAK,KAAM,iBAAkB;AAAA,UAC7B,KAAK,kBAAkB;AAAA,UACvB,QAAQ;AAAA,QACV;AAEA,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,YAAiB;AAE1D,YAAI,QAAQ,gBAAgB,QAAQ;AAClC,cAAI,KAAK,wBAAwB,KAAK,yBAAyB;AAC7D,kBAAM,IAAI;AAAA,cACR,0DACW,KAAK,uBAAuB;AAAA,YAGzC;AAAA,UACF;AAEA,eAAK;AACL,eAAK,QACF,MAAM;AAAA,YACL,QAAQ,QAAQ;AAAA,YAChB,MAAM,QAAQ;AAAA,YACd,SAAS,KAAK;AAAA,UAChB,CAAC,EACA;AAAA,YACC;AAAA,UAEF;AACF,eAAK,SAAU,wBAAwB,QAAQ,MAAM;AAAA,QACvD,OAAO;AACL,eAAK,uBAAuB;AAAA,QAC9B;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,CAAC,OAAY;AAvS9D,YAAAA,KAAA;AAySQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,CAAC,OAAY;AAnT5E,YAAAA,KAAA;AAqTQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,eAAe,MAAM,GAAG,MAAM;AAAA,QAC1F,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AArU5D,YAAAA,KAAA;AAsUQ,aAAK,KAAK,uBAAuB;AACjC,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,WAAK,SAAS,GAAG,qBAAqB,CAAC,YAAkC;AACvE,aAAK,KAAK,qBAAqB,OAAO;AAAA,MACxC,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,YAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AA5aJ;AA6aI,QACE,YAAY,WAAW,YAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,SAAK,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC1D,0BAAoB,IAAI,YAAY,OAAO,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW,CAAC,EACtF,KAAK,OAAO,EACZ,MAAM,MAAM;AAAA,IACjB,CAAC;AAAA,EACH;AAAA,EAEA,oBAAmC;AACjC,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,iBAAiB,iBAAiB,KAAK,MAAM,KAAK,KAAK,iBAAkB,QAAS;AAAA,IACzF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,sBACE,qBACA,UACA,MACA,SACA,IACM;AAvdV;AAwdI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AA7f/B;AA8fI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAY,qBAAqB;AACjF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
|
|
@@ -323,7 +323,7 @@ class VoicePipelineAgent extends import_node_events.default {
|
|
|
323
323
|
}
|
|
324
324
|
});
|
|
325
325
|
this.#humanInput.on(import_human_input.HumanInputEvent.END_OF_SPEECH, (event) => {
|
|
326
|
-
this.emit(
|
|
326
|
+
this.emit(1 /* USER_STOPPED_SPEAKING */);
|
|
327
327
|
this.#deferredValidation.onHumanEndOfSpeech(event);
|
|
328
328
|
});
|
|
329
329
|
this.#humanInput.on(import_human_input.HumanInputEvent.INTERIM_TRANSCRIPT, (event) => {
|
|
@@ -683,6 +683,7 @@ class DeferredReplyValidation {
|
|
|
683
683
|
#speaking = false;
|
|
684
684
|
#endOfSpeechDelay;
|
|
685
685
|
#finalTranscriptDelay;
|
|
686
|
+
#abort;
|
|
686
687
|
constructor(validateFunc, minEndpointingDelay) {
|
|
687
688
|
this.#validateFunc = validateFunc;
|
|
688
689
|
this.#endOfSpeechDelay = minEndpointingDelay;
|
|
@@ -701,14 +702,18 @@ class DeferredReplyValidation {
|
|
|
701
702
|
}
|
|
702
703
|
// eslint-disable-next-line @typescript-eslint/no-unused-vars
|
|
703
704
|
onHumanStartOfSpeech(_) {
|
|
705
|
+
var _a;
|
|
704
706
|
this.#speaking = true;
|
|
707
|
+
if (this.validating) {
|
|
708
|
+
(_a = this.#abort) == null ? void 0 : _a.abort();
|
|
709
|
+
}
|
|
705
710
|
}
|
|
706
711
|
// eslint-disable-next-line @typescript-eslint/no-unused-vars
|
|
707
712
|
onHumanEndOfSpeech(_) {
|
|
708
713
|
this.#speaking = false;
|
|
709
714
|
this.#lastRecvEndOfSpeechTime = Date.now();
|
|
710
715
|
if (this.#lastFinalTranscript) {
|
|
711
|
-
const delay = this.#endWithPunctuation() ? this.#endOfSpeechDelay * this.PUNCTUATION_REDUCE_FACTOR :
|
|
716
|
+
const delay = this.#endWithPunctuation() ? this.#endOfSpeechDelay * this.PUNCTUATION_REDUCE_FACTOR : 1e3;
|
|
712
717
|
this.#run(delay);
|
|
713
718
|
}
|
|
714
719
|
}
|
|
@@ -721,13 +726,20 @@ class DeferredReplyValidation {
|
|
|
721
726
|
this.#lastRecvEndOfSpeechTime = 0;
|
|
722
727
|
}
|
|
723
728
|
#run(delay) {
|
|
724
|
-
|
|
725
|
-
|
|
726
|
-
|
|
727
|
-
|
|
729
|
+
var _a;
|
|
730
|
+
const runTask = async (delay2, signal) => {
|
|
731
|
+
const timeout = setTimeout(() => {
|
|
732
|
+
this.#resetStates();
|
|
733
|
+
this.#validateFunc();
|
|
734
|
+
}, delay2);
|
|
735
|
+
signal.addEventListener("abort", () => {
|
|
736
|
+
clearTimeout(timeout);
|
|
737
|
+
});
|
|
728
738
|
};
|
|
739
|
+
(_a = this.#abort) == null ? void 0 : _a.abort();
|
|
740
|
+
this.#abort = new AbortController();
|
|
729
741
|
this.#validatingFuture = new import_utils.Future();
|
|
730
|
-
this.#validatingPromise = runTask(delay);
|
|
742
|
+
this.#validatingPromise = runTask(delay, this.#abort.signal);
|
|
731
743
|
}
|
|
732
744
|
}
|
|
733
745
|
// Annotate the CommonJS export names for ESM import in node:
|