@pagelines/sdk 1.0.822 → 1.0.824

package/dist/provider.js

@@ -1,5 +1,8 @@
 import "./rolldown-runtime.js";
 //#region agent/voice/provider.ts
+function isVoiceAudible(e) {
+	return e >= .08;
+}
 var createVoiceProviderClient = async (e) => {
 	switch (e.provider) {
 		case "elevenlabs": {
@@ -35,6 +38,6 @@ function voiceRangeSpectrum(e, t = 9) {
 	});
 }
 //#endregion
-export { voiceRangeSpectrum as n, voiceSpectrum as r, createVoiceProviderClient as t };
+export { voiceSpectrum as i, isVoiceAudible as n, voiceRangeSpectrum as r, createVoiceProviderClient as t };
 
 //# sourceMappingURL=provider.js.map

package/dist/provider.js.map

@@ -1 +1 @@
-{"version":3,"file":"provider.js","names":[],"sources":["../agent/voice/provider.ts"],"sourcesContent":["import type { VoiceSessionMint } from '@pagelines/core'\n\nexport interface VoiceLevelEvent {\n  source: 'mic' | 'playback'\n  /** Normalized 0..1 signal level. */\n  level: number\n  /** Speech-spectrum energies, low pitch first. */\n  bands?: readonly number[]\n}\n\nexport interface VoiceSessionError {\n  code: string\n  message: string\n}\n\nexport type VoiceDisconnectCause = 'agent' | 'error' | 'network' | 'unknown'\n\nexport interface VoiceProviderClient {\n  connect: (handlers: {\n    onDisconnected: (cause: VoiceDisconnectCause) => void\n    onError: (error: VoiceSessionError) => void\n    onThinkingChange?: (thinking: boolean) => void\n  }) => Promise<void>\n  disconnect: () => Promise<void>\n  sampleLevel: () => VoiceLevelEvent | undefined\n  /** Self-mute without ending the live session. */\n  setMuted?: (muted: boolean) => Promise<void>\n  /**\n   * Hand the provider text the caller typed during a live call, as though\n   * they had said it — the assistant answers aloud and the turn lands in the\n   * transcript through the normal provider path. Optional: a provider that\n   * cannot inject text simply omits it and the surface hides the affordance.\n   */\n  sendUserMessage?: (text: string) => void\n}\n\nexport type CreateVoiceProviderClient = (mint: VoiceSessionMint) => Promise<VoiceProviderClient>\n\n/** Provider selection is a closed union; each SDK remains in its lazy chunk. */\nexport const createVoiceProviderClient: CreateVoiceProviderClient = async (mint) => {\n  switch (mint.provider) {\n    case 'elevenlabs': {\n      const { ElevenLabsVoiceProvider } = await import('./providers/elevenlabs/provider')\n      return new ElevenLabsVoiceProvider(mint)\n    }\n    case 'livekit': {\n      const { LiveKitVoiceProvider } = await import('./providers/livekit/provider')\n      return new LiveKitVoiceProvider(mint)\n    }\n  }\n}\n\nexport const VOICE_SPECTRUM_BANDS = 9\n\n/** Log-spaced speech spectrum for transports that expose FFT bins. */\nexport function voiceSpectrum(frequencies: Uint8Array, sampleRate: number, fftSize: number, bandCount = VOICE_SPECTRUM_BANDS): readonly number[] {\n  if (!Number.isFinite(sampleRate) || sampleRate <= 0 || !Number.isFinite(fftSize) || fftSize <= 0 || bandCount <= 0)\n    return []\n\n  const hzPerBin = sampleRate / fftSize\n  const fromHz = 80\n  const ratio = 4_000 / fromHz\n  const bands: number[] = []\n  for (let band = 0; band < bandCount; band++) {\n    const start = Math.max(0, Math.ceil((fromHz * ratio ** (band / bandCount)) / hzPerBin))\n    const end = Math.min(frequencies.length, Math.ceil((fromHz * ratio ** ((band + 1) / bandCount)) / hzPerBin))\n    if (end <= start) {\n      bands.push(0)\n      continue\n    }\n    let squared = 0\n    for (let index = start; index < end; index++)\n      squared += (frequencies[index] / 255) ** 2\n    bands.push(Math.sqrt(squared / (end - start)))\n  }\n  return bands\n}\n\n/** Equal-width bands for SDKs that already expose voice-range FFT bins. */\nexport function voiceRangeSpectrum(frequencies: Uint8Array, bandCount = VOICE_SPECTRUM_BANDS): readonly number[] {\n  if (!frequencies.length || bandCount <= 0)\n    return []\n  return Array.from({ length: bandCount }, (_, band) => {\n    const start = Math.floor((band * frequencies.length) / bandCount)\n    const end = Math.max(start + 1, Math.floor(((band + 1) * frequencies.length) / bandCount))\n    let squared = 0\n    for (let index = start; index < Math.min(end, frequencies.length); index++)\n      squared += (frequencies[index] / 255) ** 2\n    return Math.sqrt(squared / Math.max(1, Math.min(end, frequencies.length) - start))\n  })\n}\n"],"mappings":";;AAuCA,IAAa,4BAAuD,OAAO,MAAS;CAClF,QAAQ,EAAK,UAAb;EACE,KAAK,cAAc;GACjB,IAAM,EAAE,+BAA4B,MAAM,OAAO;GACjD,OAAO,IAAI,EAAwB,CAAI;EACzC;EACA,KAAK,WAAW;GACd,IAAM,EAAE,4BAAyB,MAAM,OAAO;GAC9C,OAAO,IAAI,EAAqB,CAAI;EACtC;CACF;AACF;AAKA,SAAgB,cAAc,GAAyB,GAAoB,GAAiB,IAAA,GAAqD;CAC/I,IAAI,CAAC,OAAO,SAAS,CAAU,KAAK,KAAc,KAAK,CAAC,OAAO,SAAS,CAAO,KAAK,KAAW,KAAK,KAAa,GAC/G,OAAO,CAAC;CAEV,IAAM,IAAW,IAAa,GAGxB,IAAkB,CAAC;CACzB,KAAK,IAAI,IAAO,GAAG,IAAO,GAAW,KAAQ;EAC3C,IAAM,IAAQ,KAAK,IAAI,GAAG,KAAK,KAAM,KAAS,OAAU,IAAO,KAAc,CAAQ,CAAC,GAChF,IAAM,KAAK,IAAI,EAAY,QAAQ,KAAK,KAAM,KAAS,QAAW,IAAO,KAAK,KAAc,CAAQ,CAAC;EAC3G,IAAI,KAAO,GAAO;GAChB,EAAM,KAAK,CAAC;GACZ;EACF;EACA,IAAI,IAAU;EACd,KAAK,IAAI,IAAQ,GAAO,IAAQ,GAAK,KACnC,MAAY,EAAY,KAAS,QAAQ;EAC3C,EAAM,KAAK,KAAK,KAAK,KAAW,IAAM,EAAM,CAAC;CAC/C;CACA,OAAO;AACT;AAGA,SAAgB,mBAAmB,GAAyB,IAAA,GAAqD;CAG/G,OAFI,CAAC,EAAY,UAAU,KAAa,IAC/B,CAAC,IACH,MAAM,KAAK,EAAE,QAAQ,EAAU,IAAI,GAAG,MAAS;EACpD,IAAM,IAAQ,KAAK,MAAO,IAAO,EAAY,SAAU,CAAS,GAC1D,IAAM,KAAK,IAAI,IAAQ,GAAG,KAAK,OAAQ,IAAO,KAAK,EAAY,SAAU,CAAS,CAAC,GACrF,IAAU;EACd,KAAK,IAAI,IAAQ,GAAO,IAAQ,KAAK,IAAI,GAAK,EAAY,MAAM,GAAG,KACjE,MAAY,EAAY,KAAS,QAAQ;EAC3C,OAAO,KAAK,KAAK,IAAU,KAAK,IAAI,GAAG,KAAK,IAAI,GAAK,EAAY,MAAM,IAAI,CAAK,CAAC;CACnF,CAAC;AACH"}
+{"version":3,"file":"provider.js","names":[],"sources":["../agent/voice/provider.ts"],"sourcesContent":["import type { VoiceSessionMint } from '@pagelines/core'\n\nexport interface VoiceLevelEvent {\n  source: 'mic' | 'playback'\n  /** Normalized 0..1 signal level. */\n  level: number\n  /** Speech-spectrum energies, low pitch first. */\n  bands?: readonly number[]\n}\n\n/** One speech/noise boundary shared by audible UI state and work-cue ducking. */\nexport function isVoiceAudible(level: number): boolean {\n  return level >= 0.08\n}\n\nexport interface VoiceSessionError {\n  code: string\n  message: string\n}\n\nexport type VoiceDisconnectCause = 'agent' | 'error' | 'network' | 'unknown'\n\n/**\n * Who the transport says is talking. Raw energy is not a speech signal — room\n * tone and the speaker's own playback both read as audible — so ducking,\n * tone, and which meter the surface follows come from here.\n */\nexport interface VoiceSpeakingState {\n  user: boolean\n  agent: boolean\n}\n\nexport interface VoiceProviderClient {\n  connect: (handlers: {\n    onDisconnected: (cause: VoiceDisconnectCause) => void\n    onError: (error: VoiceSessionError) => void\n    onThinkingChange?: (thinking: boolean) => void\n    onSpeakingChange?: (state: VoiceSpeakingState) => void\n  }) => Promise<void>\n  disconnect: () => Promise<void>\n  /** Loudest source by default; `prefer` pins a meter while someone is known to be speaking. */\n  sampleLevel: (prefer?: VoiceLevelEvent['source']) => VoiceLevelEvent | undefined\n  /** Self-mute without ending the live session. */\n  setMuted?: (muted: boolean) => Promise<void>\n  /**\n   * Hand the provider text the caller typed during a live call, as though\n   * they had said it — the assistant answers aloud and the turn lands in the\n   * transcript through the normal provider path. Optional: a provider that\n   * cannot inject text simply omits it and the surface hides the affordance.\n   */\n  sendUserMessage?: (text: string) => void\n}\n\nexport type CreateVoiceProviderClient = (mint: VoiceSessionMint) => Promise<VoiceProviderClient>\n\n/** Provider selection is a closed union; each SDK remains in its lazy chunk. */\nexport const createVoiceProviderClient: CreateVoiceProviderClient = async (mint) => {\n  switch (mint.provider) {\n    case 'elevenlabs': {\n      const { ElevenLabsVoiceProvider } = await import('./providers/elevenlabs/provider')\n      return new ElevenLabsVoiceProvider(mint)\n    }\n    case 'livekit': {\n      const { LiveKitVoiceProvider } = await import('./providers/livekit/provider')\n      return new LiveKitVoiceProvider(mint)\n    }\n  }\n}\n\nexport const VOICE_SPECTRUM_BANDS = 9\n\n/** Log-spaced speech spectrum for transports that expose FFT bins. */\nexport function voiceSpectrum(frequencies: Uint8Array, sampleRate: number, fftSize: number, bandCount = VOICE_SPECTRUM_BANDS): readonly number[] {\n  if (!Number.isFinite(sampleRate) || sampleRate <= 0 || !Number.isFinite(fftSize) || fftSize <= 0 || bandCount <= 0)\n    return []\n\n  const hzPerBin = sampleRate / fftSize\n  const fromHz = 80\n  const ratio = 4_000 / fromHz\n  const bands: number[] = []\n  for (let band = 0; band < bandCount; band++) {\n    const start = Math.max(0, Math.ceil((fromHz * ratio ** (band / bandCount)) / hzPerBin))\n    const end = Math.min(frequencies.length, Math.ceil((fromHz * ratio ** ((band + 1) / bandCount)) / hzPerBin))\n    if (end <= start) {\n      bands.push(0)\n      continue\n    }\n    let squared = 0\n    for (let index = start; index < end; index++)\n      squared += (frequencies[index] / 255) ** 2\n    bands.push(Math.sqrt(squared / (end - start)))\n  }\n  return bands\n}\n\n/** Equal-width bands for SDKs that already expose voice-range FFT bins. */\nexport function voiceRangeSpectrum(frequencies: Uint8Array, bandCount = VOICE_SPECTRUM_BANDS): readonly number[] {\n  if (!frequencies.length || bandCount <= 0)\n    return []\n  return Array.from({ length: bandCount }, (_, band) => {\n    const start = Math.floor((band * frequencies.length) / bandCount)\n    const end = Math.max(start + 1, Math.floor(((band + 1) * frequencies.length) / bandCount))\n    let squared = 0\n    for (let index = start; index < Math.min(end, frequencies.length); index++)\n      squared += (frequencies[index] / 255) ** 2\n    return Math.sqrt(squared / Math.max(1, Math.min(end, frequencies.length) - start))\n  })\n}\n"],"mappings":";;AAWA,SAAgB,eAAe,GAAwB;CACrD,OAAO,KAAS;AAClB;AA2CA,IAAa,4BAAuD,OAAO,MAAS;CAClF,QAAQ,EAAK,UAAb;EACE,KAAK,cAAc;GACjB,IAAM,EAAE,+BAA4B,MAAM,OAAO;GACjD,OAAO,IAAI,EAAwB,CAAI;EACzC;EACA,KAAK,WAAW;GACd,IAAM,EAAE,4BAAyB,MAAM,OAAO;GAC9C,OAAO,IAAI,EAAqB,CAAI;EACtC;CACF;AACF;AAKA,SAAgB,cAAc,GAAyB,GAAoB,GAAiB,IAAA,GAAqD;CAC/I,IAAI,CAAC,OAAO,SAAS,CAAU,KAAK,KAAc,KAAK,CAAC,OAAO,SAAS,CAAO,KAAK,KAAW,KAAK,KAAa,GAC/G,OAAO,CAAC;CAEV,IAAM,IAAW,IAAa,GAGxB,IAAkB,CAAC;CACzB,KAAK,IAAI,IAAO,GAAG,IAAO,GAAW,KAAQ;EAC3C,IAAM,IAAQ,KAAK,IAAI,GAAG,KAAK,KAAM,KAAS,OAAU,IAAO,KAAc,CAAQ,CAAC,GAChF,IAAM,KAAK,IAAI,EAAY,QAAQ,KAAK,KAAM,KAAS,QAAW,IAAO,KAAK,KAAc,CAAQ,CAAC;EAC3G,IAAI,KAAO,GAAO;GAChB,EAAM,KAAK,CAAC;GACZ;EACF;EACA,IAAI,IAAU;EACd,KAAK,IAAI,IAAQ,GAAO,IAAQ,GAAK,KACnC,MAAY,EAAY,KAAS,QAAQ;EAC3C,EAAM,KAAK,KAAK,KAAK,KAAW,IAAM,EAAM,CAAC;CAC/C;CACA,OAAO;AACT;AAGA,SAAgB,mBAAmB,GAAyB,IAAA,GAAqD;CAG/G,OAFI,CAAC,EAAY,UAAU,KAAa,IAC/B,CAAC,IACH,MAAM,KAAK,EAAE,QAAQ,EAAU,IAAI,GAAG,MAAS;EACpD,IAAM,IAAQ,KAAK,MAAO,IAAO,EAAY,SAAU,CAAS,GAC1D,IAAM,KAAK,IAAI,IAAQ,GAAG,KAAK,OAAQ,IAAO,KAAK,EAAY,SAAU,CAAS,CAAC,GACrF,IAAU;EACd,KAAK,IAAI,IAAQ,GAAO,IAAQ,KAAK,IAAI,GAAK,EAAY,MAAM,GAAG,KACjE,MAAY,EAAY,KAAS,QAAQ;EAC3C,OAAO,KAAK,KAAK,IAAU,KAAK,IAAI,GAAG,KAAK,IAAI,GAAK,EAAY,MAAM,IAAI,CAAK,CAAC;CACnF,CAAC;AACH"}

package/dist/provider2.js

@@ -1,6 +1,6 @@
 import "./rolldown-runtime.js";
 import { s as e } from "./sdkClient.js";
-import { n as t } from "./provider.js";
+import { r as t } from "./provider.js";
 //#region ../../node_modules/.pnpm/@elevenlabs+client@1.17.0_@types+dom-mediacapture-record@1.0.22/node_modules/@elevenlabs/client/dist/utils/BaseConnection.js
 var BaseConnection = class {
 	constructor(t = {}) {

package/dist/provider3.js

@@ -1,6 +1,6 @@
 import "./rolldown-runtime.js";
 import { s as e } from "./sdkClient.js";
-import { r as t } from "./provider.js";
+import { i as t } from "./provider.js";
 //#region ../../node_modules/.pnpm/livekit-client@2.21.0_@types+dom-mediacapture-record@1.0.22/node_modules/livekit-client/dist/livekit-client.esm.mjs
 function _mergeNamespaces(e, t) {
 	return t.forEach(function(t) {
@@ -18333,19 +18333,23 @@ _defineProperty(JWSSignatureVerificationFailed, "code", "ERR_JWS_SIGNATURE_VERIF
 //#region agent/voice/providers/livekit/provider.ts
 var $a = 1e4, eo = "lk.agent.state", LiveKitVoiceProvider = class {
 	constructor(t, n = {}) {
-		e(this, "mint", void 0), e(this, "room", void 0), e(this, "createAnalyser", void 0), e(this, "handlers", void 0), e(this, "cancelReadyWait", void 0), e(this, "analysers", []), e(this, "levelSources", []), e(this, "attachedElements", []), e(this, "closing", !1), this.mint = t, this.room = n.createRoom?.() ?? new Za(), this.createAnalyser = n.createAnalyser;
+		e(this, "mint", void 0), e(this, "room", void 0), e(this, "createAnalyser", void 0), e(this, "handlers", void 0), e(this, "cancelReadyWait", void 0), e(this, "analysers", []), e(this, "levelSources", []), e(this, "attachedElements", []), e(this, "closing", !1), e(this, "userSpeaking", !1), e(this, "agentActiveSpeaker", !1), e(this, "agentStateSpeaking", !1), e(this, "lastSpeaking", void 0), this.mint = t, this.room = n.createRoom?.() ?? new Za(), this.createAnalyser = n.createAnalyser;
 	}
 	async connect(e) {
 		this.handlers = e, this.room.on(P.TrackSubscribed, (e) => {
 			e.kind === z.Kind.Audio && this.attachRemoteAudio(e);
 		}), this.room.on(P.ParticipantAttributesChanged, (e, t) => {
-			t.isAgent && this.handlers?.onThinkingChange?.(t.attributes[eo] === "thinking");
+			if (!t.isAgent) return;
+			let n = t.attributes[eo];
+			this.handlers?.onThinkingChange?.(n === "thinking"), this.agentStateSpeaking = n === "speaking", this.publishSpeaking();
+		}), this.room.on(P.ActiveSpeakersChanged, (e) => {
+			this.userSpeaking = e.some((e) => e.isLocal), this.agentActiveSpeaker = e.some((e) => e.isAgent), this.publishSpeaking();
 		}), this.room.on(P.Disconnected, (e) => {
 			this.closing || this.handlers?.onDisconnected(disconnectCause(e));
 		});
 		let t = "VOICE_CONNECT_FAILED", n;
 		try {
-			if (await this.room.connect(this.mint.url, this.mint.token), this.closing || (t = "VOICE_AGENT_UNAVAILABLE", await this.waitForAgentReady(), this.closing)) return;
+			if (await this.room.connect(this.mint.url, this.mint.token), this.closing) return;
 			if (t = "VOICE_MIC_UNAVAILABLE", n = await this.room.localParticipant.setMicrophoneEnabled(!0, {
 				echoCancellation: !0,
 				noiseSuppression: !0
@@ -18353,6 +18357,10 @@ var $a = 1e4, eo = "lk.agent.state", LiveKitVoiceProvider = class {
 				await this.room.localParticipant.setMicrophoneEnabled(!1).catch(() => {});
 				return;
 			}
+			if (t = "VOICE_AGENT_UNAVAILABLE", await this.waitForAgentReady(), this.closing) {
+				await this.room.localParticipant.setMicrophoneEnabled(!1).catch(() => {});
+				return;
+			}
 			t = "VOICE_PLAYBACK_UNAVAILABLE", await this.room.startAudio();
 		} catch (e) {
 			let n = e instanceof Error ? e : /* @__PURE__ */ Error("Could not open the voice connection.");
@@ -18377,19 +18385,26 @@ var $a = 1e4, eo = "lk.agent.state", LiveKitVoiceProvider = class {
 			noiseSuppression: !0
 		});
 	}
-	sampleLevel() {
-		let e;
-		for (let { source: t, analyser: n } of this.levelSources) {
-			let r = n.calculateVolume();
-			if (!Number.isFinite(r)) continue;
+	sampleLevel(e) {
+		let t, n = e && this.levelSources.some((t) => t.source === e) ? this.levelSources.filter((t) => t.source === e) : this.levelSources;
+		for (let { source: e, analyser: r } of n) {
+			let n = r.calculateVolume();
+			if (!Number.isFinite(n)) continue;
 			let i = {
-				source: t,
-				level: Math.min(1, Math.max(0, r)),
-				...n.calculateBands ? { bands: n.calculateBands() } : {}
+				source: e,
+				level: Math.min(1, Math.max(0, n)),
+				...r.calculateBands ? { bands: r.calculateBands() } : {}
 			};
-			(!e || i.level > e.level) && (e = i);
+			(!t || i.level > t.level) && (t = i);
 		}
-		return e;
+		return t;
+	}
+	publishSpeaking() {
+		let e = {
+			user: this.userSpeaking,
+			agent: this.agentActiveSpeaker || this.agentStateSpeaking
+		};
+		this.lastSpeaking?.user === e.user && this.lastSpeaking?.agent === e.agent || (this.lastSpeaking = e, this.handlers?.onSpeakingChange?.(e));
 	}
 	async waitForAgentReady() {
 		let ready = (e) => {