@runtypelabs/persona 4.24.0 → 4.25.0

package/dist/theme-editor.d.cts

@@ -6582,6 +6582,22 @@ type AgentWidgetStatusIndicatorConfig = {
     /** Status text while a reconnect attempt is in flight. */
     resumingText?: string;
 };
+/** Built-in `voiceRecognition.prewarm` triggers. See {@link AgentWidgetVoiceRecognitionConfig.prewarm}. */
+type AgentWidgetVoicePrewarmTrigger = 'hover';
+/**
+ * Custom `voiceRecognition.prewarm` trigger. Runs while the composer has a mic
+ * button, and again when the button is rebuilt or `update()` swaps in a
+ * different function. Call `warm()` whenever the visitor shows intent to talk
+ * (it is safe to call repeatedly). Return a cleanup to run on replacement.
+ */
+type AgentWidgetVoicePrewarmHook = (context: {
+    /** Fire-and-forget voice warm-up; latched and error-swallowing. */
+    warm: () => void;
+    /** The composer's current mic button. */
+    micButton: HTMLButtonElement;
+    /** The widget's mount element. */
+    mount: HTMLElement;
+}) => void | (() => void);
 type AgentWidgetVoiceRecognitionConfig = {
     /**
      * Enable the mic button and voice input. The mic still only renders when the
@@ -6647,6 +6663,29 @@ type AgentWidgetVoiceRecognitionConfig = {
      * Fires once per turn after the reply completes.
      */
     onMetrics?: (metrics: VoiceMetrics) => void;
+    /**
+     * Warm the voice path when the visitor shows intent to talk, before the mic
+     * click, so the call starts faster.
+     *
+     * - `'hover'`: when the pointer enters the mic button, it receives focus, or
+     *   (on touch) the first pointer-down on it.
+     * - `false`: never.
+     * - A function: custom trigger. Called with `{ warm, micButton, mount }`
+     *   while the mic button exists (and again when the button is rebuilt or
+     *   `update()` swaps in a different function); call `warm()` whenever you
+     *   decide, e.g. when the panel opens. May return a cleanup that runs on
+     *   destroy, rebuild, or replacement. Replaces the built-in trigger.
+     *
+     * What `warm()` does depends on the provider: `runtype` fires its prewarm
+     * (see `provider.runtype.prewarmMode`), a `custom` provider's optional
+     * {@link VoiceProvider.prewarm} runs, and the browser (Web Speech) path
+     * starts the client-token session init (as `sessionInit` does), since speech
+     * recognition itself cannot start before the click. Warming is
+     * fire-and-forget, latched, and never surfaces an error or UI state.
+     *
+     * @default 'hover' in client token mode, otherwise `false`
+     */
+    prewarm?: AgentWidgetVoicePrewarmTrigger | AgentWidgetVoicePrewarmHook | false;
     provider?: {
         type: 'browser' | 'runtype' | 'custom';
         browser?: {
@@ -6668,6 +6707,23 @@ type AgentWidgetVoiceRecognitionConfig = {
              * `@runtypelabs/persona/voice-worklet-player` for a jitter-buffered engine.
              */
             createPlaybackEngine?: () => VoicePlaybackEngine | Promise<VoicePlaybackEngine>;
+            /**
+             * How `prewarm()` warms the voice path (see `voiceRecognition.prewarm`).
+             *
+             * - `'request'` (default): a fire-and-forget
+             *   `POST {host}/v1/client/agents/{agentId}/voice/prewarm`.
+             * - `'attach'`: open the voice WebSocket ahead of the click without
+             *   starting a call; the click then starts the call on that socket. It can
+             *   create server work before the visitor clicks, so it is opt-in.
+             * @default 'request'
+             */
+            prewarmMode?: 'request' | 'attach';
+            /**
+             * `'attach'` mode: how long (ms) an attached socket waits for the click
+             * before it closes. The server clamps it to 30000..600000.
+             * @default 30000
+             */
+            attachIdleMs?: number;
             /** @deprecated No-op on the realtime path: the server's STT owns turn-taking. */
             pauseDuration?: number;
             /** @deprecated No-op on the realtime path: the server's STT owns turn-taking. */
@@ -6968,6 +7024,13 @@ interface VoiceProvider {
     deactivateBargeIn?(): Promise<void>;
     /** Stop playback / cancel in-flight request without starting recording */
     stopPlayback?(): void;
+    /**
+     * Warm the provider ahead of `startListening()` (the widget calls it on mic
+     * intent, see `voiceRecognition.prewarm`). Must be fire-and-forget: never
+     * throw, emit errors or status, or acquire the microphone. May be called
+     * repeatedly; throttle as needed.
+     */
+    prewarm?(): void;
 }
 /**
  * Extra context for an approval decision, surfaced to `onDecision` and to the

package/dist/theme-editor.d.ts

@@ -6582,6 +6582,22 @@ type AgentWidgetStatusIndicatorConfig = {
     /** Status text while a reconnect attempt is in flight. */
     resumingText?: string;
 };
+/** Built-in `voiceRecognition.prewarm` triggers. See {@link AgentWidgetVoiceRecognitionConfig.prewarm}. */
+type AgentWidgetVoicePrewarmTrigger = 'hover';
+/**
+ * Custom `voiceRecognition.prewarm` trigger. Runs while the composer has a mic
+ * button, and again when the button is rebuilt or `update()` swaps in a
+ * different function. Call `warm()` whenever the visitor shows intent to talk
+ * (it is safe to call repeatedly). Return a cleanup to run on replacement.
+ */
+type AgentWidgetVoicePrewarmHook = (context: {
+    /** Fire-and-forget voice warm-up; latched and error-swallowing. */
+    warm: () => void;
+    /** The composer's current mic button. */
+    micButton: HTMLButtonElement;
+    /** The widget's mount element. */
+    mount: HTMLElement;
+}) => void | (() => void);
 type AgentWidgetVoiceRecognitionConfig = {
     /**
      * Enable the mic button and voice input. The mic still only renders when the
@@ -6647,6 +6663,29 @@ type AgentWidgetVoiceRecognitionConfig = {
      * Fires once per turn after the reply completes.
      */
     onMetrics?: (metrics: VoiceMetrics) => void;
+    /**
+     * Warm the voice path when the visitor shows intent to talk, before the mic
+     * click, so the call starts faster.
+     *
+     * - `'hover'`: when the pointer enters the mic button, it receives focus, or
+     *   (on touch) the first pointer-down on it.
+     * - `false`: never.
+     * - A function: custom trigger. Called with `{ warm, micButton, mount }`
+     *   while the mic button exists (and again when the button is rebuilt or
+     *   `update()` swaps in a different function); call `warm()` whenever you
+     *   decide, e.g. when the panel opens. May return a cleanup that runs on
+     *   destroy, rebuild, or replacement. Replaces the built-in trigger.
+     *
+     * What `warm()` does depends on the provider: `runtype` fires its prewarm
+     * (see `provider.runtype.prewarmMode`), a `custom` provider's optional
+     * {@link VoiceProvider.prewarm} runs, and the browser (Web Speech) path
+     * starts the client-token session init (as `sessionInit` does), since speech
+     * recognition itself cannot start before the click. Warming is
+     * fire-and-forget, latched, and never surfaces an error or UI state.
+     *
+     * @default 'hover' in client token mode, otherwise `false`
+     */
+    prewarm?: AgentWidgetVoicePrewarmTrigger | AgentWidgetVoicePrewarmHook | false;
     provider?: {
         type: 'browser' | 'runtype' | 'custom';
         browser?: {
@@ -6668,6 +6707,23 @@ type AgentWidgetVoiceRecognitionConfig = {
              * `@runtypelabs/persona/voice-worklet-player` for a jitter-buffered engine.
              */
             createPlaybackEngine?: () => VoicePlaybackEngine | Promise<VoicePlaybackEngine>;
+            /**
+             * How `prewarm()` warms the voice path (see `voiceRecognition.prewarm`).
+             *
+             * - `'request'` (default): a fire-and-forget
+             *   `POST {host}/v1/client/agents/{agentId}/voice/prewarm`.
+             * - `'attach'`: open the voice WebSocket ahead of the click without
+             *   starting a call; the click then starts the call on that socket. It can
+             *   create server work before the visitor clicks, so it is opt-in.
+             * @default 'request'
+             */
+            prewarmMode?: 'request' | 'attach';
+            /**
+             * `'attach'` mode: how long (ms) an attached socket waits for the click
+             * before it closes. The server clamps it to 30000..600000.
+             * @default 30000
+             */
+            attachIdleMs?: number;
             /** @deprecated No-op on the realtime path: the server's STT owns turn-taking. */
             pauseDuration?: number;
             /** @deprecated No-op on the realtime path: the server's STT owns turn-taking. */
@@ -6968,6 +7024,13 @@ interface VoiceProvider {
     deactivateBargeIn?(): Promise<void>;
     /** Stop playback / cancel in-flight request without starting recording */
     stopPlayback?(): void;
+    /**
+     * Warm the provider ahead of `startListening()` (the widget calls it on mic
+     * intent, see `voiceRecognition.prewarm`). Must be fire-and-forget: never
+     * throw, emit errors or status, or acquire the microphone. May be called
+     * repeatedly; throttle as needed.
+     */
+    prewarm?(): void;
 }
 /**
  * Extra context for an approval decision, surfaced to `onDecision` and to the

package/dist/voice-runtime-7DA7MJH4.js

@@ -0,0 +1 @@
+import{a as v}from"./chunk-5EIIHQLQ.js";var g=16e3,k=24e3,S=4096,C=4,E=1380533830,A=3e4,M=5e3,P=3e4,T=3e4,V=6e5;function _(r){return r.byteLength>=44&&new DataView(r).getUint32(0,!1)===E?new Uint8Array(r,44):new Uint8Array(r)}function w(r){let t=r.replace(/\/+$/,"");return/^wss?:\/\//i.test(t)?t:/^https?:\/\//i.test(t)?t.replace(/^http/i,"ws"):`${typeof window<"u"&&window.location?.protocol==="https:"?"wss:":"ws:"}//${t}`}function R(r){return w(r).replace(/^ws/i,"http")}function L(r){return typeof r!="number"||!Number.isFinite(r)?P:Math.min(V,Math.max(T,Math.round(r)))}var f=class{constructor(t){this.config=t;this.type="runtype";this.ws=null;this.captureContext=null;this.levelCallbacks=[];this.mediaStream=null;this.sourceNode=null;this.processor=null;this.playback=null;this.callLive=!1;this.isSpeaking=!1;this.callGeneration=0;this.intentionalClose=!1;this.lastPrewarmAt=Number.NEGATIVE_INFINITY;this.attached=null;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.transcriptCallbacks=[];this.metricsCallbacks=[]}async connect(){}prewarm(){if(this.callLive||this.attached)return;let t=this.config?.agentId,e=this.config?.clientToken,i=this.config?.host;if(!t||!e||!i)return;let s=Date.now();if(!(s-this.lastPrewarmAt<A)){this.lastPrewarmAt=s;try{this.config?.prewarmMode==="attach"?this.attach(i,t,e):this.requestPrewarm(i,t,e)}catch{}}}requestPrewarm(t,e,i){if(typeof fetch!="function")return;let s=`${R(t)}/v1/client/agents/${encodeURIComponent(e)}/voice/prewarm`;fetch(s,{method:"POST",headers:{Authorization:`Bearer ${i}`},keepalive:!0}).catch(()=>{})}attach(t,e,i){let s=new URLSearchParams({clientCapabilities:"attach"}),o=this.config?.attachIdleMs,h=L(o);o!==void 0&&s.set("attachIdleMs",String(h));let c=new WebSocket(`${this.voiceSocketUrl(t,e)}?${s}`,["runtype.bearer","runtype.attach",i]);c.binaryType="arraybuffer";let u=()=>{},d=new Promise(n=>{u=n}),a={ws:c,state:"connecting",ready:d,settle:u,openedAt:0,idleTimer:void 0};this.attached=a,c.onopen=()=>{this.attached===a&&(a.state=c.protocol==="runtype.attach"?"attached":"live",a.openedAt=Date.now(),this.armAttachIdle(a,h),a.state==="attached"&&c.send('{"type":"ping"}'),a.settle())},c.onmessage=n=>{if(this.attached!==a||typeof n.data!="string")return;let l;try{l=JSON.parse(n.data)}catch{return}l.type==="attached"&&typeof l.idleMs=="number"&&this.armAttachIdle(a,l.idleMs-(Date.now()-a.openedAt))},c.onerror=()=>{},c.onclose=()=>{this.attached===a&&this.releaseAttached()}}armAttachIdle(t,e){clearTimeout(t.idleTimer),t.idleTimer=setTimeout(()=>{this.attached===t&&this.releaseAttached()},Math.max(0,e))}releaseAttached(){let t=this.attached;if(!t)return;this.attached=null,clearTimeout(t.idleTimer);let{ws:e}=t;if(e.onopen=null,e.onmessage=null,e.onerror=null,e.onclose=null,t.settle(),e.readyState===WebSocket.CONNECTING||e.readyState===WebSocket.OPEN)try{e.close(1e3,"prewarm released")}catch{}}async takeAttachedSocket(){let t=this.attached;if(!t)return null;if(t.state==="connecting"){let e;await Promise.race([t.ready,new Promise(i=>{e=setTimeout(i,M)})]),clearTimeout(e)}return this.attached!==t?null:t.state==="connecting"||t.ws.readyState!==WebSocket.OPEN?(this.releaseAttached(),null):(this.attached=null,clearTimeout(t.idleTimer),{ws:t.ws,live:t.state==="live"})}voiceSocketUrl(t,e){return`${w(t)}/ws/agents/${encodeURIComponent(e)}/voice`}async startListening(){if(this.callLive)return;let t=this.config?.agentId,e=this.config?.clientToken,i=this.config?.host;if(!t)throw new Error("Runtype voice requires an agentId");if(!e)throw new Error("Runtype voice requires a clientToken");if(!i)throw new Error("Runtype voice requires a host (or widget apiUrl)");let s=++this.callGeneration;this.intentionalClose=!1,this.callLive=!0;try{let o=await navigator.mediaDevices.getUserMedia({audio:{sampleRate:g,channelCount:1,echoCancellation:!0}});if(s!==this.callGeneration){o.getTracks().forEach(n=>n.stop());return}this.mediaStream=o;let h=window.AudioContext||window.webkitAudioContext,c=new h({sampleRate:g});c.state==="suspended"&&await c.resume().catch(()=>{}),this.captureContext=c;let u=this.config?.createPlaybackEngine?await this.config.createPlaybackEngine():new v(k);if(s!==this.callGeneration){u.destroy(),o.getTracks().forEach(n=>n.stop()),c.close().catch(()=>{});return}this.playback=u,u.onFinished(()=>{s===this.callGeneration&&(this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening"))});let d=await this.takeAttachedSocket();if(s!==this.callGeneration){d?.ws.close(1e3,"client ended call");return}if(d){let{ws:n}=d;this.ws=n,this.bindCallSocket(n,s),d.live||n.send('{"type":"start"}'),this.emitStatus("listening"),this.startCapture(c,o,n,s);return}let a=new WebSocket(this.voiceSocketUrl(i,t),["runtype.bearer",e]);a.binaryType="arraybuffer",this.ws=a,a.onopen=()=>{s===this.callGeneration&&(this.emitStatus("listening"),this.startCapture(c,o,a,s))},this.bindCallSocket(a,s)}catch(o){throw this.cleanup(),this.emitError(o),this.emitStatus("error"),o}}bindCallSocket(t,e){t.onmessage=i=>this.handleMessage(i,e),t.onerror=()=>{e===this.callGeneration&&(this.emitError(new Error("Voice connection failed")),this.emitStatus("error"),this.cleanup())},t.onclose=i=>{if(this.intentionalClose){this.intentionalClose=!1;return}if(e===this.callGeneration){if(i.code!==1e3){let s=i.code?` (code ${i.code})`:"";this.emitError(new Error(`Voice connection closed${s}`)),this.emitStatus("error")}else this.emitStatus("idle");this.cleanup()}}}async stopListening(){this.cleanup(),this.emitStatus("idle")}async disconnect(){this.cleanup(),this.emitStatus("disconnected"),this.resultCallbacks=[],this.errorCallbacks=[],this.statusCallbacks=[],this.transcriptCallbacks=[],this.metricsCallbacks=[]}stopPlayback(){this.playback&&this.playback.flush(),this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening")}getInterruptionMode(){return"barge-in"}isBargeInActive(){return this.callLive}onLevel(t){this.levelCallbacks.push(t)}async deactivateBargeIn(){this.cleanup(),this.emitStatus("idle")}startCapture(t,e,i,s){let o=t.createMediaStreamSource(e);this.sourceNode=o;let h=t.createScriptProcessor(S,1,1);this.processor=h,h.onaudioprocess=c=>{if(s!==this.callGeneration||i.readyState!==WebSocket.OPEN)return;let u=c.inputBuffer.getChannelData(0),d=new Int16Array(u.length),a=0;for(let n=0;n<u.length;n++){let l=Math.max(-1,Math.min(1,u[n]));a+=l*l,d[n]=l<0?l*32768:l*32767}if(i.send(d.buffer),this.levelCallbacks.length>0){let n=Math.sqrt(a/u.length),l=Math.max(0,Math.min(1,n*C));for(let b of this.levelCallbacks)b(l)}},o.connect(h),h.connect(t.destination)}handleMessage(t,e){if(e!==this.callGeneration)return;if(t.data instanceof ArrayBuffer){this.handleAudioFrame(t.data,e);return}let i;try{i=JSON.parse(t.data)}catch{return}switch(i.type){case"transcript_interim":this.emitStatus("listening"),this.emitTranscript("user",i.text??"",!1);break;case"transcript_final":{let s=i.role==="assistant"?"assistant":"user";this.emitStatus(s==="user"?"processing":"speaking"),this.emitTranscript(s,i.text??"",!0);break}case"audio_end":this.playback?this.playback.markStreamEnd():(this.isSpeaking=!1,this.emitStatus("listening"));break;case"metrics":this.emitMetrics({llmMs:i.llm_ms,ttsMs:i.tts_ms,firstAudioMs:i.first_audio_ms,totalMs:i.total_ms});break;case"error":this.emitError(new Error(i.error||"Voice error")),this.emitStatus("error");break}}handleAudioFrame(t,e){if(e!==this.callGeneration||!this.playback)return;let i=_(t);i.length!==0&&(this.isSpeaking||(this.isSpeaking=!0,this.emitStatus("speaking")),this.playback.enqueue(i))}cleanup(){if(this.callGeneration+=1,this.callLive=!1,this.isSpeaking=!1,this.releaseAttached(),this.processor&&(this.processor.onaudioprocess=null,this.processor.disconnect(),this.processor=null),this.sourceNode&&(this.sourceNode.disconnect(),this.sourceNode=null),this.mediaStream&&(this.mediaStream.getTracks().forEach(t=>t.stop()),this.mediaStream=null),this.captureContext&&(this.captureContext.close().catch(()=>{}),this.captureContext=null),this.playback&&(this.playback.destroy(),this.playback=null),this.ws){this.intentionalClose=!0;try{this.ws.close(1e3,"client ended call")}catch{}this.ws=null}}onResult(t){this.resultCallbacks.push(t)}onError(t){this.errorCallbacks.push(t)}onStatusChange(t){this.statusCallbacks.push(t)}onTranscript(t){this.transcriptCallbacks.push(t)}onMetrics(t){this.metricsCallbacks.push(t)}emitStatus(t){this.statusCallbacks.forEach(e=>e(t))}emitError(t){this.errorCallbacks.forEach(e=>e(t))}emitTranscript(t,e,i){this.transcriptCallbacks.forEach(s=>s(t,e,i))}emitMetrics(t){this.metricsCallbacks.forEach(e=>e(t))}};var p=class{constructor(t={}){this.config=t;this.type="browser";this.recognition=null;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.isListening=!1;this.w=typeof window<"u"?window:void 0}async connect(){this.statusCallbacks.forEach(t=>t("connected"))}async startListening(){try{if(this.isListening)throw new Error("Already listening");if(!this.w)throw new Error("Window object not available");let t=this.w.SpeechRecognition||this.w.webkitSpeechRecognition;if(!t)throw new Error("Browser speech recognition not supported");this.recognition=new t,this.recognition.lang=this.config?.language||"en-US",this.recognition.continuous=this.config?.continuous||!1,this.recognition.interimResults=!0,this.recognition.onresult=e=>{let i=Array.from(e.results).map(o=>o[0]).map(o=>o.transcript).join(""),s=e.results[e.results.length-1].isFinal;this.resultCallbacks.forEach(o=>o({text:i,confidence:s?.8:.5,provider:"browser"})),s&&!this.config?.continuous&&this.stopListening()},this.recognition.onerror=e=>{this.errorCallbacks.forEach(i=>i(new Error(e.error))),this.statusCallbacks.forEach(i=>i("error"))},this.recognition.onstart=()=>{this.isListening=!0,this.statusCallbacks.forEach(e=>e("listening"))},this.recognition.onend=()=>{this.isListening=!1,this.statusCallbacks.forEach(e=>e("idle"))},this.recognition.start()}catch(t){throw this.errorCallbacks.forEach(e=>e(t)),this.statusCallbacks.forEach(e=>e("error")),t}}async stopListening(){this.recognition&&(this.recognition.stop(),this.recognition=null),this.isListening=!1,this.statusCallbacks.forEach(t=>t("idle"))}onResult(t){this.resultCallbacks.push(t)}onError(t){this.errorCallbacks.push(t)}onStatusChange(t){this.statusCallbacks.push(t)}async disconnect(){await this.stopListening(),this.statusCallbacks.forEach(t=>t("disconnected"))}static isSupported(){return"SpeechRecognition"in window||"webkitSpeechRecognition"in window}};function m(r){switch(r.type){case"runtype":if(!r.runtype)throw new Error("Runtype voice provider requires configuration");return new f(r.runtype);case"browser":if(!p.isSupported())throw new Error("Browser speech recognition not supported");return new p(r.browser||{});case"custom":{let t=r.custom;if(!t)throw new Error("Custom voice provider requires a `custom` provider instance or factory");let e=typeof t=="function"?t():t;if(!e||typeof e.startListening!="function")throw new Error("Custom voice provider `custom` must be a VoiceProvider (or a factory returning one)");return e}default:throw new Error(`Unknown voice provider type: ${r.type}`)}}function y(r){if(r?.type==="custom"&&r.custom)return m({type:"custom",custom:r.custom});if(r?.type==="runtype"&&r.runtype)return m({type:"runtype",runtype:r.runtype});if(p.isSupported())return m({type:"browser",browser:r?.browser||{language:"en-US"}});throw new Error("No supported voice providers available")}function I(r){try{return y(r),!0}catch{return!1}}export{p as BrowserVoiceProvider,f as RuntypeVoiceProvider,y as createBestAvailableVoiceProvider,m as createVoiceProvider,I as isVoiceSupported};

package/dist/voice-runtime.cjs

@@ -1 +1 @@
-"use strict";var g=Object.defineProperty;var C=Object.getOwnPropertyDescriptor;var k=Object.getOwnPropertyNames;var E=Object.prototype.hasOwnProperty;var x=(r,e)=>{for(var t in e)g(r,t,{get:e[t],enumerable:!0})},A=(r,e,t,i)=>{if(e&&typeof e=="object"||typeof e=="function")for(let s of k(e))!E.call(r,s)&&s!==t&&g(r,s,{get:()=>e[s],enumerable:!(i=C(e,s))||i.enumerable});return r};var V=r=>A(g({},"__esModule",{value:!0}),r);var F={};x(F,{BrowserVoiceProvider:()=>p,RuntypeVoiceProvider:()=>d,createBestAvailableVoiceProvider:()=>b,createVoiceProvider:()=>f,isVoiceSupported:()=>w});module.exports=V(F);var v=class{constructor(e=24e3,t={}){this.ctx=null;this.nextStartTime=0;this.activeSources=[];this.finishedCallbacks=[];this.startedCallbacks=[];this.playing=!1;this.streamEnded=!1;this.pendingCount=0;this.started=!1;this.userPaused=!1;this.pendingBuffers=[];this.pendingSamples=0;this.remainder=null;this.sampleRate=e;let i=Math.max(0,t.prebufferMs??0);this.waterlineSamples=Math.round(e*i/1e3),this.buffering=this.waterlineSamples>0}ensureContext(){if(!this.ctx){let t=typeof window<"u"?window:void 0;if(!t)throw new Error("AudioPlaybackManager requires a browser environment");let i=t.AudioContext||t.webkitAudioContext;this.ctx=new i({sampleRate:this.sampleRate})}let e=this.ctx;return e.state==="suspended"&&!this.userPaused&&e.resume(),e}enqueue(e){if(e.length===0)return;let t=e;if(this.remainder){let s=new Uint8Array(this.remainder.length+e.length);s.set(this.remainder),s.set(e,this.remainder.length),t=s,this.remainder=null}if(t.length%2!==0&&(this.remainder=new Uint8Array([t[t.length-1]]),t=t.subarray(0,t.length-1)),t.length===0)return;let i=this.pcmToFloat32(t);i.length!==0&&(this.buffering?(this.pendingBuffers.push(i),this.pendingSamples+=i.length,this.pendingSamples>=this.waterlineSamples&&this.releaseBuffer()):this.scheduleSamples(i))}markStreamEnd(){this.pendingBuffers.length>0&&this.releaseBuffer(),this.streamEnded=!0,this.checkFinished()}flush(){for(let e of this.activeSources)try{e.stop(),e.disconnect()}catch{}this.activeSources=[],this.pendingCount=0,this.nextStartTime=0,this.playing=!1,this.streamEnded=!1,this.finishedCallbacks=[],this.startedCallbacks=[],this.remainder=null,this.pendingBuffers=[],this.pendingSamples=0,this.buffering=this.waterlineSamples>0,this.started=!1}isPlaying(){return this.playing}onFinished(e){this.finishedCallbacks.push(e)}onStarted(e){this.startedCallbacks.push(e)}pause(){this.userPaused=!0,this.ctx&&this.ctx.state==="running"&&this.ctx.suspend()}resume(){this.userPaused=!1,this.ctx&&this.ctx.state==="suspended"&&this.ctx.resume()}async destroy(){this.flush(),this.ctx&&(await this.ctx.close(),this.ctx=null)}releaseBuffer(){this.buffering=!1;let e=this.pendingBuffers;this.pendingBuffers=[],this.pendingSamples=0;for(let t of e)this.scheduleSamples(t)}scheduleSamples(e){if(e.length===0)return;let t=this.ensureContext(),i=t.createBuffer(1,e.length,this.sampleRate);i.getChannelData(0).set(e);let s=t.createBufferSource();s.buffer=i,s.connect(t.destination);let n=t.currentTime;if(this.nextStartTime===0?this.nextStartTime=n:this.nextStartTime<n&&(this.nextStartTime=n,this.waterlineSamples>0&&(this.buffering=!0)),s.start(this.nextStartTime),this.nextStartTime+=i.duration,this.activeSources.push(s),this.pendingCount++,this.playing=!0,!this.started){this.started=!0;let a=this.startedCallbacks.slice();this.startedCallbacks=[];for(let l of a)l()}s.onended=()=>{let a=this.activeSources.indexOf(s);a!==-1&&this.activeSources.splice(a,1),this.pendingCount--,this.checkFinished()}}checkFinished(){if(this.streamEnded&&this.pendingCount<=0&&this.pendingBuffers.length===0){this.playing=!1,this.streamEnded=!1;let e=this.finishedCallbacks.slice();this.finishedCallbacks=[];for(let t of e)t()}}pcmToFloat32(e){let t=Math.floor(e.length/2),i=new Float32Array(t),s=new DataView(e.buffer,e.byteOffset,e.byteLength);for(let n=0;n<t;n++){let a=s.getInt16(n*2,!0);i[n]=a/32768}return i}};var y=16e3,P=24e3,M=4096,R=4,B=1380533830;function L(r){return r.byteLength>=44&&new DataView(r).getUint32(0,!1)===B?new Uint8Array(r,44):new Uint8Array(r)}function T(r){let e=r.replace(/\/+$/,"");return/^wss?:\/\//i.test(e)?e:/^https?:\/\//i.test(e)?e.replace(/^http/i,"ws"):`${typeof window<"u"&&window.location?.protocol==="https:"?"wss:":"ws:"}//${e}`}var d=class{constructor(e){this.config=e;this.type="runtype";this.ws=null;this.captureContext=null;this.levelCallbacks=[];this.mediaStream=null;this.sourceNode=null;this.processor=null;this.playback=null;this.callLive=!1;this.isSpeaking=!1;this.callGeneration=0;this.intentionalClose=!1;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.transcriptCallbacks=[];this.metricsCallbacks=[]}async connect(){}async startListening(){if(this.callLive)return;let e=this.config?.agentId,t=this.config?.clientToken,i=this.config?.host;if(!e)throw new Error("Runtype voice requires an agentId");if(!t)throw new Error("Runtype voice requires a clientToken");if(!i)throw new Error("Runtype voice requires a host (or widget apiUrl)");let s=++this.callGeneration;this.intentionalClose=!1,this.callLive=!0;try{let n=await navigator.mediaDevices.getUserMedia({audio:{sampleRate:y,channelCount:1,echoCancellation:!0}});if(s!==this.callGeneration){n.getTracks().forEach(o=>o.stop());return}this.mediaStream=n;let a=window.AudioContext||window.webkitAudioContext,l=new a({sampleRate:y});l.state==="suspended"&&await l.resume().catch(()=>{}),this.captureContext=l;let h=this.config?.createPlaybackEngine?await this.config.createPlaybackEngine():new v(P);if(s!==this.callGeneration){h.destroy(),n.getTracks().forEach(o=>o.stop()),l.close().catch(()=>{});return}this.playback=h,h.onFinished(()=>{s===this.callGeneration&&(this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening"))});let m=`${T(i)}/ws/agents/${encodeURIComponent(e)}/voice`,c=new WebSocket(m,["runtype.bearer",t]);c.binaryType="arraybuffer",this.ws=c,c.onopen=()=>{s===this.callGeneration&&(this.emitStatus("listening"),this.startCapture(l,n,c,s))},c.onmessage=o=>this.handleMessage(o,s),c.onerror=()=>{s===this.callGeneration&&(this.emitError(new Error("Voice connection failed")),this.emitStatus("error"),this.cleanup())},c.onclose=o=>{if(this.intentionalClose){this.intentionalClose=!1;return}if(s===this.callGeneration){if(o.code!==1e3){let u=o.code?` (code ${o.code})`:"";this.emitError(new Error(`Voice connection closed${u}`)),this.emitStatus("error")}else this.emitStatus("idle");this.cleanup()}}}catch(n){throw this.cleanup(),this.emitError(n),this.emitStatus("error"),n}}async stopListening(){this.cleanup(),this.emitStatus("idle")}async disconnect(){this.cleanup(),this.emitStatus("disconnected"),this.resultCallbacks=[],this.errorCallbacks=[],this.statusCallbacks=[],this.transcriptCallbacks=[],this.metricsCallbacks=[]}stopPlayback(){this.playback&&this.playback.flush(),this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening")}getInterruptionMode(){return"barge-in"}isBargeInActive(){return this.callLive}onLevel(e){this.levelCallbacks.push(e)}async deactivateBargeIn(){this.cleanup(),this.emitStatus("idle")}startCapture(e,t,i,s){let n=e.createMediaStreamSource(t);this.sourceNode=n;let a=e.createScriptProcessor(M,1,1);this.processor=a,a.onaudioprocess=l=>{if(s!==this.callGeneration||i.readyState!==WebSocket.OPEN)return;let h=l.inputBuffer.getChannelData(0),m=new Int16Array(h.length),c=0;for(let o=0;o<h.length;o++){let u=Math.max(-1,Math.min(1,h[o]));c+=u*u,m[o]=u<0?u*32768:u*32767}if(i.send(m.buffer),this.levelCallbacks.length>0){let o=Math.sqrt(c/h.length),u=Math.max(0,Math.min(1,o*R));for(let S of this.levelCallbacks)S(u)}},n.connect(a),a.connect(e.destination)}handleMessage(e,t){if(t!==this.callGeneration)return;if(e.data instanceof ArrayBuffer){this.handleAudioFrame(e.data,t);return}let i;try{i=JSON.parse(e.data)}catch{return}switch(i.type){case"transcript_interim":this.emitStatus("listening"),this.emitTranscript("user",i.text??"",!1);break;case"transcript_final":{let s=i.role==="assistant"?"assistant":"user";this.emitStatus(s==="user"?"processing":"speaking"),this.emitTranscript(s,i.text??"",!0);break}case"audio_end":this.playback?this.playback.markStreamEnd():(this.isSpeaking=!1,this.emitStatus("listening"));break;case"metrics":this.emitMetrics({llmMs:i.llm_ms,ttsMs:i.tts_ms,firstAudioMs:i.first_audio_ms,totalMs:i.total_ms});break;case"error":this.emitError(new Error(i.error||"Voice error")),this.emitStatus("error");break}}handleAudioFrame(e,t){if(t!==this.callGeneration||!this.playback)return;let i=L(e);i.length!==0&&(this.isSpeaking||(this.isSpeaking=!0,this.emitStatus("speaking")),this.playback.enqueue(i))}cleanup(){if(this.callGeneration+=1,this.callLive=!1,this.isSpeaking=!1,this.processor&&(this.processor.onaudioprocess=null,this.processor.disconnect(),this.processor=null),this.sourceNode&&(this.sourceNode.disconnect(),this.sourceNode=null),this.mediaStream&&(this.mediaStream.getTracks().forEach(e=>e.stop()),this.mediaStream=null),this.captureContext&&(this.captureContext.close().catch(()=>{}),this.captureContext=null),this.playback&&(this.playback.destroy(),this.playback=null),this.ws){this.intentionalClose=!0;try{this.ws.close(1e3,"client ended call")}catch{}this.ws=null}}onResult(e){this.resultCallbacks.push(e)}onError(e){this.errorCallbacks.push(e)}onStatusChange(e){this.statusCallbacks.push(e)}onTranscript(e){this.transcriptCallbacks.push(e)}onMetrics(e){this.metricsCallbacks.push(e)}emitStatus(e){this.statusCallbacks.forEach(t=>t(e))}emitError(e){this.errorCallbacks.forEach(t=>t(e))}emitTranscript(e,t,i){this.transcriptCallbacks.forEach(s=>s(e,t,i))}emitMetrics(e){this.metricsCallbacks.forEach(t=>t(e))}};var p=class{constructor(e={}){this.config=e;this.type="browser";this.recognition=null;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.isListening=!1;this.w=typeof window<"u"?window:void 0}async connect(){this.statusCallbacks.forEach(e=>e("connected"))}async startListening(){try{if(this.isListening)throw new Error("Already listening");if(!this.w)throw new Error("Window object not available");let e=this.w.SpeechRecognition||this.w.webkitSpeechRecognition;if(!e)throw new Error("Browser speech recognition not supported");this.recognition=new e,this.recognition.lang=this.config?.language||"en-US",this.recognition.continuous=this.config?.continuous||!1,this.recognition.interimResults=!0,this.recognition.onresult=t=>{let i=Array.from(t.results).map(n=>n[0]).map(n=>n.transcript).join(""),s=t.results[t.results.length-1].isFinal;this.resultCallbacks.forEach(n=>n({text:i,confidence:s?.8:.5,provider:"browser"})),s&&!this.config?.continuous&&this.stopListening()},this.recognition.onerror=t=>{this.errorCallbacks.forEach(i=>i(new Error(t.error))),this.statusCallbacks.forEach(i=>i("error"))},this.recognition.onstart=()=>{this.isListening=!0,this.statusCallbacks.forEach(t=>t("listening"))},this.recognition.onend=()=>{this.isListening=!1,this.statusCallbacks.forEach(t=>t("idle"))},this.recognition.start()}catch(e){throw this.errorCallbacks.forEach(t=>t(e)),this.statusCallbacks.forEach(t=>t("error")),e}}async stopListening(){this.recognition&&(this.recognition.stop(),this.recognition=null),this.isListening=!1,this.statusCallbacks.forEach(e=>e("idle"))}onResult(e){this.resultCallbacks.push(e)}onError(e){this.errorCallbacks.push(e)}onStatusChange(e){this.statusCallbacks.push(e)}async disconnect(){await this.stopListening(),this.statusCallbacks.forEach(e=>e("disconnected"))}static isSupported(){return"SpeechRecognition"in window||"webkitSpeechRecognition"in window}};function f(r){switch(r.type){case"runtype":if(!r.runtype)throw new Error("Runtype voice provider requires configuration");return new d(r.runtype);case"browser":if(!p.isSupported())throw new Error("Browser speech recognition not supported");return new p(r.browser||{});case"custom":{let e=r.custom;if(!e)throw new Error("Custom voice provider requires a `custom` provider instance or factory");let t=typeof e=="function"?e():e;if(!t||typeof t.startListening!="function")throw new Error("Custom voice provider `custom` must be a VoiceProvider (or a factory returning one)");return t}default:throw new Error(`Unknown voice provider type: ${r.type}`)}}function b(r){if(r?.type==="custom"&&r.custom)return f({type:"custom",custom:r.custom});if(r?.type==="runtype"&&r.runtype)return f({type:"runtype",runtype:r.runtype});if(p.isSupported())return f({type:"browser",browser:r?.browser||{language:"en-US"}});throw new Error("No supported voice providers available")}function w(r){try{return b(r),!0}catch{return!1}}0&&(module.exports={BrowserVoiceProvider,RuntypeVoiceProvider,createBestAvailableVoiceProvider,createVoiceProvider,isVoiceSupported});
+"use strict";var g=Object.defineProperty;var C=Object.getOwnPropertyDescriptor;var E=Object.getOwnPropertyNames;var A=Object.prototype.hasOwnProperty;var x=(r,t)=>{for(var e in t)g(r,e,{get:t[e],enumerable:!0})},P=(r,t,e,i)=>{if(t&&typeof t=="object"||typeof t=="function")for(let s of E(t))!A.call(r,s)&&s!==e&&g(r,s,{get:()=>t[s],enumerable:!(i=C(t,s))||i.enumerable});return r};var T=r=>P(g({},"__esModule",{value:!0}),r);var O={};x(O,{BrowserVoiceProvider:()=>p,RuntypeVoiceProvider:()=>f,createBestAvailableVoiceProvider:()=>b,createVoiceProvider:()=>m,isVoiceSupported:()=>S});module.exports=T(O);var v=class{constructor(t=24e3,e={}){this.ctx=null;this.nextStartTime=0;this.activeSources=[];this.finishedCallbacks=[];this.startedCallbacks=[];this.playing=!1;this.streamEnded=!1;this.pendingCount=0;this.started=!1;this.userPaused=!1;this.pendingBuffers=[];this.pendingSamples=0;this.remainder=null;this.sampleRate=t;let i=Math.max(0,e.prebufferMs??0);this.waterlineSamples=Math.round(t*i/1e3),this.buffering=this.waterlineSamples>0}ensureContext(){if(!this.ctx){let e=typeof window<"u"?window:void 0;if(!e)throw new Error("AudioPlaybackManager requires a browser environment");let i=e.AudioContext||e.webkitAudioContext;this.ctx=new i({sampleRate:this.sampleRate})}let t=this.ctx;return t.state==="suspended"&&!this.userPaused&&t.resume(),t}enqueue(t){if(t.length===0)return;let e=t;if(this.remainder){let s=new Uint8Array(this.remainder.length+t.length);s.set(this.remainder),s.set(t,this.remainder.length),e=s,this.remainder=null}if(e.length%2!==0&&(this.remainder=new Uint8Array([e[e.length-1]]),e=e.subarray(0,e.length-1)),e.length===0)return;let i=this.pcmToFloat32(e);i.length!==0&&(this.buffering?(this.pendingBuffers.push(i),this.pendingSamples+=i.length,this.pendingSamples>=this.waterlineSamples&&this.releaseBuffer()):this.scheduleSamples(i))}markStreamEnd(){this.pendingBuffers.length>0&&this.releaseBuffer(),this.streamEnded=!0,this.checkFinished()}flush(){for(let t of this.activeSources)try{t.stop(),t.disconnect()}catch{}this.activeSources=[],this.pendingCount=0,this.nextStartTime=0,this.playing=!1,this.streamEnded=!1,this.finishedCallbacks=[],this.startedCallbacks=[],this.remainder=null,this.pendingBuffers=[],this.pendingSamples=0,this.buffering=this.waterlineSamples>0,this.started=!1}isPlaying(){return this.playing}onFinished(t){this.finishedCallbacks.push(t)}onStarted(t){this.startedCallbacks.push(t)}pause(){this.userPaused=!0,this.ctx&&this.ctx.state==="running"&&this.ctx.suspend()}resume(){this.userPaused=!1,this.ctx&&this.ctx.state==="suspended"&&this.ctx.resume()}async destroy(){this.flush(),this.ctx&&(await this.ctx.close(),this.ctx=null)}releaseBuffer(){this.buffering=!1;let t=this.pendingBuffers;this.pendingBuffers=[],this.pendingSamples=0;for(let e of t)this.scheduleSamples(e)}scheduleSamples(t){if(t.length===0)return;let e=this.ensureContext(),i=e.createBuffer(1,t.length,this.sampleRate);i.getChannelData(0).set(t);let s=e.createBufferSource();s.buffer=i,s.connect(e.destination);let n=e.currentTime;if(this.nextStartTime===0?this.nextStartTime=n:this.nextStartTime<n&&(this.nextStartTime=n,this.waterlineSamples>0&&(this.buffering=!0)),s.start(this.nextStartTime),this.nextStartTime+=i.duration,this.activeSources.push(s),this.pendingCount++,this.playing=!0,!this.started){this.started=!0;let l=this.startedCallbacks.slice();this.startedCallbacks=[];for(let c of l)c()}s.onended=()=>{let l=this.activeSources.indexOf(s);l!==-1&&this.activeSources.splice(l,1),this.pendingCount--,this.checkFinished()}}checkFinished(){if(this.streamEnded&&this.pendingCount<=0&&this.pendingBuffers.length===0){this.playing=!1,this.streamEnded=!1;let t=this.finishedCallbacks.slice();this.finishedCallbacks=[];for(let e of t)e()}}pcmToFloat32(t){let e=Math.floor(t.length/2),i=new Float32Array(e),s=new DataView(t.buffer,t.byteOffset,t.byteLength);for(let n=0;n<e;n++){let l=s.getInt16(n*2,!0);i[n]=l/32768}return i}};var y=16e3,M=24e3,V=4096,R=4,_=1380533830,L=3e4,I=5e3,B=3e4,N=3e4,U=6e5;function F(r){return r.byteLength>=44&&new DataView(r).getUint32(0,!1)===_?new Uint8Array(r,44):new Uint8Array(r)}function w(r){let t=r.replace(/\/+$/,"");return/^wss?:\/\//i.test(t)?t:/^https?:\/\//i.test(t)?t.replace(/^http/i,"ws"):`${typeof window<"u"&&window.location?.protocol==="https:"?"wss:":"ws:"}//${t}`}function W(r){return w(r).replace(/^ws/i,"http")}function G(r){return typeof r!="number"||!Number.isFinite(r)?B:Math.min(U,Math.max(N,Math.round(r)))}var f=class{constructor(t){this.config=t;this.type="runtype";this.ws=null;this.captureContext=null;this.levelCallbacks=[];this.mediaStream=null;this.sourceNode=null;this.processor=null;this.playback=null;this.callLive=!1;this.isSpeaking=!1;this.callGeneration=0;this.intentionalClose=!1;this.lastPrewarmAt=Number.NEGATIVE_INFINITY;this.attached=null;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.transcriptCallbacks=[];this.metricsCallbacks=[]}async connect(){}prewarm(){if(this.callLive||this.attached)return;let t=this.config?.agentId,e=this.config?.clientToken,i=this.config?.host;if(!t||!e||!i)return;let s=Date.now();if(!(s-this.lastPrewarmAt<L)){this.lastPrewarmAt=s;try{this.config?.prewarmMode==="attach"?this.attach(i,t,e):this.requestPrewarm(i,t,e)}catch{}}}requestPrewarm(t,e,i){if(typeof fetch!="function")return;let s=`${W(t)}/v1/client/agents/${encodeURIComponent(e)}/voice/prewarm`;fetch(s,{method:"POST",headers:{Authorization:`Bearer ${i}`},keepalive:!0}).catch(()=>{})}attach(t,e,i){let s=new URLSearchParams({clientCapabilities:"attach"}),n=this.config?.attachIdleMs,l=G(n);n!==void 0&&s.set("attachIdleMs",String(l));let c=new WebSocket(`${this.voiceSocketUrl(t,e)}?${s}`,["runtype.bearer","runtype.attach",i]);c.binaryType="arraybuffer";let u=()=>{},d=new Promise(o=>{u=o}),a={ws:c,state:"connecting",ready:d,settle:u,openedAt:0,idleTimer:void 0};this.attached=a,c.onopen=()=>{this.attached===a&&(a.state=c.protocol==="runtype.attach"?"attached":"live",a.openedAt=Date.now(),this.armAttachIdle(a,l),a.state==="attached"&&c.send('{"type":"ping"}'),a.settle())},c.onmessage=o=>{if(this.attached!==a||typeof o.data!="string")return;let h;try{h=JSON.parse(o.data)}catch{return}h.type==="attached"&&typeof h.idleMs=="number"&&this.armAttachIdle(a,h.idleMs-(Date.now()-a.openedAt))},c.onerror=()=>{},c.onclose=()=>{this.attached===a&&this.releaseAttached()}}armAttachIdle(t,e){clearTimeout(t.idleTimer),t.idleTimer=setTimeout(()=>{this.attached===t&&this.releaseAttached()},Math.max(0,e))}releaseAttached(){let t=this.attached;if(!t)return;this.attached=null,clearTimeout(t.idleTimer);let{ws:e}=t;if(e.onopen=null,e.onmessage=null,e.onerror=null,e.onclose=null,t.settle(),e.readyState===WebSocket.CONNECTING||e.readyState===WebSocket.OPEN)try{e.close(1e3,"prewarm released")}catch{}}async takeAttachedSocket(){let t=this.attached;if(!t)return null;if(t.state==="connecting"){let e;await Promise.race([t.ready,new Promise(i=>{e=setTimeout(i,I)})]),clearTimeout(e)}return this.attached!==t?null:t.state==="connecting"||t.ws.readyState!==WebSocket.OPEN?(this.releaseAttached(),null):(this.attached=null,clearTimeout(t.idleTimer),{ws:t.ws,live:t.state==="live"})}voiceSocketUrl(t,e){return`${w(t)}/ws/agents/${encodeURIComponent(e)}/voice`}async startListening(){if(this.callLive)return;let t=this.config?.agentId,e=this.config?.clientToken,i=this.config?.host;if(!t)throw new Error("Runtype voice requires an agentId");if(!e)throw new Error("Runtype voice requires a clientToken");if(!i)throw new Error("Runtype voice requires a host (or widget apiUrl)");let s=++this.callGeneration;this.intentionalClose=!1,this.callLive=!0;try{let n=await navigator.mediaDevices.getUserMedia({audio:{sampleRate:y,channelCount:1,echoCancellation:!0}});if(s!==this.callGeneration){n.getTracks().forEach(o=>o.stop());return}this.mediaStream=n;let l=window.AudioContext||window.webkitAudioContext,c=new l({sampleRate:y});c.state==="suspended"&&await c.resume().catch(()=>{}),this.captureContext=c;let u=this.config?.createPlaybackEngine?await this.config.createPlaybackEngine():new v(M);if(s!==this.callGeneration){u.destroy(),n.getTracks().forEach(o=>o.stop()),c.close().catch(()=>{});return}this.playback=u,u.onFinished(()=>{s===this.callGeneration&&(this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening"))});let d=await this.takeAttachedSocket();if(s!==this.callGeneration){d?.ws.close(1e3,"client ended call");return}if(d){let{ws:o}=d;this.ws=o,this.bindCallSocket(o,s),d.live||o.send('{"type":"start"}'),this.emitStatus("listening"),this.startCapture(c,n,o,s);return}let a=new WebSocket(this.voiceSocketUrl(i,t),["runtype.bearer",e]);a.binaryType="arraybuffer",this.ws=a,a.onopen=()=>{s===this.callGeneration&&(this.emitStatus("listening"),this.startCapture(c,n,a,s))},this.bindCallSocket(a,s)}catch(n){throw this.cleanup(),this.emitError(n),this.emitStatus("error"),n}}bindCallSocket(t,e){t.onmessage=i=>this.handleMessage(i,e),t.onerror=()=>{e===this.callGeneration&&(this.emitError(new Error("Voice connection failed")),this.emitStatus("error"),this.cleanup())},t.onclose=i=>{if(this.intentionalClose){this.intentionalClose=!1;return}if(e===this.callGeneration){if(i.code!==1e3){let s=i.code?` (code ${i.code})`:"";this.emitError(new Error(`Voice connection closed${s}`)),this.emitStatus("error")}else this.emitStatus("idle");this.cleanup()}}}async stopListening(){this.cleanup(),this.emitStatus("idle")}async disconnect(){this.cleanup(),this.emitStatus("disconnected"),this.resultCallbacks=[],this.errorCallbacks=[],this.statusCallbacks=[],this.transcriptCallbacks=[],this.metricsCallbacks=[]}stopPlayback(){this.playback&&this.playback.flush(),this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening")}getInterruptionMode(){return"barge-in"}isBargeInActive(){return this.callLive}onLevel(t){this.levelCallbacks.push(t)}async deactivateBargeIn(){this.cleanup(),this.emitStatus("idle")}startCapture(t,e,i,s){let n=t.createMediaStreamSource(e);this.sourceNode=n;let l=t.createScriptProcessor(V,1,1);this.processor=l,l.onaudioprocess=c=>{if(s!==this.callGeneration||i.readyState!==WebSocket.OPEN)return;let u=c.inputBuffer.getChannelData(0),d=new Int16Array(u.length),a=0;for(let o=0;o<u.length;o++){let h=Math.max(-1,Math.min(1,u[o]));a+=h*h,d[o]=h<0?h*32768:h*32767}if(i.send(d.buffer),this.levelCallbacks.length>0){let o=Math.sqrt(a/u.length),h=Math.max(0,Math.min(1,o*R));for(let k of this.levelCallbacks)k(h)}},n.connect(l),l.connect(t.destination)}handleMessage(t,e){if(e!==this.callGeneration)return;if(t.data instanceof ArrayBuffer){this.handleAudioFrame(t.data,e);return}let i;try{i=JSON.parse(t.data)}catch{return}switch(i.type){case"transcript_interim":this.emitStatus("listening"),this.emitTranscript("user",i.text??"",!1);break;case"transcript_final":{let s=i.role==="assistant"?"assistant":"user";this.emitStatus(s==="user"?"processing":"speaking"),this.emitTranscript(s,i.text??"",!0);break}case"audio_end":this.playback?this.playback.markStreamEnd():(this.isSpeaking=!1,this.emitStatus("listening"));break;case"metrics":this.emitMetrics({llmMs:i.llm_ms,ttsMs:i.tts_ms,firstAudioMs:i.first_audio_ms,totalMs:i.total_ms});break;case"error":this.emitError(new Error(i.error||"Voice error")),this.emitStatus("error");break}}handleAudioFrame(t,e){if(e!==this.callGeneration||!this.playback)return;let i=F(t);i.length!==0&&(this.isSpeaking||(this.isSpeaking=!0,this.emitStatus("speaking")),this.playback.enqueue(i))}cleanup(){if(this.callGeneration+=1,this.callLive=!1,this.isSpeaking=!1,this.releaseAttached(),this.processor&&(this.processor.onaudioprocess=null,this.processor.disconnect(),this.processor=null),this.sourceNode&&(this.sourceNode.disconnect(),this.sourceNode=null),this.mediaStream&&(this.mediaStream.getTracks().forEach(t=>t.stop()),this.mediaStream=null),this.captureContext&&(this.captureContext.close().catch(()=>{}),this.captureContext=null),this.playback&&(this.playback.destroy(),this.playback=null),this.ws){this.intentionalClose=!0;try{this.ws.close(1e3,"client ended call")}catch{}this.ws=null}}onResult(t){this.resultCallbacks.push(t)}onError(t){this.errorCallbacks.push(t)}onStatusChange(t){this.statusCallbacks.push(t)}onTranscript(t){this.transcriptCallbacks.push(t)}onMetrics(t){this.metricsCallbacks.push(t)}emitStatus(t){this.statusCallbacks.forEach(e=>e(t))}emitError(t){this.errorCallbacks.forEach(e=>e(t))}emitTranscript(t,e,i){this.transcriptCallbacks.forEach(s=>s(t,e,i))}emitMetrics(t){this.metricsCallbacks.forEach(e=>e(t))}};var p=class{constructor(t={}){this.config=t;this.type="browser";this.recognition=null;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.isListening=!1;this.w=typeof window<"u"?window:void 0}async connect(){this.statusCallbacks.forEach(t=>t("connected"))}async startListening(){try{if(this.isListening)throw new Error("Already listening");if(!this.w)throw new Error("Window object not available");let t=this.w.SpeechRecognition||this.w.webkitSpeechRecognition;if(!t)throw new Error("Browser speech recognition not supported");this.recognition=new t,this.recognition.lang=this.config?.language||"en-US",this.recognition.continuous=this.config?.continuous||!1,this.recognition.interimResults=!0,this.recognition.onresult=e=>{let i=Array.from(e.results).map(n=>n[0]).map(n=>n.transcript).join(""),s=e.results[e.results.length-1].isFinal;this.resultCallbacks.forEach(n=>n({text:i,confidence:s?.8:.5,provider:"browser"})),s&&!this.config?.continuous&&this.stopListening()},this.recognition.onerror=e=>{this.errorCallbacks.forEach(i=>i(new Error(e.error))),this.statusCallbacks.forEach(i=>i("error"))},this.recognition.onstart=()=>{this.isListening=!0,this.statusCallbacks.forEach(e=>e("listening"))},this.recognition.onend=()=>{this.isListening=!1,this.statusCallbacks.forEach(e=>e("idle"))},this.recognition.start()}catch(t){throw this.errorCallbacks.forEach(e=>e(t)),this.statusCallbacks.forEach(e=>e("error")),t}}async stopListening(){this.recognition&&(this.recognition.stop(),this.recognition=null),this.isListening=!1,this.statusCallbacks.forEach(t=>t("idle"))}onResult(t){this.resultCallbacks.push(t)}onError(t){this.errorCallbacks.push(t)}onStatusChange(t){this.statusCallbacks.push(t)}async disconnect(){await this.stopListening(),this.statusCallbacks.forEach(t=>t("disconnected"))}static isSupported(){return"SpeechRecognition"in window||"webkitSpeechRecognition"in window}};function m(r){switch(r.type){case"runtype":if(!r.runtype)throw new Error("Runtype voice provider requires configuration");return new f(r.runtype);case"browser":if(!p.isSupported())throw new Error("Browser speech recognition not supported");return new p(r.browser||{});case"custom":{let t=r.custom;if(!t)throw new Error("Custom voice provider requires a `custom` provider instance or factory");let e=typeof t=="function"?t():t;if(!e||typeof e.startListening!="function")throw new Error("Custom voice provider `custom` must be a VoiceProvider (or a factory returning one)");return e}default:throw new Error(`Unknown voice provider type: ${r.type}`)}}function b(r){if(r?.type==="custom"&&r.custom)return m({type:"custom",custom:r.custom});if(r?.type==="runtype"&&r.runtype)return m({type:"runtype",runtype:r.runtype});if(p.isSupported())return m({type:"browser",browser:r?.browser||{language:"en-US"}});throw new Error("No supported voice providers available")}function S(r){try{return b(r),!0}catch{return!1}}0&&(module.exports={BrowserVoiceProvider,RuntypeVoiceProvider,createBestAvailableVoiceProvider,createVoiceProvider,isVoiceSupported});

package/dist/voice-runtime.d.cts

@@ -77,6 +77,10 @@ type VoiceConfig = {
          * `@runtypelabs/persona/voice-worklet-player` for a jitter-buffered engine.
          */
         createPlaybackEngine?: () => VoicePlaybackEngine | Promise<VoicePlaybackEngine>;
+        /** See `voiceRecognition.provider.runtype.prewarmMode`. @default 'request' */
+        prewarmMode?: 'request' | 'attach';
+        /** See `voiceRecognition.provider.runtype.attachIdleMs`. @default 30000 */
+        attachIdleMs?: number;
         /** @deprecated No-op on the realtime path: the server's STT owns turn-taking. */
         pauseDuration?: number;
         /** @deprecated No-op on the realtime path: the server's STT owns turn-taking. */
@@ -137,6 +141,13 @@ interface VoiceProvider {
     deactivateBargeIn?(): Promise<void>;
     /** Stop playback / cancel in-flight request without starting recording */
     stopPlayback?(): void;
+    /**
+     * Warm the provider ahead of `startListening()` (the widget calls it on mic
+     * intent, see `voiceRecognition.prewarm`). Must be fire-and-forget: never
+     * throw, emit errors or status, or acquire the microphone. May be called
+     * repeatedly; throttle as needed.
+     */
+    prewarm?(): void;
 }
 
 declare function createVoiceProvider(config: VoiceConfig): VoiceProvider;
@@ -157,6 +168,8 @@ declare class RuntypeVoiceProvider implements VoiceProvider {
     private isSpeaking;
     private callGeneration;
     private intentionalClose;
+    private lastPrewarmAt;
+    private attached;
     private resultCallbacks;
     private errorCallbacks;
     private statusCallbacks;
@@ -165,8 +178,28 @@ declare class RuntypeVoiceProvider implements VoiceProvider {
     constructor(config: VoiceConfig["runtype"]);
     /** No-op: the WS session opens lazily in `startListening` (the "call"). */
     connect(): Promise<void>;
+    /**
+     * Warm the voice path ahead of the click. Fire-and-forget and throttled to
+     * once per {@link PREWARM_THROTTLE_MS}; never emits an error or a status.
+     */
+    prewarm(): void;
+    private requestPrewarm;
+    private attach;
+    private armAttachIdle;
+    /** Close and forget the prewarmed socket, if any. Silent by design. */
+    private releaseAttached;
+    /**
+     * Hand the prewarmed socket to the call, waiting up to
+     * {@link ATTACH_HANDSHAKE_TIMEOUT_MS} for a handshake still in flight.
+     * Resolves `null` when there is none, it stalled, or it dropped, so the
+     * caller opens a fresh socket.
+     */
+    private takeAttachedSocket;
+    private voiceSocketUrl;
     /** Start the call: acquire mic, open the WS, stream PCM until hang-up. */
     startListening(): Promise<void>;
+    /** Route a call socket's frames, errors, and close into this call. */
+    private bindCallSocket;
     /** End the call (hang up). */
     stopListening(): Promise<void>;
     /** Tear down the call and drop all callbacks (used by `cleanupVoice`). */

package/dist/voice-runtime.d.ts

@@ -77,6 +77,10 @@ type VoiceConfig = {
          * `@runtypelabs/persona/voice-worklet-player` for a jitter-buffered engine.
          */
         createPlaybackEngine?: () => VoicePlaybackEngine | Promise<VoicePlaybackEngine>;
+        /** See `voiceRecognition.provider.runtype.prewarmMode`. @default 'request' */
+        prewarmMode?: 'request' | 'attach';
+        /** See `voiceRecognition.provider.runtype.attachIdleMs`. @default 30000 */
+        attachIdleMs?: number;
         /** @deprecated No-op on the realtime path: the server's STT owns turn-taking. */
         pauseDuration?: number;
         /** @deprecated No-op on the realtime path: the server's STT owns turn-taking. */
@@ -137,6 +141,13 @@ interface VoiceProvider {
     deactivateBargeIn?(): Promise<void>;
     /** Stop playback / cancel in-flight request without starting recording */
     stopPlayback?(): void;
+    /**
+     * Warm the provider ahead of `startListening()` (the widget calls it on mic
+     * intent, see `voiceRecognition.prewarm`). Must be fire-and-forget: never
+     * throw, emit errors or status, or acquire the microphone. May be called
+     * repeatedly; throttle as needed.
+     */
+    prewarm?(): void;
 }
 
 declare function createVoiceProvider(config: VoiceConfig): VoiceProvider;
@@ -157,6 +168,8 @@ declare class RuntypeVoiceProvider implements VoiceProvider {
     private isSpeaking;
     private callGeneration;
     private intentionalClose;
+    private lastPrewarmAt;
+    private attached;
     private resultCallbacks;
     private errorCallbacks;
     private statusCallbacks;
@@ -165,8 +178,28 @@ declare class RuntypeVoiceProvider implements VoiceProvider {
     constructor(config: VoiceConfig["runtype"]);
     /** No-op: the WS session opens lazily in `startListening` (the "call"). */
     connect(): Promise<void>;
+    /**
+     * Warm the voice path ahead of the click. Fire-and-forget and throttled to
+     * once per {@link PREWARM_THROTTLE_MS}; never emits an error or a status.
+     */
+    prewarm(): void;
+    private requestPrewarm;
+    private attach;
+    private armAttachIdle;
+    /** Close and forget the prewarmed socket, if any. Silent by design. */
+    private releaseAttached;
+    /**
+     * Hand the prewarmed socket to the call, waiting up to
+     * {@link ATTACH_HANDSHAKE_TIMEOUT_MS} for a handshake still in flight.
+     * Resolves `null` when there is none, it stalled, or it dropped, so the
+     * caller opens a fresh socket.
+     */
+    private takeAttachedSocket;
+    private voiceSocketUrl;
     /** Start the call: acquire mic, open the WS, stream PCM until hang-up. */
     startListening(): Promise<void>;
+    /** Route a call socket's frames, errors, and close into this call. */
+    private bindCallSocket;
     /** End the call (hang up). */
     stopListening(): Promise<void>;
     /** Tear down the call and drop all callbacks (used by `cleanupVoice`). */

package/dist/voice-runtime.js

@@ -1 +1 @@
-var m=class{constructor(e=24e3,t={}){this.ctx=null;this.nextStartTime=0;this.activeSources=[];this.finishedCallbacks=[];this.startedCallbacks=[];this.playing=!1;this.streamEnded=!1;this.pendingCount=0;this.started=!1;this.userPaused=!1;this.pendingBuffers=[];this.pendingSamples=0;this.remainder=null;this.sampleRate=e;let i=Math.max(0,t.prebufferMs??0);this.waterlineSamples=Math.round(e*i/1e3),this.buffering=this.waterlineSamples>0}ensureContext(){if(!this.ctx){let t=typeof window<"u"?window:void 0;if(!t)throw new Error("AudioPlaybackManager requires a browser environment");let i=t.AudioContext||t.webkitAudioContext;this.ctx=new i({sampleRate:this.sampleRate})}let e=this.ctx;return e.state==="suspended"&&!this.userPaused&&e.resume(),e}enqueue(e){if(e.length===0)return;let t=e;if(this.remainder){let s=new Uint8Array(this.remainder.length+e.length);s.set(this.remainder),s.set(e,this.remainder.length),t=s,this.remainder=null}if(t.length%2!==0&&(this.remainder=new Uint8Array([t[t.length-1]]),t=t.subarray(0,t.length-1)),t.length===0)return;let i=this.pcmToFloat32(t);i.length!==0&&(this.buffering?(this.pendingBuffers.push(i),this.pendingSamples+=i.length,this.pendingSamples>=this.waterlineSamples&&this.releaseBuffer()):this.scheduleSamples(i))}markStreamEnd(){this.pendingBuffers.length>0&&this.releaseBuffer(),this.streamEnded=!0,this.checkFinished()}flush(){for(let e of this.activeSources)try{e.stop(),e.disconnect()}catch{}this.activeSources=[],this.pendingCount=0,this.nextStartTime=0,this.playing=!1,this.streamEnded=!1,this.finishedCallbacks=[],this.startedCallbacks=[],this.remainder=null,this.pendingBuffers=[],this.pendingSamples=0,this.buffering=this.waterlineSamples>0,this.started=!1}isPlaying(){return this.playing}onFinished(e){this.finishedCallbacks.push(e)}onStarted(e){this.startedCallbacks.push(e)}pause(){this.userPaused=!0,this.ctx&&this.ctx.state==="running"&&this.ctx.suspend()}resume(){this.userPaused=!1,this.ctx&&this.ctx.state==="suspended"&&this.ctx.resume()}async destroy(){this.flush(),this.ctx&&(await this.ctx.close(),this.ctx=null)}releaseBuffer(){this.buffering=!1;let e=this.pendingBuffers;this.pendingBuffers=[],this.pendingSamples=0;for(let t of e)this.scheduleSamples(t)}scheduleSamples(e){if(e.length===0)return;let t=this.ensureContext(),i=t.createBuffer(1,e.length,this.sampleRate);i.getChannelData(0).set(e);let s=t.createBufferSource();s.buffer=i,s.connect(t.destination);let n=t.currentTime;if(this.nextStartTime===0?this.nextStartTime=n:this.nextStartTime<n&&(this.nextStartTime=n,this.waterlineSamples>0&&(this.buffering=!0)),s.start(this.nextStartTime),this.nextStartTime+=i.duration,this.activeSources.push(s),this.pendingCount++,this.playing=!0,!this.started){this.started=!0;let a=this.startedCallbacks.slice();this.startedCallbacks=[];for(let l of a)l()}s.onended=()=>{let a=this.activeSources.indexOf(s);a!==-1&&this.activeSources.splice(a,1),this.pendingCount--,this.checkFinished()}}checkFinished(){if(this.streamEnded&&this.pendingCount<=0&&this.pendingBuffers.length===0){this.playing=!1,this.streamEnded=!1;let e=this.finishedCallbacks.slice();this.finishedCallbacks=[];for(let t of e)t()}}pcmToFloat32(e){let t=Math.floor(e.length/2),i=new Float32Array(t),s=new DataView(e.buffer,e.byteOffset,e.byteLength);for(let n=0;n<t;n++){let a=s.getInt16(n*2,!0);i[n]=a/32768}return i}};var g=16e3,w=24e3,S=4096,C=4,k=1380533830;function E(r){return r.byteLength>=44&&new DataView(r).getUint32(0,!1)===k?new Uint8Array(r,44):new Uint8Array(r)}function x(r){let e=r.replace(/\/+$/,"");return/^wss?:\/\//i.test(e)?e:/^https?:\/\//i.test(e)?e.replace(/^http/i,"ws"):`${typeof window<"u"&&window.location?.protocol==="https:"?"wss:":"ws:"}//${e}`}var d=class{constructor(e){this.config=e;this.type="runtype";this.ws=null;this.captureContext=null;this.levelCallbacks=[];this.mediaStream=null;this.sourceNode=null;this.processor=null;this.playback=null;this.callLive=!1;this.isSpeaking=!1;this.callGeneration=0;this.intentionalClose=!1;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.transcriptCallbacks=[];this.metricsCallbacks=[]}async connect(){}async startListening(){if(this.callLive)return;let e=this.config?.agentId,t=this.config?.clientToken,i=this.config?.host;if(!e)throw new Error("Runtype voice requires an agentId");if(!t)throw new Error("Runtype voice requires a clientToken");if(!i)throw new Error("Runtype voice requires a host (or widget apiUrl)");let s=++this.callGeneration;this.intentionalClose=!1,this.callLive=!0;try{let n=await navigator.mediaDevices.getUserMedia({audio:{sampleRate:g,channelCount:1,echoCancellation:!0}});if(s!==this.callGeneration){n.getTracks().forEach(o=>o.stop());return}this.mediaStream=n;let a=window.AudioContext||window.webkitAudioContext,l=new a({sampleRate:g});l.state==="suspended"&&await l.resume().catch(()=>{}),this.captureContext=l;let h=this.config?.createPlaybackEngine?await this.config.createPlaybackEngine():new m(w);if(s!==this.callGeneration){h.destroy(),n.getTracks().forEach(o=>o.stop()),l.close().catch(()=>{});return}this.playback=h,h.onFinished(()=>{s===this.callGeneration&&(this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening"))});let f=`${x(i)}/ws/agents/${encodeURIComponent(e)}/voice`,c=new WebSocket(f,["runtype.bearer",t]);c.binaryType="arraybuffer",this.ws=c,c.onopen=()=>{s===this.callGeneration&&(this.emitStatus("listening"),this.startCapture(l,n,c,s))},c.onmessage=o=>this.handleMessage(o,s),c.onerror=()=>{s===this.callGeneration&&(this.emitError(new Error("Voice connection failed")),this.emitStatus("error"),this.cleanup())},c.onclose=o=>{if(this.intentionalClose){this.intentionalClose=!1;return}if(s===this.callGeneration){if(o.code!==1e3){let u=o.code?` (code ${o.code})`:"";this.emitError(new Error(`Voice connection closed${u}`)),this.emitStatus("error")}else this.emitStatus("idle");this.cleanup()}}}catch(n){throw this.cleanup(),this.emitError(n),this.emitStatus("error"),n}}async stopListening(){this.cleanup(),this.emitStatus("idle")}async disconnect(){this.cleanup(),this.emitStatus("disconnected"),this.resultCallbacks=[],this.errorCallbacks=[],this.statusCallbacks=[],this.transcriptCallbacks=[],this.metricsCallbacks=[]}stopPlayback(){this.playback&&this.playback.flush(),this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening")}getInterruptionMode(){return"barge-in"}isBargeInActive(){return this.callLive}onLevel(e){this.levelCallbacks.push(e)}async deactivateBargeIn(){this.cleanup(),this.emitStatus("idle")}startCapture(e,t,i,s){let n=e.createMediaStreamSource(t);this.sourceNode=n;let a=e.createScriptProcessor(S,1,1);this.processor=a,a.onaudioprocess=l=>{if(s!==this.callGeneration||i.readyState!==WebSocket.OPEN)return;let h=l.inputBuffer.getChannelData(0),f=new Int16Array(h.length),c=0;for(let o=0;o<h.length;o++){let u=Math.max(-1,Math.min(1,h[o]));c+=u*u,f[o]=u<0?u*32768:u*32767}if(i.send(f.buffer),this.levelCallbacks.length>0){let o=Math.sqrt(c/h.length),u=Math.max(0,Math.min(1,o*C));for(let y of this.levelCallbacks)y(u)}},n.connect(a),a.connect(e.destination)}handleMessage(e,t){if(t!==this.callGeneration)return;if(e.data instanceof ArrayBuffer){this.handleAudioFrame(e.data,t);return}let i;try{i=JSON.parse(e.data)}catch{return}switch(i.type){case"transcript_interim":this.emitStatus("listening"),this.emitTranscript("user",i.text??"",!1);break;case"transcript_final":{let s=i.role==="assistant"?"assistant":"user";this.emitStatus(s==="user"?"processing":"speaking"),this.emitTranscript(s,i.text??"",!0);break}case"audio_end":this.playback?this.playback.markStreamEnd():(this.isSpeaking=!1,this.emitStatus("listening"));break;case"metrics":this.emitMetrics({llmMs:i.llm_ms,ttsMs:i.tts_ms,firstAudioMs:i.first_audio_ms,totalMs:i.total_ms});break;case"error":this.emitError(new Error(i.error||"Voice error")),this.emitStatus("error");break}}handleAudioFrame(e,t){if(t!==this.callGeneration||!this.playback)return;let i=E(e);i.length!==0&&(this.isSpeaking||(this.isSpeaking=!0,this.emitStatus("speaking")),this.playback.enqueue(i))}cleanup(){if(this.callGeneration+=1,this.callLive=!1,this.isSpeaking=!1,this.processor&&(this.processor.onaudioprocess=null,this.processor.disconnect(),this.processor=null),this.sourceNode&&(this.sourceNode.disconnect(),this.sourceNode=null),this.mediaStream&&(this.mediaStream.getTracks().forEach(e=>e.stop()),this.mediaStream=null),this.captureContext&&(this.captureContext.close().catch(()=>{}),this.captureContext=null),this.playback&&(this.playback.destroy(),this.playback=null),this.ws){this.intentionalClose=!0;try{this.ws.close(1e3,"client ended call")}catch{}this.ws=null}}onResult(e){this.resultCallbacks.push(e)}onError(e){this.errorCallbacks.push(e)}onStatusChange(e){this.statusCallbacks.push(e)}onTranscript(e){this.transcriptCallbacks.push(e)}onMetrics(e){this.metricsCallbacks.push(e)}emitStatus(e){this.statusCallbacks.forEach(t=>t(e))}emitError(e){this.errorCallbacks.forEach(t=>t(e))}emitTranscript(e,t,i){this.transcriptCallbacks.forEach(s=>s(e,t,i))}emitMetrics(e){this.metricsCallbacks.forEach(t=>t(e))}};var p=class{constructor(e={}){this.config=e;this.type="browser";this.recognition=null;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.isListening=!1;this.w=typeof window<"u"?window:void 0}async connect(){this.statusCallbacks.forEach(e=>e("connected"))}async startListening(){try{if(this.isListening)throw new Error("Already listening");if(!this.w)throw new Error("Window object not available");let e=this.w.SpeechRecognition||this.w.webkitSpeechRecognition;if(!e)throw new Error("Browser speech recognition not supported");this.recognition=new e,this.recognition.lang=this.config?.language||"en-US",this.recognition.continuous=this.config?.continuous||!1,this.recognition.interimResults=!0,this.recognition.onresult=t=>{let i=Array.from(t.results).map(n=>n[0]).map(n=>n.transcript).join(""),s=t.results[t.results.length-1].isFinal;this.resultCallbacks.forEach(n=>n({text:i,confidence:s?.8:.5,provider:"browser"})),s&&!this.config?.continuous&&this.stopListening()},this.recognition.onerror=t=>{this.errorCallbacks.forEach(i=>i(new Error(t.error))),this.statusCallbacks.forEach(i=>i("error"))},this.recognition.onstart=()=>{this.isListening=!0,this.statusCallbacks.forEach(t=>t("listening"))},this.recognition.onend=()=>{this.isListening=!1,this.statusCallbacks.forEach(t=>t("idle"))},this.recognition.start()}catch(e){throw this.errorCallbacks.forEach(t=>t(e)),this.statusCallbacks.forEach(t=>t("error")),e}}async stopListening(){this.recognition&&(this.recognition.stop(),this.recognition=null),this.isListening=!1,this.statusCallbacks.forEach(e=>e("idle"))}onResult(e){this.resultCallbacks.push(e)}onError(e){this.errorCallbacks.push(e)}onStatusChange(e){this.statusCallbacks.push(e)}async disconnect(){await this.stopListening(),this.statusCallbacks.forEach(e=>e("disconnected"))}static isSupported(){return"SpeechRecognition"in window||"webkitSpeechRecognition"in window}};function v(r){switch(r.type){case"runtype":if(!r.runtype)throw new Error("Runtype voice provider requires configuration");return new d(r.runtype);case"browser":if(!p.isSupported())throw new Error("Browser speech recognition not supported");return new p(r.browser||{});case"custom":{let e=r.custom;if(!e)throw new Error("Custom voice provider requires a `custom` provider instance or factory");let t=typeof e=="function"?e():e;if(!t||typeof t.startListening!="function")throw new Error("Custom voice provider `custom` must be a VoiceProvider (or a factory returning one)");return t}default:throw new Error(`Unknown voice provider type: ${r.type}`)}}function b(r){if(r?.type==="custom"&&r.custom)return v({type:"custom",custom:r.custom});if(r?.type==="runtype"&&r.runtype)return v({type:"runtype",runtype:r.runtype});if(p.isSupported())return v({type:"browser",browser:r?.browser||{language:"en-US"}});throw new Error("No supported voice providers available")}function A(r){try{return b(r),!0}catch{return!1}}export{p as BrowserVoiceProvider,d as RuntypeVoiceProvider,b as createBestAvailableVoiceProvider,v as createVoiceProvider,A as isVoiceSupported};
+var m=class{constructor(t=24e3,e={}){this.ctx=null;this.nextStartTime=0;this.activeSources=[];this.finishedCallbacks=[];this.startedCallbacks=[];this.playing=!1;this.streamEnded=!1;this.pendingCount=0;this.started=!1;this.userPaused=!1;this.pendingBuffers=[];this.pendingSamples=0;this.remainder=null;this.sampleRate=t;let i=Math.max(0,e.prebufferMs??0);this.waterlineSamples=Math.round(t*i/1e3),this.buffering=this.waterlineSamples>0}ensureContext(){if(!this.ctx){let e=typeof window<"u"?window:void 0;if(!e)throw new Error("AudioPlaybackManager requires a browser environment");let i=e.AudioContext||e.webkitAudioContext;this.ctx=new i({sampleRate:this.sampleRate})}let t=this.ctx;return t.state==="suspended"&&!this.userPaused&&t.resume(),t}enqueue(t){if(t.length===0)return;let e=t;if(this.remainder){let s=new Uint8Array(this.remainder.length+t.length);s.set(this.remainder),s.set(t,this.remainder.length),e=s,this.remainder=null}if(e.length%2!==0&&(this.remainder=new Uint8Array([e[e.length-1]]),e=e.subarray(0,e.length-1)),e.length===0)return;let i=this.pcmToFloat32(e);i.length!==0&&(this.buffering?(this.pendingBuffers.push(i),this.pendingSamples+=i.length,this.pendingSamples>=this.waterlineSamples&&this.releaseBuffer()):this.scheduleSamples(i))}markStreamEnd(){this.pendingBuffers.length>0&&this.releaseBuffer(),this.streamEnded=!0,this.checkFinished()}flush(){for(let t of this.activeSources)try{t.stop(),t.disconnect()}catch{}this.activeSources=[],this.pendingCount=0,this.nextStartTime=0,this.playing=!1,this.streamEnded=!1,this.finishedCallbacks=[],this.startedCallbacks=[],this.remainder=null,this.pendingBuffers=[],this.pendingSamples=0,this.buffering=this.waterlineSamples>0,this.started=!1}isPlaying(){return this.playing}onFinished(t){this.finishedCallbacks.push(t)}onStarted(t){this.startedCallbacks.push(t)}pause(){this.userPaused=!0,this.ctx&&this.ctx.state==="running"&&this.ctx.suspend()}resume(){this.userPaused=!1,this.ctx&&this.ctx.state==="suspended"&&this.ctx.resume()}async destroy(){this.flush(),this.ctx&&(await this.ctx.close(),this.ctx=null)}releaseBuffer(){this.buffering=!1;let t=this.pendingBuffers;this.pendingBuffers=[],this.pendingSamples=0;for(let e of t)this.scheduleSamples(e)}scheduleSamples(t){if(t.length===0)return;let e=this.ensureContext(),i=e.createBuffer(1,t.length,this.sampleRate);i.getChannelData(0).set(t);let s=e.createBufferSource();s.buffer=i,s.connect(e.destination);let n=e.currentTime;if(this.nextStartTime===0?this.nextStartTime=n:this.nextStartTime<n&&(this.nextStartTime=n,this.waterlineSamples>0&&(this.buffering=!0)),s.start(this.nextStartTime),this.nextStartTime+=i.duration,this.activeSources.push(s),this.pendingCount++,this.playing=!0,!this.started){this.started=!0;let l=this.startedCallbacks.slice();this.startedCallbacks=[];for(let c of l)c()}s.onended=()=>{let l=this.activeSources.indexOf(s);l!==-1&&this.activeSources.splice(l,1),this.pendingCount--,this.checkFinished()}}checkFinished(){if(this.streamEnded&&this.pendingCount<=0&&this.pendingBuffers.length===0){this.playing=!1,this.streamEnded=!1;let t=this.finishedCallbacks.slice();this.finishedCallbacks=[];for(let e of t)e()}}pcmToFloat32(t){let e=Math.floor(t.length/2),i=new Float32Array(e),s=new DataView(t.buffer,t.byteOffset,t.byteLength);for(let n=0;n<e;n++){let l=s.getInt16(n*2,!0);i[n]=l/32768}return i}};var g=16e3,S=24e3,k=4096,C=4,E=1380533830,A=3e4,x=5e3,P=3e4,T=3e4,M=6e5;function V(r){return r.byteLength>=44&&new DataView(r).getUint32(0,!1)===E?new Uint8Array(r,44):new Uint8Array(r)}function b(r){let t=r.replace(/\/+$/,"");return/^wss?:\/\//i.test(t)?t:/^https?:\/\//i.test(t)?t.replace(/^http/i,"ws"):`${typeof window<"u"&&window.location?.protocol==="https:"?"wss:":"ws:"}//${t}`}function R(r){return b(r).replace(/^ws/i,"http")}function _(r){return typeof r!="number"||!Number.isFinite(r)?P:Math.min(M,Math.max(T,Math.round(r)))}var f=class{constructor(t){this.config=t;this.type="runtype";this.ws=null;this.captureContext=null;this.levelCallbacks=[];this.mediaStream=null;this.sourceNode=null;this.processor=null;this.playback=null;this.callLive=!1;this.isSpeaking=!1;this.callGeneration=0;this.intentionalClose=!1;this.lastPrewarmAt=Number.NEGATIVE_INFINITY;this.attached=null;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.transcriptCallbacks=[];this.metricsCallbacks=[]}async connect(){}prewarm(){if(this.callLive||this.attached)return;let t=this.config?.agentId,e=this.config?.clientToken,i=this.config?.host;if(!t||!e||!i)return;let s=Date.now();if(!(s-this.lastPrewarmAt<A)){this.lastPrewarmAt=s;try{this.config?.prewarmMode==="attach"?this.attach(i,t,e):this.requestPrewarm(i,t,e)}catch{}}}requestPrewarm(t,e,i){if(typeof fetch!="function")return;let s=`${R(t)}/v1/client/agents/${encodeURIComponent(e)}/voice/prewarm`;fetch(s,{method:"POST",headers:{Authorization:`Bearer ${i}`},keepalive:!0}).catch(()=>{})}attach(t,e,i){let s=new URLSearchParams({clientCapabilities:"attach"}),n=this.config?.attachIdleMs,l=_(n);n!==void 0&&s.set("attachIdleMs",String(l));let c=new WebSocket(`${this.voiceSocketUrl(t,e)}?${s}`,["runtype.bearer","runtype.attach",i]);c.binaryType="arraybuffer";let u=()=>{},d=new Promise(o=>{u=o}),a={ws:c,state:"connecting",ready:d,settle:u,openedAt:0,idleTimer:void 0};this.attached=a,c.onopen=()=>{this.attached===a&&(a.state=c.protocol==="runtype.attach"?"attached":"live",a.openedAt=Date.now(),this.armAttachIdle(a,l),a.state==="attached"&&c.send('{"type":"ping"}'),a.settle())},c.onmessage=o=>{if(this.attached!==a||typeof o.data!="string")return;let h;try{h=JSON.parse(o.data)}catch{return}h.type==="attached"&&typeof h.idleMs=="number"&&this.armAttachIdle(a,h.idleMs-(Date.now()-a.openedAt))},c.onerror=()=>{},c.onclose=()=>{this.attached===a&&this.releaseAttached()}}armAttachIdle(t,e){clearTimeout(t.idleTimer),t.idleTimer=setTimeout(()=>{this.attached===t&&this.releaseAttached()},Math.max(0,e))}releaseAttached(){let t=this.attached;if(!t)return;this.attached=null,clearTimeout(t.idleTimer);let{ws:e}=t;if(e.onopen=null,e.onmessage=null,e.onerror=null,e.onclose=null,t.settle(),e.readyState===WebSocket.CONNECTING||e.readyState===WebSocket.OPEN)try{e.close(1e3,"prewarm released")}catch{}}async takeAttachedSocket(){let t=this.attached;if(!t)return null;if(t.state==="connecting"){let e;await Promise.race([t.ready,new Promise(i=>{e=setTimeout(i,x)})]),clearTimeout(e)}return this.attached!==t?null:t.state==="connecting"||t.ws.readyState!==WebSocket.OPEN?(this.releaseAttached(),null):(this.attached=null,clearTimeout(t.idleTimer),{ws:t.ws,live:t.state==="live"})}voiceSocketUrl(t,e){return`${b(t)}/ws/agents/${encodeURIComponent(e)}/voice`}async startListening(){if(this.callLive)return;let t=this.config?.agentId,e=this.config?.clientToken,i=this.config?.host;if(!t)throw new Error("Runtype voice requires an agentId");if(!e)throw new Error("Runtype voice requires a clientToken");if(!i)throw new Error("Runtype voice requires a host (or widget apiUrl)");let s=++this.callGeneration;this.intentionalClose=!1,this.callLive=!0;try{let n=await navigator.mediaDevices.getUserMedia({audio:{sampleRate:g,channelCount:1,echoCancellation:!0}});if(s!==this.callGeneration){n.getTracks().forEach(o=>o.stop());return}this.mediaStream=n;let l=window.AudioContext||window.webkitAudioContext,c=new l({sampleRate:g});c.state==="suspended"&&await c.resume().catch(()=>{}),this.captureContext=c;let u=this.config?.createPlaybackEngine?await this.config.createPlaybackEngine():new m(S);if(s!==this.callGeneration){u.destroy(),n.getTracks().forEach(o=>o.stop()),c.close().catch(()=>{});return}this.playback=u,u.onFinished(()=>{s===this.callGeneration&&(this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening"))});let d=await this.takeAttachedSocket();if(s!==this.callGeneration){d?.ws.close(1e3,"client ended call");return}if(d){let{ws:o}=d;this.ws=o,this.bindCallSocket(o,s),d.live||o.send('{"type":"start"}'),this.emitStatus("listening"),this.startCapture(c,n,o,s);return}let a=new WebSocket(this.voiceSocketUrl(i,t),["runtype.bearer",e]);a.binaryType="arraybuffer",this.ws=a,a.onopen=()=>{s===this.callGeneration&&(this.emitStatus("listening"),this.startCapture(c,n,a,s))},this.bindCallSocket(a,s)}catch(n){throw this.cleanup(),this.emitError(n),this.emitStatus("error"),n}}bindCallSocket(t,e){t.onmessage=i=>this.handleMessage(i,e),t.onerror=()=>{e===this.callGeneration&&(this.emitError(new Error("Voice connection failed")),this.emitStatus("error"),this.cleanup())},t.onclose=i=>{if(this.intentionalClose){this.intentionalClose=!1;return}if(e===this.callGeneration){if(i.code!==1e3){let s=i.code?` (code ${i.code})`:"";this.emitError(new Error(`Voice connection closed${s}`)),this.emitStatus("error")}else this.emitStatus("idle");this.cleanup()}}}async stopListening(){this.cleanup(),this.emitStatus("idle")}async disconnect(){this.cleanup(),this.emitStatus("disconnected"),this.resultCallbacks=[],this.errorCallbacks=[],this.statusCallbacks=[],this.transcriptCallbacks=[],this.metricsCallbacks=[]}stopPlayback(){this.playback&&this.playback.flush(),this.isSpeaking=!1,this.ws&&this.ws.readyState===WebSocket.OPEN&&this.emitStatus("listening")}getInterruptionMode(){return"barge-in"}isBargeInActive(){return this.callLive}onLevel(t){this.levelCallbacks.push(t)}async deactivateBargeIn(){this.cleanup(),this.emitStatus("idle")}startCapture(t,e,i,s){let n=t.createMediaStreamSource(e);this.sourceNode=n;let l=t.createScriptProcessor(k,1,1);this.processor=l,l.onaudioprocess=c=>{if(s!==this.callGeneration||i.readyState!==WebSocket.OPEN)return;let u=c.inputBuffer.getChannelData(0),d=new Int16Array(u.length),a=0;for(let o=0;o<u.length;o++){let h=Math.max(-1,Math.min(1,u[o]));a+=h*h,d[o]=h<0?h*32768:h*32767}if(i.send(d.buffer),this.levelCallbacks.length>0){let o=Math.sqrt(a/u.length),h=Math.max(0,Math.min(1,o*C));for(let w of this.levelCallbacks)w(h)}},n.connect(l),l.connect(t.destination)}handleMessage(t,e){if(e!==this.callGeneration)return;if(t.data instanceof ArrayBuffer){this.handleAudioFrame(t.data,e);return}let i;try{i=JSON.parse(t.data)}catch{return}switch(i.type){case"transcript_interim":this.emitStatus("listening"),this.emitTranscript("user",i.text??"",!1);break;case"transcript_final":{let s=i.role==="assistant"?"assistant":"user";this.emitStatus(s==="user"?"processing":"speaking"),this.emitTranscript(s,i.text??"",!0);break}case"audio_end":this.playback?this.playback.markStreamEnd():(this.isSpeaking=!1,this.emitStatus("listening"));break;case"metrics":this.emitMetrics({llmMs:i.llm_ms,ttsMs:i.tts_ms,firstAudioMs:i.first_audio_ms,totalMs:i.total_ms});break;case"error":this.emitError(new Error(i.error||"Voice error")),this.emitStatus("error");break}}handleAudioFrame(t,e){if(e!==this.callGeneration||!this.playback)return;let i=V(t);i.length!==0&&(this.isSpeaking||(this.isSpeaking=!0,this.emitStatus("speaking")),this.playback.enqueue(i))}cleanup(){if(this.callGeneration+=1,this.callLive=!1,this.isSpeaking=!1,this.releaseAttached(),this.processor&&(this.processor.onaudioprocess=null,this.processor.disconnect(),this.processor=null),this.sourceNode&&(this.sourceNode.disconnect(),this.sourceNode=null),this.mediaStream&&(this.mediaStream.getTracks().forEach(t=>t.stop()),this.mediaStream=null),this.captureContext&&(this.captureContext.close().catch(()=>{}),this.captureContext=null),this.playback&&(this.playback.destroy(),this.playback=null),this.ws){this.intentionalClose=!0;try{this.ws.close(1e3,"client ended call")}catch{}this.ws=null}}onResult(t){this.resultCallbacks.push(t)}onError(t){this.errorCallbacks.push(t)}onStatusChange(t){this.statusCallbacks.push(t)}onTranscript(t){this.transcriptCallbacks.push(t)}onMetrics(t){this.metricsCallbacks.push(t)}emitStatus(t){this.statusCallbacks.forEach(e=>e(t))}emitError(t){this.errorCallbacks.forEach(e=>e(t))}emitTranscript(t,e,i){this.transcriptCallbacks.forEach(s=>s(t,e,i))}emitMetrics(t){this.metricsCallbacks.forEach(e=>e(t))}};var p=class{constructor(t={}){this.config=t;this.type="browser";this.recognition=null;this.resultCallbacks=[];this.errorCallbacks=[];this.statusCallbacks=[];this.isListening=!1;this.w=typeof window<"u"?window:void 0}async connect(){this.statusCallbacks.forEach(t=>t("connected"))}async startListening(){try{if(this.isListening)throw new Error("Already listening");if(!this.w)throw new Error("Window object not available");let t=this.w.SpeechRecognition||this.w.webkitSpeechRecognition;if(!t)throw new Error("Browser speech recognition not supported");this.recognition=new t,this.recognition.lang=this.config?.language||"en-US",this.recognition.continuous=this.config?.continuous||!1,this.recognition.interimResults=!0,this.recognition.onresult=e=>{let i=Array.from(e.results).map(n=>n[0]).map(n=>n.transcript).join(""),s=e.results[e.results.length-1].isFinal;this.resultCallbacks.forEach(n=>n({text:i,confidence:s?.8:.5,provider:"browser"})),s&&!this.config?.continuous&&this.stopListening()},this.recognition.onerror=e=>{this.errorCallbacks.forEach(i=>i(new Error(e.error))),this.statusCallbacks.forEach(i=>i("error"))},this.recognition.onstart=()=>{this.isListening=!0,this.statusCallbacks.forEach(e=>e("listening"))},this.recognition.onend=()=>{this.isListening=!1,this.statusCallbacks.forEach(e=>e("idle"))},this.recognition.start()}catch(t){throw this.errorCallbacks.forEach(e=>e(t)),this.statusCallbacks.forEach(e=>e("error")),t}}async stopListening(){this.recognition&&(this.recognition.stop(),this.recognition=null),this.isListening=!1,this.statusCallbacks.forEach(t=>t("idle"))}onResult(t){this.resultCallbacks.push(t)}onError(t){this.errorCallbacks.push(t)}onStatusChange(t){this.statusCallbacks.push(t)}async disconnect(){await this.stopListening(),this.statusCallbacks.forEach(t=>t("disconnected"))}static isSupported(){return"SpeechRecognition"in window||"webkitSpeechRecognition"in window}};function v(r){switch(r.type){case"runtype":if(!r.runtype)throw new Error("Runtype voice provider requires configuration");return new f(r.runtype);case"browser":if(!p.isSupported())throw new Error("Browser speech recognition not supported");return new p(r.browser||{});case"custom":{let t=r.custom;if(!t)throw new Error("Custom voice provider requires a `custom` provider instance or factory");let e=typeof t=="function"?t():t;if(!e||typeof e.startListening!="function")throw new Error("Custom voice provider `custom` must be a VoiceProvider (or a factory returning one)");return e}default:throw new Error(`Unknown voice provider type: ${r.type}`)}}function y(r){if(r?.type==="custom"&&r.custom)return v({type:"custom",custom:r.custom});if(r?.type==="runtype"&&r.runtype)return v({type:"runtype",runtype:r.runtype});if(p.isSupported())return v({type:"browser",browser:r?.browser||{language:"en-US"}});throw new Error("No supported voice providers available")}function L(r){try{return y(r),!0}catch{return!1}}export{p as BrowserVoiceProvider,f as RuntypeVoiceProvider,y as createBestAvailableVoiceProvider,v as createVoiceProvider,L as isVoiceSupported};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@runtypelabs/persona",
-  "version": "4.24.0",
+  "version": "4.25.0",
   "description": "Themeable, pluggable streaming agent widget for websites, in plain JS with support for voice input and reasoning / tool output.",
   "type": "module",
   "main": "dist/index.cjs",

package/src/index-core.ts

@@ -11,6 +11,8 @@ export type {
   AgentWidgetConfigPatch,
   AgentWidgetSessionInitTrigger,
   AgentWidgetSessionInitHook,
+  AgentWidgetVoicePrewarmTrigger,
+  AgentWidgetVoicePrewarmHook,
   TargetResolver,
   ResolvedTarget,
   AgentWidgetFeatureFlags,

package/src/session.ts

@@ -910,6 +910,33 @@ export class AgentWidgetSession {
     }
   }
 
+  /**
+   * Warm the voice path ahead of the mic click (`voiceRecognition.prewarm`).
+   * Session-owned providers (`runtype`, `custom`) get `provider.prewarm()` once
+   * the lazy voice runtime has installed them; the browser (Web Speech) path
+   * warms the client-token session instead. Fire-and-forget and silent.
+   */
+  public prewarmVoice(): void {
+    if (!usesSessionVoice(this.config.voiceRecognition?.provider)) {
+      this.warmClientSession();
+      return;
+    }
+    const generation = this.voiceSetupGeneration;
+    const run = () => {
+      const provider = this.voiceProvider;
+      if (!provider?.prewarm || generation !== this.voiceSetupGeneration) return;
+      if (this.voiceActive || this.isBargeInActive()) return;
+      void Promise.resolve()
+        .then(() => provider.prewarm?.())
+        .catch(() => {});
+    };
+    if (!this.voiceProvider && this.voiceSetupPromise) {
+      void this.voiceSetupPromise.then(run, () => {});
+      return;
+    }
+    run();
+  }
+
   /**
    * Cleanup voice resources
    */
@@ -975,7 +1002,9 @@ export class AgentWidgetSession {
             clientToken: providerConfig.runtype?.clientToken ?? this.config.clientToken,
             host: providerConfig.runtype?.host ?? this.config.apiUrl,
             voiceId: providerConfig.runtype?.voiceId,
-            createPlaybackEngine: providerConfig.runtype?.createPlaybackEngine
+            createPlaybackEngine: providerConfig.runtype?.createPlaybackEngine,
+            prewarmMode: providerConfig.runtype?.prewarmMode,
+            attachIdleMs: providerConfig.runtype?.attachIdleMs
           }
         };
       

package/src/session.voice.test.ts

@@ -13,7 +13,8 @@ const h = vi.hoisted(() => {
       | ((role: 'user' | 'assistant', text: string, isFinal: boolean) => void)
       | null;
     metricsCb: ((m: VoiceMetrics) => void) | null;
-  } = { transcriptCb: null, metricsCb: null };
+    prewarms: number;
+  } = { transcriptCb: null, metricsCb: null, prewarms: 0 };
 
   const fakeProvider = {
     type: 'runtype' as const,
@@ -30,6 +31,9 @@ const h = vi.hoisted(() => {
     onMetrics: (cb: typeof state.metricsCb) => {
       state.metricsCb = cb;
     },
+    prewarm: () => {
+      state.prewarms += 1;
+    },
   };
 
   return { state, fakeProvider };
@@ -212,3 +216,70 @@ describe('AgentWidgetSession - Runtype TTS config', () => {
     }
   });
 });
+
+describe('AgentWidgetSession - prewarmVoice', () => {
+  const callbacks = {
+    onMessagesChanged: () => {},
+    onStatusChanged: () => {},
+    onStreamingChanged: () => {},
+    onError: () => {},
+  };
+  const flush = () => new Promise((resolve) => setTimeout(resolve, 0));
+  const runtypeSession = () =>
+    new AgentWidgetSession(
+      {
+        apiUrl: 'http://localhost:8000',
+        voiceRecognition: {
+          enabled: true,
+          provider: { type: 'runtype', runtype: { agentId: 'a1' } },
+        },
+      },
+      callbacks,
+    );
+
+  beforeEach(() => {
+    h.state.prewarms = 0;
+  });
+
+  it("calls the provider's prewarm once setup has installed it", async () => {
+    const session = runtypeSession();
+    session.setupVoice();
+    await flush();
+    session.prewarmVoice();
+    await flush();
+    expect(h.state.prewarms).toBe(1);
+  });
+
+  it('waits for a setup still loading the voice runtime', async () => {
+    const session = runtypeSession();
+    session.setupVoice();
+    session.prewarmVoice();
+    expect(h.state.prewarms).toBe(0);
+    await flush();
+    expect(h.state.prewarms).toBe(1);
+  });
+
+  it('drops a prewarm whose setup was torn down meanwhile', async () => {
+    const session = runtypeSession();
+    session.setupVoice();
+    session.prewarmVoice();
+    session.cleanupVoice();
+    await flush();
+    expect(h.state.prewarms).toBe(0);
+  });
+
+  it('warms the client session on the browser (Web Speech) path', () => {
+    const session = new AgentWidgetSession(
+      {
+        apiUrl: 'http://localhost:8000',
+        clientToken: 'ct_test',
+        voiceRecognition: { enabled: true, provider: { type: 'browser' } },
+      },
+      callbacks,
+    );
+    const warm = vi.spyOn(session, 'warmClientSession').mockImplementation(() => {});
+    session.prewarmVoice();
+    expect(warm).toHaveBeenCalledOnce();
+    expect(h.state.prewarms).toBe(0);
+  });
+});