@craftedxp/voice-js 0.3.2 → 0.4.0

package/dist/browser.js

@@ -1,24 +1,26 @@
-"use strict";
-var __defProp = Object.defineProperty;
-var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
-var __getOwnPropNames = Object.getOwnPropertyNames;
-var __hasOwnProp = Object.prototype.hasOwnProperty;
+'use strict'
+var __defProp = Object.defineProperty
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor
+var __getOwnPropNames = Object.getOwnPropertyNames
+var __hasOwnProp = Object.prototype.hasOwnProperty
 var __export = (target, all) => {
-  for (var name in all)
-    __defProp(target, name, { get: all[name], enumerable: true });
-};
+  for (var name in all) __defProp(target, name, { get: all[name], enumerable: true })
+}
 var __copyProps = (to, from, except, desc) => {
-  if (from && typeof from === "object" || typeof from === "function") {
+  if ((from && typeof from === 'object') || typeof from === 'function') {
     for (let key of __getOwnPropNames(from))
       if (!__hasOwnProp.call(to, key) && key !== except)
-        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+        __defProp(to, key, {
+          get: () => from[key],
+          enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable,
+        })
   }
-  return to;
-};
-var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+  return to
+}
+var __toCommonJS = (mod) => __copyProps(__defProp({}, '__esModule', { value: true }), mod)
 
 // src/browser.ts
-var browser_exports = {};
+var browser_exports = {}
 __export(browser_exports, {
   buildWsUrl: () => buildWsUrl,
   configureVoiceClient: () => configureVoiceClient,
@@ -26,57 +28,64 @@ __export(browser_exports, {
   createAudioPlayback: () => createAudioPlayback,
   createProtocolState: () => createProtocolState,
   createReconnectingWebSocket: () => createReconnectingWebSocket,
-  handleServerMessage: () => handleServerMessage
-});
-module.exports = __toCommonJS(browser_exports);
+  handleServerMessage: () => handleServerMessage,
+})
+module.exports = __toCommonJS(browser_exports)
 
 // src/config.ts
 function normalizeConfig(config) {
-  if (!config) throw new Error("configureVoiceClient: config is required");
-  if ("apiKey" in config) {
+  if (!config) throw new Error('configureVoiceClient: config is required')
+  if ('apiKey' in config) {
     throw new Error(
-      "configureVoiceClient: `apiKey` is no longer supported. Embedding sk_ in JS code ships server-grade credentials to every client. Pass `fetchToken: async ({ agentId }) => { /* call YOUR backend mint */ }` instead \u2014 see the @craftedxp/voice-js README for the migration recipe."
-    );
+      'configureVoiceClient: `apiKey` is no longer supported. Embedding sk_ in JS code ships server-grade credentials to every client. Pass `fetchToken: async ({ agentId }) => { /* call YOUR backend mint */ }` instead \u2014 see the @craftedxp/voice-js README for the migration recipe.',
+    )
   }
   if (!config.apiBase) {
-    throw new Error("configureVoiceClient: apiBase is required");
+    throw new Error('configureVoiceClient: apiBase is required')
   }
-  if (typeof config.fetchToken !== "function") {
-    throw new Error("configureVoiceClient: fetchToken must be a function");
+  if (typeof config.fetchToken !== 'function') {
+    throw new Error('configureVoiceClient: fetchToken must be a function')
   }
   return {
     ...config,
-    apiBase: config.apiBase.replace(/\/+$/, "")
-  };
+    apiBase: config.apiBase.replace(/\/+$/, ''),
+  }
 }
 function mergeStartCallContext(factory, call) {
-  const context = factory.defaultContext || call.context ? { ...factory.defaultContext ?? {}, ...call.context ?? {} } : void 0;
-  const metadata = factory.defaultMetadata || call.metadata ? { ...factory.defaultMetadata ?? {}, ...call.metadata ?? {} } : void 0;
-  return { context, metadata };
+  const context =
+    factory.defaultContext || call.context
+      ? { ...(factory.defaultContext ?? {}), ...(call.context ?? {}) }
+      : void 0
+  const metadata =
+    factory.defaultMetadata || call.metadata
+      ? { ...(factory.defaultMetadata ?? {}), ...(call.metadata ?? {}) }
+      : void 0
+  return { context, metadata }
 }
 
 // src/worklets/mic-downsampler.worklet.js
-var mic_downsampler_worklet_default = "// AudioWorklet \u2014 runs off the main thread in the audio rendering graph.\n//\n// Input: Float32 samples at the AudioContext's native sampleRate (typically\n// 48000 Hz on desktop, 44100 Hz on some iOS devices).\n// Output: 16 kHz mono Int16 PCM, shipped to the main thread via\n// `port.postMessage(ArrayBuffer, [ArrayBuffer])` (transferred, not copied).\n//\n// Why AudioWorklet instead of ScriptProcessorNode: ScriptProcessorNode is\n// deprecated + main-thread-bound, so any JS jank produces audible audio\n// glitches. AudioWorklet's `process()` runs on the audio rendering thread\n// at the graph's block cadence (128 frames by default) and backpressures\n// via returning `true` / `false`.\n//\n// This file is loaded as text (see tsup.config.ts loader) and registered\n// at runtime via `audioWorklet.addModule(blobUrl)`.\n\nclass MicDownsampler extends AudioWorkletProcessor {\n  constructor() {\n    super()\n    // Target sample rate for STT. Matches Deepgram Nova-3 + the platform's\n    // server-side SAMPLE_RATE constant in AgentCallHandler.\n    this.targetRate = 16000\n    // Accumulator for the downsample. We collect incoming samples and emit\n    // an Int16 chunk when we've accumulated ~1024 target-rate samples\n    // (~64 ms at 16 kHz) \u2014 matches the mobile SDK's chunk size so both\n    // platforms have the same server-side framing.\n    this.outputFrames = 1024\n    this.acc = []\n    // Running index used for fractional resampling.\n    this.readCursor = 0\n  }\n\n  // `inputs[0][0]` = first channel of first input. 128 Float32 samples per\n  // call at the context's sampleRate. Return true = keep processing.\n  process(inputs) {\n    const input = inputs[0]\n    if (!input || input.length === 0) return true\n    const channel = input[0]\n    if (!channel || channel.length === 0) return true\n\n    const ctxRate = sampleRate // global inside AudioWorkletProcessor\n    const ratio = ctxRate / this.targetRate\n\n    // Simple linear-interp downsample. For 48000 \u2192 16000 that's 3:1, which\n    // linear handles fine for voice. Anti-alias filtering would be\n    // theoretically better but inaudible for speech.\n    for (let i = 0; i < channel.length; i++) {\n      this.acc.push(channel[i])\n    }\n\n    while (this.acc.length - this.readCursor >= ratio * this.outputFrames) {\n      const out = new Int16Array(this.outputFrames)\n      let readIdx = this.readCursor\n      for (let i = 0; i < this.outputFrames; i++) {\n        // Linear interp between floor(readIdx) and ceil(readIdx)\n        const low = Math.floor(readIdx)\n        const high = Math.min(low + 1, this.acc.length - 1)\n        const frac = readIdx - low\n        const sample = this.acc[low] * (1 - frac) + this.acc[high] * frac\n        // Clip + convert to int16\n        const clipped = Math.max(-1, Math.min(1, sample))\n        out[i] = clipped < 0 ? clipped * 0x8000 : clipped * 0x7fff\n        readIdx += ratio\n      }\n      // Transfer the ArrayBuffer (zero-copy) to the main thread.\n      this.port.postMessage(out.buffer, [out.buffer])\n      this.readCursor = readIdx\n    }\n\n    // Garbage-collect the consumed portion of `acc` every so often so it\n    // doesn't grow without bound. Leave ~one chunk of headroom.\n    if (this.readCursor > ratio * this.outputFrames) {\n      this.acc = this.acc.slice(Math.floor(this.readCursor))\n      this.readCursor -= Math.floor(this.readCursor)\n    }\n\n    return true\n  }\n}\n\nregisterProcessor('mic-downsampler', MicDownsampler)\n";
+var mic_downsampler_worklet_default =
+  "// AudioWorklet \u2014 runs off the main thread in the audio rendering graph.\n//\n// Input: Float32 samples at the AudioContext's native sampleRate (typically\n// 48000 Hz on desktop, 44100 Hz on some iOS devices).\n// Output: 16 kHz mono Int16 PCM, shipped to the main thread via\n// `port.postMessage(ArrayBuffer, [ArrayBuffer])` (transferred, not copied).\n//\n// Why AudioWorklet instead of ScriptProcessorNode: ScriptProcessorNode is\n// deprecated + main-thread-bound, so any JS jank produces audible audio\n// glitches. AudioWorklet's `process()` runs on the audio rendering thread\n// at the graph's block cadence (128 frames by default) and backpressures\n// via returning `true` / `false`.\n//\n// This file is loaded as text (see tsup.config.ts loader) and registered\n// at runtime via `audioWorklet.addModule(blobUrl)`.\n\nclass MicDownsampler extends AudioWorkletProcessor {\n  constructor() {\n    super()\n    // Target sample rate for STT. Matches Deepgram Nova-3 + the platform's\n    // server-side SAMPLE_RATE constant in AgentCallHandler.\n    this.targetRate = 16000\n    // Accumulator for the downsample. We collect incoming samples and emit\n    // an Int16 chunk when we've accumulated ~1024 target-rate samples\n    // (~64 ms at 16 kHz) \u2014 matches the mobile SDK's chunk size so both\n    // platforms have the same server-side framing.\n    this.outputFrames = 1024\n    this.acc = []\n    // Running index used for fractional resampling.\n    this.readCursor = 0\n  }\n\n  // `inputs[0][0]` = first channel of first input. 128 Float32 samples per\n  // call at the context's sampleRate. Return true = keep processing.\n  process(inputs) {\n    const input = inputs[0]\n    if (!input || input.length === 0) return true\n    const channel = input[0]\n    if (!channel || channel.length === 0) return true\n\n    const ctxRate = sampleRate // global inside AudioWorkletProcessor\n    const ratio = ctxRate / this.targetRate\n\n    // Simple linear-interp downsample. For 48000 \u2192 16000 that's 3:1, which\n    // linear handles fine for voice. Anti-alias filtering would be\n    // theoretically better but inaudible for speech.\n    for (let i = 0; i < channel.length; i++) {\n      this.acc.push(channel[i])\n    }\n\n    while (this.acc.length - this.readCursor >= ratio * this.outputFrames) {\n      const out = new Int16Array(this.outputFrames)\n      let readIdx = this.readCursor\n      for (let i = 0; i < this.outputFrames; i++) {\n        // Linear interp between floor(readIdx) and ceil(readIdx)\n        const low = Math.floor(readIdx)\n        const high = Math.min(low + 1, this.acc.length - 1)\n        const frac = readIdx - low\n        const sample = this.acc[low] * (1 - frac) + this.acc[high] * frac\n        // Clip + convert to int16\n        const clipped = Math.max(-1, Math.min(1, sample))\n        out[i] = clipped < 0 ? clipped * 0x8000 : clipped * 0x7fff\n        readIdx += ratio\n      }\n      // Transfer the ArrayBuffer (zero-copy) to the main thread.\n      this.port.postMessage(out.buffer, [out.buffer])\n      this.readCursor = readIdx\n    }\n\n    // Garbage-collect the consumed portion of `acc` every so often so it\n    // doesn't grow without bound. Leave ~one chunk of headroom.\n    if (this.readCursor > ratio * this.outputFrames) {\n      this.acc = this.acc.slice(Math.floor(this.readCursor))\n      this.readCursor -= Math.floor(this.readCursor)\n    }\n\n    return true\n  }\n}\n\nregisterProcessor('mic-downsampler', MicDownsampler)\n"
 
 // src/AudioCapture.ts
-var VOLUME_INTERVAL_MS = 100;
+var VOLUME_INTERVAL_MS = 100
 var createAudioCapture = (options) => {
-  let audioContext = null;
-  let mediaStream = null;
-  let sourceNode = null;
-  let workletNode = null;
-  let analyser = null;
-  let volumeTimer = null;
-  let muted = false;
-  let capturing = false;
+  let audioContext = null
+  let mediaStream = null
+  let sourceNode = null
+  let workletNode = null
+  let analyser = null
+  let volumeTimer = null
+  let muted = false
+  let capturing = false
   const computeRms = (buf) => {
-    let sum = 0;
-    for (let i = 0; i < buf.length; i++) sum += buf[i] * buf[i];
-    const rms = Math.sqrt(sum / buf.length);
-    return Math.min(1, rms * 1.8);
-  };
+    let sum = 0
+    for (let i = 0; i < buf.length; i++) sum += buf[i] * buf[i]
+    const rms = Math.sqrt(sum / buf.length)
+    return Math.min(1, rms * 1.8)
+  }
   const start = async () => {
-    if (capturing) return;
+    if (capturing) return
     try {
       mediaStream = await navigator.mediaDevices.getUserMedia({
         audio: {
@@ -87,657 +96,748 @@ var createAudioCapture = (options) => {
           echoCancellation: true,
           noiseSuppression: true,
           autoGainControl: true,
-          channelCount: 1
-        }
-      });
-      audioContext = new AudioContext();
-      if (audioContext.state === "suspended") await audioContext.resume();
-      const blob = new Blob([mic_downsampler_worklet_default], { type: "application/javascript" });
-      const url = URL.createObjectURL(blob);
+          channelCount: 1,
+        },
+      })
+      audioContext = new AudioContext()
+      if (audioContext.state === 'suspended') await audioContext.resume()
+      const blob = new Blob([mic_downsampler_worklet_default], { type: 'application/javascript' })
+      const url = URL.createObjectURL(blob)
       try {
-        await audioContext.audioWorklet.addModule(url);
+        await audioContext.audioWorklet.addModule(url)
       } finally {
-        URL.revokeObjectURL(url);
+        URL.revokeObjectURL(url)
       }
-      sourceNode = audioContext.createMediaStreamSource(mediaStream);
-      workletNode = new AudioWorkletNode(audioContext, "mic-downsampler");
+      sourceNode = audioContext.createMediaStreamSource(mediaStream)
+      workletNode = new AudioWorkletNode(audioContext, 'mic-downsampler')
       workletNode.port.onmessage = (event) => {
-        if (muted) return;
-        options.onChunk(event.data);
-      };
+        if (muted) return
+        options.onChunk(event.data)
+      }
       if (options.onVolume) {
-        analyser = audioContext.createAnalyser();
-        analyser.fftSize = 256;
-        sourceNode.connect(analyser);
-        const buf = new Float32Array(analyser.fftSize);
+        analyser = audioContext.createAnalyser()
+        analyser.fftSize = 256
+        sourceNode.connect(analyser)
+        const buf = new Float32Array(analyser.fftSize)
         volumeTimer = setInterval(() => {
-          if (!analyser) return;
-          analyser.getFloatTimeDomainData(buf);
-          options.onVolume?.(computeRms(buf));
-        }, VOLUME_INTERVAL_MS);
+          if (!analyser) return
+          analyser.getFloatTimeDomainData(buf)
+          options.onVolume?.(computeRms(buf))
+        }, VOLUME_INTERVAL_MS)
       }
-      sourceNode.connect(workletNode);
-      const sink = audioContext.createGain();
-      sink.gain.value = 0;
-      workletNode.connect(sink).connect(audioContext.destination);
-      capturing = true;
+      sourceNode.connect(workletNode)
+      const sink = audioContext.createGain()
+      sink.gain.value = 0
+      workletNode.connect(sink).connect(audioContext.destination)
+      capturing = true
     } catch (err) {
-      const wrapped = err instanceof Error ? err : new Error(typeof err === "string" ? err : "capture failed");
-      options.onError?.(wrapped);
-      throw wrapped;
+      const wrapped =
+        err instanceof Error ? err : new Error(typeof err === 'string' ? err : 'capture failed')
+      options.onError?.(wrapped)
+      throw wrapped
     }
-  };
+  }
   const stop = () => {
-    if (!capturing) return;
-    capturing = false;
+    if (!capturing) return
+    capturing = false
     if (volumeTimer) {
-      clearInterval(volumeTimer);
-      volumeTimer = null;
+      clearInterval(volumeTimer)
+      volumeTimer = null
     }
     try {
-      workletNode?.disconnect();
-      analyser?.disconnect();
-      sourceNode?.disconnect();
-    } catch {
-    }
-    workletNode = null;
-    analyser = null;
-    sourceNode = null;
+      workletNode?.disconnect()
+      analyser?.disconnect()
+      sourceNode?.disconnect()
+    } catch {}
+    workletNode = null
+    analyser = null
+    sourceNode = null
     if (mediaStream) {
-      for (const track of mediaStream.getTracks()) track.stop();
-      mediaStream = null;
+      for (const track of mediaStream.getTracks()) track.stop()
+      mediaStream = null
     }
-    if (audioContext && audioContext.state !== "closed") {
-      void audioContext.close().catch(() => void 0);
+    if (audioContext && audioContext.state !== 'closed') {
+      void audioContext.close().catch(() => void 0)
     }
-    audioContext = null;
-  };
+    audioContext = null
+  }
   return {
     start,
     stop,
     mute: (v) => {
-      muted = v;
+      muted = v
     },
-    isCapturing: () => capturing
-  };
-};
+    isCapturing: () => capturing,
+  }
+}
 
 // src/AudioPlayback.ts
-var DEFAULT_SAMPLE_RATE = 16e3;
-var VOLUME_INTERVAL_MS2 = 100;
+var DEFAULT_SAMPLE_RATE = 16e3
+var VOLUME_INTERVAL_MS2 = 100
 var createAudioPlayback = (options = {}) => {
-  const sampleRate = options.sampleRate ?? DEFAULT_SAMPLE_RATE;
-  let audioContext = null;
-  let gainNode = null;
-  let analyser = null;
-  let volumeTimer = null;
-  let nextStartTime = 0;
-  let scheduledNodes = [];
-  let speaking = false;
+  const sampleRate = options.sampleRate ?? DEFAULT_SAMPLE_RATE
+  let audioContext = null
+  let gainNode = null
+  let analyser = null
+  let volumeTimer = null
+  let nextStartTime = 0
+  let scheduledNodes = []
+  let speaking = false
   const ensureContext = async () => {
     if (audioContext) {
-      if (audioContext.state === "suspended") await audioContext.resume();
-      return;
+      if (audioContext.state === 'suspended') await audioContext.resume()
+      return
     }
-    audioContext = new AudioContext({ sampleRate });
-    gainNode = audioContext.createGain();
+    audioContext = new AudioContext({ sampleRate })
+    gainNode = audioContext.createGain()
     if (options.onVolume) {
-      analyser = audioContext.createAnalyser();
-      analyser.fftSize = 256;
-      gainNode.connect(analyser);
-      const buf = new Float32Array(analyser.fftSize);
+      analyser = audioContext.createAnalyser()
+      analyser.fftSize = 256
+      gainNode.connect(analyser)
+      const buf = new Float32Array(analyser.fftSize)
       volumeTimer = setInterval(() => {
-        if (!analyser) return;
-        analyser.getFloatTimeDomainData(buf);
-        let sum = 0;
-        for (let i = 0; i < buf.length; i++) sum += buf[i] * buf[i];
-        const rms = Math.sqrt(sum / buf.length);
-        options.onVolume?.(Math.min(1, rms * 1.8));
-      }, VOLUME_INTERVAL_MS2);
-    }
-    gainNode.connect(audioContext.destination);
-    nextStartTime = audioContext.currentTime;
-  };
+        if (!analyser) return
+        analyser.getFloatTimeDomainData(buf)
+        let sum = 0
+        for (let i = 0; i < buf.length; i++) sum += buf[i] * buf[i]
+        const rms = Math.sqrt(sum / buf.length)
+        options.onVolume?.(Math.min(1, rms * 1.8))
+      }, VOLUME_INTERVAL_MS2)
+    }
+    gainNode.connect(audioContext.destination)
+    nextStartTime = audioContext.currentTime
+  }
   const setSpeaking = (v) => {
-    if (v === speaking) return;
-    speaking = v;
-    options.onSpeakingChange?.(v);
-  };
+    if (v === speaking) return
+    speaking = v
+    options.onSpeakingChange?.(v)
+  }
   const pruneFinished = () => {
-    const now = audioContext?.currentTime ?? 0;
+    const now = audioContext?.currentTime ?? 0
     scheduledNodes = scheduledNodes.filter((n) => {
-      const node = n;
-      return (node._endsAt ?? 0) > now;
-    });
-    if (scheduledNodes.length === 0) setSpeaking(false);
-  };
+      const node = n
+      return (node._endsAt ?? 0) > now
+    })
+    if (scheduledNodes.length === 0) setSpeaking(false)
+  }
   const enqueue = (pcm) => {
     if (!audioContext) {
-      void ensureContext().then(() => enqueue(pcm));
-      return;
-    }
-    if (!audioContext || !gainNode) return;
-    const int16 = new Int16Array(pcm);
-    if (int16.length === 0) return;
-    const audioBuffer = audioContext.createBuffer(1, int16.length, sampleRate);
-    const float32 = audioBuffer.getChannelData(0);
+      void ensureContext().then(() => enqueue(pcm))
+      return
+    }
+    if (!audioContext || !gainNode) return
+    const int16 = new Int16Array(pcm)
+    if (int16.length === 0) return
+    const audioBuffer = audioContext.createBuffer(1, int16.length, sampleRate)
+    const float32 = audioBuffer.getChannelData(0)
     for (let i = 0; i < int16.length; i++) {
-      float32[i] = int16[i] / 32768;
-    }
-    const node = audioContext.createBufferSource();
-    node.buffer = audioBuffer;
-    node.connect(gainNode);
-    const now = audioContext.currentTime;
-    const startAt = Math.max(now, nextStartTime);
-    node.start(startAt);
-    const duration = int16.length / sampleRate;
-    node._endsAt = startAt + duration;
-    nextStartTime = startAt + duration;
-    scheduledNodes.push(node);
-    setSpeaking(true);
-    node.onended = () => pruneFinished();
-  };
+      float32[i] = int16[i] / 32768
+    }
+    const node = audioContext.createBufferSource()
+    node.buffer = audioBuffer
+    node.connect(gainNode)
+    const now = audioContext.currentTime
+    const startAt = Math.max(now, nextStartTime)
+    node.start(startAt)
+    const duration = int16.length / sampleRate
+    node._endsAt = startAt + duration
+    nextStartTime = startAt + duration
+    scheduledNodes.push(node)
+    setSpeaking(true)
+    node.onended = () => pruneFinished()
+  }
   const flush = () => {
-    if (!audioContext || !gainNode) return;
+    if (!audioContext || !gainNode) return
     for (const node of scheduledNodes) {
       try {
-        node.stop();
-      } catch {
-      }
+        node.stop()
+      } catch {}
     }
-    scheduledNodes = [];
-    gainNode.disconnect();
-    gainNode = audioContext.createGain();
+    scheduledNodes = []
+    gainNode.disconnect()
+    gainNode = audioContext.createGain()
     if (analyser) {
-      analyser.disconnect();
-      gainNode.connect(analyser);
+      analyser.disconnect()
+      gainNode.connect(analyser)
     }
-    gainNode.connect(audioContext.destination);
-    nextStartTime = audioContext.currentTime;
-    setSpeaking(false);
-  };
+    gainNode.connect(audioContext.destination)
+    nextStartTime = audioContext.currentTime
+    setSpeaking(false)
+  }
   const close = () => {
-    flush();
+    flush()
     if (volumeTimer) {
-      clearInterval(volumeTimer);
-      volumeTimer = null;
+      clearInterval(volumeTimer)
+      volumeTimer = null
     }
-    if (audioContext && audioContext.state !== "closed") {
-      void audioContext.close().catch(() => void 0);
+    if (audioContext && audioContext.state !== 'closed') {
+      void audioContext.close().catch(() => void 0)
     }
-    audioContext = null;
-    gainNode = null;
-    analyser = null;
-  };
+    audioContext = null
+    gainNode = null
+    analyser = null
+  }
   const resume = async () => {
-    await ensureContext();
-  };
-  return { enqueue, flush, close, resume };
-};
+    await ensureContext()
+  }
+  return { enqueue, flush, close, resume }
+}
 
 // src/ReconnectingWebSocket.ts
-var READYSTATE_OPEN = 1;
-var READYSTATE_CLOSED = 3;
+var READYSTATE_OPEN = 1
+var READYSTATE_CLOSED = 3
 var createReconnectingWebSocket = (options, onEvent) => {
-  const maxRetries = options.maxRetries ?? 3;
-  const initialBackoff = options.initialBackoffMs ?? 500;
-  const maxBackoff = options.maxBackoffMs ?? 8e3;
-  let ws = null;
-  let intentionalClose = false;
-  let retries = 0;
-  let backoff = initialBackoff;
-  let reconnectTimer = null;
+  const maxRetries = options.maxRetries ?? 3
+  const initialBackoff = options.initialBackoffMs ?? 500
+  const maxBackoff = options.maxBackoffMs ?? 8e3
+  let ws = null
+  let intentionalClose = false
+  let retries = 0
+  let backoff = initialBackoff
+  let reconnectTimer = null
   const openOnce = () => {
-    ws = options.wsFactory(options.url);
-    ws.binaryType = "arraybuffer";
+    ws = options.wsFactory(options.url)
+    ws.binaryType = 'arraybuffer'
     ws.onopen = () => {
-      if (retries === 0) onEvent({ type: "open" });
-      else onEvent({ type: "reconnected" });
-      retries = 0;
-      backoff = initialBackoff;
-    };
+      if (retries === 0) onEvent({ type: 'open' })
+      else onEvent({ type: 'reconnected' })
+      retries = 0
+      backoff = initialBackoff
+    }
     ws.onmessage = (ev) => {
-      onEvent({ type: "message", data: ev.data });
-    };
+      onEvent({ type: 'message', data: ev.data })
+    }
     ws.onerror = () => {
-      onEvent({ type: "error", error: new Error("WebSocket error") });
-    };
+      onEvent({ type: 'error', error: new Error('WebSocket error') })
+    }
     ws.onclose = (ev) => {
-      ws = null;
-      const shouldRetry = !intentionalClose && retries < maxRetries;
+      ws = null
+      const shouldRetry = !intentionalClose && retries < maxRetries
       if (!shouldRetry) {
         onEvent({
-          type: "close",
+          type: 'close',
           code: ev.code,
           reason: ev.reason,
-          permanent: true
-        });
-        return;
+          permanent: true,
+        })
+        return
       }
       onEvent({
-        type: "close",
+        type: 'close',
         code: ev.code,
         reason: ev.reason,
-        permanent: false
-      });
-      retries++;
-      const delay = Math.min(backoff, maxBackoff);
-      backoff = Math.min(backoff * 2, maxBackoff);
-      reconnectTimer = setTimeout(openOnce, delay);
-    };
-  };
-  openOnce();
+        permanent: false,
+      })
+      retries++
+      const delay = Math.min(backoff, maxBackoff)
+      backoff = Math.min(backoff * 2, maxBackoff)
+      reconnectTimer = setTimeout(openOnce, delay)
+    }
+  }
+  openOnce()
   return {
     send: (data) => {
-      if (ws && ws.readyState === READYSTATE_OPEN) ws.send(data);
+      if (ws && ws.readyState === READYSTATE_OPEN) ws.send(data)
     },
-    close: (code = 1e3, reason = "client-requested") => {
-      intentionalClose = true;
+    close: (code = 1e3, reason = 'client-requested') => {
+      intentionalClose = true
       if (reconnectTimer) {
-        clearTimeout(reconnectTimer);
-        reconnectTimer = null;
+        clearTimeout(reconnectTimer)
+        reconnectTimer = null
       }
       try {
-        ws?.close(code, reason);
-      } catch {
-      }
+        ws?.close(code, reason)
+      } catch {}
     },
-    readyState: () => ws?.readyState ?? READYSTATE_CLOSED
-  };
-};
+    readyState: () => ws?.readyState ?? READYSTATE_CLOSED,
+  }
+}
 
 // src/protocol.ts
 var createProtocolState = () => ({
-  state: "idle",
+  state: 'idle',
   transcript: [],
   agentBubbleId: null,
   idCounter: 0,
-  endReason: null
-});
+  endReason: null,
+})
 var mapEndReason = (raw) => {
-  if (raw === "agent_ended") return "agent_ended";
-  if (raw === "caller_hung_up") return "user_hangup";
-  if (raw === "silence_timeout" || raw === "max_duration") return "timeout";
-  return "error";
-};
+  if (raw === 'agent_ended') return 'agent_ended'
+  if (raw === 'caller_hung_up') return 'user_hangup'
+  if (raw === 'silence_timeout' || raw === 'max_duration') return 'timeout'
+  return 'error'
+}
 function handleServerMessage(raw, state, cb) {
-  let msg;
+  let msg
   try {
-    msg = JSON.parse(raw);
+    msg = JSON.parse(raw)
   } catch {
-    return;
+    return
   }
   switch (msg.type) {
-    case "connected":
-      cb.onConnected();
-      setState(state, "listening", cb);
-      return;
-    case "transcript": {
-      const text = msg.text ?? "";
-      if (!text) return;
-      const isFinal = !!msg.isFinal;
-      if (!isFinal) setState(state, "user_speaking", cb);
-      upsertUserPartial(state, text, isFinal);
-      cb.onTranscript(state.transcript);
-      return;
-    }
-    case "agent_turn_start": {
-      const id = `m${state.idCounter++}`;
-      state.agentBubbleId = id;
-      state.transcript = [...state.transcript, { id, role: "agent", text: "" }];
-      cb.onTranscript(state.transcript);
-      cb.onAgentTurnStart();
-      setState(state, "agent_speaking", cb);
-      return;
-    }
-    case "agent_text": {
-      const delta = msg.text ?? "";
-      if (!delta || !state.agentBubbleId) return;
-      const id = state.agentBubbleId;
-      state.transcript = state.transcript.map(
-        (e) => e.id === id && e.role === "agent" ? { ...e, text: e.text + delta } : e
-      );
-      cb.onTranscript(state.transcript);
-      return;
-    }
-    case "agent_turn_end":
-      state.agentBubbleId = null;
-      setState(state, "listening", cb);
-      return;
-    case "interrupt":
-      cb.onInterrupt();
-      return;
-    case "agent_turn_abort": {
-      const committed = (msg.committedText ?? "").trim();
+    case 'connected':
+      cb.onConnected()
+      setState(state, 'listening', cb)
+      return
+    case 'transcript': {
+      const text = msg.text ?? ''
+      if (!text) return
+      const isFinal = !!msg.isFinal
+      if (!isFinal) setState(state, 'user_speaking', cb)
+      upsertUserPartial(state, text, isFinal)
+      cb.onTranscript(state.transcript)
+      return
+    }
+    case 'agent_turn_start': {
+      const id = `m${state.idCounter++}`
+      state.agentBubbleId = id
+      state.transcript = [...state.transcript, { id, role: 'agent', text: '' }]
+      cb.onTranscript(state.transcript)
+      const seq = typeof msg.seq === 'number' ? msg.seq : void 0
+      cb.onAgentTurnStart(seq)
+      setState(state, 'agent_speaking', cb)
+      return
+    }
+    case 'agent_text': {
+      const delta = msg.text ?? ''
+      if (!delta || !state.agentBubbleId) return
+      const id = state.agentBubbleId
+      state.transcript = state.transcript.map((e) =>
+        e.id === id && e.role === 'agent' ? { ...e, text: e.text + delta } : e,
+      )
+      cb.onTranscript(state.transcript)
+      return
+    }
+    case 'agent_turn_end': {
+      state.agentBubbleId = null
+      const seq = typeof msg.seq === 'number' ? msg.seq : void 0
+      cb.onAgentTurnEnd(seq)
+      setState(state, 'listening', cb)
+      return
+    }
+    case 'interrupt':
+      cb.onInterrupt()
+      return
+    case 'agent_turn_abort': {
+      const committed = (msg.committedText ?? '').trim()
       if (state.agentBubbleId) {
-        const id = state.agentBubbleId;
+        const id = state.agentBubbleId
         if (committed) {
-          state.transcript = state.transcript.map(
-            (e) => e.id === id && e.role === "agent" ? { ...e, text: committed, interrupted: true } : e
-          );
+          state.transcript = state.transcript.map((e) =>
+            e.id === id && e.role === 'agent' ? { ...e, text: committed, interrupted: true } : e,
+          )
         } else {
-          state.transcript = state.transcript.filter((e) => e.id !== id);
+          state.transcript = state.transcript.filter((e) => e.id !== id)
         }
-        cb.onTranscript(state.transcript);
+        cb.onTranscript(state.transcript)
       }
-      state.agentBubbleId = null;
-      return;
+      state.agentBubbleId = null
+      return
     }
-    case "tool_call":
+    case 'tool_call':
       state.transcript = [
         ...state.transcript,
         {
           id: `m${state.idCounter++}`,
-          role: "tool",
-          text: `\u2192 ${String(msg.tool ?? "?")}(${msg.args ? JSON.stringify(msg.args) : ""})`
-        }
-      ];
-      cb.onTranscript(state.transcript);
-      return;
-    case "tool_result":
+          role: 'tool',
+          text: `\u2192 ${String(msg.tool ?? '?')}(${msg.args ? JSON.stringify(msg.args) : ''})`,
+        },
+      ]
+      cb.onTranscript(state.transcript)
+      return
+    case 'tool_result':
       state.transcript = [
         ...state.transcript,
         {
           id: `m${state.idCounter++}`,
-          role: "tool",
-          text: `${msg.ok ? "\u2713" : "\u2717"} ${String(msg.tool ?? "?")}`
-        }
-      ];
-      cb.onTranscript(state.transcript);
-      return;
-    case "client_tool_call": {
-      const toolCallId = String(msg.toolCallId ?? "");
-      const name = String(msg.name ?? "");
-      const args = msg.args ?? {};
-      if (!toolCallId || !name) return;
-      cb.onClientToolCall({ toolCallId, name, args });
-      return;
-    }
-    case "call_end": {
-      const reasonRaw = String(msg.reason ?? "");
-      const reason = mapEndReason(reasonRaw);
-      state.endReason = reason;
+          role: 'tool',
+          text: `${msg.ok ? '\u2713' : '\u2717'} ${String(msg.tool ?? '?')}`,
+        },
+      ]
+      cb.onTranscript(state.transcript)
+      return
+    case 'client_tool_call': {
+      const toolCallId = String(msg.toolCallId ?? '')
+      const name = String(msg.name ?? '')
+      const args = msg.args ?? {}
+      if (!toolCallId || !name) return
+      cb.onClientToolCall({ toolCallId, name, args })
+      return
+    }
+    case 'call_end': {
+      const reasonRaw = String(msg.reason ?? '')
+      const reason = mapEndReason(reasonRaw)
+      state.endReason = reason
       state.transcript = [
         ...state.transcript,
         {
           id: `m${state.idCounter++}`,
-          role: "system",
-          text: `call ended${reasonRaw ? ` (${reasonRaw})` : ""}`
-        }
-      ];
-      cb.onTranscript(state.transcript);
-      cb.onCallEnd(reason);
-      return;
+          role: 'system',
+          text: `call ended${reasonRaw ? ` (${reasonRaw})` : ''}`,
+        },
+      ]
+      cb.onTranscript(state.transcript)
+      cb.onCallEnd(reason)
+      return
     }
-    case "error": {
-      const code = msg.code ?? "server_error";
-      const message = msg.message ?? "server error";
-      cb.onError({ code, message });
-      return;
+    case 'error': {
+      const code = msg.code ?? 'server_error'
+      const message = msg.message ?? 'server error'
+      cb.onError({ code, message })
+      return
     }
   }
 }
 var setState = (state, next, cb) => {
-  if (state.state === next) return;
-  cb.onState(next);
-};
+  if (state.state === next) return
+  cb.onState(next)
+}
 var upsertUserPartial = (state, text, isFinal) => {
-  let idx = -1;
+  let idx = -1
   for (let i = state.transcript.length - 1; i >= 0; i--) {
-    const e = state.transcript[i];
-    if (e.role === "user" && e.committed === false) {
-      idx = i;
-      break;
+    const e = state.transcript[i]
+    if (e.role === 'user' && e.committed === false) {
+      idx = i
+      break
     }
   }
   if (idx === -1) {
     state.transcript = [
       ...state.transcript,
-      { id: `m${state.idCounter++}`, role: "user", text, committed: isFinal }
-    ];
-    return;
-  }
-  const target = state.transcript[idx];
-  const next = [...state.transcript];
-  next[idx] = { ...target, text, committed: isFinal };
-  state.transcript = next;
-};
+      { id: `m${state.idCounter++}`, role: 'user', text, committed: isFinal },
+    ]
+    return
+  }
+  const target = state.transcript[idx]
+  const next = [...state.transcript]
+  next[idx] = { ...target, text, committed: isFinal }
+  state.transcript = next
+}
 function buildWsUrl(args) {
-  const base = new URL(args.apiBase);
-  const proto = base.protocol === "https:" ? "wss:" : "ws:";
-  const bargeQS = args.bargeIn === false ? "&barge=off" : "";
-  return `${proto}//${base.host}/v1/agents/${encodeURIComponent(args.agentId)}/call?token=${encodeURIComponent(args.token)}${bargeQS}`;
+  const base = new URL(args.apiBase)
+  const proto = base.protocol === 'https:' ? 'wss:' : 'ws:'
+  const bargeQS = args.bargeIn === false ? '&barge=off' : ''
+  return `${proto}//${base.host}/v1/agents/${encodeURIComponent(args.agentId)}/call?token=${encodeURIComponent(args.token)}${bargeQS}`
 }
 
 // src/clientTools.ts
-var NAME_RE = /^[a-zA-Z_][a-zA-Z0-9_]*$/;
-var MAX_TOOLS = 64;
-var MAX_USAGE = 500;
-var MAX_TIMEOUT_MS = 3e4;
+var NAME_RE = /^[a-zA-Z_][a-zA-Z0-9_]*$/
+var MAX_TOOLS = 64
+var MAX_USAGE = 500
+var MAX_TIMEOUT_MS = 3e4
 var validateClientToolMap = (tools) => {
-  if (tools === void 0) return;
-  if (typeof tools !== "object" || tools === null || Array.isArray(tools)) {
-    throw new Error("clientTools must be an object keyed by tool name");
+  if (tools === void 0) return
+  if (typeof tools !== 'object' || tools === null || Array.isArray(tools)) {
+    throw new Error('clientTools must be an object keyed by tool name')
   }
-  const entries = Object.entries(tools);
+  const entries = Object.entries(tools)
   if (entries.length > MAX_TOOLS) {
-    throw new Error(`clientTools may declare at most 64 tools (got ${entries.length})`);
+    throw new Error(`clientTools may declare at most 64 tools (got ${entries.length})`)
   }
   for (const [name, def] of entries) {
     if (!NAME_RE.test(name)) {
       throw new Error(
-        `clientTools["${name}"]: name must be a valid identifier (^[a-zA-Z_][a-zA-Z0-9_]*$)`
-      );
+        `clientTools["${name}"]: name must be a valid identifier (^[a-zA-Z_][a-zA-Z0-9_]*$)`,
+      )
     }
-    if (!def || typeof def !== "object") {
-      throw new Error(`clientTools["${name}"]: must be an object`);
+    if (!def || typeof def !== 'object') {
+      throw new Error(`clientTools["${name}"]: must be an object`)
     }
-    if (typeof def.description !== "string" || def.description.length === 0) {
-      throw new Error(`clientTools["${name}"]: must have a description`);
+    if (typeof def.description !== 'string' || def.description.length === 0) {
+      throw new Error(`clientTools["${name}"]: must have a description`)
     }
-    if (typeof def.handler !== "function") {
-      throw new Error(`clientTools["${name}"]: must have a handler function`);
+    if (typeof def.handler !== 'function') {
+      throw new Error(`clientTools["${name}"]: must have a handler function`)
     }
     if (def.usage !== void 0 && def.usage.length > MAX_USAGE) {
-      throw new Error(`clientTools["${name}"]: usage must be \u2264500 chars`);
+      throw new Error(`clientTools["${name}"]: usage must be \u2264500 chars`)
     }
-    if (def.timeoutMs !== void 0 && (!Number.isFinite(def.timeoutMs) || def.timeoutMs <= 0 || def.timeoutMs > MAX_TIMEOUT_MS)) {
-      throw new Error(`clientTools["${name}"]: timeoutMs must be in (0, 30000]`);
+    if (
+      def.timeoutMs !== void 0 &&
+      (!Number.isFinite(def.timeoutMs) || def.timeoutMs <= 0 || def.timeoutMs > MAX_TIMEOUT_MS)
+    ) {
+      throw new Error(`clientTools["${name}"]: timeoutMs must be in (0, 30000]`)
     }
   }
-};
+}
 var buildRegisterFrame = (tools) => ({
-  type: "client_tools_register",
+  type: 'client_tools_register',
   tools: Object.entries(tools).map(([name, def]) => ({
     name,
     description: def.description,
     parameters: def.parameters,
-    ...def.usage !== void 0 ? { usage: def.usage } : {},
-    ...def.timeoutMs !== void 0 ? { timeoutMs: def.timeoutMs } : {}
-  }))
-});
+    ...(def.usage !== void 0 ? { usage: def.usage } : {}),
+    ...(def.timeoutMs !== void 0 ? { timeoutMs: def.timeoutMs } : {}),
+  })),
+})
 var dispatchClientToolCall = (send, tools, frame) => {
   const safeSend = (payload) => {
     try {
-      send(payload);
-    } catch {
-    }
-  };
-  const tool = tools[frame.name];
+      send(payload)
+    } catch {}
+  }
+  const tool = tools[frame.name]
   if (!tool) {
     safeSend({
-      type: "client_tool_result",
+      type: 'client_tool_result',
       toolCallId: frame.toolCallId,
-      error: `No handler for ${frame.name}`
-    });
-    return;
+      error: `No handler for ${frame.name}`,
+    })
+    return
   }
   void (async () => {
     try {
-      const out = await tool.handler(frame.args);
+      const out = await tool.handler(frame.args)
       safeSend({
-        type: "client_tool_result",
+        type: 'client_tool_result',
         toolCallId: frame.toolCallId,
-        result: typeof out === "string" ? out : JSON.stringify(out)
-      });
+        result: typeof out === 'string' ? out : JSON.stringify(out),
+      })
     } catch (err) {
       safeSend({
-        type: "client_tool_result",
+        type: 'client_tool_result',
         toolCallId: frame.toolCallId,
-        error: err instanceof Error ? err.message : String(err)
-      });
+        error: err instanceof Error ? err.message : String(err),
+      })
     }
-  })();
-};
+  })()
+}
+
+// src/ClientMarksBuffer.ts
+var createClientMarksBuffer = (args) => {
+  const now = args.now ?? (() => performance.now())
+  let pendingFirstOutboundAt = null
+  const inFlight = /* @__PURE__ */ new Map()
+  const tryEmit = (seq) => {
+    const slot = inFlight.get(seq)
+    if (!slot) return
+    if (!slot.ended) return
+    const marks = {}
+    if (slot.firstOutboundAt !== null && slot.firstAudibleAt !== null) {
+      marks.client_mic_to_first_audible_ms = slot.firstAudibleAt - slot.firstOutboundAt
+    }
+    args.send({
+      type: 'client_marks',
+      seq,
+      marks,
+      clientNow: Date.now(),
+    })
+    inFlight.delete(seq)
+  }
+  const markFirstOutboundAudio = () => {
+    if (pendingFirstOutboundAt !== null) return
+    pendingFirstOutboundAt = now()
+  }
+  const markFirstAudibleOutput = () => {
+    let target
+    for (const slot of inFlight.values()) {
+      if (!slot.ended) {
+        target = slot
+      }
+    }
+    if (!target) return
+    if (target.firstAudibleAt !== null) return
+    target.firstAudibleAt = now()
+  }
+  const onAgentTurnStart = (seq) => {
+    inFlight.set(seq, {
+      firstOutboundAt: pendingFirstOutboundAt,
+      firstAudibleAt: null,
+      ended: false,
+    })
+    pendingFirstOutboundAt = null
+  }
+  const onAgentTurnEnd = (seq) => {
+    const slot = inFlight.get(seq)
+    if (!slot) {
+      args.send({ type: 'client_marks', seq, marks: {}, clientNow: Date.now() })
+      return
+    }
+    slot.ended = true
+    tryEmit(seq)
+  }
+  const flush = () => {
+    for (const seq of [...inFlight.keys()]) {
+      const slot = inFlight.get(seq)
+      slot.ended = true
+      tryEmit(seq)
+    }
+    pendingFirstOutboundAt = null
+  }
+  return {
+    markFirstOutboundAudio,
+    markFirstAudibleOutput,
+    onAgentTurnStart,
+    onAgentTurnEnd,
+    flush,
+  }
+}
 
 // src/VoiceClient.ts
 var BrowserVoiceClient = class {
   constructor(args) {
-    this.rws = null;
-    this.capture = null;
-    this.playback = null;
-    this.muted = false;
-    this.inputVolume = 0;
-    this.outputVolume = 0;
-    this.startedAt = null;
-    this.endedFired = false;
-    this.lastError = null;
+    this.rws = null
+    this.capture = null
+    this.playback = null
+    this.muted = false
+    this.inputVolume = 0
+    this.outputVolume = 0
+    this.startedAt = null
+    this.endedFired = false
+    this.lastError = null
     this.end = () => {
-      this.teardown("user_hangup");
-    };
+      this.teardown('user_hangup')
+    }
     this.mute = () => {
-      if (this.muted) return;
-      this.muted = true;
-      this.capture?.mute(true);
-    };
+      if (this.muted) return
+      this.muted = true
+      this.capture?.mute(true)
+    }
     this.unmute = () => {
-      if (!this.muted) return;
-      this.muted = false;
-      this.capture?.mute(false);
-    };
+      if (!this.muted) return
+      this.muted = false
+      this.capture?.mute(false)
+    }
     // ---------------------------------------------------------------
     // Internal
     // ---------------------------------------------------------------
     this.sendClientToolsRegister = () => {
-      const frame = buildRegisterFrame(this.args.options.clientTools ?? {});
-      this.rws?.send(JSON.stringify(frame));
-    };
+      const frame = buildRegisterFrame(this.args.options.clientTools ?? {})
+      this.rws?.send(JSON.stringify(frame))
+    }
     this.setState = (next) => {
-      if (this.proto.state === next) return;
-      this.proto.state = next;
-      this.args.options.onStateChange?.(next);
-    };
+      if (this.proto.state === next) return
+      this.proto.state = next
+      this.args.options.onStateChange?.(next)
+    }
     this.emitError = (err) => {
-      this.lastError = err;
-      this.args.options.onError?.(err);
-    };
+      this.lastError = err
+      this.args.options.onError?.(err)
+    }
     this.handleSocketEvent = (ev) => {
       switch (ev.type) {
-        case "open":
-          void this.startCapture();
-          break;
-        case "reconnected":
-          this.proto.transcript = [];
-          this.proto.agentBubbleId = null;
-          this.args.options.onTranscript?.(this.proto.transcript);
-          void this.startCapture();
-          this.setState("listening");
-          break;
-        case "message":
-          if (typeof ev.data === "string") {
+        case 'open':
+          void this.startCapture()
+          break
+        case 'reconnected':
+          this.proto.transcript = []
+          this.proto.agentBubbleId = null
+          this.args.options.onTranscript?.(this.proto.transcript)
+          void this.startCapture()
+          this.setState('listening')
+          break
+        case 'message':
+          if (typeof ev.data === 'string') {
             handleServerMessage(ev.data, this.proto, {
               onState: this.setState,
               onTranscript: (entries) => this.args.options.onTranscript?.(entries),
               onError: this.emitError,
               onInterrupt: () => {
-                this.playback?.flush();
-                this.args.options.onInterrupt?.();
+                this.playback?.flush()
+                this.args.options.onInterrupt?.()
+              },
+              onAgentTurnStart: (seq) => {
+                if (typeof seq === 'number') this.marks.onAgentTurnStart(seq)
+                this.args.options.onAgentTurnStart?.()
+              },
+              onAgentTurnEnd: (seq) => {
+                if (typeof seq === 'number') this.marks.onAgentTurnEnd(seq)
               },
-              onAgentTurnStart: () => this.args.options.onAgentTurnStart?.(),
               onCallEnd: (reason) => this.teardown(reason),
               onConnected: () => this.sendClientToolsRegister(),
-              onClientToolCall: (frame) => dispatchClientToolCall(
-                (f) => this.rws?.send(JSON.stringify(f)),
-                this.args.options.clientTools ?? {},
-                frame
-              )
-            });
+              onClientToolCall: (frame) =>
+                dispatchClientToolCall(
+                  (f) => this.rws?.send(JSON.stringify(f)),
+                  this.args.options.clientTools ?? {},
+                  frame,
+                ),
+            })
           } else {
-            this.playback?.enqueue(ev.data);
+            this.marks.markFirstAudibleOutput()
+            this.playback?.enqueue(ev.data)
           }
-          break;
-        case "close":
+          break
+        case 'close':
           if (ev.permanent) {
-            const reason = this.proto.endReason ?? (this.lastError ? "error" : "user_hangup");
-            this.teardown(reason);
+            const reason = this.proto.endReason ?? (this.lastError ? 'error' : 'user_hangup')
+            this.teardown(reason)
           }
-          break;
-        case "error":
-          this.emitError({ code: "socket_error", message: ev.error.message });
-          break;
+          break
+        case 'error':
+          this.emitError({ code: 'socket_error', message: ev.error.message })
+          break
       }
-    };
+    }
     this.startCapture = async () => {
-      if (this.capture?.isCapturing()) return;
+      if (this.capture?.isCapturing()) return
       this.capture = createAudioCapture({
         onChunk: (pcm) => {
-          this.rws?.send(pcm);
+          this.marks.markFirstOutboundAudio()
+          this.rws?.send(pcm)
         },
         onVolume: (v) => {
-          this.inputVolume = v;
-          this.args.options.onVolume?.({ input: v, output: this.outputVolume });
+          this.inputVolume = v
+          this.args.options.onVolume?.({ input: v, output: this.outputVolume })
         },
         onError: (err) => {
           this.emitError({
-            code: err.name === "NotAllowedError" ? "mic_denied" : "mic_start_failed",
-            message: err.message
-          });
-        }
-      });
-      if (this.muted) this.capture.mute(true);
+            code: err.name === 'NotAllowedError' ? 'mic_denied' : 'mic_start_failed',
+            message: err.message,
+          })
+        },
+      })
+      if (this.muted) this.capture.mute(true)
       try {
-        await this.capture.start();
-      } catch {
-      }
-    };
+        await this.capture.start()
+      } catch {}
+    }
     this.teardown = (reason) => {
-      this.capture?.stop();
-      this.capture = null;
-      this.playback?.close();
-      this.playback = null;
       try {
-        this.rws?.close(1e3, reason);
-      } catch {
-      }
-      this.rws = null;
-      this.setState("ended");
-      this.fireEndOnce(reason);
-    };
+        this.marks.flush()
+      } catch {}
+      this.capture?.stop()
+      this.capture = null
+      this.playback?.close()
+      this.playback = null
+      try {
+        this.rws?.close(1e3, reason)
+      } catch {}
+      this.rws = null
+      this.setState('ended')
+      this.fireEndOnce(reason)
+    }
     this.fireEndOnce = (reason) => {
-      if (this.endedFired) return;
-      this.endedFired = true;
-      const startedAt = this.startedAt ?? Date.now();
+      if (this.endedFired) return
+      this.endedFired = true
+      const startedAt = this.startedAt ?? Date.now()
       this.args.options.onEnd?.({
         reason,
-        errorCode: reason === "error" ? this.lastError?.code : void 0,
-        durationMs: Date.now() - startedAt
-      });
-    };
-    this.args = args;
-    this.proto = createProtocolState();
-    validateClientToolMap(args.options.clientTools);
+        errorCode: reason === 'error' ? this.lastError?.code : void 0,
+        durationMs: Date.now() - startedAt,
+      })
+    }
+    this.args = args
+    this.proto = createProtocolState()
+    validateClientToolMap(args.options.clientTools)
+    this.marks = createClientMarksBuffer({
+      send: (frame) => {
+        try {
+          this.rws?.send(JSON.stringify(frame))
+        } catch {}
+      },
+    })
   }
   // ---------------------------------------------------------------
   // Call interface
   // ---------------------------------------------------------------
   get state() {
-    return this.proto.state;
+    return this.proto.state
   }
   get transcript() {
-    return this.proto.transcript.slice();
+    return this.proto.transcript.slice()
   }
   get isMuted() {
-    return this.muted;
+    return this.muted
   }
   // ---------------------------------------------------------------
   // Lifecycle — called by the factory immediately after construction.
@@ -745,84 +845,262 @@ var BrowserVoiceClient = class {
   // failures arrive via `onError`.
   // ---------------------------------------------------------------
   async start() {
-    this.setState("connecting");
-    this.startedAt = Date.now();
+    this.setState('connecting')
+    this.startedAt = Date.now()
     const url = buildWsUrl({
       apiBase: this.args.config.apiBase,
       agentId: this.args.options.agentId,
       token: this.args.token,
-      bargeIn: this.args.options.bargeIn
-    });
+      bargeIn: this.args.options.bargeIn,
+    })
     this.playback = createAudioPlayback({
       onVolume: (v) => {
-        this.outputVolume = v;
-        this.args.options.onVolume?.({ input: this.inputVolume, output: v });
-      }
-    });
+        this.outputVolume = v
+        this.args.options.onVolume?.({ input: this.inputVolume, output: v })
+      },
+    })
     try {
-      await this.playback.resume();
-    } catch {
-    }
+      await this.playback.resume()
+    } catch {}
     this.rws = createReconnectingWebSocket(
       {
         url,
         wsFactory: this.args.wsFactory,
-        maxRetries: 3
+        maxRetries: 3,
       },
-      (ev) => this.handleSocketEvent(ev)
-    );
+      (ev) => this.handleSocketEvent(ev),
+    )
+  }
+}
+
+// src/webrtc/createWebRtcCall.ts
+async function createWebRtcCall(opts) {
+  const proto = createProtocolState()
+  let muted = false
+  let ended = false
+  const fireState = (next) => {
+    if (proto.state === next) return
+    proto.state = next
+    opts.onStateChange?.(next)
+  }
+  const dispatch = (raw) => {
+    handleServerMessage(raw, proto, {
+      onState: fireState,
+      onTranscript: (entries) => opts.onTranscript?.(entries),
+      onError: (err) => opts.onError?.(err),
+      onInterrupt: () => opts.onInterrupt?.(),
+      onAgentTurnStart: () => opts.onAgentTurnStart?.(),
+      onAgentTurnEnd: () => {},
+      onCallEnd: () => teardown(),
+      onConnected: () => {},
+      onClientToolCall: () => {},
+    })
   }
-};
+  fireState('connecting')
+  const pc = new RTCPeerConnection({
+    iceServers: [{ urls: 'stun:stun.l.google.com:19302' }],
+  })
+  const audioEl = document.createElement('audio')
+  audioEl.autoplay = true
+  audioEl.style.display = 'none'
+  document.body.appendChild(audioEl)
+  pc.ontrack = (event) => {
+    audioEl.srcObject = event.streams[0] ?? new MediaStream([event.track])
+  }
+  let mic
+  try {
+    mic = await navigator.mediaDevices.getUserMedia({ audio: true })
+  } catch (err) {
+    const code =
+      err instanceof DOMException && err.name === 'NotAllowedError'
+        ? 'mic_denied'
+        : 'mic_start_failed'
+    opts.onError?.({
+      code,
+      message: err instanceof Error ? err.message : 'getUserMedia failed',
+    })
+    fireState('error')
+    pc.close()
+    audioEl.remove()
+    throw err
+  }
+  for (const track of mic.getAudioTracks()) pc.addTrack(track, mic)
+  const dc = pc.createDataChannel('control', { ordered: true })
+  dc.onmessage = (e) => {
+    if (typeof e.data === 'string') dispatch(e.data)
+  }
+  dc.onerror = () => {
+    opts.onError?.({ code: 'socket_error', message: 'control channel error' })
+  }
+  const gateway = opts.webrtcGatewayBase || ''
+  const offerUrl = gateway
+    ? `${gateway}/webrtc/offer?token=${encodeURIComponent(opts.token)}`
+    : `${opts.apiBase}/v1/agents/${encodeURIComponent(opts.agentId)}/webrtc/offer?token=${encodeURIComponent(opts.token)}`
+  const iceUrl = gateway
+    ? `${gateway}/webrtc/ice?token=${encodeURIComponent(opts.token)}`
+    : `${opts.apiBase}/v1/agents/${encodeURIComponent(opts.agentId)}/webrtc/ice?token=${encodeURIComponent(opts.token)}`
+  await pc.setLocalDescription(await pc.createOffer())
+  let callId
+  try {
+    const offerRes = await fetch(offerUrl, {
+      method: 'POST',
+      headers: { 'content-type': 'application/json' },
+      body: JSON.stringify({ sdp: pc.localDescription.sdp, type: 'offer', agentId: opts.agentId }),
+    })
+    if (!offerRes.ok) {
+      const code = offerRes.status === 401 ? 'unauthorized' : 'server_error'
+      opts.onError?.({ code, message: `signaling failed: HTTP ${offerRes.status}` })
+      fireState('error')
+      mic.getTracks().forEach((t) => t.stop())
+      pc.close()
+      audioEl.remove()
+      throw new Error(`webrtc offer failed: ${offerRes.status}`)
+    }
+    const body = await offerRes.json()
+    callId = body.callId
+    await pc.setRemoteDescription({ type: 'answer', sdp: body.sdp })
+  } catch (err) {
+    if (!ended) {
+      opts.onError?.({
+        code: 'network_unreachable',
+        message: err instanceof Error ? err.message : 'signaling failed',
+      })
+      fireState('error')
+      mic.getTracks().forEach((t) => t.stop())
+      pc.close()
+      audioEl.remove()
+    }
+    throw err
+  }
+  pc.onicecandidate = (e) => {
+    if (!e.candidate) return
+    void fetch(iceUrl, {
+      method: 'POST',
+      headers: { 'content-type': 'application/json' },
+      body: JSON.stringify({ callId, candidate: e.candidate }),
+    }).catch(() => {})
+  }
+  pc.onconnectionstatechange = () => {
+    const s = pc.connectionState
+    if (s === 'connected') fireState('listening')
+    if (s === 'failed' || s === 'disconnected') {
+      opts.onError?.({ code: 'socket_error', message: `webrtc connection ${s}` })
+      teardown()
+    }
+    if (s === 'closed' && !ended) teardown()
+  }
+  const teardown = () => {
+    if (ended) return
+    ended = true
+    try {
+      mic.getTracks().forEach((t) => t.stop())
+    } catch {}
+    try {
+      pc.close()
+    } catch {}
+    try {
+      audioEl.remove()
+    } catch {}
+    fireState('ended')
+    opts.onEnd?.()
+  }
+  return {
+    get state() {
+      return proto.state
+    },
+    get transcript() {
+      return proto.transcript.slice()
+    },
+    get isMuted() {
+      return muted
+    },
+    end: () => teardown(),
+    mute: () => {
+      if (muted) return
+      muted = true
+      mic.getAudioTracks().forEach((t) => (t.enabled = false))
+    },
+    unmute: () => {
+      if (!muted) return
+      muted = false
+      mic.getAudioTracks().forEach((t) => (t.enabled = true))
+    },
+  }
+}
 
 // src/browser.ts
-var browserWsFactory = (url) => new globalThis.WebSocket(url);
+var browserWsFactory = (url) => new globalThis.WebSocket(url)
 var BrowserVoiceFactory = class {
   constructor(config) {
     this.startCall = async (options) => {
       if (!options.agentId) {
-        throw new Error("startCall: agentId is required");
+        throw new Error('startCall: agentId is required')
       }
-      const { context, metadata } = mergeStartCallContext(this.config, options);
+      const { context, metadata } = mergeStartCallContext(this.config, options)
       const fetchArgs = {
         agentId: options.agentId,
         userId: options.userId,
         context,
-        metadata
-      };
-      let token;
+        metadata,
+      }
+      let resolved
       if (options.token) {
-        token = options.token;
+        resolved = { token: options.token, transport: 'ws' }
       } else {
-        token = await this.config.fetchToken(fetchArgs);
-        if (!token) {
-          throw new Error("configureVoiceClient.fetchToken returned empty token");
+        const r = await this.config.fetchToken(fetchArgs)
+        if (!r) {
+          throw new Error('configureVoiceClient.fetchToken returned empty token')
         }
+        resolved = typeof r === 'string' ? { token: r, transport: 'ws' } : r
+        if (!resolved.token) {
+          throw new Error('configureVoiceClient.fetchToken returned an object without `token`')
+        }
+      }
+      if (resolved.transport === 'webrtc') {
+        return createWebRtcCall({
+          agentId: options.agentId,
+          apiBase: this.config.apiBase,
+          token: resolved.token,
+          webrtcGatewayBase: resolved.webrtcGatewayBase,
+          onStateChange: options.onStateChange,
+          onTranscript: options.onTranscript,
+          onError: options.onError,
+          // Synthesise a minimal CallEndEvent. WebRTC doesn't carry an end reason
+          // from the server yet — use 'agent_ended' as placeholder. durationMs is
+          // tracked at 0 until the followup lands (see spec Followups section).
+          onEnd: options.onEnd
+            ? () => options.onEnd({ reason: 'agent_ended', durationMs: 0 })
+            : void 0,
+          onInterrupt: options.onInterrupt,
+          onAgentTurnStart: options.onAgentTurnStart,
+        })
       }
       const client = new BrowserVoiceClient({
         config: this.config,
         // Carry merged context/metadata through to startCall so server can
         // see what the SDK saw.
         options: { ...options, context, metadata },
-        token,
-        wsFactory: browserWsFactory
-      });
-      await client.start();
-      return client;
-    };
-    this.config = config;
-  }
-};
+        token: resolved.token,
+        wsFactory: browserWsFactory,
+      })
+      await client.start()
+      return client
+    }
+    this.config = config
+  }
+}
 function configureVoiceClient(config) {
-  return new BrowserVoiceFactory(normalizeConfig(config));
+  return new BrowserVoiceFactory(normalizeConfig(config))
 }
 // Annotate the CommonJS export names for ESM import in node:
-0 && (module.exports = {
-  buildWsUrl,
-  configureVoiceClient,
-  createAudioCapture,
-  createAudioPlayback,
-  createProtocolState,
-  createReconnectingWebSocket,
-  handleServerMessage
-});
-//# sourceMappingURL=browser.js.map
+0 &&
+  (module.exports = {
+    buildWsUrl,
+    configureVoiceClient,
+    createAudioCapture,
+    createAudioPlayback,
+    createProtocolState,
+    createReconnectingWebSocket,
+    handleServerMessage,
+  })
+//# sourceMappingURL=browser.js.map