@mulmoclaude/core 0.12.1 → 0.13.0

package/dist/whisper/client.cjs

@@ -80,12 +80,13 @@ function blobToDataUrl(blob) {
 function toMessage(err) {
 	return err instanceof Error ? err.message : String(err);
 }
-function createVoiceCapture(transport, language, callbacks) {
+function createCaptureState(onState) {
 	let available = false;
 	let listening = false;
 	let transcribing = false;
+	let pending = 0;
 	function emit() {
-		callbacks.onState?.({
+		onState?.({
 			available,
 			listening,
 			transcribing
@@ -103,23 +104,6 @@ function createVoiceCapture(transport, language, callbacks) {
 			emit();
 		}
 	}
-	let stream = null;
-	let audioCtx = null;
-	let analyser = null;
-	let vadBuffer = /* @__PURE__ */ new Float32Array(0);
-	let monitorHandle = null;
-	let recorder = null;
-	let chunks = [];
-	let mimeType = "";
-	let segmentHasSpeech = false;
-	let silenceStart = null;
-	let segmentStart = 0;
-	let pending = 0;
-	let queue = Promise.resolve();
-	let availabilityPollHandle = null;
-	let generation = 0;
-	let segmentGeneration = 0;
-	let startInFlight = false;
 	function setPending(delta) {
 		pending += delta;
 		const next = pending > 0;
@@ -128,145 +112,252 @@ function createVoiceCapture(transport, language, callbacks) {
 			emit();
 		}
 	}
-	function stopAvailabilityPoll() {
+	return {
+		setAvailable,
+		setListening,
+		setPending,
+		isListening: () => listening
+	};
+}
+function createAvailabilityPoller(transport, setAvailable) {
+	let availabilityPollHandle = null;
+	function stop() {
 		if (availabilityPollHandle !== null) {
 			window.clearInterval(availabilityPollHandle);
 			availabilityPollHandle = null;
 		}
 	}
-	async function refreshAvailability() {
+	async function refresh() {
 		let status;
 		try {
 			status = await transport.getStatus();
 		} catch {
 			setAvailable(false);
-			stopAvailabilityPoll();
+			stop();
 			return;
 		}
 		setAvailable(status.ready);
 		if (status.downloading) {
 			if (availabilityPollHandle === null) availabilityPollHandle = window.setInterval(() => {
-				refreshAvailability();
+				refresh();
 			}, AVAILABILITY_POLL_MS);
-		} else stopAvailabilityPoll();
+		} else stop();
 	}
+	return {
+		refresh,
+		stop
+	};
+}
+function createSegmentQueue(options) {
+	const { transport, language, callbacks, setPending, getGeneration } = options;
+	let queue = Promise.resolve();
 	async function sendSegment(blob, gen) {
-		if (gen !== generation) return;
+		if (gen !== getGeneration()) return;
 		try {
 			const dataUrl = await blobToDataUrl(blob);
 			const result = await transport.transcribe(dataUrl, language());
-			if (gen !== generation) return;
+			if (gen !== getGeneration()) return;
 			const text = result.text.trim();
 			if (text.length === 0) callbacks.onEmpty?.();
 			else callbacks.onTranscript(text);
 		} catch (err) {
-			if (gen === generation) callbacks.onError?.(toMessage(err));
+			if (gen === getGeneration()) callbacks.onError?.(toMessage(err));
 		}
 	}
 	function enqueue(blob, gen) {
 		setPending(1);
 		queue = queue.then(() => sendSegment(blob, gen)).catch(() => void 0).finally(() => setPending(-1));
 	}
-	function onSegmentStop() {
+	return { enqueue };
+}
+function createAudioGraph() {
+	let stream = null;
+	let audioCtx = null;
+	let analyser = null;
+	let vadBuffer = /* @__PURE__ */ new Float32Array(0);
+	function attach(nextStream) {
+		stream = nextStream;
+		audioCtx = new AudioContext();
+		analyser = audioCtx.createAnalyser();
+		analyser.fftSize = 2048;
+		vadBuffer = new Float32Array(analyser.fftSize);
+		audioCtx.createMediaStreamSource(nextStream).connect(analyser);
+	}
+	function sample() {
+		analyser?.getFloatTimeDomainData(vadBuffer);
+		return vadBuffer;
+	}
+	function teardown() {
+		if (audioCtx) {
+			audioCtx.close().catch(() => void 0);
+			audioCtx = null;
+		}
+		analyser = null;
+		stream?.getTracks().forEach((track) => track.stop());
+		stream = null;
+	}
+	return {
+		isAttached: () => analyser !== null,
+		sample,
+		getStream: () => stream,
+		attach,
+		teardown
+	};
+}
+async function bootstrapCapture(callbacks, generationAtStart, generationNow) {
+	const mimeType = pickRecorderMime() ?? "";
+	if (!mimeType || !navigator.mediaDevices?.getUserMedia) {
+		callbacks.onError?.("unsupported");
+		return null;
+	}
+	let acquired;
+	try {
+		acquired = await navigator.mediaDevices.getUserMedia({ audio: true });
+	} catch {
+		callbacks.onError?.("permission-denied");
+		return null;
+	}
+	if (generationAtStart !== generationNow()) {
+		acquired.getTracks().forEach((track) => track.stop());
+		return null;
+	}
+	const audioGraph = createAudioGraph();
+	try {
+		audioGraph.attach(acquired);
+	} catch (err) {
+		audioGraph.teardown();
+		throw err;
+	}
+	return {
+		audioGraph,
+		mimeType
+	};
+}
+function createRecorderSession(deps) {
+	let recorder = null;
+	let chunks = [];
+	let segmentHasSpeech = false;
+	let sessionGen = 0;
+	let sessionMimeType = "";
+	function onStop() {
 		const hadSpeech = segmentHasSpeech;
-		const gen = segmentGeneration;
-		const blob = new Blob(chunks, { type: containerTypeFromMime(mimeType) });
-		if (listening) startRecorder();
-		if (hadSpeech && blob.size > 0 && gen === generation) enqueue(blob, gen);
+		const gen = sessionGen;
+		const blob = new Blob(chunks, { type: containerTypeFromMime(sessionMimeType) });
+		deps.onSegmentEnd(blob, hadSpeech, gen);
 	}
-	function startRecorder() {
-		if (!stream) return;
+	function startNext(stream, mimeType, gen) {
 		chunks = [];
 		segmentHasSpeech = false;
-		silenceStart = null;
-		segmentStart = Date.now();
-		segmentGeneration = generation;
+		sessionGen = gen;
+		sessionMimeType = mimeType;
 		recorder = new MediaRecorder(stream, { mimeType });
 		recorder.ondataavailable = (event) => {
 			if (event.data.size > 0) chunks.push(event.data);
 		};
-		recorder.onstop = onSegmentStop;
+		recorder.onstop = onStop;
 		recorder.start();
 	}
-	function cutSegment() {
-		if (recorder && recorder.state === "recording") recorder.stop();
+	return {
+		isRecording: () => recorder?.state === "recording",
+		hadSpeech: () => segmentHasSpeech,
+		markSpeech: () => {
+			segmentHasSpeech = true;
+		},
+		startNext,
+		cut: () => {
+			if (recorder?.state === "recording") recorder.stop();
+		}
+	};
+}
+var INITIAL_RUNTIME = {
+	audioGraph: null,
+	monitorHandle: null,
+	mimeType: "",
+	silenceStart: null,
+	segmentStart: 0,
+	generation: 0,
+	startInFlight: false
+};
+function stopCapture(runtime, state, session) {
+	runtime.generation += 1;
+	state.setListening(false);
+	if (runtime.monitorHandle !== null) {
+		window.clearInterval(runtime.monitorHandle);
+		runtime.monitorHandle = null;
 	}
-	function monitorTick() {
-		if (!analyser) return;
-		analyser.getFloatTimeDomainData(vadBuffer);
-		const rms = computeRms(vadBuffer);
-		const { hasSpeech, silenceStart: nextSilence, cut } = evaluateVad({
-			hasSpeech: segmentHasSpeech,
-			silenceStart
-		}, segmentStart, rms, Date.now(), VAD_CONFIG);
-		segmentHasSpeech = hasSpeech;
-		silenceStart = nextSilence;
-		if (cut) cutSegment();
+	session.cut();
+	runtime.audioGraph?.teardown();
+	runtime.audioGraph = null;
+}
+function monitorTick(runtime, session) {
+	if (!runtime.audioGraph?.isAttached()) return;
+	const rms = computeRms(runtime.audioGraph.sample());
+	const { hasSpeech, silenceStart: nextSilence, cut } = evaluateVad({
+		hasSpeech: session.hadSpeech(),
+		silenceStart: runtime.silenceStart
+	}, runtime.segmentStart, rms, Date.now(), VAD_CONFIG);
+	if (hasSpeech) session.markSpeech();
+	runtime.silenceStart = nextSilence;
+	if (cut) session.cut();
+}
+function createVoiceCapture(transport, language, callbacks) {
+	const state = createCaptureState(callbacks.onState);
+	const poller = createAvailabilityPoller(transport, state.setAvailable);
+	const runtime = { ...INITIAL_RUNTIME };
+	const segments = createSegmentQueue({
+		transport,
+		language,
+		callbacks,
+		setPending: state.setPending,
+		getGeneration: () => runtime.generation
+	});
+	const session = createRecorderSession({ onSegmentEnd: handleSegmentEnd });
+	function handleSegmentEnd(blob, hadSpeech, gen) {
+		if (state.isListening()) startRecorder();
+		if (hadSpeech && blob.size > 0 && gen === runtime.generation) segments.enqueue(blob, gen);
+	}
+	function startRecorder() {
+		const stream = runtime.audioGraph?.getStream();
+		if (!stream) return;
+		runtime.silenceStart = null;
+		runtime.segmentStart = Date.now();
+		session.startNext(stream, runtime.mimeType, runtime.generation);
 	}
 	async function start() {
-		if (startInFlight || listening) return true;
-		startInFlight = true;
-		const startGen = generation;
+		if (runtime.startInFlight || state.isListening()) return true;
+		runtime.startInFlight = true;
+		const startGen = runtime.generation;
 		try {
-			mimeType = pickRecorderMime() ?? "";
-			if (!mimeType || !navigator.mediaDevices?.getUserMedia) {
-				callbacks.onError?.("unsupported");
-				return false;
-			}
-			let acquired;
-			try {
-				acquired = await navigator.mediaDevices.getUserMedia({ audio: true });
-			} catch {
-				callbacks.onError?.("permission-denied");
-				return false;
-			}
-			if (startGen !== generation) {
-				acquired.getTracks().forEach((track) => track.stop());
-				return false;
-			}
-			stream = acquired;
-			audioCtx = new AudioContext();
-			analyser = audioCtx.createAnalyser();
-			analyser.fftSize = 2048;
-			vadBuffer = new Float32Array(analyser.fftSize);
-			audioCtx.createMediaStreamSource(stream).connect(analyser);
-			setListening(true);
+			const capture = await bootstrapCapture(callbacks, startGen, () => runtime.generation);
+			if (!capture) return false;
+			runtime.audioGraph = capture.audioGraph;
+			runtime.mimeType = capture.mimeType;
+			state.setListening(true);
 			startRecorder();
-			monitorHandle = window.setInterval(monitorTick, MONITOR_INTERVAL_MS);
+			runtime.monitorHandle = window.setInterval(() => monitorTick(runtime, session), MONITOR_INTERVAL_MS);
 			return true;
 		} finally {
-			startInFlight = false;
+			runtime.startInFlight = false;
 		}
 	}
 	function stop() {
-		generation += 1;
-		setListening(false);
-		if (monitorHandle !== null) {
-			window.clearInterval(monitorHandle);
-			monitorHandle = null;
-		}
-		if (recorder && recorder.state === "recording") recorder.stop();
-		recorder = null;
-		if (audioCtx) {
-			audioCtx.close().catch(() => void 0);
-			audioCtx = null;
-		}
-		analyser = null;
-		stream?.getTracks().forEach((track) => track.stop());
-		stream = null;
+		stopCapture(runtime, state, session);
 	}
 	function dispose() {
-		stopAvailabilityPoll();
+		poller.stop();
 		stop();
 	}
 	return {
-		refreshAvailability,
+		refreshAvailability: poller.refresh,
 		start,
 		stop,
 		dispose
 	};
 }
 //#endregion
+exports.createAvailabilityPoller = createAvailabilityPoller;
+exports.createCaptureState = createCaptureState;
+exports.createSegmentQueue = createSegmentQueue;
 exports.createVoiceCapture = createVoiceCapture;
 exports.localeToWhisperLanguage = localeToWhisperLanguage;
 

package/dist/whisper/client.cjs.map

@@ -1 +1 @@
-{"version":3,"file":"client.cjs","names":[],"sources":["../../src/whisper/client-helpers.ts","../../src/whisper/client.ts"],"sourcesContent":["// Pure, framework-neutral helpers for the browser voice-capture controller.\n// Kept separate from `client.ts` so the audio math and VAD state machine can be\n// unit-tested without a DOM / MediaRecorder / Web Audio environment.\n\nexport function computeRms(buffer: Float32Array): number {\n  let sum = 0;\n  for (const sample of buffer) sum += sample * sample;\n  return Math.sqrt(sum / buffer.length);\n}\n\n/** The recorder mime can carry a codec suffix (`audio/webm;codecs=opus`); the\n *  Blob container type is just the part before the `;`. */\nexport function containerTypeFromMime(mimeType: string): string {\n  return mimeType.split(\";\")[0] || \"audio/webm\";\n}\n\nexport interface VadState {\n  readonly hasSpeech: boolean;\n  readonly silenceStart: number | null;\n}\n\nexport interface VadConfig {\n  readonly speechRms: number;\n  readonly silenceMs: number;\n  readonly maxSegmentMs: number;\n}\n\nexport interface VadDecision {\n  readonly hasSpeech: boolean;\n  readonly silenceStart: number | null;\n  readonly cut: boolean;\n}\n\nfunction isSpeech(rms: number, config: VadConfig): boolean {\n  return rms > config.speechRms;\n}\n\nfunction nextSilenceStart(state: VadState, rms: number, nowMs: number, config: VadConfig): number | null {\n  if (isSpeech(rms, config)) return null;\n  if (state.hasSpeech && state.silenceStart === null) return nowMs;\n  return state.silenceStart;\n}\n\nfunction silenceExceeded(state: VadState, rms: number, nowMs: number, config: VadConfig): boolean {\n  if (isSpeech(rms, config) || !state.hasSpeech || state.silenceStart === null) return false;\n  return nowMs - state.silenceStart > config.silenceMs;\n}\n\n/** Advance the pause-detector one tick: fold the current RMS/time into the\n *  segment state and decide whether the segment should be force-cut (either a\n *  long-enough silence after speech, or the max-segment length reached). */\nexport function evaluateVad(state: VadState, segmentStartMs: number, rms: number, nowMs: number, config: VadConfig): VadDecision {\n  const hasSpeech = isSpeech(rms, config) || state.hasSpeech;\n  const silenceStart = nextSilenceStart(state, rms, nowMs, config);\n  const maxCut = hasSpeech && nowMs - segmentStartMs > config.maxSegmentMs;\n  return { hasSpeech, silenceStart, cut: silenceExceeded(state, rms, nowMs, config) || maxCut };\n}\n","// @mulmoclaude/core/whisper/client — framework-neutral browser capture controller.\n// Records one utterance at a time with MediaRecorder, segments on pauses via a\n// Web Audio VAD, and sends each segment through an injected transport. State is\n// pushed via `onState`; the host wraps this into its own reactivity (Vue refs,\n// React state, …). No framework dependency. See plans/done/feat-extract-whisper-package.md.\n\nimport { computeRms, containerTypeFromMime, evaluateVad, type VadConfig } from \"./client-helpers.ts\";\n\n// Map a UI locale to a Whisper language code. UI language is a strong prior for\n// the spoken language; \"auto\" lets Whisper detect it from the audio.\nconst LOCALE_TO_WHISPER: Record<string, string> = {\n  en: \"en\",\n  ja: \"ja\",\n  zh: \"zh\",\n  ko: \"ko\",\n  es: \"es\",\n  \"pt-BR\": \"pt\",\n  fr: \"fr\",\n  de: \"de\",\n};\n\nexport function localeToWhisperLanguage(locale: string): string {\n  return LOCALE_TO_WHISPER[locale] ?? \"auto\";\n}\n\n// VAD tuning. RMS over [-1,1] float samples; a pause is SILENCE_MS of\n// sub-threshold level after speech. MAX_SEGMENT_MS force-cuts a long unbroken\n// utterance so no clip exceeds Whisper's 30s window or the server's size cap.\nconst SPEECH_RMS = 0.015;\nconst SILENCE_MS = 800;\nconst MAX_SEGMENT_MS = 20_000;\nconst MONITOR_INTERVAL_MS = 100;\nconst AVAILABILITY_POLL_MS = 2_000;\nconst VAD_CONFIG: VadConfig = { speechRms: SPEECH_RMS, silenceMs: SILENCE_MS, maxSegmentMs: MAX_SEGMENT_MS };\n\nfunction pickRecorderMime(): string | undefined {\n  const candidates = [\"audio/webm;codecs=opus\", \"audio/webm\", \"audio/mp4\"];\n  if (typeof MediaRecorder === \"undefined\") return undefined;\n  return candidates.find((type) => MediaRecorder.isTypeSupported(type));\n}\n\nfunction blobToDataUrl(blob: Blob): Promise<string> {\n  return new Promise((resolve, reject) => {\n    const reader = new FileReader();\n    reader.onload = () => resolve(String(reader.result));\n    reader.onerror = () => reject(reader.error ?? new Error(\"FileReader failed\"));\n    reader.readAsDataURL(blob);\n  });\n}\n\nfunction toMessage(err: unknown): string {\n  return err instanceof Error ? err.message : String(err);\n}\n\nexport interface VoiceCaptureTransport {\n  /** Transcribe one segment. Throws on failure. */\n  transcribe: (dataUrl: string, language: string) => Promise<{ text: string }>;\n  /** Current readiness. `downloading` true keeps the controller polling so it\n   *  flips to ready as soon as a model download finishes. */\n  getStatus: () => Promise<{ ready: boolean; downloading: boolean }>;\n}\n\nexport interface VoiceCaptureState {\n  available: boolean;\n  listening: boolean;\n  transcribing: boolean;\n}\n\nexport interface VoiceCaptureCallbacks {\n  /** A recognized (non-empty) segment transcript. */\n  onTranscript: (text: string) => void;\n  /** A segment produced no speech. */\n  onEmpty?: () => void;\n  /** A recoverable error message (transport failure, permission denied, etc.). */\n  onError?: (message: string) => void;\n  /** Pushed whenever available/listening/transcribing changes. */\n  onState?: (state: VoiceCaptureState) => void;\n}\n\nexport interface VoiceCapture {\n  refreshAvailability: () => Promise<void>;\n  start: () => Promise<boolean>;\n  stop: () => void;\n  dispose: () => void;\n}\n\nexport function createVoiceCapture(transport: VoiceCaptureTransport, language: () => string, callbacks: VoiceCaptureCallbacks): VoiceCapture {\n  let available = false;\n  let listening = false;\n  let transcribing = false;\n\n  function emit(): void {\n    callbacks.onState?.({ available, listening, transcribing });\n  }\n  function setAvailable(value: boolean): void {\n    if (available !== value) {\n      available = value;\n      emit();\n    }\n  }\n  function setListening(value: boolean): void {\n    if (listening !== value) {\n      listening = value;\n      emit();\n    }\n  }\n\n  let stream: MediaStream | null = null;\n  let audioCtx: AudioContext | null = null;\n  let analyser: AnalyserNode | null = null;\n  let vadBuffer = new Float32Array(0);\n  let monitorHandle: number | null = null;\n  let recorder: MediaRecorder | null = null;\n  let chunks: Blob[] = [];\n  let mimeType = \"\";\n  let segmentHasSpeech = false;\n  let silenceStart: number | null = null;\n  let segmentStart = 0;\n  let pending = 0;\n  let queue: Promise<void> = Promise.resolve();\n  let availabilityPollHandle: number | null = null;\n  // Bumped on stop(). Segments captured / sends resolved under an older\n  // generation are dropped, so a late transcript never leaks across sessions.\n  let generation = 0;\n  let segmentGeneration = 0;\n  // Single-flight guard for start(): true between entry and the moment capture\n  // is set up (or the attempt aborts), so a second start can't race the first.\n  let startInFlight = false;\n\n  function setPending(delta: number): void {\n    pending += delta;\n    const next = pending > 0;\n    if (transcribing !== next) {\n      transcribing = next;\n      emit();\n    }\n  }\n\n  function stopAvailabilityPoll(): void {\n    if (availabilityPollHandle !== null) {\n      window.clearInterval(availabilityPollHandle);\n      availabilityPollHandle = null;\n    }\n  }\n\n  async function refreshAvailability(): Promise<void> {\n    let status: { ready: boolean; downloading: boolean };\n    try {\n      status = await transport.getStatus();\n    } catch {\n      setAvailable(false);\n      stopAvailabilityPoll();\n      return;\n    }\n    setAvailable(status.ready);\n    if (status.downloading) {\n      if (availabilityPollHandle === null) {\n        availabilityPollHandle = window.setInterval(() => {\n          void refreshAvailability();\n        }, AVAILABILITY_POLL_MS);\n      }\n    } else {\n      stopAvailabilityPoll();\n    }\n  }\n\n  async function sendSegment(blob: Blob, gen: number): Promise<void> {\n    if (gen !== generation) return;\n    try {\n      const dataUrl = await blobToDataUrl(blob);\n      const result = await transport.transcribe(dataUrl, language());\n      if (gen !== generation) return;\n      const text = result.text.trim();\n      if (text.length === 0) callbacks.onEmpty?.();\n      else callbacks.onTranscript(text);\n    } catch (err) {\n      // Generation-guard the failure path too: a send rejected after stop()/\n      // session change belongs to a session the user already left.\n      if (gen === generation) callbacks.onError?.(toMessage(err));\n    }\n  }\n\n  // Serialize sends so transcripts append in capture order; `pending` keeps\n  // `transcribing` true from enqueue until the send resolves.\n  function enqueue(blob: Blob, gen: number): void {\n    setPending(1);\n    queue = queue\n      .then(() => sendSegment(blob, gen))\n      .catch(() => undefined)\n      .finally(() => setPending(-1));\n  }\n\n  function onSegmentStop(): void {\n    const hadSpeech = segmentHasSpeech;\n    const gen = segmentGeneration;\n    const blob = new Blob(chunks, { type: containerTypeFromMime(mimeType) });\n    if (listening) startRecorder();\n    if (hadSpeech && blob.size > 0 && gen === generation) enqueue(blob, gen);\n  }\n\n  function startRecorder(): void {\n    if (!stream) return;\n    chunks = [];\n    segmentHasSpeech = false;\n    silenceStart = null;\n    segmentStart = Date.now();\n    segmentGeneration = generation;\n    recorder = new MediaRecorder(stream, { mimeType });\n    recorder.ondataavailable = (event) => {\n      if (event.data.size > 0) chunks.push(event.data);\n    };\n    recorder.onstop = onSegmentStop;\n    recorder.start();\n  }\n\n  function cutSegment(): void {\n    if (recorder && recorder.state === \"recording\") recorder.stop();\n  }\n\n  function monitorTick(): void {\n    if (!analyser) return;\n    analyser.getFloatTimeDomainData(vadBuffer);\n    const rms = computeRms(vadBuffer);\n    const { hasSpeech, silenceStart: nextSilence, cut } = evaluateVad({ hasSpeech: segmentHasSpeech, silenceStart }, segmentStart, rms, Date.now(), VAD_CONFIG);\n    segmentHasSpeech = hasSpeech;\n    silenceStart = nextSilence;\n    if (cut) cutSegment();\n  }\n\n  async function start(): Promise<boolean> {\n    // Single-flight: `listening` only flips true AFTER getUserMedia resolves, so\n    // a benign skip returns true (a start is already active / in progress).\n    if (startInFlight || listening) return true;\n    startInFlight = true;\n    const startGen = generation;\n    try {\n      mimeType = pickRecorderMime() ?? \"\";\n      if (!mimeType || !navigator.mediaDevices?.getUserMedia) {\n        callbacks.onError?.(\"unsupported\");\n        return false;\n      }\n      let acquired: MediaStream;\n      try {\n        acquired = await navigator.mediaDevices.getUserMedia({ audio: true });\n      } catch {\n        callbacks.onError?.(\"permission-denied\");\n        return false;\n      }\n      // stop() bumps the generation; if it fired while we awaited permission\n      // this start is stale — release the mic and abort.\n      if (startGen !== generation) {\n        acquired.getTracks().forEach((track) => track.stop());\n        return false;\n      }\n      stream = acquired;\n      audioCtx = new AudioContext();\n      analyser = audioCtx.createAnalyser();\n      analyser.fftSize = 2048;\n      vadBuffer = new Float32Array(analyser.fftSize);\n      audioCtx.createMediaStreamSource(stream).connect(analyser);\n      setListening(true);\n      startRecorder();\n      monitorHandle = window.setInterval(monitorTick, MONITOR_INTERVAL_MS);\n      return true;\n    } finally {\n      startInFlight = false;\n    }\n  }\n\n  function stop(): void {\n    // Bump the generation so any in-flight/queued segment is dropped rather than\n    // applied after the user stops or switches sessions.\n    generation += 1;\n    setListening(false);\n    if (monitorHandle !== null) {\n      window.clearInterval(monitorHandle);\n      monitorHandle = null;\n    }\n    if (recorder && recorder.state === \"recording\") recorder.stop();\n    recorder = null;\n    if (audioCtx) {\n      audioCtx.close().catch(() => undefined);\n      audioCtx = null;\n    }\n    analyser = null;\n    stream?.getTracks().forEach((track) => track.stop());\n    stream = null;\n  }\n\n  function dispose(): void {\n    stopAvailabilityPoll();\n    stop();\n  }\n\n  return { refreshAvailability, start, stop, dispose };\n}\n"],"mappings":";;AAIA,SAAgB,WAAW,QAA8B;CACvD,IAAI,MAAM;CACV,KAAK,MAAM,UAAU,QAAQ,OAAO,SAAS;CAC7C,OAAO,KAAK,KAAK,MAAM,OAAO,MAAM;AACtC;;;AAIA,SAAgB,sBAAsB,UAA0B;CAC9D,OAAO,SAAS,MAAM,GAAG,CAAC,CAAC,MAAM;AACnC;AAmBA,SAAS,SAAS,KAAa,QAA4B;CACzD,OAAO,MAAM,OAAO;AACtB;AAEA,SAAS,iBAAiB,OAAiB,KAAa,OAAe,QAAkC;CACvG,IAAI,SAAS,KAAK,MAAM,GAAG,OAAO;CAClC,IAAI,MAAM,aAAa,MAAM,iBAAiB,MAAM,OAAO;CAC3D,OAAO,MAAM;AACf;AAEA,SAAS,gBAAgB,OAAiB,KAAa,OAAe,QAA4B;CAChG,IAAI,SAAS,KAAK,MAAM,KAAK,CAAC,MAAM,aAAa,MAAM,iBAAiB,MAAM,OAAO;CACrF,OAAO,QAAQ,MAAM,eAAe,OAAO;AAC7C;;;;AAKA,SAAgB,YAAY,OAAiB,gBAAwB,KAAa,OAAe,QAAgC;CAC/H,MAAM,YAAY,SAAS,KAAK,MAAM,KAAK,MAAM;CACjD,MAAM,eAAe,iBAAiB,OAAO,KAAK,OAAO,MAAM;CAC/D,MAAM,SAAS,aAAa,QAAQ,iBAAiB,OAAO;CAC5D,OAAO;EAAE;EAAW;EAAc,KAAK,gBAAgB,OAAO,KAAK,OAAO,MAAM,KAAK;CAAO;AAC9F;;;AC9CA,IAAM,oBAA4C;CAChD,IAAI;CACJ,IAAI;CACJ,IAAI;CACJ,IAAI;CACJ,IAAI;CACJ,SAAS;CACT,IAAI;CACJ,IAAI;AACN;AAEA,SAAgB,wBAAwB,QAAwB;CAC9D,OAAO,kBAAkB,WAAW;AACtC;AAKA,IAAM,aAAa;AACnB,IAAM,aAAa;AACnB,IAAM,iBAAiB;AACvB,IAAM,sBAAsB;AAC5B,IAAM,uBAAuB;AAC7B,IAAM,aAAwB;CAAE,WAAW;CAAY,WAAW;CAAY,cAAc;AAAe;AAE3G,SAAS,mBAAuC;CAC9C,MAAM,aAAa;EAAC;EAA0B;EAAc;CAAW;CACvE,IAAI,OAAO,kBAAkB,aAAa,OAAO,KAAA;CACjD,OAAO,WAAW,MAAM,SAAS,cAAc,gBAAgB,IAAI,CAAC;AACtE;AAEA,SAAS,cAAc,MAA6B;CAClD,OAAO,IAAI,SAAS,SAAS,WAAW;EACtC,MAAM,SAAS,IAAI,WAAW;EAC9B,OAAO,eAAe,QAAQ,OAAO,OAAO,MAAM,CAAC;EACnD,OAAO,gBAAgB,OAAO,OAAO,yBAAS,IAAI,MAAM,mBAAmB,CAAC;EAC5E,OAAO,cAAc,IAAI;CAC3B,CAAC;AACH;AAEA,SAAS,UAAU,KAAsB;CACvC,OAAO,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AACxD;AAkCA,SAAgB,mBAAmB,WAAkC,UAAwB,WAAgD;CAC3I,IAAI,YAAY;CAChB,IAAI,YAAY;CAChB,IAAI,eAAe;CAEnB,SAAS,OAAa;EACpB,UAAU,UAAU;GAAE;GAAW;GAAW;EAAa,CAAC;CAC5D;CACA,SAAS,aAAa,OAAsB;EAC1C,IAAI,cAAc,OAAO;GACvB,YAAY;GACZ,KAAK;EACP;CACF;CACA,SAAS,aAAa,OAAsB;EAC1C,IAAI,cAAc,OAAO;GACvB,YAAY;GACZ,KAAK;EACP;CACF;CAEA,IAAI,SAA6B;CACjC,IAAI,WAAgC;CACpC,IAAI,WAAgC;CACpC,IAAI,4BAAY,IAAI,aAAa,CAAC;CAClC,IAAI,gBAA+B;CACnC,IAAI,WAAiC;CACrC,IAAI,SAAiB,CAAC;CACtB,IAAI,WAAW;CACf,IAAI,mBAAmB;CACvB,IAAI,eAA8B;CAClC,IAAI,eAAe;CACnB,IAAI,UAAU;CACd,IAAI,QAAuB,QAAQ,QAAQ;CAC3C,IAAI,yBAAwC;CAG5C,IAAI,aAAa;CACjB,IAAI,oBAAoB;CAGxB,IAAI,gBAAgB;CAEpB,SAAS,WAAW,OAAqB;EACvC,WAAW;EACX,MAAM,OAAO,UAAU;EACvB,IAAI,iBAAiB,MAAM;GACzB,eAAe;GACf,KAAK;EACP;CACF;CAEA,SAAS,uBAA6B;EACpC,IAAI,2BAA2B,MAAM;GACnC,OAAO,cAAc,sBAAsB;GAC3C,yBAAyB;EAC3B;CACF;CAEA,eAAe,sBAAqC;EAClD,IAAI;EACJ,IAAI;GACF,SAAS,MAAM,UAAU,UAAU;EACrC,QAAQ;GACN,aAAa,KAAK;GAClB,qBAAqB;GACrB;EACF;EACA,aAAa,OAAO,KAAK;EACzB,IAAI,OAAO;OACL,2BAA2B,MAC7B,yBAAyB,OAAO,kBAAkB;IAChD,oBAAyB;GAC3B,GAAG,oBAAoB;EAAA,OAGzB,qBAAqB;CAEzB;CAEA,eAAe,YAAY,MAAY,KAA4B;EACjE,IAAI,QAAQ,YAAY;EACxB,IAAI;GACF,MAAM,UAAU,MAAM,cAAc,IAAI;GACxC,MAAM,SAAS,MAAM,UAAU,WAAW,SAAS,SAAS,CAAC;GAC7D,IAAI,QAAQ,YAAY;GACxB,MAAM,OAAO,OAAO,KAAK,KAAK;GAC9B,IAAI,KAAK,WAAW,GAAG,UAAU,UAAU;QACtC,UAAU,aAAa,IAAI;EAClC,SAAS,KAAK;GAGZ,IAAI,QAAQ,YAAY,UAAU,UAAU,UAAU,GAAG,CAAC;EAC5D;CACF;CAIA,SAAS,QAAQ,MAAY,KAAmB;EAC9C,WAAW,CAAC;EACZ,QAAQ,MACL,WAAW,YAAY,MAAM,GAAG,CAAC,CAAC,CAClC,YAAY,KAAA,CAAS,CAAC,CACtB,cAAc,WAAW,EAAE,CAAC;CACjC;CAEA,SAAS,gBAAsB;EAC7B,MAAM,YAAY;EAClB,MAAM,MAAM;EACZ,MAAM,OAAO,IAAI,KAAK,QAAQ,EAAE,MAAM,sBAAsB,QAAQ,EAAE,CAAC;EACvE,IAAI,WAAW,cAAc;EAC7B,IAAI,aAAa,KAAK,OAAO,KAAK,QAAQ,YAAY,QAAQ,MAAM,GAAG;CACzE;CAEA,SAAS,gBAAsB;EAC7B,IAAI,CAAC,QAAQ;EACb,SAAS,CAAC;EACV,mBAAmB;EACnB,eAAe;EACf,eAAe,KAAK,IAAI;EACxB,oBAAoB;EACpB,WAAW,IAAI,cAAc,QAAQ,EAAE,SAAS,CAAC;EACjD,SAAS,mBAAmB,UAAU;GACpC,IAAI,MAAM,KAAK,OAAO,GAAG,OAAO,KAAK,MAAM,IAAI;EACjD;EACA,SAAS,SAAS;EAClB,SAAS,MAAM;CACjB;CAEA,SAAS,aAAmB;EAC1B,IAAI,YAAY,SAAS,UAAU,aAAa,SAAS,KAAK;CAChE;CAEA,SAAS,cAAoB;EAC3B,IAAI,CAAC,UAAU;EACf,SAAS,uBAAuB,SAAS;EACzC,MAAM,MAAM,WAAW,SAAS;EAChC,MAAM,EAAE,WAAW,cAAc,aAAa,QAAQ,YAAY;GAAE,WAAW;GAAkB;EAAa,GAAG,cAAc,KAAK,KAAK,IAAI,GAAG,UAAU;EAC1J,mBAAmB;EACnB,eAAe;EACf,IAAI,KAAK,WAAW;CACtB;CAEA,eAAe,QAA0B;EAGvC,IAAI,iBAAiB,WAAW,OAAO;EACvC,gBAAgB;EAChB,MAAM,WAAW;EACjB,IAAI;GACF,WAAW,iBAAiB,KAAK;GACjC,IAAI,CAAC,YAAY,CAAC,UAAU,cAAc,cAAc;IACtD,UAAU,UAAU,aAAa;IACjC,OAAO;GACT;GACA,IAAI;GACJ,IAAI;IACF,WAAW,MAAM,UAAU,aAAa,aAAa,EAAE,OAAO,KAAK,CAAC;GACtE,QAAQ;IACN,UAAU,UAAU,mBAAmB;IACvC,OAAO;GACT;GAGA,IAAI,aAAa,YAAY;IAC3B,SAAS,UAAU,CAAC,CAAC,SAAS,UAAU,MAAM,KAAK,CAAC;IACpD,OAAO;GACT;GACA,SAAS;GACT,WAAW,IAAI,aAAa;GAC5B,WAAW,SAAS,eAAe;GACnC,SAAS,UAAU;GACnB,YAAY,IAAI,aAAa,SAAS,OAAO;GAC7C,SAAS,wBAAwB,MAAM,CAAC,CAAC,QAAQ,QAAQ;GACzD,aAAa,IAAI;GACjB,cAAc;GACd,gBAAgB,OAAO,YAAY,aAAa,mBAAmB;GACnE,OAAO;EACT,UAAU;GACR,gBAAgB;EAClB;CACF;CAEA,SAAS,OAAa;EAGpB,cAAc;EACd,aAAa,KAAK;EAClB,IAAI,kBAAkB,MAAM;GAC1B,OAAO,cAAc,aAAa;GAClC,gBAAgB;EAClB;EACA,IAAI,YAAY,SAAS,UAAU,aAAa,SAAS,KAAK;EAC9D,WAAW;EACX,IAAI,UAAU;GACZ,SAAS,MAAM,CAAC,CAAC,YAAY,KAAA,CAAS;GACtC,WAAW;EACb;EACA,WAAW;EACX,QAAQ,UAAU,CAAC,CAAC,SAAS,UAAU,MAAM,KAAK,CAAC;EACnD,SAAS;CACX;CAEA,SAAS,UAAgB;EACvB,qBAAqB;EACrB,KAAK;CACP;CAEA,OAAO;EAAE;EAAqB;EAAO;EAAM;CAAQ;AACrD"}
+{"version":3,"file":"client.cjs","names":[],"sources":["../../src/whisper/client-helpers.ts","../../src/whisper/client.ts"],"sourcesContent":["// Pure, framework-neutral helpers for the browser voice-capture controller.\n// Kept separate from `client.ts` so the audio math and VAD state machine can be\n// unit-tested without a DOM / MediaRecorder / Web Audio environment.\n\nexport function computeRms(buffer: Float32Array): number {\n  let sum = 0;\n  for (const sample of buffer) sum += sample * sample;\n  return Math.sqrt(sum / buffer.length);\n}\n\n/** The recorder mime can carry a codec suffix (`audio/webm;codecs=opus`); the\n *  Blob container type is just the part before the `;`. */\nexport function containerTypeFromMime(mimeType: string): string {\n  return mimeType.split(\";\")[0] || \"audio/webm\";\n}\n\nexport interface VadState {\n  readonly hasSpeech: boolean;\n  readonly silenceStart: number | null;\n}\n\nexport interface VadConfig {\n  readonly speechRms: number;\n  readonly silenceMs: number;\n  readonly maxSegmentMs: number;\n}\n\nexport interface VadDecision {\n  readonly hasSpeech: boolean;\n  readonly silenceStart: number | null;\n  readonly cut: boolean;\n}\n\nfunction isSpeech(rms: number, config: VadConfig): boolean {\n  return rms > config.speechRms;\n}\n\nfunction nextSilenceStart(state: VadState, rms: number, nowMs: number, config: VadConfig): number | null {\n  if (isSpeech(rms, config)) return null;\n  if (state.hasSpeech && state.silenceStart === null) return nowMs;\n  return state.silenceStart;\n}\n\nfunction silenceExceeded(state: VadState, rms: number, nowMs: number, config: VadConfig): boolean {\n  if (isSpeech(rms, config) || !state.hasSpeech || state.silenceStart === null) return false;\n  return nowMs - state.silenceStart > config.silenceMs;\n}\n\n/** Advance the pause-detector one tick: fold the current RMS/time into the\n *  segment state and decide whether the segment should be force-cut (either a\n *  long-enough silence after speech, or the max-segment length reached). */\nexport function evaluateVad(state: VadState, segmentStartMs: number, rms: number, nowMs: number, config: VadConfig): VadDecision {\n  const hasSpeech = isSpeech(rms, config) || state.hasSpeech;\n  const silenceStart = nextSilenceStart(state, rms, nowMs, config);\n  const maxCut = hasSpeech && nowMs - segmentStartMs > config.maxSegmentMs;\n  return { hasSpeech, silenceStart, cut: silenceExceeded(state, rms, nowMs, config) || maxCut };\n}\n","// @mulmoclaude/core/whisper/client — framework-neutral browser capture controller.\n// Records one utterance at a time with MediaRecorder, segments on pauses via a\n// Web Audio VAD, and sends each segment through an injected transport. State is\n// pushed via `onState`; the host wraps this into its own reactivity (Vue refs,\n// React state, …). No framework dependency. See plans/done/feat-extract-whisper-package.md.\n\nimport { computeRms, containerTypeFromMime, evaluateVad, type VadConfig } from \"./client-helpers.ts\";\n\n// Map a UI locale to a Whisper language code. UI language is a strong prior for\n// the spoken language; \"auto\" lets Whisper detect it from the audio.\nconst LOCALE_TO_WHISPER: Record<string, string> = {\n  en: \"en\",\n  ja: \"ja\",\n  zh: \"zh\",\n  ko: \"ko\",\n  es: \"es\",\n  \"pt-BR\": \"pt\",\n  fr: \"fr\",\n  de: \"de\",\n};\n\nexport function localeToWhisperLanguage(locale: string): string {\n  return LOCALE_TO_WHISPER[locale] ?? \"auto\";\n}\n\n// VAD tuning. RMS over [-1,1] float samples; a pause is SILENCE_MS of\n// sub-threshold level after speech. MAX_SEGMENT_MS force-cuts a long unbroken\n// utterance so no clip exceeds Whisper's 30s window or the server's size cap.\nconst SPEECH_RMS = 0.015;\nconst SILENCE_MS = 800;\nconst MAX_SEGMENT_MS = 20_000;\nconst MONITOR_INTERVAL_MS = 100;\nconst AVAILABILITY_POLL_MS = 2_000;\nconst VAD_CONFIG: VadConfig = { speechRms: SPEECH_RMS, silenceMs: SILENCE_MS, maxSegmentMs: MAX_SEGMENT_MS };\n\nfunction pickRecorderMime(): string | undefined {\n  const candidates = [\"audio/webm;codecs=opus\", \"audio/webm\", \"audio/mp4\"];\n  if (typeof MediaRecorder === \"undefined\") return undefined;\n  return candidates.find((type) => MediaRecorder.isTypeSupported(type));\n}\n\nfunction blobToDataUrl(blob: Blob): Promise<string> {\n  return new Promise((resolve, reject) => {\n    const reader = new FileReader();\n    reader.onload = () => resolve(String(reader.result));\n    reader.onerror = () => reject(reader.error ?? new Error(\"FileReader failed\"));\n    reader.readAsDataURL(blob);\n  });\n}\n\nfunction toMessage(err: unknown): string {\n  return err instanceof Error ? err.message : String(err);\n}\n\nexport interface VoiceCaptureTransport {\n  /** Transcribe one segment. Throws on failure. */\n  transcribe: (dataUrl: string, language: string) => Promise<{ text: string }>;\n  /** Current readiness. `downloading` true keeps the controller polling so it\n   *  flips to ready as soon as a model download finishes. */\n  getStatus: () => Promise<{ ready: boolean; downloading: boolean }>;\n}\n\nexport interface VoiceCaptureState {\n  available: boolean;\n  listening: boolean;\n  transcribing: boolean;\n}\n\nexport interface VoiceCaptureCallbacks {\n  /** A recognized (non-empty) segment transcript. */\n  onTranscript: (text: string) => void;\n  /** A segment produced no speech. */\n  onEmpty?: () => void;\n  /** A recoverable error message (transport failure, permission denied, etc.). */\n  onError?: (message: string) => void;\n  /** Pushed whenever available/listening/transcribing changes. */\n  onState?: (state: VoiceCaptureState) => void;\n}\n\nexport interface VoiceCapture {\n  refreshAvailability: () => Promise<void>;\n  start: () => Promise<boolean>;\n  stop: () => void;\n  dispose: () => void;\n}\n\nexport interface CaptureStateController {\n  setAvailable: (value: boolean) => void;\n  setListening: (value: boolean) => void;\n  setPending: (delta: number) => void;\n  isListening: () => boolean;\n}\n\n// Owns the three observable flags. Each setter emits only on an actual change so\n// the host's reactivity never churns on a no-op write. `pending` is a private\n// counter; `transcribing` is true exactly while at least one send is in flight.\nexport function createCaptureState(onState?: (state: VoiceCaptureState) => void): CaptureStateController {\n  let available = false;\n  let listening = false;\n  let transcribing = false;\n  let pending = 0;\n\n  function emit(): void {\n    onState?.({ available, listening, transcribing });\n  }\n  function setAvailable(value: boolean): void {\n    if (available !== value) {\n      available = value;\n      emit();\n    }\n  }\n  function setListening(value: boolean): void {\n    if (listening !== value) {\n      listening = value;\n      emit();\n    }\n  }\n  function setPending(delta: number): void {\n    pending += delta;\n    const next = pending > 0;\n    if (transcribing !== next) {\n      transcribing = next;\n      emit();\n    }\n  }\n\n  return { setAvailable, setListening, setPending, isListening: () => listening };\n}\n\nexport interface AvailabilityPoller {\n  refresh: () => Promise<void>;\n  stop: () => void;\n}\n\n// Owns the polling timer. `refresh` mirrors transport readiness into\n// `setAvailable`; while a model is downloading it self-schedules so `available`\n// flips as soon as the download finishes. A `getStatus` throw fails closed\n// (unavailable) and tears the timer down.\nexport function createAvailabilityPoller(transport: VoiceCaptureTransport, setAvailable: (value: boolean) => void): AvailabilityPoller {\n  let availabilityPollHandle: number | null = null;\n\n  function stop(): void {\n    if (availabilityPollHandle !== null) {\n      window.clearInterval(availabilityPollHandle);\n      availabilityPollHandle = null;\n    }\n  }\n\n  async function refresh(): Promise<void> {\n    let status: { ready: boolean; downloading: boolean };\n    try {\n      status = await transport.getStatus();\n    } catch {\n      setAvailable(false);\n      stop();\n      return;\n    }\n    setAvailable(status.ready);\n    if (status.downloading) {\n      if (availabilityPollHandle === null) {\n        availabilityPollHandle = window.setInterval(() => {\n          void refresh();\n        }, AVAILABILITY_POLL_MS);\n      }\n    } else {\n      stop();\n    }\n  }\n\n  return { refresh, stop };\n}\n\nexport interface SegmentQueueOptions {\n  transport: VoiceCaptureTransport;\n  language: () => string;\n  callbacks: VoiceCaptureCallbacks;\n  setPending: (delta: number) => void;\n  // Read-only epoch seam: the queue never mutates the generation, it only reads\n  // the parent's current value to drop segments captured under an older session.\n  getGeneration: () => number;\n}\n\nexport interface SegmentQueue {\n  enqueue: (blob: Blob, gen: number) => void;\n}\n\n// Owns the serialized send chain. Each blob is transcribed in capture order; the\n// generation is checked at entry and again after transcription so a segment\n// belonging to a session the user already left is dropped silently.\nexport function createSegmentQueue(options: SegmentQueueOptions): SegmentQueue {\n  const { transport, language, callbacks, setPending, getGeneration } = options;\n  let queue: Promise<void> = Promise.resolve();\n\n  async function sendSegment(blob: Blob, gen: number): Promise<void> {\n    if (gen !== getGeneration()) return;\n    try {\n      const dataUrl = await blobToDataUrl(blob);\n      const result = await transport.transcribe(dataUrl, language());\n      if (gen !== getGeneration()) return;\n      const text = result.text.trim();\n      if (text.length === 0) callbacks.onEmpty?.();\n      else callbacks.onTranscript(text);\n    } catch (err) {\n      // Generation-guard the failure path too: a send rejected after stop()/\n      // session change belongs to a session the user already left.\n      if (gen === getGeneration()) callbacks.onError?.(toMessage(err));\n    }\n  }\n\n  // Serialize sends so transcripts append in capture order; `pending` keeps\n  // `transcribing` true from enqueue until the send resolves.\n  function enqueue(blob: Blob, gen: number): void {\n    setPending(1);\n    queue = queue\n      .then(() => sendSegment(blob, gen))\n      .catch(() => undefined)\n      .finally(() => setPending(-1));\n  }\n\n  return { enqueue };\n}\n\nexport interface AudioGraph {\n  /** True between `attach()` and the next `teardown()`. */\n  isAttached: () => boolean;\n  /** Read the current PCM window into an internal buffer and return it.\n   *  Callers must not mutate the returned array. */\n  sample: () => Float32Array;\n  /** Live mic stream — the outer needs it to hand to MediaRecorder. */\n  getStream: () => MediaStream | null;\n  /** Attach a fresh mic stream: create AudioContext + AnalyserNode + VAD buffer,\n   *  connect the source. Called only by `bootstrapCapture` after the caller has\n   *  passed the stale-generation / permission checks. */\n  attach: (stream: MediaStream) => void;\n  /** Close the AudioContext, stop the mic tracks, null everything. Safe to\n   *  call before `attach()` or twice in a row. */\n  teardown: () => void;\n}\n\n// Owns the browser-side audio pipeline: mic stream + AudioContext + AnalyserNode\n// + the reusable Float32 buffer VAD reads samples into. Separated from\n// `createVoiceCapture` so its four browser-API fields don't inflate the parent\n// factory's line count and so the setup/teardown ordering is in one place.\nfunction createAudioGraph(): AudioGraph {\n  let stream: MediaStream | null = null;\n  let audioCtx: AudioContext | null = null;\n  let analyser: AnalyserNode | null = null;\n  let vadBuffer = new Float32Array(0);\n\n  function attach(nextStream: MediaStream): void {\n    stream = nextStream;\n    audioCtx = new AudioContext();\n    analyser = audioCtx.createAnalyser();\n    analyser.fftSize = 2048;\n    vadBuffer = new Float32Array(analyser.fftSize);\n    audioCtx.createMediaStreamSource(nextStream).connect(analyser);\n  }\n\n  function sample(): Float32Array {\n    analyser?.getFloatTimeDomainData(vadBuffer);\n    return vadBuffer;\n  }\n\n  function teardown(): void {\n    if (audioCtx) {\n      audioCtx.close().catch(() => undefined);\n      audioCtx = null;\n    }\n    analyser = null;\n    stream?.getTracks().forEach((track) => track.stop());\n    stream = null;\n  }\n\n  return { isAttached: () => analyser !== null, sample, getStream: () => stream, attach, teardown };\n}\n\ninterface BootstrappedCapture {\n  audioGraph: AudioGraph;\n  mimeType: string;\n}\n\n// Acquire the mic + probe MediaRecorder support + wire an AudioGraph, returning\n// null (with `callbacks.onError` fired) on any recoverable failure. Deliberately\n// does NOT touch `state` or start the recorder — the caller commits the session\n// only after this resolves successfully, so a stale `generation` detected mid-\n// flight can drop the tracks and bail without partially publishing state.\nasync function bootstrapCapture(callbacks: VoiceCaptureCallbacks, generationAtStart: number, generationNow: () => number): Promise<BootstrappedCapture | null> {\n  const mimeType = pickRecorderMime() ?? \"\";\n  if (!mimeType || !navigator.mediaDevices?.getUserMedia) {\n    callbacks.onError?.(\"unsupported\");\n    return null;\n  }\n  let acquired: MediaStream;\n  try {\n    acquired = await navigator.mediaDevices.getUserMedia({ audio: true });\n  } catch {\n    callbacks.onError?.(\"permission-denied\");\n    return null;\n  }\n  if (generationAtStart !== generationNow()) {\n    acquired.getTracks().forEach((track) => track.stop());\n    return null;\n  }\n  const audioGraph = createAudioGraph();\n  try {\n    audioGraph.attach(acquired);\n  } catch (err) {\n    // teardown() is null-safe and closes any half-created AudioContext AND\n    // stops the mic tracks — the pre-refactor code only stopped tracks and\n    // leaked the context (codex/CodeRabbit findings on this PR).\n    audioGraph.teardown();\n    throw err;\n  }\n  return { audioGraph, mimeType };\n}\n\nexport interface RecorderSession {\n  isRecording: () => boolean;\n  hadSpeech: () => boolean;\n  markSpeech: () => void;\n  /** Start a new segment on `stream`, tagged with `gen`. Resets per-segment\n   *  state (chunks + hadSpeech) synchronously so a subsequent VAD tick sees\n   *  a fresh window. */\n  startNext: (stream: MediaStream, mimeType: string, gen: number) => void;\n  /** Force-close the current segment (silence exceeded or MAX_SEGMENT_MS). */\n  cut: () => void;\n}\n\ninterface RecorderSessionDeps {\n  /** Fires each time a segment finishes. The parent decides whether to\n   *  restart (via `session.startNext`) and whether to enqueue the blob. */\n  onSegmentEnd: (blob: Blob, hadSpeech: boolean, gen: number) => void;\n}\n\n// Owns MediaRecorder + its per-segment scratchpad (chunks, hadSpeech, gen,\n// mimeType). The parent keeps VAD state and the global generation, and calls\n// `startNext` on both initial-start and post-onstop restart paths.\nfunction createRecorderSession(deps: RecorderSessionDeps): RecorderSession {\n  let recorder: MediaRecorder | null = null;\n  let chunks: Blob[] = [];\n  let segmentHasSpeech = false;\n  let sessionGen = 0;\n  let sessionMimeType = \"\";\n\n  function onStop(): void {\n    // Snapshot BEFORE the parent's onSegmentEnd callback runs; `startNext`\n    // (if the parent chooses to restart) will clear chunks synchronously.\n    const hadSpeech = segmentHasSpeech;\n    const gen = sessionGen;\n    const blob = new Blob(chunks, { type: containerTypeFromMime(sessionMimeType) });\n    deps.onSegmentEnd(blob, hadSpeech, gen);\n  }\n\n  function startNext(stream: MediaStream, mimeType: string, gen: number): void {\n    chunks = [];\n    segmentHasSpeech = false;\n    sessionGen = gen;\n    sessionMimeType = mimeType;\n    recorder = new MediaRecorder(stream, { mimeType });\n    recorder.ondataavailable = (event) => {\n      if (event.data.size > 0) chunks.push(event.data);\n    };\n    recorder.onstop = onStop;\n    recorder.start();\n  }\n\n  return {\n    isRecording: () => recorder?.state === \"recording\",\n    hadSpeech: () => segmentHasSpeech,\n    markSpeech: () => {\n      segmentHasSpeech = true;\n    },\n    startNext,\n    cut: () => {\n      if (recorder?.state === \"recording\") recorder.stop();\n    },\n  };\n}\n\n// Zero-values used to seed a fresh `CaptureRuntime` at the top of every\n// `createVoiceCapture` call. Kept module-scope so the seed doesn't inflate the\n// factory's line count.\nconst INITIAL_RUNTIME: Readonly<CaptureRuntime> = {\n  audioGraph: null,\n  monitorHandle: null,\n  mimeType: \"\",\n  silenceStart: null,\n  segmentStart: 0,\n  generation: 0,\n  startInFlight: false,\n};\n\n// Mutable capture-time state, kept as one object so module-scope helpers\n// (`stopCapture`, `monitorTick`) can be typed against one shape rather than\n// threading each field through their signatures.\ninterface CaptureRuntime {\n  audioGraph: AudioGraph | null;\n  monitorHandle: number | null;\n  mimeType: string;\n  silenceStart: number | null;\n  segmentStart: number;\n  /** Bumped on stop(); segments captured under an older value are dropped so a\n   *  late transcript never leaks across sessions. */\n  generation: number;\n  /** Single-flight guard on start() — true between entry and completion so a\n   *  second start can't race the first before `listening` flips. */\n  startInFlight: boolean;\n}\n\n// Tear down the browser-side resources. `state.setListening(false)` MUST happen\n// BEFORE `session.cut()` so the parent's onSegmentEnd callback sees\n// isListening()===false and does NOT restart. The monitor interval MUST clear\n// BEFORE `audioGraph.teardown()` so a scheduled tick can't run against a\n// half-closed AudioContext.\nfunction stopCapture(runtime: CaptureRuntime, state: CaptureStateController, session: RecorderSession): void {\n  runtime.generation += 1;\n  state.setListening(false);\n  if (runtime.monitorHandle !== null) {\n    window.clearInterval(runtime.monitorHandle);\n    runtime.monitorHandle = null;\n  }\n  session.cut();\n  runtime.audioGraph?.teardown();\n  runtime.audioGraph = null;\n}\n\n// One VAD tick against the audio graph's current window. No-op when the graph\n// is not attached (defensive: `stopCapture` clears `monitorHandle` before\n// `audioGraph.teardown()`, so this is the belt-and-braces for any late fire).\nfunction monitorTick(runtime: CaptureRuntime, session: RecorderSession): void {\n  if (!runtime.audioGraph?.isAttached()) return;\n  const rms = computeRms(runtime.audioGraph.sample());\n  const {\n    hasSpeech,\n    silenceStart: nextSilence,\n    cut,\n  } = evaluateVad({ hasSpeech: session.hadSpeech(), silenceStart: runtime.silenceStart }, runtime.segmentStart, rms, Date.now(), VAD_CONFIG);\n  if (hasSpeech) session.markSpeech();\n  runtime.silenceStart = nextSilence;\n  if (cut) session.cut();\n}\n\nexport function createVoiceCapture(transport: VoiceCaptureTransport, language: () => string, callbacks: VoiceCaptureCallbacks): VoiceCapture {\n  const state = createCaptureState(callbacks.onState);\n  const poller = createAvailabilityPoller(transport, state.setAvailable);\n  const runtime: CaptureRuntime = { ...INITIAL_RUNTIME };\n  const segments = createSegmentQueue({ transport, language, callbacks, setPending: state.setPending, getGeneration: () => runtime.generation });\n\n  const session = createRecorderSession({ onSegmentEnd: handleSegmentEnd });\n\n  function handleSegmentEnd(blob: Blob, hadSpeech: boolean, gen: number): void {\n    // Restart FIRST so audio-loss between segments is minimal; the snapshot\n    // in blob/hadSpeech/gen is already independent of session state.\n    if (state.isListening()) startRecorder();\n    if (hadSpeech && blob.size > 0 && gen === runtime.generation) segments.enqueue(blob, gen);\n  }\n\n  function startRecorder(): void {\n    const stream = runtime.audioGraph?.getStream();\n    if (!stream) return;\n    runtime.silenceStart = null;\n    runtime.segmentStart = Date.now();\n    session.startNext(stream, runtime.mimeType, runtime.generation);\n  }\n\n  async function start(): Promise<boolean> {\n    if (runtime.startInFlight || state.isListening()) return true;\n    runtime.startInFlight = true;\n    const startGen = runtime.generation;\n    try {\n      const capture = await bootstrapCapture(callbacks, startGen, () => runtime.generation);\n      if (!capture) return false;\n      runtime.audioGraph = capture.audioGraph;\n      runtime.mimeType = capture.mimeType;\n      state.setListening(true);\n      startRecorder();\n      runtime.monitorHandle = window.setInterval(() => monitorTick(runtime, session), MONITOR_INTERVAL_MS);\n      return true;\n    } finally {\n      runtime.startInFlight = false;\n    }\n  }\n\n  function stop(): void {\n    stopCapture(runtime, state, session);\n  }\n\n  function dispose(): void {\n    poller.stop();\n    stop();\n  }\n\n  return { refreshAvailability: poller.refresh, start, stop, dispose };\n}\n"],"mappings":";;AAIA,SAAgB,WAAW,QAA8B;CACvD,IAAI,MAAM;CACV,KAAK,MAAM,UAAU,QAAQ,OAAO,SAAS;CAC7C,OAAO,KAAK,KAAK,MAAM,OAAO,MAAM;AACtC;;;AAIA,SAAgB,sBAAsB,UAA0B;CAC9D,OAAO,SAAS,MAAM,GAAG,CAAC,CAAC,MAAM;AACnC;AAmBA,SAAS,SAAS,KAAa,QAA4B;CACzD,OAAO,MAAM,OAAO;AACtB;AAEA,SAAS,iBAAiB,OAAiB,KAAa,OAAe,QAAkC;CACvG,IAAI,SAAS,KAAK,MAAM,GAAG,OAAO;CAClC,IAAI,MAAM,aAAa,MAAM,iBAAiB,MAAM,OAAO;CAC3D,OAAO,MAAM;AACf;AAEA,SAAS,gBAAgB,OAAiB,KAAa,OAAe,QAA4B;CAChG,IAAI,SAAS,KAAK,MAAM,KAAK,CAAC,MAAM,aAAa,MAAM,iBAAiB,MAAM,OAAO;CACrF,OAAO,QAAQ,MAAM,eAAe,OAAO;AAC7C;;;;AAKA,SAAgB,YAAY,OAAiB,gBAAwB,KAAa,OAAe,QAAgC;CAC/H,MAAM,YAAY,SAAS,KAAK,MAAM,KAAK,MAAM;CACjD,MAAM,eAAe,iBAAiB,OAAO,KAAK,OAAO,MAAM;CAC/D,MAAM,SAAS,aAAa,QAAQ,iBAAiB,OAAO;CAC5D,OAAO;EAAE;EAAW;EAAc,KAAK,gBAAgB,OAAO,KAAK,OAAO,MAAM,KAAK;CAAO;AAC9F;;;AC9CA,IAAM,oBAA4C;CAChD,IAAI;CACJ,IAAI;CACJ,IAAI;CACJ,IAAI;CACJ,IAAI;CACJ,SAAS;CACT,IAAI;CACJ,IAAI;AACN;AAEA,SAAgB,wBAAwB,QAAwB;CAC9D,OAAO,kBAAkB,WAAW;AACtC;AAKA,IAAM,aAAa;AACnB,IAAM,aAAa;AACnB,IAAM,iBAAiB;AACvB,IAAM,sBAAsB;AAC5B,IAAM,uBAAuB;AAC7B,IAAM,aAAwB;CAAE,WAAW;CAAY,WAAW;CAAY,cAAc;AAAe;AAE3G,SAAS,mBAAuC;CAC9C,MAAM,aAAa;EAAC;EAA0B;EAAc;CAAW;CACvE,IAAI,OAAO,kBAAkB,aAAa,OAAO,KAAA;CACjD,OAAO,WAAW,MAAM,SAAS,cAAc,gBAAgB,IAAI,CAAC;AACtE;AAEA,SAAS,cAAc,MAA6B;CAClD,OAAO,IAAI,SAAS,SAAS,WAAW;EACtC,MAAM,SAAS,IAAI,WAAW;EAC9B,OAAO,eAAe,QAAQ,OAAO,OAAO,MAAM,CAAC;EACnD,OAAO,gBAAgB,OAAO,OAAO,yBAAS,IAAI,MAAM,mBAAmB,CAAC;EAC5E,OAAO,cAAc,IAAI;CAC3B,CAAC;AACH;AAEA,SAAS,UAAU,KAAsB;CACvC,OAAO,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AACxD;AA4CA,SAAgB,mBAAmB,SAAsE;CACvG,IAAI,YAAY;CAChB,IAAI,YAAY;CAChB,IAAI,eAAe;CACnB,IAAI,UAAU;CAEd,SAAS,OAAa;EACpB,UAAU;GAAE;GAAW;GAAW;EAAa,CAAC;CAClD;CACA,SAAS,aAAa,OAAsB;EAC1C,IAAI,cAAc,OAAO;GACvB,YAAY;GACZ,KAAK;EACP;CACF;CACA,SAAS,aAAa,OAAsB;EAC1C,IAAI,cAAc,OAAO;GACvB,YAAY;GACZ,KAAK;EACP;CACF;CACA,SAAS,WAAW,OAAqB;EACvC,WAAW;EACX,MAAM,OAAO,UAAU;EACvB,IAAI,iBAAiB,MAAM;GACzB,eAAe;GACf,KAAK;EACP;CACF;CAEA,OAAO;EAAE;EAAc;EAAc;EAAY,mBAAmB;CAAU;AAChF;AAWA,SAAgB,yBAAyB,WAAkC,cAA4D;CACrI,IAAI,yBAAwC;CAE5C,SAAS,OAAa;EACpB,IAAI,2BAA2B,MAAM;GACnC,OAAO,cAAc,sBAAsB;GAC3C,yBAAyB;EAC3B;CACF;CAEA,eAAe,UAAyB;EACtC,IAAI;EACJ,IAAI;GACF,SAAS,MAAM,UAAU,UAAU;EACrC,QAAQ;GACN,aAAa,KAAK;GAClB,KAAK;GACL;EACF;EACA,aAAa,OAAO,KAAK;EACzB,IAAI,OAAO;OACL,2BAA2B,MAC7B,yBAAyB,OAAO,kBAAkB;IAChD,QAAa;GACf,GAAG,oBAAoB;EAAA,OAGzB,KAAK;CAET;CAEA,OAAO;EAAE;EAAS;CAAK;AACzB;AAmBA,SAAgB,mBAAmB,SAA4C;CAC7E,MAAM,EAAE,WAAW,UAAU,WAAW,YAAY,kBAAkB;CACtE,IAAI,QAAuB,QAAQ,QAAQ;CAE3C,eAAe,YAAY,MAAY,KAA4B;EACjE,IAAI,QAAQ,cAAc,GAAG;EAC7B,IAAI;GACF,MAAM,UAAU,MAAM,cAAc,IAAI;GACxC,MAAM,SAAS,MAAM,UAAU,WAAW,SAAS,SAAS,CAAC;GAC7D,IAAI,QAAQ,cAAc,GAAG;GAC7B,MAAM,OAAO,OAAO,KAAK,KAAK;GAC9B,IAAI,KAAK,WAAW,GAAG,UAAU,UAAU;QACtC,UAAU,aAAa,IAAI;EAClC,SAAS,KAAK;GAGZ,IAAI,QAAQ,cAAc,GAAG,UAAU,UAAU,UAAU,GAAG,CAAC;EACjE;CACF;CAIA,SAAS,QAAQ,MAAY,KAAmB;EAC9C,WAAW,CAAC;EACZ,QAAQ,MACL,WAAW,YAAY,MAAM,GAAG,CAAC,CAAC,CAClC,YAAY,KAAA,CAAS,CAAC,CACtB,cAAc,WAAW,EAAE,CAAC;CACjC;CAEA,OAAO,EAAE,QAAQ;AACnB;AAuBA,SAAS,mBAA+B;CACtC,IAAI,SAA6B;CACjC,IAAI,WAAgC;CACpC,IAAI,WAAgC;CACpC,IAAI,4BAAY,IAAI,aAAa,CAAC;CAElC,SAAS,OAAO,YAA+B;EAC7C,SAAS;EACT,WAAW,IAAI,aAAa;EAC5B,WAAW,SAAS,eAAe;EACnC,SAAS,UAAU;EACnB,YAAY,IAAI,aAAa,SAAS,OAAO;EAC7C,SAAS,wBAAwB,UAAU,CAAC,CAAC,QAAQ,QAAQ;CAC/D;CAEA,SAAS,SAAuB;EAC9B,UAAU,uBAAuB,SAAS;EAC1C,OAAO;CACT;CAEA,SAAS,WAAiB;EACxB,IAAI,UAAU;GACZ,SAAS,MAAM,CAAC,CAAC,YAAY,KAAA,CAAS;GACtC,WAAW;EACb;EACA,WAAW;EACX,QAAQ,UAAU,CAAC,CAAC,SAAS,UAAU,MAAM,KAAK,CAAC;EACnD,SAAS;CACX;CAEA,OAAO;EAAE,kBAAkB,aAAa;EAAM;EAAQ,iBAAiB;EAAQ;EAAQ;CAAS;AAClG;AAYA,eAAe,iBAAiB,WAAkC,mBAA2B,eAAkE;CAC7J,MAAM,WAAW,iBAAiB,KAAK;CACvC,IAAI,CAAC,YAAY,CAAC,UAAU,cAAc,cAAc;EACtD,UAAU,UAAU,aAAa;EACjC,OAAO;CACT;CACA,IAAI;CACJ,IAAI;EACF,WAAW,MAAM,UAAU,aAAa,aAAa,EAAE,OAAO,KAAK,CAAC;CACtE,QAAQ;EACN,UAAU,UAAU,mBAAmB;EACvC,OAAO;CACT;CACA,IAAI,sBAAsB,cAAc,GAAG;EACzC,SAAS,UAAU,CAAC,CAAC,SAAS,UAAU,MAAM,KAAK,CAAC;EACpD,OAAO;CACT;CACA,MAAM,aAAa,iBAAiB;CACpC,IAAI;EACF,WAAW,OAAO,QAAQ;CAC5B,SAAS,KAAK;EAIZ,WAAW,SAAS;EACpB,MAAM;CACR;CACA,OAAO;EAAE;EAAY;CAAS;AAChC;AAuBA,SAAS,sBAAsB,MAA4C;CACzE,IAAI,WAAiC;CACrC,IAAI,SAAiB,CAAC;CACtB,IAAI,mBAAmB;CACvB,IAAI,aAAa;CACjB,IAAI,kBAAkB;CAEtB,SAAS,SAAe;EAGtB,MAAM,YAAY;EAClB,MAAM,MAAM;EACZ,MAAM,OAAO,IAAI,KAAK,QAAQ,EAAE,MAAM,sBAAsB,eAAe,EAAE,CAAC;EAC9E,KAAK,aAAa,MAAM,WAAW,GAAG;CACxC;CAEA,SAAS,UAAU,QAAqB,UAAkB,KAAmB;EAC3E,SAAS,CAAC;EACV,mBAAmB;EACnB,aAAa;EACb,kBAAkB;EAClB,WAAW,IAAI,cAAc,QAAQ,EAAE,SAAS,CAAC;EACjD,SAAS,mBAAmB,UAAU;GACpC,IAAI,MAAM,KAAK,OAAO,GAAG,OAAO,KAAK,MAAM,IAAI;EACjD;EACA,SAAS,SAAS;EAClB,SAAS,MAAM;CACjB;CAEA,OAAO;EACL,mBAAmB,UAAU,UAAU;EACvC,iBAAiB;EACjB,kBAAkB;GAChB,mBAAmB;EACrB;EACA;EACA,WAAW;GACT,IAAI,UAAU,UAAU,aAAa,SAAS,KAAK;EACrD;CACF;AACF;AAKA,IAAM,kBAA4C;CAChD,YAAY;CACZ,eAAe;CACf,UAAU;CACV,cAAc;CACd,cAAc;CACd,YAAY;CACZ,eAAe;AACjB;AAwBA,SAAS,YAAY,SAAyB,OAA+B,SAAgC;CAC3G,QAAQ,cAAc;CACtB,MAAM,aAAa,KAAK;CACxB,IAAI,QAAQ,kBAAkB,MAAM;EAClC,OAAO,cAAc,QAAQ,aAAa;EAC1C,QAAQ,gBAAgB;CAC1B;CACA,QAAQ,IAAI;CACZ,QAAQ,YAAY,SAAS;CAC7B,QAAQ,aAAa;AACvB;AAKA,SAAS,YAAY,SAAyB,SAAgC;CAC5E,IAAI,CAAC,QAAQ,YAAY,WAAW,GAAG;CACvC,MAAM,MAAM,WAAW,QAAQ,WAAW,OAAO,CAAC;CAClD,MAAM,EACJ,WACA,cAAc,aACd,QACE,YAAY;EAAE,WAAW,QAAQ,UAAU;EAAG,cAAc,QAAQ;CAAa,GAAG,QAAQ,cAAc,KAAK,KAAK,IAAI,GAAG,UAAU;CACzI,IAAI,WAAW,QAAQ,WAAW;CAClC,QAAQ,eAAe;CACvB,IAAI,KAAK,QAAQ,IAAI;AACvB;AAEA,SAAgB,mBAAmB,WAAkC,UAAwB,WAAgD;CAC3I,MAAM,QAAQ,mBAAmB,UAAU,OAAO;CAClD,MAAM,SAAS,yBAAyB,WAAW,MAAM,YAAY;CACrE,MAAM,UAA0B,EAAE,GAAG,gBAAgB;CACrD,MAAM,WAAW,mBAAmB;EAAE;EAAW;EAAU;EAAW,YAAY,MAAM;EAAY,qBAAqB,QAAQ;CAAW,CAAC;CAE7I,MAAM,UAAU,sBAAsB,EAAE,cAAc,iBAAiB,CAAC;CAExE,SAAS,iBAAiB,MAAY,WAAoB,KAAmB;EAG3E,IAAI,MAAM,YAAY,GAAG,cAAc;EACvC,IAAI,aAAa,KAAK,OAAO,KAAK,QAAQ,QAAQ,YAAY,SAAS,QAAQ,MAAM,GAAG;CAC1F;CAEA,SAAS,gBAAsB;EAC7B,MAAM,SAAS,QAAQ,YAAY,UAAU;EAC7C,IAAI,CAAC,QAAQ;EACb,QAAQ,eAAe;EACvB,QAAQ,eAAe,KAAK,IAAI;EAChC,QAAQ,UAAU,QAAQ,QAAQ,UAAU,QAAQ,UAAU;CAChE;CAEA,eAAe,QAA0B;EACvC,IAAI,QAAQ,iBAAiB,MAAM,YAAY,GAAG,OAAO;EACzD,QAAQ,gBAAgB;EACxB,MAAM,WAAW,QAAQ;EACzB,IAAI;GACF,MAAM,UAAU,MAAM,iBAAiB,WAAW,gBAAgB,QAAQ,UAAU;GACpF,IAAI,CAAC,SAAS,OAAO;GACrB,QAAQ,aAAa,QAAQ;GAC7B,QAAQ,WAAW,QAAQ;GAC3B,MAAM,aAAa,IAAI;GACvB,cAAc;GACd,QAAQ,gBAAgB,OAAO,kBAAkB,YAAY,SAAS,OAAO,GAAG,mBAAmB;GACnG,OAAO;EACT,UAAU;GACR,QAAQ,gBAAgB;EAC1B;CACF;CAEA,SAAS,OAAa;EACpB,YAAY,SAAS,OAAO,OAAO;CACrC;CAEA,SAAS,UAAgB;EACvB,OAAO,KAAK;EACZ,KAAK;CACP;CAEA,OAAO;EAAE,qBAAqB,OAAO;EAAS;EAAO;EAAM;CAAQ;AACrE"}

package/dist/whisper/client.d.ts

@@ -32,4 +32,54 @@ export interface VoiceCapture {
     stop: () => void;
     dispose: () => void;
 }
+export interface CaptureStateController {
+    setAvailable: (value: boolean) => void;
+    setListening: (value: boolean) => void;
+    setPending: (delta: number) => void;
+    isListening: () => boolean;
+}
+export declare function createCaptureState(onState?: (state: VoiceCaptureState) => void): CaptureStateController;
+export interface AvailabilityPoller {
+    refresh: () => Promise<void>;
+    stop: () => void;
+}
+export declare function createAvailabilityPoller(transport: VoiceCaptureTransport, setAvailable: (value: boolean) => void): AvailabilityPoller;
+export interface SegmentQueueOptions {
+    transport: VoiceCaptureTransport;
+    language: () => string;
+    callbacks: VoiceCaptureCallbacks;
+    setPending: (delta: number) => void;
+    getGeneration: () => number;
+}
+export interface SegmentQueue {
+    enqueue: (blob: Blob, gen: number) => void;
+}
+export declare function createSegmentQueue(options: SegmentQueueOptions): SegmentQueue;
+export interface AudioGraph {
+    /** True between `attach()` and the next `teardown()`. */
+    isAttached: () => boolean;
+    /** Read the current PCM window into an internal buffer and return it.
+     *  Callers must not mutate the returned array. */
+    sample: () => Float32Array;
+    /** Live mic stream — the outer needs it to hand to MediaRecorder. */
+    getStream: () => MediaStream | null;
+    /** Attach a fresh mic stream: create AudioContext + AnalyserNode + VAD buffer,
+     *  connect the source. Called only by `bootstrapCapture` after the caller has
+     *  passed the stale-generation / permission checks. */
+    attach: (stream: MediaStream) => void;
+    /** Close the AudioContext, stop the mic tracks, null everything. Safe to
+     *  call before `attach()` or twice in a row. */
+    teardown: () => void;
+}
+export interface RecorderSession {
+    isRecording: () => boolean;
+    hadSpeech: () => boolean;
+    markSpeech: () => void;
+    /** Start a new segment on `stream`, tagged with `gen`. Resets per-segment
+     *  state (chunks + hadSpeech) synchronously so a subsequent VAD tick sees
+     *  a fresh window. */
+    startNext: (stream: MediaStream, mimeType: string, gen: number) => void;
+    /** Force-close the current segment (silence exceeded or MAX_SEGMENT_MS). */
+    cut: () => void;
+}
 export declare function createVoiceCapture(transport: VoiceCaptureTransport, language: () => string, callbacks: VoiceCaptureCallbacks): VoiceCapture;