@mulmoclaude/core 0.1.0

package/dist/whisper/client.cjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"client.cjs","names":[],"sources":["../../src/whisper/client.ts"],"sourcesContent":["// @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/feat-extract-whisper-package.md.\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;\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 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\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 containerType(): string {\n    return mimeType.split(\";\")[0] || \"audio/webm\";\n  }\n\n  function onSegmentStop(): void {\n    const hadSpeech = segmentHasSpeech;\n    const gen = segmentGeneration;\n    const blob = new Blob(chunks, { type: containerType() });\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 now = Date.now();\n    if (rms > SPEECH_RMS) {\n      segmentHasSpeech = true;\n      silenceStart = null;\n    } else if (segmentHasSpeech) {\n      if (silenceStart === null) silenceStart = now;\n      else if (now - silenceStart > SILENCE_MS) cutSegment();\n    }\n    if (segmentHasSpeech && now - segmentStart > MAX_SEGMENT_MS) 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":";;AAQA,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;AAE7B,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,WAAW,QAA8B;CAChD,IAAI,MAAM;CACV,KAAK,MAAM,UAAU,QAAQ,OAAO,SAAS;CAC7C,OAAO,KAAK,KAAK,MAAM,OAAO,MAAM;AACtC;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,YAAY,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,EACjC,YAAY,KAAA,CAAS,EACrB,cAAc,WAAW,EAAE,CAAC;CACjC;CAEA,SAAS,gBAAwB;EAC/B,OAAO,SAAS,MAAM,GAAG,EAAE,MAAM;CACnC;CAEA,SAAS,gBAAsB;EAC7B,MAAM,YAAY;EAClB,MAAM,MAAM;EACZ,MAAM,OAAO,IAAI,KAAK,QAAQ,EAAE,MAAM,cAAc,EAAE,CAAC;EACvD,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,MAAM,KAAK,IAAI;EACrB,IAAI,MAAM,YAAY;GACpB,mBAAmB;GACnB,eAAe;EACjB,OAAO,IAAI;OACL,iBAAiB,MAAM,eAAe;QACrC,IAAI,MAAM,eAAe,YAAY,WAAW;EAAA;EAEvD,IAAI,oBAAoB,MAAM,eAAe,gBAAgB,WAAW;CAC1E;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,EAAE,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,EAAE,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,EAAE,YAAY,KAAA,CAAS;GACtC,WAAW;EACb;EACA,WAAW;EACX,QAAQ,UAAU,EAAE,SAAS,UAAU,MAAM,KAAK,CAAC;EACnD,SAAS;CACX;CAEA,SAAS,UAAgB;EACvB,qBAAqB;EACrB,KAAK;CACP;CAEA,OAAO;EAAE;EAAqB;EAAO;EAAM;CAAQ;AACrD"}

package/dist/whisper/client.d.ts

@@ -0,0 +1,35 @@
+export declare function localeToWhisperLanguage(locale: string): string;
+export interface VoiceCaptureTransport {
+    /** Transcribe one segment. Throws on failure. */
+    transcribe: (dataUrl: string, language: string) => Promise<{
+        text: string;
+    }>;
+    /** Current readiness. `downloading` true keeps the controller polling so it
+     *  flips to ready as soon as a model download finishes. */
+    getStatus: () => Promise<{
+        ready: boolean;
+        downloading: boolean;
+    }>;
+}
+export interface VoiceCaptureState {
+    available: boolean;
+    listening: boolean;
+    transcribing: boolean;
+}
+export interface VoiceCaptureCallbacks {
+    /** A recognized (non-empty) segment transcript. */
+    onTranscript: (text: string) => void;
+    /** A segment produced no speech. */
+    onEmpty?: () => void;
+    /** A recoverable error message (transport failure, permission denied, etc.). */
+    onError?: (message: string) => void;
+    /** Pushed whenever available/listening/transcribing changes. */
+    onState?: (state: VoiceCaptureState) => void;
+}
+export interface VoiceCapture {
+    refreshAvailability: () => Promise<void>;
+    start: () => Promise<boolean>;
+    stop: () => void;
+    dispose: () => void;
+}
+export declare function createVoiceCapture(transport: VoiceCaptureTransport, language: () => string, callbacks: VoiceCaptureCallbacks): VoiceCapture;

package/dist/whisper/client.js

@@ -0,0 +1,239 @@
+//#region src/whisper/client.ts
+var LOCALE_TO_WHISPER = {
+	en: "en",
+	ja: "ja",
+	zh: "zh",
+	ko: "ko",
+	es: "es",
+	"pt-BR": "pt",
+	fr: "fr",
+	de: "de"
+};
+function localeToWhisperLanguage(locale) {
+	return LOCALE_TO_WHISPER[locale] ?? "auto";
+}
+var SPEECH_RMS = .015;
+var SILENCE_MS = 800;
+var MAX_SEGMENT_MS = 2e4;
+var MONITOR_INTERVAL_MS = 100;
+var AVAILABILITY_POLL_MS = 2e3;
+function pickRecorderMime() {
+	const candidates = [
+		"audio/webm;codecs=opus",
+		"audio/webm",
+		"audio/mp4"
+	];
+	if (typeof MediaRecorder === "undefined") return void 0;
+	return candidates.find((type) => MediaRecorder.isTypeSupported(type));
+}
+function blobToDataUrl(blob) {
+	return new Promise((resolve, reject) => {
+		const reader = new FileReader();
+		reader.onload = () => resolve(String(reader.result));
+		reader.onerror = () => reject(reader.error ?? /* @__PURE__ */ new Error("FileReader failed"));
+		reader.readAsDataURL(blob);
+	});
+}
+function computeRms(buffer) {
+	let sum = 0;
+	for (const sample of buffer) sum += sample * sample;
+	return Math.sqrt(sum / buffer.length);
+}
+function toMessage(err) {
+	return err instanceof Error ? err.message : String(err);
+}
+function createVoiceCapture(transport, language, callbacks) {
+	let available = false;
+	let listening = false;
+	let transcribing = false;
+	function emit() {
+		callbacks.onState?.({
+			available,
+			listening,
+			transcribing
+		});
+	}
+	function setAvailable(value) {
+		if (available !== value) {
+			available = value;
+			emit();
+		}
+	}
+	function setListening(value) {
+		if (listening !== value) {
+			listening = value;
+			emit();
+		}
+	}
+	let stream = null;
+	let audioCtx = null;
+	let analyser = null;
+	let vadBuffer = 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;
+		if (transcribing !== next) {
+			transcribing = next;
+			emit();
+		}
+	}
+	function stopAvailabilityPoll() {
+		if (availabilityPollHandle !== null) {
+			window.clearInterval(availabilityPollHandle);
+			availabilityPollHandle = null;
+		}
+	}
+	async function refreshAvailability() {
+		let status;
+		try {
+			status = await transport.getStatus();
+		} catch {
+			setAvailable(false);
+			stopAvailabilityPoll();
+			return;
+		}
+		setAvailable(status.ready);
+		if (status.downloading) {
+			if (availabilityPollHandle === null) availabilityPollHandle = window.setInterval(() => {
+				refreshAvailability();
+			}, AVAILABILITY_POLL_MS);
+		} else stopAvailabilityPoll();
+	}
+	async function sendSegment(blob, gen) {
+		if (gen !== generation) return;
+		try {
+			const dataUrl = await blobToDataUrl(blob);
+			const result = await transport.transcribe(dataUrl, language());
+			if (gen !== generation) 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));
+		}
+	}
+	function enqueue(blob, gen) {
+		setPending(1);
+		queue = queue.then(() => sendSegment(blob, gen)).catch(() => void 0).finally(() => setPending(-1));
+	}
+	function containerType() {
+		return mimeType.split(";")[0] || "audio/webm";
+	}
+	function onSegmentStop() {
+		const hadSpeech = segmentHasSpeech;
+		const gen = segmentGeneration;
+		const blob = new Blob(chunks, { type: containerType() });
+		if (listening) startRecorder();
+		if (hadSpeech && blob.size > 0 && gen === generation) enqueue(blob, gen);
+	}
+	function startRecorder() {
+		if (!stream) return;
+		chunks = [];
+		segmentHasSpeech = false;
+		silenceStart = null;
+		segmentStart = Date.now();
+		segmentGeneration = generation;
+		recorder = new MediaRecorder(stream, { mimeType });
+		recorder.ondataavailable = (event) => {
+			if (event.data.size > 0) chunks.push(event.data);
+		};
+		recorder.onstop = onSegmentStop;
+		recorder.start();
+	}
+	function cutSegment() {
+		if (recorder && recorder.state === "recording") recorder.stop();
+	}
+	function monitorTick() {
+		if (!analyser) return;
+		analyser.getFloatTimeDomainData(vadBuffer);
+		const rms = computeRms(vadBuffer);
+		const now = Date.now();
+		if (rms > SPEECH_RMS) {
+			segmentHasSpeech = true;
+			silenceStart = null;
+		} else if (segmentHasSpeech) {
+			if (silenceStart === null) silenceStart = now;
+			else if (now - silenceStart > SILENCE_MS) cutSegment();
+		}
+		if (segmentHasSpeech && now - segmentStart > MAX_SEGMENT_MS) cutSegment();
+	}
+	async function start() {
+		if (startInFlight || listening) return true;
+		startInFlight = true;
+		const startGen = 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);
+			startRecorder();
+			monitorHandle = window.setInterval(monitorTick, MONITOR_INTERVAL_MS);
+			return true;
+		} finally {
+			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;
+	}
+	function dispose() {
+		stopAvailabilityPoll();
+		stop();
+	}
+	return {
+		refreshAvailability,
+		start,
+		stop,
+		dispose
+	};
+}
+//#endregion
+export { createVoiceCapture, localeToWhisperLanguage };
+
+//# sourceMappingURL=client.js.map

package/dist/whisper/client.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"client.js","names":[],"sources":["../../src/whisper/client.ts"],"sourcesContent":["// @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/feat-extract-whisper-package.md.\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;\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 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\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 containerType(): string {\n    return mimeType.split(\";\")[0] || \"audio/webm\";\n  }\n\n  function onSegmentStop(): void {\n    const hadSpeech = segmentHasSpeech;\n    const gen = segmentGeneration;\n    const blob = new Blob(chunks, { type: containerType() });\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 now = Date.now();\n    if (rms > SPEECH_RMS) {\n      segmentHasSpeech = true;\n      silenceStart = null;\n    } else if (segmentHasSpeech) {\n      if (silenceStart === null) silenceStart = now;\n      else if (now - silenceStart > SILENCE_MS) cutSegment();\n    }\n    if (segmentHasSpeech && now - segmentStart > MAX_SEGMENT_MS) 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":";AAQA,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;AAE7B,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,WAAW,QAA8B;CAChD,IAAI,MAAM;CACV,KAAK,MAAM,UAAU,QAAQ,OAAO,SAAS;CAC7C,OAAO,KAAK,KAAK,MAAM,OAAO,MAAM;AACtC;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,YAAY,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,EACjC,YAAY,KAAA,CAAS,EACrB,cAAc,WAAW,EAAE,CAAC;CACjC;CAEA,SAAS,gBAAwB;EAC/B,OAAO,SAAS,MAAM,GAAG,EAAE,MAAM;CACnC;CAEA,SAAS,gBAAsB;EAC7B,MAAM,YAAY;EAClB,MAAM,MAAM;EACZ,MAAM,OAAO,IAAI,KAAK,QAAQ,EAAE,MAAM,cAAc,EAAE,CAAC;EACvD,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,MAAM,KAAK,IAAI;EACrB,IAAI,MAAM,YAAY;GACpB,mBAAmB;GACnB,eAAe;EACjB,OAAO,IAAI;OACL,iBAAiB,MAAM,eAAe;QACrC,IAAI,MAAM,eAAe,YAAY,WAAW;EAAA;EAEvD,IAAI,oBAAoB,MAAM,eAAe,gBAAgB,WAAW;CAC1E;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,EAAE,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,EAAE,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,EAAE,YAAY,KAAA,CAAS;GACtC,WAAW;EACb;EACA,WAAW;EACX,QAAQ,UAAU,EAAE,SAAS,UAAU,MAAM,KAAK,CAAC;EACnD,SAAS;CACX;CAEA,SAAS,UAAgB;EACvB,qBAAqB;EACrB,KAAK;CACP;CAEA,OAAO;EAAE;EAAqB;EAAO;EAAM;CAAQ;AACrD"}

package/dist/whisper/ffmpeg.d.ts

@@ -0,0 +1,6 @@
+/** ffmpeg args to decode any input to 16 kHz mono signed-16-bit WAV. Pure +
+ *  exported for unit tests. */
+export declare function buildWav16kArgs(inputPath: string, outputPath: string): string[];
+/** Convert `inputPath` to a 16 kHz mono WAV at `outputPath`. Throws on ffmpeg
+ *  failure or timeout. */
+export declare function convertToWav16k(inputPath: string, outputPath: string, ffmpegBinary?: string): Promise<void>;