@ai-ecoverse/pi-audio 1.3.0 → 1.5.0

package/dist/transcription.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["../src/transcription.ts"],
-  "sourcesContent": ["import type { ProviderRequestOptions, Usage } from '@earendil-works/pi-ai';\nimport { EventStream } from '@earendil-works/pi-ai';\nimport type { AudioSource } from './audio.ts';\nimport { blankUsage } from './blank-usage.ts';\nimport { VoiceRequestError } from './errors.ts';\nimport type { TranscriptionModel } from './models.ts';\nimport { streamResult } from './stream-result.ts';\n\nexport type AudioInput = { data: Uint8Array; mimeType: string } | { url: string };\n\nexport interface TranscriptSegment {\n  text: string;\n  start: number;\n  end: number;\n  speaker?: string;\n}\n\nexport interface Transcript {\n  text: string;\n  language?: string;\n  segments?: TranscriptSegment[];\n  usage: Usage;\n}\n\nexport type TranscriptionEvent =\n  | { type: 'speech_start'; at: number }\n  | { type: 'partial'; segmentId: string; text: string }\n  | { type: 'final'; segmentId: string; text: string; confidence?: number }\n  | { type: 'turn_end'; text: string; at: number }\n  | { type: 'done'; transcript: Transcript }\n  | { type: 'error'; reason: 'aborted' | 'error'; error: string };\n\nexport interface ListenOptions extends ProviderRequestOptions {\n  language?: string;\n  vocabulary?: string[];\n}\n\nfunction blankTranscript(): Transcript {\n  return { text: '', usage: blankUsage() };\n}\n\nexport class TranscriptionEventStream extends EventStream<TranscriptionEvent, Transcript> {\n  readonly #settled = streamResult<Transcript>();\n\n  constructor() {\n    super(\n      (event) => event.type === 'done' || event.type === 'error',\n      (event) => (event.type === 'done' ? event.transcript : blankTranscript())\n    );\n  }\n\n  override push(event: TranscriptionEvent): void {\n    super.push(event);\n    if (event.type === 'done') this.#settled.finish(event.transcript);\n    else if (event.type === 'error')\n      this.#settled.fail(new VoiceRequestError(event.reason, event.error));\n  }\n\n  override end(result?: Transcript): void {\n    super.end(result);\n    if (result !== undefined) this.#settled.finish(result);\n    else\n      this.#settled.fail(\n        new VoiceRequestError('error', 'Transcription ended without a transcript')\n      );\n  }\n\n  override result(): Promise<Transcript> {\n    return this.#settled.result;\n  }\n}\n\nexport interface TranscriptionApiProvider<TApi extends string = string> {\n  api: TApi;\n  transcribe?: (\n    model: TranscriptionModel<TApi>,\n    input: AudioInput,\n    options?: ProviderRequestOptions\n  ) => Promise<Transcript>;\n  listen?: (\n    model: TranscriptionModel<TApi>,\n    audio: AudioSource,\n    options?: ListenOptions\n  ) => TranscriptionEventStream;\n}\n"],
-  "mappings": "AACA,SAAS,mBAAmB;AAE5B,SAAS,kBAAkB;AAC3B,SAAS,yBAAyB;AAElC,SAAS,oBAAoB;AA+B7B,SAAS,kBAA8B;AACrC,SAAO,EAAE,MAAM,IAAI,OAAO,WAAW,EAAE;AACzC;AAEO,MAAM,iCAAiC,YAA4C;AAAA,EAC/E,WAAW,aAAyB;AAAA,EAE7C,cAAc;AACZ;AAAA,MACE,CAAC,UAAU,MAAM,SAAS,UAAU,MAAM,SAAS;AAAA,MACnD,CAAC,UAAW,MAAM,SAAS,SAAS,MAAM,aAAa,gBAAgB;AAAA,IACzE;AAAA,EACF;AAAA,EAES,KAAK,OAAiC;AAC7C,UAAM,KAAK,KAAK;AAChB,QAAI,MAAM,SAAS,OAAQ,MAAK,SAAS,OAAO,MAAM,UAAU;AAAA,aACvD,MAAM,SAAS;AACtB,WAAK,SAAS,KAAK,IAAI,kBAAkB,MAAM,QAAQ,MAAM,KAAK,CAAC;AAAA,EACvE;AAAA,EAES,IAAI,QAA2B;AACtC,UAAM,IAAI,MAAM;AAChB,QAAI,WAAW,OAAW,MAAK,SAAS,OAAO,MAAM;AAAA;AAEnD,WAAK,SAAS;AAAA,QACZ,IAAI,kBAAkB,SAAS,0CAA0C;AAAA,MAC3E;AAAA,EACJ;AAAA,EAES,SAA8B;AACrC,WAAO,KAAK,SAAS;AAAA,EACvB;AACF;",
+  "sourcesContent": ["import type { ProviderRequestOptions, Usage } from '@earendil-works/pi-ai';\nimport { EventStream } from '@earendil-works/pi-ai';\nimport type { AudioSource } from './audio.ts';\nimport { blankUsage } from './blank-usage.ts';\nimport { VoiceRequestError } from './errors.ts';\nimport type { TranscriptionModel } from './models.ts';\nimport { streamResult } from './stream-result.ts';\nimport type { VadHook } from './vad.ts';\n\nexport type AudioInput = { data: Uint8Array; mimeType: string } | { url: string };\n\nexport interface TranscriptSegment {\n  text: string;\n  start: number;\n  end: number;\n  speaker?: string;\n}\n\nexport interface Transcript {\n  text: string;\n  language?: string;\n  segments?: TranscriptSegment[];\n  usage: Usage;\n}\n\nexport type TranscriptionEvent =\n  | { type: 'speech_start'; at: number }\n  | { type: 'partial'; segmentId: string; text: string }\n  | { type: 'final'; segmentId: string; text: string; confidence?: number }\n  | { type: 'turn_end'; text: string; at: number }\n  | { type: 'done'; transcript: Transcript }\n  | { type: 'error'; reason: 'aborted' | 'error'; error: string };\n\nexport interface ListenOptions extends ProviderRequestOptions {\n  language?: string;\n  vocabulary?: string[];\n  vad?: VadHook;\n}\n\nfunction blankTranscript(): Transcript {\n  return { text: '', usage: blankUsage() };\n}\n\nexport class TranscriptionEventStream extends EventStream<TranscriptionEvent, Transcript> {\n  readonly #settled = streamResult<Transcript>();\n\n  constructor() {\n    super(\n      (event) => event.type === 'done' || event.type === 'error',\n      (event) => (event.type === 'done' ? event.transcript : blankTranscript())\n    );\n  }\n\n  override push(event: TranscriptionEvent): void {\n    super.push(event);\n    if (event.type === 'done') this.#settled.finish(event.transcript);\n    else if (event.type === 'error')\n      this.#settled.fail(new VoiceRequestError(event.reason, event.error));\n  }\n\n  override end(result?: Transcript): void {\n    super.end(result);\n    if (result !== undefined) this.#settled.finish(result);\n    else\n      this.#settled.fail(\n        new VoiceRequestError('error', 'Transcription ended without a transcript')\n      );\n  }\n\n  override result(): Promise<Transcript> {\n    return this.#settled.result;\n  }\n}\n\nexport interface TranscriptionApiProvider<TApi extends string = string> {\n  api: TApi;\n  transcribe?: (\n    model: TranscriptionModel<TApi>,\n    input: AudioInput,\n    options?: ProviderRequestOptions\n  ) => Promise<Transcript>;\n  listen?: (\n    model: TranscriptionModel<TApi>,\n    audio: AudioSource,\n    options?: ListenOptions\n  ) => TranscriptionEventStream;\n}\n"],
+  "mappings": "AACA,SAAS,mBAAmB;AAE5B,SAAS,kBAAkB;AAC3B,SAAS,yBAAyB;AAElC,SAAS,oBAAoB;AAiC7B,SAAS,kBAA8B;AACrC,SAAO,EAAE,MAAM,IAAI,OAAO,WAAW,EAAE;AACzC;AAEO,MAAM,iCAAiC,YAA4C;AAAA,EAC/E,WAAW,aAAyB;AAAA,EAE7C,cAAc;AACZ;AAAA,MACE,CAAC,UAAU,MAAM,SAAS,UAAU,MAAM,SAAS;AAAA,MACnD,CAAC,UAAW,MAAM,SAAS,SAAS,MAAM,aAAa,gBAAgB;AAAA,IACzE;AAAA,EACF;AAAA,EAES,KAAK,OAAiC;AAC7C,UAAM,KAAK,KAAK;AAChB,QAAI,MAAM,SAAS,OAAQ,MAAK,SAAS,OAAO,MAAM,UAAU;AAAA,aACvD,MAAM,SAAS;AACtB,WAAK,SAAS,KAAK,IAAI,kBAAkB,MAAM,QAAQ,MAAM,KAAK,CAAC;AAAA,EACvE;AAAA,EAES,IAAI,QAA2B;AACtC,UAAM,IAAI,MAAM;AAChB,QAAI,WAAW,OAAW,MAAK,SAAS,OAAO,MAAM;AAAA;AAEnD,WAAK,SAAS;AAAA,QACZ,IAAI,kBAAkB,SAAS,0CAA0C;AAAA,MAC3E;AAAA,EACJ;AAAA,EAES,SAA8B;AACrC,WAAO,KAAK,SAAS;AAAA,EACvB;AACF;",
   "names": []
 }

package/dist/types/chunk-speech.d.ts

@@ -0,0 +1,4 @@
+import type { ProviderRequestOptions } from '@earendil-works/pi-ai';
+import type { SpeechModel } from './models.ts';
+import { type SpeechApiProvider, SpeechEventStream, type SpeechInput } from './speech.ts';
+export declare function speakInSentences(speak: SpeechApiProvider['speak'], model: SpeechModel<string>, input: SpeechInput, options?: ProviderRequestOptions): SpeechEventStream;

package/dist/types/index.d.ts

@@ -13,3 +13,5 @@ export type { AudioInput, ListenOptions, Transcript, TranscriptionApiProvider, T
 export { TranscriptionEventStream } from './transcription.ts';
 export type { AudioCost, AudioCounters, AudioUsage, RealtimeAudioCost, SpeechAudioCost, TranscriptionAudioCost, } from './usage.ts';
 export { addAudioCounters, priceAudio } from './usage.ts';
+export type { VadHook, VadUpdate } from './vad.ts';
+export { energyVad } from './vad.ts';

package/dist/types/openai/catalog.d.ts

@@ -0,0 +1 @@
+export declare function registerOpenAI(): () => void;

package/dist/types/openai/index.d.ts

@@ -0,0 +1 @@
+export { registerOpenAI } from './catalog.ts';

package/dist/types/openai/listen.d.ts

@@ -0,0 +1,4 @@
+import type { AudioChunk } from '../audio.ts';
+import type { TranscriptionModel } from '../models.ts';
+import { type ListenOptions, TranscriptionEventStream } from '../transcription.ts';
+export declare function listenOpenAI(model: TranscriptionModel<string>, audio: AsyncIterable<AudioChunk>, options?: ListenOptions): TranscriptionEventStream;

package/dist/types/openai/speak.d.ts

@@ -0,0 +1,4 @@
+import type { ProviderRequestOptions } from '@earendil-works/pi-ai';
+import type { SpeechModel } from '../models.ts';
+import { SpeechEventStream, type SpeechInput } from '../speech.ts';
+export declare function speakOpenAI(model: SpeechModel<string>, input: SpeechInput, options?: ProviderRequestOptions): SpeechEventStream;

package/dist/types/openai/transcribe.d.ts

@@ -0,0 +1,4 @@
+import type { ProviderRequestOptions } from '@earendil-works/pi-ai';
+import type { TranscriptionModel } from '../models.ts';
+import type { AudioInput, Transcript } from '../transcription.ts';
+export declare function transcribeOpenAI(model: TranscriptionModel<string>, input: AudioInput, options?: ProviderRequestOptions): Promise<Transcript>;

package/dist/types/openai/wire.d.ts

@@ -0,0 +1,52 @@
+import type { ProviderRequestOptions } from '@earendil-works/pi-ai';
+export interface AudioFile {
+    data: Uint8Array;
+    mimeType: string;
+    filename: string;
+    format: string;
+}
+export interface JsonBody {
+    [key: string]: unknown;
+}
+interface HeaderModel {
+    provider: string;
+    headers?: Record<string, string>;
+}
+export declare function isOpenRouter(model: {
+    provider: string;
+    baseUrl: string;
+}): boolean;
+export declare function endpoint(baseUrl: string, path: string): string;
+export declare function realtimeUrl(baseUrl: string): string;
+export declare function audioFile(mimeType: string): {
+    format: string;
+    filename: string;
+};
+export declare function isRecord(value: unknown): value is Record<string, unknown>;
+export declare function stringField(record: Record<string, unknown>, key: string): string | undefined;
+export declare function numberField(record: Record<string, unknown>, key: string): number | undefined;
+export declare function asVoiceError(error: unknown): {
+    reason: 'aborted' | 'error';
+    error: string;
+};
+export declare function hasAuth(headers: Record<string, string | null> | undefined): boolean;
+export declare function requestHeaders(model: HeaderModel, options: ProviderRequestOptions | undefined): Record<string, string>;
+export declare function requestSignal(options: ProviderRequestOptions | undefined): AbortSignal | undefined;
+export declare function encodeMultipart(file: AudioFile, fields: Record<string, string>): Uint8Array;
+export declare function multipartType(): string;
+export declare function readAudio(input: {
+    data: Uint8Array;
+    mimeType: string;
+} | {
+    url: string;
+}, fetchImpl: typeof fetch, signal: AbortSignal | undefined): Promise<{
+    data: Uint8Array;
+    mimeType: string;
+}>;
+export declare function audioPayload(file: AudioFile, openRouter: boolean, modelId: string): JsonBody;
+export declare function applyPayload(payload: unknown, model: unknown, options: ProviderRequestOptions | undefined): Promise<unknown>;
+export declare function voiceFetch(model: HeaderModel & {
+    id: string;
+}, options: ProviderRequestOptions | undefined, url: string, body: BodyInit, contentType: string): Promise<Response>;
+export declare function readJson(response: Response): Promise<unknown>;
+export {};

package/dist/types/sentences.d.ts

@@ -0,0 +1,5 @@
+export declare function splitSentences(buffer: string, flush: boolean): {
+    sentences: string[];
+    rest: string;
+};
+export declare function sentencesOf(text: string | AsyncIterable<string>): AsyncGenerator<string>;

package/dist/types/transcription.d.ts

@@ -2,6 +2,7 @@ import type { ProviderRequestOptions, Usage } from '@earendil-works/pi-ai';
 import { EventStream } from '@earendil-works/pi-ai';
 import type { AudioSource } from './audio.ts';
 import type { TranscriptionModel } from './models.ts';
+import type { VadHook } from './vad.ts';
 export type AudioInput = {
     data: Uint8Array;
     mimeType: string;
@@ -47,6 +48,7 @@ export type TranscriptionEvent = {
 export interface ListenOptions extends ProviderRequestOptions {
     language?: string;
     vocabulary?: string[];
+    vad?: VadHook;
 }
 export declare class TranscriptionEventStream extends EventStream<TranscriptionEvent, Transcript> {
     #private;

package/dist/types/vad-listen.d.ts

@@ -0,0 +1,7 @@
+import type { AudioSource } from './audio.ts';
+import type { TranscriptionModel } from './models.ts';
+import { type ListenOptions, TranscriptionEventStream } from './transcription.ts';
+import { type VadHook } from './vad.ts';
+type ListenModel = (model: TranscriptionModel<string>, audio: AudioSource, options?: ListenOptions) => TranscriptionEventStream;
+export declare function listenWithVad(listenModel: ListenModel, model: TranscriptionModel<string>, audio: AudioSource, options?: ListenOptions, vad?: VadHook): TranscriptionEventStream;
+export {};

package/dist/types/vad.d.ts

@@ -0,0 +1,13 @@
+import type { AudioChunk } from './audio.ts';
+export interface VadUpdate {
+    speechStart?: number;
+    turnEnd?: number;
+}
+export interface VadHook {
+    push(chunk: AudioChunk): VadUpdate | undefined;
+}
+export interface EnergyVadOptions {
+    threshold?: number;
+    silenceMs?: number;
+}
+export declare function energyVad(options?: EnergyVadOptions): VadHook;

package/dist/vad-listen.js

@@ -0,0 +1,223 @@
+import {
+  TranscriptionEventStream
+} from "./transcription.js";
+import { energyVad } from "./vad.js";
+function tick() {
+  return new Promise((resolve) => {
+    setTimeout(resolve, 0);
+  });
+}
+function audioGate() {
+  const queued = [];
+  let resolveWait;
+  const source = {
+    async *[Symbol.asyncIterator]() {
+      while (true) {
+        const ready = queued.shift();
+        if (ready !== void 0) {
+          if (ready.chunk === null) return;
+          yield ready.chunk;
+          continue;
+        }
+        const chunk = await new Promise((resolve) => {
+          resolveWait = resolve;
+        });
+        if (chunk === null) return;
+        yield chunk;
+      }
+    }
+  };
+  return {
+    source,
+    push(chunk) {
+      if (resolveWait) {
+        const resolve = resolveWait;
+        resolveWait = void 0;
+        resolve(chunk);
+        return;
+      }
+      queued.push({ chunk });
+    }
+  };
+}
+function joinSegments(parts) {
+  let text = "";
+  for (const part of parts) {
+    if (text.length === 0 || /\s$/.test(text) || /^\s/.test(part)) text += part;
+    else text += ` ${part}`;
+  }
+  return text;
+}
+function liveText(session) {
+  const parts = [...session.segments.values()];
+  if (parts.length > 0) return joinSegments(parts);
+  return session.pendingDone?.transcript.text ?? session.providerTurn ?? "";
+}
+function turnText(session) {
+  const live = liveText(session);
+  if (session.closed.length > 0 && live.startsWith(session.closed)) {
+    return live.slice(session.closed.length).trim();
+  }
+  if (session.providerTurn && !live.startsWith(session.providerTurn)) return session.providerTurn;
+  return live.trim();
+}
+function closeTurn(session, at) {
+  if (!session.turnOpen || session.stopped) return;
+  const text = turnText(session);
+  if (text.length === 0) return;
+  session.closed = liveText(session);
+  session.turnOpen = false;
+  session.started = false;
+  session.pendingEnd = void 0;
+  session.heldEnd = void 0;
+  session.providerTurn = void 0;
+  session.stream.push({ type: "turn_end", text, at });
+}
+function openTurn(session, at) {
+  if (session.started || session.stopped) return;
+  session.stream.push({ type: "speech_start", at });
+  session.started = true;
+  session.turnOpen = true;
+  session.pendingStart = void 0;
+}
+function applyEvent(session, event) {
+  if (event.type === "speech_start") {
+    if (!session.started) session.stream.push(event);
+    session.started = true;
+    session.turnOpen = true;
+    session.pendingStart = void 0;
+    return;
+  }
+  if (event.type === "turn_end") {
+    session.providerTurn = event.text;
+    session.pendingEnd = event.at;
+    return;
+  }
+  if (event.type === "partial" || event.type === "final") {
+    if (!session.started) openTurn(session, session.pendingStart ?? session.lastAt);
+    session.segments.set(event.segmentId, event.text);
+    session.stream.push(event);
+    return;
+  }
+  if (event.type === "error") {
+    session.stream.push(event);
+    session.stopped = true;
+    return;
+  }
+  if (event.type === "done") session.pendingDone = event;
+}
+function absorb(session, events, update) {
+  if (session.stopped) return;
+  if (update?.speechStart !== void 0) session.pendingStart = update.speechStart;
+  if (update?.turnEnd !== void 0) session.pendingEnd = update.turnEnd;
+  const textual = events.some(
+    (event) => event.type === "partial" || event.type === "final" || event.type === "done"
+  );
+  if (session.pendingStart !== void 0 && (textual || events.length === 0)) {
+    openTurn(session, session.pendingStart);
+  }
+  for (const event of events) {
+    applyEvent(session, event);
+    if (session.stopped) return;
+  }
+  if (session.pendingEnd !== void 0) {
+    session.heldEnd = session.pendingEnd;
+    session.pendingEnd = void 0;
+  }
+  if (session.heldEnd !== void 0) closeTurn(session, session.heldEnd);
+}
+function finish(session) {
+  if (session.stopped) return;
+  if (session.pendingStart !== void 0) openTurn(session, session.pendingStart);
+  if (session.turnOpen) closeTurn(session, session.heldEnd ?? session.pendingEnd ?? session.lastAt);
+  if (session.pendingDone) session.stream.push(session.pendingDone);
+  else {
+    session.stream.push({
+      type: "error",
+      reason: "error",
+      error: "Transcription ended without a transcript"
+    });
+  }
+}
+function listenWithVad(listenModel, model, audio, options, vad = options?.vad ?? energyVad()) {
+  const stream = new TranscriptionEventStream();
+  void pump(listenModel, model, audio, options, stream, vad);
+  return stream;
+}
+async function pump(listenModel, model, audio, options, stream, vad) {
+  if (options?.signal?.aborted) {
+    stream.push({ type: "error", reason: "aborted", error: "aborted" });
+    return;
+  }
+  const gate = audioGate();
+  let inner;
+  try {
+    inner = listenModel(model, gate.source, options);
+  } catch (error) {
+    const message = error instanceof Error ? error.message : String(error);
+    stream.push({ type: "error", reason: "error", error: message });
+    return;
+  }
+  const inbox = [];
+  let readerClosed = false;
+  const reading = (async () => {
+    for await (const event of inner) inbox.push(event);
+    readerClosed = true;
+  })();
+  const session = {
+    stream,
+    segments: /* @__PURE__ */ new Map(),
+    closed: "",
+    turnOpen: false,
+    started: false,
+    stopped: false,
+    lastAt: 0
+  };
+  const take = (update) => {
+    const events = inbox.splice(0, inbox.length);
+    absorb(session, events, update);
+  };
+  try {
+    await tick();
+    take();
+    for await (const chunk of audio) {
+      if (session.stopped) break;
+      if (options?.signal?.aborted) {
+        gate.push(null);
+        await tick();
+        take();
+        if (!session.stopped) {
+          stream.push({ type: "error", reason: "aborted", error: "aborted" });
+          session.stopped = true;
+        }
+        break;
+      }
+      session.lastAt = chunk.timestamp;
+      const update = vad.push(chunk);
+      gate.push(chunk);
+      await tick();
+      take(update);
+    }
+    gate.push(null);
+    if (!session.stopped) {
+      let attempt = 0;
+      do {
+        await tick();
+        take();
+        attempt += 1;
+      } while (!readerClosed && attempt < 50);
+      finish(session);
+    } else await tick();
+  } catch (error) {
+    gate.push(null);
+    if (!session.stopped) {
+      const message = error instanceof Error ? error.message : String(error);
+      stream.push({ type: "error", reason: "error", error: message });
+    }
+  }
+  await reading;
+}
+export {
+  listenWithVad
+};
+//# sourceMappingURL=vad-listen.js.map

package/dist/vad-listen.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../src/vad-listen.ts"],
+  "sourcesContent": ["import type { AudioChunk, AudioSource } from './audio.ts';\nimport type { TranscriptionModel } from './models.ts';\nimport {\n  type ListenOptions,\n  type TranscriptionEvent,\n  TranscriptionEventStream,\n} from './transcription.ts';\nimport { energyVad, type VadHook, type VadUpdate } from './vad.ts';\n\ntype ListenModel = (\n  model: TranscriptionModel<string>,\n  audio: AudioSource,\n  options?: ListenOptions\n) => TranscriptionEventStream;\n\ninterface Gate {\n  source: AudioSource;\n  push(chunk: AudioChunk | null): void;\n}\n\nfunction tick(): Promise<void> {\n  return new Promise((resolve) => {\n    setTimeout(resolve, 0);\n  });\n}\n\nfunction audioGate(): Gate {\n  const queued: { chunk: AudioChunk | null }[] = [];\n  let resolveWait: ((chunk: AudioChunk | null) => void) | undefined;\n  const source: AudioSource = {\n    async *[Symbol.asyncIterator]() {\n      while (true) {\n        const ready = queued.shift();\n        if (ready !== undefined) {\n          if (ready.chunk === null) return;\n          yield ready.chunk;\n          continue;\n        }\n        const chunk = await new Promise<AudioChunk | null>((resolve) => {\n          resolveWait = resolve;\n        });\n        if (chunk === null) return;\n        yield chunk;\n      }\n    },\n  };\n  return {\n    source,\n    push(chunk) {\n      if (resolveWait) {\n        const resolve = resolveWait;\n        resolveWait = undefined;\n        resolve(chunk);\n        return;\n      }\n      queued.push({ chunk });\n    },\n  };\n}\n\nfunction joinSegments(parts: string[]): string {\n  let text = '';\n  for (const part of parts) {\n    if (text.length === 0 || /\\s$/.test(text) || /^\\s/.test(part)) text += part;\n    else text += ` ${part}`;\n  }\n  return text;\n}\n\ninterface Session {\n  stream: TranscriptionEventStream;\n  segments: Map<string, string>;\n  closed: string;\n  turnOpen: boolean;\n  started: boolean;\n  pendingStart?: number;\n  pendingEnd?: number;\n  heldEnd?: number;\n  providerTurn?: string;\n  pendingDone?: Extract<TranscriptionEvent, { type: 'done' }>;\n  stopped: boolean;\n  lastAt: number;\n}\n\nfunction liveText(session: Session): string {\n  const parts = [...session.segments.values()];\n  if (parts.length > 0) return joinSegments(parts);\n  return session.pendingDone?.transcript.text ?? session.providerTurn ?? '';\n}\n\nfunction turnText(session: Session): string {\n  const live = liveText(session);\n  if (session.closed.length > 0 && live.startsWith(session.closed)) {\n    return live.slice(session.closed.length).trim();\n  }\n  if (session.providerTurn && !live.startsWith(session.providerTurn)) return session.providerTurn;\n  return live.trim();\n}\n\nfunction closeTurn(session: Session, at: number) {\n  if (!session.turnOpen || session.stopped) return;\n  const text = turnText(session);\n  if (text.length === 0) return;\n  session.closed = liveText(session);\n  session.turnOpen = false;\n  session.started = false;\n  session.pendingEnd = undefined;\n  session.heldEnd = undefined;\n  session.providerTurn = undefined;\n  session.stream.push({ type: 'turn_end', text, at });\n}\n\nfunction openTurn(session: Session, at: number) {\n  if (session.started || session.stopped) return;\n  session.stream.push({ type: 'speech_start', at });\n  session.started = true;\n  session.turnOpen = true;\n  session.pendingStart = undefined;\n}\n\nfunction applyEvent(session: Session, event: TranscriptionEvent) {\n  if (event.type === 'speech_start') {\n    if (!session.started) session.stream.push(event);\n    session.started = true;\n    session.turnOpen = true;\n    session.pendingStart = undefined;\n    return;\n  }\n  if (event.type === 'turn_end') {\n    session.providerTurn = event.text;\n    session.pendingEnd = event.at;\n    return;\n  }\n  if (event.type === 'partial' || event.type === 'final') {\n    if (!session.started) openTurn(session, session.pendingStart ?? session.lastAt);\n    session.segments.set(event.segmentId, event.text);\n    session.stream.push(event);\n    return;\n  }\n  if (event.type === 'error') {\n    session.stream.push(event);\n    session.stopped = true;\n    return;\n  }\n  if (event.type === 'done') session.pendingDone = event;\n}\n\nfunction absorb(session: Session, events: TranscriptionEvent[], update: VadUpdate | undefined) {\n  if (session.stopped) return;\n  if (update?.speechStart !== undefined) session.pendingStart = update.speechStart;\n  if (update?.turnEnd !== undefined) session.pendingEnd = update.turnEnd;\n  const textual = events.some(\n    (event) => event.type === 'partial' || event.type === 'final' || event.type === 'done'\n  );\n  if (session.pendingStart !== undefined && (textual || events.length === 0)) {\n    openTurn(session, session.pendingStart);\n  }\n  for (const event of events) {\n    applyEvent(session, event);\n    if (session.stopped) return;\n  }\n  if (session.pendingEnd !== undefined) {\n    session.heldEnd = session.pendingEnd;\n    session.pendingEnd = undefined;\n  }\n  if (session.heldEnd !== undefined) closeTurn(session, session.heldEnd);\n}\n\nfunction finish(session: Session) {\n  if (session.stopped) return;\n  if (session.pendingStart !== undefined) openTurn(session, session.pendingStart);\n  if (session.turnOpen) closeTurn(session, session.heldEnd ?? session.pendingEnd ?? session.lastAt);\n  if (session.pendingDone) session.stream.push(session.pendingDone);\n  else {\n    session.stream.push({\n      type: 'error',\n      reason: 'error',\n      error: 'Transcription ended without a transcript',\n    });\n  }\n}\n\nexport function listenWithVad(\n  listenModel: ListenModel,\n  model: TranscriptionModel<string>,\n  audio: AudioSource,\n  options?: ListenOptions,\n  vad: VadHook = options?.vad ?? energyVad()\n): TranscriptionEventStream {\n  const stream = new TranscriptionEventStream();\n  void pump(listenModel, model, audio, options, stream, vad);\n  return stream;\n}\n\nasync function pump(\n  listenModel: ListenModel,\n  model: TranscriptionModel<string>,\n  audio: AudioSource,\n  options: ListenOptions | undefined,\n  stream: TranscriptionEventStream,\n  vad: VadHook\n) {\n  if (options?.signal?.aborted) {\n    stream.push({ type: 'error', reason: 'aborted', error: 'aborted' });\n    return;\n  }\n  const gate = audioGate();\n  let inner: TranscriptionEventStream;\n  try {\n    inner = listenModel(model, gate.source, options);\n  } catch (error) {\n    const message = error instanceof Error ? error.message : String(error);\n    stream.push({ type: 'error', reason: 'error', error: message });\n    return;\n  }\n  const inbox: TranscriptionEvent[] = [];\n  let readerClosed = false;\n  const reading = (async () => {\n    for await (const event of inner) inbox.push(event);\n    readerClosed = true;\n  })();\n  const session: Session = {\n    stream,\n    segments: new Map(),\n    closed: '',\n    turnOpen: false,\n    started: false,\n    stopped: false,\n    lastAt: 0,\n  };\n  const take = (update?: VadUpdate) => {\n    const events = inbox.splice(0, inbox.length);\n    absorb(session, events, update);\n  };\n  try {\n    await tick();\n    take();\n    for await (const chunk of audio) {\n      if (session.stopped) break;\n      if (options?.signal?.aborted) {\n        gate.push(null);\n        await tick();\n        take();\n        if (!session.stopped) {\n          stream.push({ type: 'error', reason: 'aborted', error: 'aborted' });\n          session.stopped = true;\n        }\n        break;\n      }\n      session.lastAt = chunk.timestamp;\n      const update = vad.push(chunk);\n      gate.push(chunk);\n      await tick();\n      take(update);\n    }\n    gate.push(null);\n    if (!session.stopped) {\n      let attempt = 0;\n      do {\n        await tick();\n        take();\n        attempt += 1;\n      } while (!readerClosed && attempt < 50);\n      finish(session);\n    } else await tick();\n  } catch (error) {\n    gate.push(null);\n    if (!session.stopped) {\n      const message = error instanceof Error ? error.message : String(error);\n      stream.push({ type: 'error', reason: 'error', error: message });\n    }\n  }\n  await reading;\n}\n"],
+  "mappings": "AAEA;AAAA,EAGE;AAAA,OACK;AACP,SAAS,iBAA+C;AAaxD,SAAS,OAAsB;AAC7B,SAAO,IAAI,QAAQ,CAAC,YAAY;AAC9B,eAAW,SAAS,CAAC;AAAA,EACvB,CAAC;AACH;AAEA,SAAS,YAAkB;AACzB,QAAM,SAAyC,CAAC;AAChD,MAAI;AACJ,QAAM,SAAsB;AAAA,IAC1B,QAAQ,OAAO,aAAa,IAAI;AAC9B,aAAO,MAAM;AACX,cAAM,QAAQ,OAAO,MAAM;AAC3B,YAAI,UAAU,QAAW;AACvB,cAAI,MAAM,UAAU,KAAM;AAC1B,gBAAM,MAAM;AACZ;AAAA,QACF;AACA,cAAM,QAAQ,MAAM,IAAI,QAA2B,CAAC,YAAY;AAC9D,wBAAc;AAAA,QAChB,CAAC;AACD,YAAI,UAAU,KAAM;AACpB,cAAM;AAAA,MACR;AAAA,IACF;AAAA,EACF;AACA,SAAO;AAAA,IACL;AAAA,IACA,KAAK,OAAO;AACV,UAAI,aAAa;AACf,cAAM,UAAU;AAChB,sBAAc;AACd,gBAAQ,KAAK;AACb;AAAA,MACF;AACA,aAAO,KAAK,EAAE,MAAM,CAAC;AAAA,IACvB;AAAA,EACF;AACF;AAEA,SAAS,aAAa,OAAyB;AAC7C,MAAI,OAAO;AACX,aAAW,QAAQ,OAAO;AACxB,QAAI,KAAK,WAAW,KAAK,MAAM,KAAK,IAAI,KAAK,MAAM,KAAK,IAAI,EAAG,SAAQ;AAAA,QAClE,SAAQ,IAAI,IAAI;AAAA,EACvB;AACA,SAAO;AACT;AAiBA,SAAS,SAAS,SAA0B;AAC1C,QAAM,QAAQ,CAAC,GAAG,QAAQ,SAAS,OAAO,CAAC;AAC3C,MAAI,MAAM,SAAS,EAAG,QAAO,aAAa,KAAK;AAC/C,SAAO,QAAQ,aAAa,WAAW,QAAQ,QAAQ,gBAAgB;AACzE;AAEA,SAAS,SAAS,SAA0B;AAC1C,QAAM,OAAO,SAAS,OAAO;AAC7B,MAAI,QAAQ,OAAO,SAAS,KAAK,KAAK,WAAW,QAAQ,MAAM,GAAG;AAChE,WAAO,KAAK,MAAM,QAAQ,OAAO,MAAM,EAAE,KAAK;AAAA,EAChD;AACA,MAAI,QAAQ,gBAAgB,CAAC,KAAK,WAAW,QAAQ,YAAY,EAAG,QAAO,QAAQ;AACnF,SAAO,KAAK,KAAK;AACnB;AAEA,SAAS,UAAU,SAAkB,IAAY;AAC/C,MAAI,CAAC,QAAQ,YAAY,QAAQ,QAAS;AAC1C,QAAM,OAAO,SAAS,OAAO;AAC7B,MAAI,KAAK,WAAW,EAAG;AACvB,UAAQ,SAAS,SAAS,OAAO;AACjC,UAAQ,WAAW;AACnB,UAAQ,UAAU;AAClB,UAAQ,aAAa;AACrB,UAAQ,UAAU;AAClB,UAAQ,eAAe;AACvB,UAAQ,OAAO,KAAK,EAAE,MAAM,YAAY,MAAM,GAAG,CAAC;AACpD;AAEA,SAAS,SAAS,SAAkB,IAAY;AAC9C,MAAI,QAAQ,WAAW,QAAQ,QAAS;AACxC,UAAQ,OAAO,KAAK,EAAE,MAAM,gBAAgB,GAAG,CAAC;AAChD,UAAQ,UAAU;AAClB,UAAQ,WAAW;AACnB,UAAQ,eAAe;AACzB;AAEA,SAAS,WAAW,SAAkB,OAA2B;AAC/D,MAAI,MAAM,SAAS,gBAAgB;AACjC,QAAI,CAAC,QAAQ,QAAS,SAAQ,OAAO,KAAK,KAAK;AAC/C,YAAQ,UAAU;AAClB,YAAQ,WAAW;AACnB,YAAQ,eAAe;AACvB;AAAA,EACF;AACA,MAAI,MAAM,SAAS,YAAY;AAC7B,YAAQ,eAAe,MAAM;AAC7B,YAAQ,aAAa,MAAM;AAC3B;AAAA,EACF;AACA,MAAI,MAAM,SAAS,aAAa,MAAM,SAAS,SAAS;AACtD,QAAI,CAAC,QAAQ,QAAS,UAAS,SAAS,QAAQ,gBAAgB,QAAQ,MAAM;AAC9E,YAAQ,SAAS,IAAI,MAAM,WAAW,MAAM,IAAI;AAChD,YAAQ,OAAO,KAAK,KAAK;AACzB;AAAA,EACF;AACA,MAAI,MAAM,SAAS,SAAS;AAC1B,YAAQ,OAAO,KAAK,KAAK;AACzB,YAAQ,UAAU;AAClB;AAAA,EACF;AACA,MAAI,MAAM,SAAS,OAAQ,SAAQ,cAAc;AACnD;AAEA,SAAS,OAAO,SAAkB,QAA8B,QAA+B;AAC7F,MAAI,QAAQ,QAAS;AACrB,MAAI,QAAQ,gBAAgB,OAAW,SAAQ,eAAe,OAAO;AACrE,MAAI,QAAQ,YAAY,OAAW,SAAQ,aAAa,OAAO;AAC/D,QAAM,UAAU,OAAO;AAAA,IACrB,CAAC,UAAU,MAAM,SAAS,aAAa,MAAM,SAAS,WAAW,MAAM,SAAS;AAAA,EAClF;AACA,MAAI,QAAQ,iBAAiB,WAAc,WAAW,OAAO,WAAW,IAAI;AAC1E,aAAS,SAAS,QAAQ,YAAY;AAAA,EACxC;AACA,aAAW,SAAS,QAAQ;AAC1B,eAAW,SAAS,KAAK;AACzB,QAAI,QAAQ,QAAS;AAAA,EACvB;AACA,MAAI,QAAQ,eAAe,QAAW;AACpC,YAAQ,UAAU,QAAQ;AAC1B,YAAQ,aAAa;AAAA,EACvB;AACA,MAAI,QAAQ,YAAY,OAAW,WAAU,SAAS,QAAQ,OAAO;AACvE;AAEA,SAAS,OAAO,SAAkB;AAChC,MAAI,QAAQ,QAAS;AACrB,MAAI,QAAQ,iBAAiB,OAAW,UAAS,SAAS,QAAQ,YAAY;AAC9E,MAAI,QAAQ,SAAU,WAAU,SAAS,QAAQ,WAAW,QAAQ,cAAc,QAAQ,MAAM;AAChG,MAAI,QAAQ,YAAa,SAAQ,OAAO,KAAK,QAAQ,WAAW;AAAA,OAC3D;AACH,YAAQ,OAAO,KAAK;AAAA,MAClB,MAAM;AAAA,MACN,QAAQ;AAAA,MACR,OAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AAEO,SAAS,cACd,aACA,OACA,OACA,SACA,MAAe,SAAS,OAAO,UAAU,GACf;AAC1B,QAAM,SAAS,IAAI,yBAAyB;AAC5C,OAAK,KAAK,aAAa,OAAO,OAAO,SAAS,QAAQ,GAAG;AACzD,SAAO;AACT;AAEA,eAAe,KACb,aACA,OACA,OACA,SACA,QACA,KACA;AACA,MAAI,SAAS,QAAQ,SAAS;AAC5B,WAAO,KAAK,EAAE,MAAM,SAAS,QAAQ,WAAW,OAAO,UAAU,CAAC;AAClE;AAAA,EACF;AACA,QAAM,OAAO,UAAU;AACvB,MAAI;AACJ,MAAI;AACF,YAAQ,YAAY,OAAO,KAAK,QAAQ,OAAO;AAAA,EACjD,SAAS,OAAO;AACd,UAAM,UAAU,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AACrE,WAAO,KAAK,EAAE,MAAM,SAAS,QAAQ,SAAS,OAAO,QAAQ,CAAC;AAC9D;AAAA,EACF;AACA,QAAM,QAA8B,CAAC;AACrC,MAAI,eAAe;AACnB,QAAM,WAAW,YAAY;AAC3B,qBAAiB,SAAS,MAAO,OAAM,KAAK,KAAK;AACjD,mBAAe;AAAA,EACjB,GAAG;AACH,QAAM,UAAmB;AAAA,IACvB;AAAA,IACA,UAAU,oBAAI,IAAI;AAAA,IAClB,QAAQ;AAAA,IACR,UAAU;AAAA,IACV,SAAS;AAAA,IACT,SAAS;AAAA,IACT,QAAQ;AAAA,EACV;AACA,QAAM,OAAO,CAAC,WAAuB;AACnC,UAAM,SAAS,MAAM,OAAO,GAAG,MAAM,MAAM;AAC3C,WAAO,SAAS,QAAQ,MAAM;AAAA,EAChC;AACA,MAAI;AACF,UAAM,KAAK;AACX,SAAK;AACL,qBAAiB,SAAS,OAAO;AAC/B,UAAI,QAAQ,QAAS;AACrB,UAAI,SAAS,QAAQ,SAAS;AAC5B,aAAK,KAAK,IAAI;AACd,cAAM,KAAK;AACX,aAAK;AACL,YAAI,CAAC,QAAQ,SAAS;AACpB,iBAAO,KAAK,EAAE,MAAM,SAAS,QAAQ,WAAW,OAAO,UAAU,CAAC;AAClE,kBAAQ,UAAU;AAAA,QACpB;AACA;AAAA,MACF;AACA,cAAQ,SAAS,MAAM;AACvB,YAAM,SAAS,IAAI,KAAK,KAAK;AAC7B,WAAK,KAAK,KAAK;AACf,YAAM,KAAK;AACX,WAAK,MAAM;AAAA,IACb;AACA,SAAK,KAAK,IAAI;AACd,QAAI,CAAC,QAAQ,SAAS;AACpB,UAAI,UAAU;AACd,SAAG;AACD,cAAM,KAAK;AACX,aAAK;AACL,mBAAW;AAAA,MACb,SAAS,CAAC,gBAAgB,UAAU;AACpC,aAAO,OAAO;AAAA,IAChB,MAAO,OAAM,KAAK;AAAA,EACpB,SAAS,OAAO;AACd,SAAK,KAAK,IAAI;AACd,QAAI,CAAC,QAAQ,SAAS;AACpB,YAAM,UAAU,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AACrE,aAAO,KAAK,EAAE,MAAM,SAAS,QAAQ,SAAS,OAAO,QAAQ,CAAC;AAAA,IAChE;AAAA,EACF;AACA,QAAM;AACR;",
+  "names": []
+}

package/dist/vad.js

@@ -0,0 +1,57 @@
+function width(encoding) {
+  if (encoding === "pcm16") return 2;
+  if (encoding === "pcmf32") return 4;
+  return void 0;
+}
+function durationMs(chunk) {
+  const bytes = width(chunk.format.encoding);
+  if (bytes === void 0 || chunk.format.sampleRate <= 0 || chunk.format.channels <= 0) return 0;
+  const samples = chunk.data.byteLength / bytes / chunk.format.channels;
+  return samples / chunk.format.sampleRate * 1e3;
+}
+function rms(chunk) {
+  const bytes = width(chunk.format.encoding);
+  if (bytes === void 0) return void 0;
+  const samples = Math.floor(chunk.data.byteLength / bytes);
+  if (samples === 0) return 0;
+  const view = new DataView(chunk.data.buffer, chunk.data.byteOffset, chunk.data.byteLength);
+  let sum = 0;
+  for (let index = 0; index < samples; index += 1) {
+    const sample = bytes === 2 ? view.getInt16(index * 2, true) / 32768 : view.getFloat32(index * 4, true);
+    sum += sample * sample;
+  }
+  return Math.sqrt(sum / samples);
+}
+function energyVad(options = {}) {
+  const threshold = options.threshold ?? 0.02;
+  const silenceMs = options.silenceMs ?? 500;
+  let speaking = false;
+  let silence = 0;
+  return {
+    push(chunk) {
+      const level = rms(chunk);
+      if (level === void 0) {
+        if (chunk.data.byteLength === 0 || speaking) return void 0;
+        speaking = true;
+        return { speechStart: chunk.timestamp };
+      }
+      const duration = durationMs(chunk);
+      if (level >= threshold) {
+        silence = 0;
+        if (speaking) return void 0;
+        speaking = true;
+        return { speechStart: chunk.timestamp };
+      }
+      if (!speaking) return void 0;
+      silence += duration;
+      if (silence < silenceMs) return void 0;
+      speaking = false;
+      silence = 0;
+      return { turnEnd: chunk.timestamp + duration };
+    }
+  };
+}
+export {
+  energyVad
+};
+//# sourceMappingURL=vad.js.map

package/dist/vad.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../src/vad.ts"],
+  "sourcesContent": ["import type { AudioChunk } from './audio.ts';\n\nexport interface VadUpdate {\n  speechStart?: number;\n  turnEnd?: number;\n}\n\nexport interface VadHook {\n  push(chunk: AudioChunk): VadUpdate | undefined;\n}\n\nexport interface EnergyVadOptions {\n  threshold?: number;\n  silenceMs?: number;\n}\n\nfunction width(encoding: AudioChunk['format']['encoding']): number | undefined {\n  if (encoding === 'pcm16') return 2;\n  if (encoding === 'pcmf32') return 4;\n  return undefined;\n}\n\nfunction durationMs(chunk: AudioChunk): number {\n  const bytes = width(chunk.format.encoding);\n  if (bytes === undefined || chunk.format.sampleRate <= 0 || chunk.format.channels <= 0) return 0;\n  const samples = chunk.data.byteLength / bytes / chunk.format.channels;\n  return (samples / chunk.format.sampleRate) * 1000;\n}\n\nfunction rms(chunk: AudioChunk): number | undefined {\n  const bytes = width(chunk.format.encoding);\n  if (bytes === undefined) return undefined;\n  const samples = Math.floor(chunk.data.byteLength / bytes);\n  if (samples === 0) return 0;\n  const view = new DataView(chunk.data.buffer, chunk.data.byteOffset, chunk.data.byteLength);\n  let sum = 0;\n  for (let index = 0; index < samples; index += 1) {\n    const sample =\n      bytes === 2 ? view.getInt16(index * 2, true) / 32768 : view.getFloat32(index * 4, true);\n    sum += sample * sample;\n  }\n  return Math.sqrt(sum / samples);\n}\n\nexport function energyVad(options: EnergyVadOptions = {}): VadHook {\n  const threshold = options.threshold ?? 0.02;\n  const silenceMs = options.silenceMs ?? 500;\n  let speaking = false;\n  let silence = 0;\n  return {\n    push(chunk) {\n      const level = rms(chunk);\n      if (level === undefined) {\n        if (chunk.data.byteLength === 0 || speaking) return undefined;\n        speaking = true;\n        return { speechStart: chunk.timestamp };\n      }\n      const duration = durationMs(chunk);\n      if (level >= threshold) {\n        silence = 0;\n        if (speaking) return undefined;\n        speaking = true;\n        return { speechStart: chunk.timestamp };\n      }\n      if (!speaking) return undefined;\n      silence += duration;\n      if (silence < silenceMs) return undefined;\n      speaking = false;\n      silence = 0;\n      return { turnEnd: chunk.timestamp + duration };\n    },\n  };\n}\n"],
+  "mappings": "AAgBA,SAAS,MAAM,UAAgE;AAC7E,MAAI,aAAa,QAAS,QAAO;AACjC,MAAI,aAAa,SAAU,QAAO;AAClC,SAAO;AACT;AAEA,SAAS,WAAW,OAA2B;AAC7C,QAAM,QAAQ,MAAM,MAAM,OAAO,QAAQ;AACzC,MAAI,UAAU,UAAa,MAAM,OAAO,cAAc,KAAK,MAAM,OAAO,YAAY,EAAG,QAAO;AAC9F,QAAM,UAAU,MAAM,KAAK,aAAa,QAAQ,MAAM,OAAO;AAC7D,SAAQ,UAAU,MAAM,OAAO,aAAc;AAC/C;AAEA,SAAS,IAAI,OAAuC;AAClD,QAAM,QAAQ,MAAM,MAAM,OAAO,QAAQ;AACzC,MAAI,UAAU,OAAW,QAAO;AAChC,QAAM,UAAU,KAAK,MAAM,MAAM,KAAK,aAAa,KAAK;AACxD,MAAI,YAAY,EAAG,QAAO;AAC1B,QAAM,OAAO,IAAI,SAAS,MAAM,KAAK,QAAQ,MAAM,KAAK,YAAY,MAAM,KAAK,UAAU;AACzF,MAAI,MAAM;AACV,WAAS,QAAQ,GAAG,QAAQ,SAAS,SAAS,GAAG;AAC/C,UAAM,SACJ,UAAU,IAAI,KAAK,SAAS,QAAQ,GAAG,IAAI,IAAI,QAAQ,KAAK,WAAW,QAAQ,GAAG,IAAI;AACxF,WAAO,SAAS;AAAA,EAClB;AACA,SAAO,KAAK,KAAK,MAAM,OAAO;AAChC;AAEO,SAAS,UAAU,UAA4B,CAAC,GAAY;AACjE,QAAM,YAAY,QAAQ,aAAa;AACvC,QAAM,YAAY,QAAQ,aAAa;AACvC,MAAI,WAAW;AACf,MAAI,UAAU;AACd,SAAO;AAAA,IACL,KAAK,OAAO;AACV,YAAM,QAAQ,IAAI,KAAK;AACvB,UAAI,UAAU,QAAW;AACvB,YAAI,MAAM,KAAK,eAAe,KAAK,SAAU,QAAO;AACpD,mBAAW;AACX,eAAO,EAAE,aAAa,MAAM,UAAU;AAAA,MACxC;AACA,YAAM,WAAW,WAAW,KAAK;AACjC,UAAI,SAAS,WAAW;AACtB,kBAAU;AACV,YAAI,SAAU,QAAO;AACrB,mBAAW;AACX,eAAO,EAAE,aAAa,MAAM,UAAU;AAAA,MACxC;AACA,UAAI,CAAC,SAAU,QAAO;AACtB,iBAAW;AACX,UAAI,UAAU,UAAW,QAAO;AAChC,iBAAW;AACX,gBAAU;AACV,aAAO,EAAE,SAAS,MAAM,YAAY,SAAS;AAAA,IAC/C;AAAA,EACF;AACF;",
+  "names": []
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-ecoverse/pi-audio",
-  "version": "1.3.0",
+  "version": "1.5.0",
   "description": "Voice bindings for pi: speech-to-text, text-to-speech and realtime speech for pi-coding-agent plugins and slicc-agent",
   "type": "module",
   "license": "Apache-2.0",
@@ -33,6 +33,10 @@
     "./worklet": {
       "types": "./dist/types/worklet/convert.d.ts",
       "default": "./dist/worklet/convert.js"
+    },
+    "./openai": {
+      "types": "./dist/types/openai/index.d.ts",
+      "default": "./dist/openai/index.js"
     }
   },
   "files": [