@ai-ecoverse/pi-audio 1.3.0 → 1.4.0

package/dist/chunk-speech.js

@@ -0,0 +1,135 @@
+import { blankUsage } from "./blank-usage.js";
+import { VoiceRequestError } from "./errors.js";
+import { sentencesOf } from "./sentences.js";
+import {
+  SpeechEventStream
+} from "./speech.js";
+import { addAudioCounters } from "./usage.js";
+function shiftSpan(span, offset) {
+  if ("entryId" in span) {
+    return {
+      entryId: span.entryId,
+      contentIndex: span.contentIndex,
+      start: span.start + offset,
+      end: span.end + offset
+    };
+  }
+  return { itemId: span.itemId, start: span.start + offset, end: span.end + offset };
+}
+function mergeUsage(usages) {
+  const total = blankUsage();
+  for (const usage of usages) {
+    total.input += usage.input;
+    total.output += usage.output;
+    total.cacheRead += usage.cacheRead;
+    total.cacheWrite += usage.cacheWrite;
+    total.totalTokens += usage.totalTokens;
+    total.cost.input += usage.cost.input;
+    total.cost.output += usage.cost.output;
+    total.cost.cacheRead += usage.cost.cacheRead;
+    total.cost.cacheWrite += usage.cost.cacheWrite;
+    total.cost.total += usage.cost.total;
+    if (usage.reasoning !== void 0) total.reasoning = (total.reasoning ?? 0) + usage.reasoning;
+    if (usage.cacheWrite1h !== void 0) {
+      total.cacheWrite1h = (total.cacheWrite1h ?? 0) + usage.cacheWrite1h;
+    }
+    addAudioCounters(total, usage);
+  }
+  return total;
+}
+function failure(error) {
+  if (error instanceof VoiceRequestError) {
+    return { type: "error", reason: error.reason, error: error.message };
+  }
+  const message = error instanceof Error ? error.message : String(error);
+  return { type: "error", reason: "error", error: message };
+}
+async function playSentence(speak, model, input, sentence, offset, options, stream, usages, state) {
+  let inner;
+  try {
+    inner = speak(model, { ...input, text: sentence }, options);
+  } catch (error) {
+    stream.push(failure(error));
+    return false;
+  }
+  let terminal = false;
+  for await (const event of inner) {
+    if (options?.signal?.aborted && event.type !== "error" && event.type !== "done") {
+      stream.push({ type: "error", reason: "aborted", error: "aborted" });
+      return false;
+    }
+    if (event.type === "start") {
+      if (!state.started) stream.push(event);
+      state.started = true;
+      continue;
+    }
+    if (event.type === "audio") {
+      stream.push({ ...event, span: shiftSpan(event.span, offset) });
+      continue;
+    }
+    if (event.type === "done") {
+      usages.push(event.usage);
+      terminal = true;
+      continue;
+    }
+    stream.push(event);
+    return false;
+  }
+  if (terminal) return true;
+  try {
+    usages.push(await inner.result());
+    return true;
+  } catch (error) {
+    stream.push(failure(error));
+    return false;
+  }
+}
+function speakInSentences(speak, model, input, options) {
+  const stream = new SpeechEventStream();
+  void pump(speak, model, input, options, stream);
+  return stream;
+}
+async function pump(speak, model, input, options, stream) {
+  if (options?.signal?.aborted) {
+    stream.push({ type: "error", reason: "aborted", error: "aborted" });
+    return;
+  }
+  const usages = [];
+  const state = { started: false };
+  let offset = 0;
+  let played = false;
+  try {
+    for await (const sentence of sentencesOf(input.text)) {
+      if (options?.signal?.aborted) {
+        stream.push({ type: "error", reason: "aborted", error: "aborted" });
+        return;
+      }
+      played = true;
+      const ok = await playSentence(
+        speak,
+        model,
+        input,
+        sentence,
+        offset,
+        options,
+        stream,
+        usages,
+        state
+      );
+      if (!ok) return;
+      offset += sentence.length;
+    }
+    if (!played) {
+      const text = typeof input.text === "string" ? input.text : "";
+      const ok = await playSentence(speak, model, input, text, 0, options, stream, usages, state);
+      if (!ok) return;
+    }
+    stream.push({ type: "done", usage: mergeUsage(usages) });
+  } catch (error) {
+    stream.push(failure(error));
+  }
+}
+export {
+  speakInSentences
+};
+//# sourceMappingURL=chunk-speech.js.map

package/dist/chunk-speech.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../src/chunk-speech.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions, Usage } from '@earendil-works/pi-ai';\nimport type { TextSpan } from './audio.ts';\nimport { blankUsage } from './blank-usage.ts';\nimport { VoiceRequestError } from './errors.ts';\nimport type { SpeechModel } from './models.ts';\nimport { sentencesOf } from './sentences.ts';\nimport {\n  type SpeechApiProvider,\n  type SpeechEvent,\n  SpeechEventStream,\n  type SpeechInput,\n} from './speech.ts';\nimport { addAudioCounters } from './usage.ts';\n\nfunction shiftSpan(span: TextSpan, offset: number): TextSpan {\n  if ('entryId' in span) {\n    return {\n      entryId: span.entryId,\n      contentIndex: span.contentIndex,\n      start: span.start + offset,\n      end: span.end + offset,\n    };\n  }\n  return { itemId: span.itemId, start: span.start + offset, end: span.end + offset };\n}\n\nfunction mergeUsage(usages: Usage[]): Usage {\n  const total = blankUsage();\n  for (const usage of usages) {\n    total.input += usage.input;\n    total.output += usage.output;\n    total.cacheRead += usage.cacheRead;\n    total.cacheWrite += usage.cacheWrite;\n    total.totalTokens += usage.totalTokens;\n    total.cost.input += usage.cost.input;\n    total.cost.output += usage.cost.output;\n    total.cost.cacheRead += usage.cost.cacheRead;\n    total.cost.cacheWrite += usage.cost.cacheWrite;\n    total.cost.total += usage.cost.total;\n    if (usage.reasoning !== undefined) total.reasoning = (total.reasoning ?? 0) + usage.reasoning;\n    if (usage.cacheWrite1h !== undefined) {\n      total.cacheWrite1h = (total.cacheWrite1h ?? 0) + usage.cacheWrite1h;\n    }\n    addAudioCounters(total, usage);\n  }\n  return total;\n}\n\nfunction failure(error: unknown): SpeechEvent {\n  if (error instanceof VoiceRequestError) {\n    return { type: 'error', reason: error.reason, error: error.message };\n  }\n  const message = error instanceof Error ? error.message : String(error);\n  return { type: 'error', reason: 'error', error: message };\n}\n\nasync function playSentence(\n  speak: SpeechApiProvider['speak'],\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  sentence: string,\n  offset: number,\n  options: ProviderRequestOptions | undefined,\n  stream: SpeechEventStream,\n  usages: Usage[],\n  state: { started: boolean }\n): Promise<boolean> {\n  let inner: SpeechEventStream;\n  try {\n    inner = speak(model, { ...input, text: sentence }, options);\n  } catch (error) {\n    stream.push(failure(error));\n    return false;\n  }\n  let terminal = false;\n  for await (const event of inner) {\n    if (options?.signal?.aborted && event.type !== 'error' && event.type !== 'done') {\n      stream.push({ type: 'error', reason: 'aborted', error: 'aborted' });\n      return false;\n    }\n    if (event.type === 'start') {\n      if (!state.started) stream.push(event);\n      state.started = true;\n      continue;\n    }\n    if (event.type === 'audio') {\n      stream.push({ ...event, span: shiftSpan(event.span, offset) });\n      continue;\n    }\n    if (event.type === 'done') {\n      usages.push(event.usage);\n      terminal = true;\n      continue;\n    }\n    stream.push(event);\n    return false;\n  }\n  if (terminal) return true;\n  try {\n    usages.push(await inner.result());\n    return true;\n  } catch (error) {\n    stream.push(failure(error));\n    return false;\n  }\n}\n\nexport function speakInSentences(\n  speak: SpeechApiProvider['speak'],\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options?: ProviderRequestOptions\n): SpeechEventStream {\n  const stream = new SpeechEventStream();\n  void pump(speak, model, input, options, stream);\n  return stream;\n}\n\nasync function pump(\n  speak: SpeechApiProvider['speak'],\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options: ProviderRequestOptions | undefined,\n  stream: SpeechEventStream\n) {\n  if (options?.signal?.aborted) {\n    stream.push({ type: 'error', reason: 'aborted', error: 'aborted' });\n    return;\n  }\n  const usages: Usage[] = [];\n  const state = { started: false };\n  let offset = 0;\n  let played = false;\n  try {\n    for await (const sentence of sentencesOf(input.text)) {\n      if (options?.signal?.aborted) {\n        stream.push({ type: 'error', reason: 'aborted', error: 'aborted' });\n        return;\n      }\n      played = true;\n      const ok = await playSentence(\n        speak,\n        model,\n        input,\n        sentence,\n        offset,\n        options,\n        stream,\n        usages,\n        state\n      );\n      if (!ok) return;\n      offset += sentence.length;\n    }\n    if (!played) {\n      const text = typeof input.text === 'string' ? input.text : '';\n      const ok = await playSentence(speak, model, input, text, 0, options, stream, usages, state);\n      if (!ok) return;\n    }\n    stream.push({ type: 'done', usage: mergeUsage(usages) });\n  } catch (error) {\n    stream.push(failure(error));\n  }\n}\n"],
+  "mappings": "AAEA,SAAS,kBAAkB;AAC3B,SAAS,yBAAyB;AAElC,SAAS,mBAAmB;AAC5B;AAAA,EAGE;AAAA,OAEK;AACP,SAAS,wBAAwB;AAEjC,SAAS,UAAU,MAAgB,QAA0B;AAC3D,MAAI,aAAa,MAAM;AACrB,WAAO;AAAA,MACL,SAAS,KAAK;AAAA,MACd,cAAc,KAAK;AAAA,MACnB,OAAO,KAAK,QAAQ;AAAA,MACpB,KAAK,KAAK,MAAM;AAAA,IAClB;AAAA,EACF;AACA,SAAO,EAAE,QAAQ,KAAK,QAAQ,OAAO,KAAK,QAAQ,QAAQ,KAAK,KAAK,MAAM,OAAO;AACnF;AAEA,SAAS,WAAW,QAAwB;AAC1C,QAAM,QAAQ,WAAW;AACzB,aAAW,SAAS,QAAQ;AAC1B,UAAM,SAAS,MAAM;AACrB,UAAM,UAAU,MAAM;AACtB,UAAM,aAAa,MAAM;AACzB,UAAM,cAAc,MAAM;AAC1B,UAAM,eAAe,MAAM;AAC3B,UAAM,KAAK,SAAS,MAAM,KAAK;AAC/B,UAAM,KAAK,UAAU,MAAM,KAAK;AAChC,UAAM,KAAK,aAAa,MAAM,KAAK;AACnC,UAAM,KAAK,cAAc,MAAM,KAAK;AACpC,UAAM,KAAK,SAAS,MAAM,KAAK;AAC/B,QAAI,MAAM,cAAc,OAAW,OAAM,aAAa,MAAM,aAAa,KAAK,MAAM;AACpF,QAAI,MAAM,iBAAiB,QAAW;AACpC,YAAM,gBAAgB,MAAM,gBAAgB,KAAK,MAAM;AAAA,IACzD;AACA,qBAAiB,OAAO,KAAK;AAAA,EAC/B;AACA,SAAO;AACT;AAEA,SAAS,QAAQ,OAA6B;AAC5C,MAAI,iBAAiB,mBAAmB;AACtC,WAAO,EAAE,MAAM,SAAS,QAAQ,MAAM,QAAQ,OAAO,MAAM,QAAQ;AAAA,EACrE;AACA,QAAM,UAAU,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AACrE,SAAO,EAAE,MAAM,SAAS,QAAQ,SAAS,OAAO,QAAQ;AAC1D;AAEA,eAAe,aACb,OACA,OACA,OACA,UACA,QACA,SACA,QACA,QACA,OACkB;AAClB,MAAI;AACJ,MAAI;AACF,YAAQ,MAAM,OAAO,EAAE,GAAG,OAAO,MAAM,SAAS,GAAG,OAAO;AAAA,EAC5D,SAAS,OAAO;AACd,WAAO,KAAK,QAAQ,KAAK,CAAC;AAC1B,WAAO;AAAA,EACT;AACA,MAAI,WAAW;AACf,mBAAiB,SAAS,OAAO;AAC/B,QAAI,SAAS,QAAQ,WAAW,MAAM,SAAS,WAAW,MAAM,SAAS,QAAQ;AAC/E,aAAO,KAAK,EAAE,MAAM,SAAS,QAAQ,WAAW,OAAO,UAAU,CAAC;AAClE,aAAO;AAAA,IACT;AACA,QAAI,MAAM,SAAS,SAAS;AAC1B,UAAI,CAAC,MAAM,QAAS,QAAO,KAAK,KAAK;AACrC,YAAM,UAAU;AAChB;AAAA,IACF;AACA,QAAI,MAAM,SAAS,SAAS;AAC1B,aAAO,KAAK,EAAE,GAAG,OAAO,MAAM,UAAU,MAAM,MAAM,MAAM,EAAE,CAAC;AAC7D;AAAA,IACF;AACA,QAAI,MAAM,SAAS,QAAQ;AACzB,aAAO,KAAK,MAAM,KAAK;AACvB,iBAAW;AACX;AAAA,IACF;AACA,WAAO,KAAK,KAAK;AACjB,WAAO;AAAA,EACT;AACA,MAAI,SAAU,QAAO;AACrB,MAAI;AACF,WAAO,KAAK,MAAM,MAAM,OAAO,CAAC;AAChC,WAAO;AAAA,EACT,SAAS,OAAO;AACd,WAAO,KAAK,QAAQ,KAAK,CAAC;AAC1B,WAAO;AAAA,EACT;AACF;AAEO,SAAS,iBACd,OACA,OACA,OACA,SACmB;AACnB,QAAM,SAAS,IAAI,kBAAkB;AACrC,OAAK,KAAK,OAAO,OAAO,OAAO,SAAS,MAAM;AAC9C,SAAO;AACT;AAEA,eAAe,KACb,OACA,OACA,OACA,SACA,QACA;AACA,MAAI,SAAS,QAAQ,SAAS;AAC5B,WAAO,KAAK,EAAE,MAAM,SAAS,QAAQ,WAAW,OAAO,UAAU,CAAC;AAClE;AAAA,EACF;AACA,QAAM,SAAkB,CAAC;AACzB,QAAM,QAAQ,EAAE,SAAS,MAAM;AAC/B,MAAI,SAAS;AACb,MAAI,SAAS;AACb,MAAI;AACF,qBAAiB,YAAY,YAAY,MAAM,IAAI,GAAG;AACpD,UAAI,SAAS,QAAQ,SAAS;AAC5B,eAAO,KAAK,EAAE,MAAM,SAAS,QAAQ,WAAW,OAAO,UAAU,CAAC;AAClE;AAAA,MACF;AACA,eAAS;AACT,YAAM,KAAK,MAAM;AAAA,QACf;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF;AACA,UAAI,CAAC,GAAI;AACT,gBAAU,SAAS;AAAA,IACrB;AACA,QAAI,CAAC,QAAQ;AACX,YAAM,OAAO,OAAO,MAAM,SAAS,WAAW,MAAM,OAAO;AAC3D,YAAM,KAAK,MAAM,aAAa,OAAO,OAAO,OAAO,MAAM,GAAG,SAAS,QAAQ,QAAQ,KAAK;AAC1F,UAAI,CAAC,GAAI;AAAA,IACX;AACA,WAAO,KAAK,EAAE,MAAM,QAAQ,OAAO,WAAW,MAAM,EAAE,CAAC;AAAA,EACzD,SAAS,OAAO;AACd,WAAO,KAAK,QAAQ,KAAK,CAAC;AAAA,EAC5B;AACF;",
+  "names": []
+}

package/dist/index.js

@@ -24,6 +24,7 @@ import {
 import { SpeechEventStream } from "./speech.js";
 import { TranscriptionEventStream } from "./transcription.js";
 import { addAudioCounters, priceAudio } from "./usage.js";
+import { energyVad } from "./vad.js";
 export {
   SpeechEventStream,
   TranscriptionEventStream,
@@ -35,6 +36,7 @@ export {
   convertAudioSource,
   decodeAudioBase64,
   encodeAudioBase64,
+  energyVad,
   getVoiceModel,
   listVoiceModels,
   listen,

package/dist/index.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["../src/index.ts"],
-  "sourcesContent": ["export type {\n  AudioChunk,\n  AudioEncoding,\n  AudioFormat,\n  AudioSink,\n  AudioSource,\n  AudioTag,\n  CharAlignment,\n  TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport {\n  convertAudio,\n  convertAudioSource,\n  decodeAudioBase64,\n  encodeAudioBase64,\n} from './convert.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n  Delivery,\n  DeliverySupport,\n  KnownRealtimeApi,\n  KnownSpeechApi,\n  KnownTranscriptionApi,\n  RealtimeApi,\n  RealtimeModel,\n  SpeechApi,\n  SpeechModel,\n  TranscriptionApi,\n  TranscriptionModel,\n  VoiceInfo,\n  VoiceModel,\n  VoiceModelType,\n} from './models.ts';\nexport {\n  getVoiceModel,\n  listVoiceModels,\n  registerVoiceModel,\n} from './models.ts';\nexport type {\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeEvent,\n  RealtimeOptions,\n  RealtimeSession,\n  RealtimeToken,\n} from './realtime.ts';\nexport {\n  connectRealtime,\n  listen,\n  registerRealtimeApiProvider,\n  registerSpeechApiProvider,\n  registerTranscriptionApiProvider,\n  speak,\n  transcribe,\n} from './registry.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEvent,\n  TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n  AudioCost,\n  AudioCounters,\n  AudioUsage,\n  RealtimeAudioCost,\n  SpeechAudioCost,\n  TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\n"],
-  "mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAC3B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AASP;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,yBAAyB;AASlC,SAAS,gCAAgC;AASzC,SAAS,kBAAkB,kBAAkB;",
+  "sourcesContent": ["export type {\n  AudioChunk,\n  AudioEncoding,\n  AudioFormat,\n  AudioSink,\n  AudioSource,\n  AudioTag,\n  CharAlignment,\n  TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport {\n  convertAudio,\n  convertAudioSource,\n  decodeAudioBase64,\n  encodeAudioBase64,\n} from './convert.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n  Delivery,\n  DeliverySupport,\n  KnownRealtimeApi,\n  KnownSpeechApi,\n  KnownTranscriptionApi,\n  RealtimeApi,\n  RealtimeModel,\n  SpeechApi,\n  SpeechModel,\n  TranscriptionApi,\n  TranscriptionModel,\n  VoiceInfo,\n  VoiceModel,\n  VoiceModelType,\n} from './models.ts';\nexport {\n  getVoiceModel,\n  listVoiceModels,\n  registerVoiceModel,\n} from './models.ts';\nexport type {\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeEvent,\n  RealtimeOptions,\n  RealtimeSession,\n  RealtimeToken,\n} from './realtime.ts';\nexport {\n  connectRealtime,\n  listen,\n  registerRealtimeApiProvider,\n  registerSpeechApiProvider,\n  registerTranscriptionApiProvider,\n  speak,\n  transcribe,\n} from './registry.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEvent,\n  TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n  AudioCost,\n  AudioCounters,\n  AudioUsage,\n  RealtimeAudioCost,\n  SpeechAudioCost,\n  TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\nexport type { VadHook, VadUpdate } from './vad.ts';\nexport { energyVad } from './vad.ts';\n"],
+  "mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAC3B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AASP;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,yBAAyB;AASlC,SAAS,gCAAgC;AASzC,SAAS,kBAAkB,kBAAkB;AAE7C,SAAS,iBAAiB;",
   "names": []
 }

package/dist/registry.js

@@ -1,3 +1,5 @@
+import { speakInSentences } from "./chunk-speech.js";
+import { listenWithVad } from "./vad-listen.js";
 const transcriptionApis = /* @__PURE__ */ new Map();
 const speechApis = /* @__PURE__ */ new Map();
 const realtimeApis = /* @__PURE__ */ new Map();
@@ -33,11 +35,13 @@ function transcribe(model, input, options) {
 function listen(model, audio, options) {
   const listenModel = transcriptionApis.get(model.api)?.provider.listen;
   if (!listenModel) throw missing(model.api);
+  if (model.turnDetection === "none") return listenWithVad(listenModel, model, audio, options);
   return listenModel(model, audio, options);
 }
 function speak(model, input, options) {
   const speakModel = speechApis.get(model.api)?.provider.speak;
   if (!speakModel) throw missing(model.api);
+  if (!model.textStreaming) return speakInSentences(speakModel, model, input, options);
   return speakModel(model, input, options);
 }
 function connectRealtime(model, context, options) {

package/dist/registry.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["../src/registry.ts"],
-  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport type { AudioSource } from './audio.ts';\nimport type { RealtimeModel, SpeechModel, TranscriptionModel } from './models.ts';\nimport type {\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeOptions,\n  RealtimeSession,\n} from './realtime.ts';\nimport type { SpeechApiProvider, SpeechEventStream, SpeechInput } from './speech.ts';\nimport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEventStream,\n} from './transcription.ts';\n\ninterface TranscriptionEntry {\n  provider: TranscriptionApiProvider;\n}\n\ninterface SpeechEntry {\n  provider: SpeechApiProvider;\n}\n\ninterface RealtimeEntry {\n  provider: RealtimeApiProvider;\n}\n\nconst transcriptionApis = new Map<string, TranscriptionEntry>();\nconst speechApis = new Map<string, SpeechEntry>();\nconst realtimeApis = new Map<string, RealtimeEntry>();\n\nfunction missing(api: string): Error {\n  return new Error(`No API provider registered for api: ${api}`);\n}\n\nexport function registerTranscriptionApiProvider(provider: TranscriptionApiProvider): () => void {\n  const entry = { provider };\n  transcriptionApis.set(provider.api, entry);\n  return () => {\n    if (transcriptionApis.get(provider.api) === entry) transcriptionApis.delete(provider.api);\n  };\n}\n\nexport function registerSpeechApiProvider(provider: SpeechApiProvider): () => void {\n  const entry = { provider };\n  speechApis.set(provider.api, entry);\n  return () => {\n    if (speechApis.get(provider.api) === entry) speechApis.delete(provider.api);\n  };\n}\n\nexport function registerRealtimeApiProvider(provider: RealtimeApiProvider): () => void {\n  const entry = { provider };\n  realtimeApis.set(provider.api, entry);\n  return () => {\n    if (realtimeApis.get(provider.api) === entry) realtimeApis.delete(provider.api);\n  };\n}\n\nexport function transcribe(\n  model: TranscriptionModel<string>,\n  input: AudioInput,\n  options?: ProviderRequestOptions\n): Promise<Transcript> {\n  const transcribeModel = transcriptionApis.get(model.api)?.provider.transcribe;\n  if (!transcribeModel) return Promise.reject(missing(model.api));\n  return transcribeModel(model, input, options);\n}\n\nexport function listen(\n  model: TranscriptionModel<string>,\n  audio: AudioSource,\n  options?: ListenOptions\n): TranscriptionEventStream {\n  const listenModel = transcriptionApis.get(model.api)?.provider.listen;\n  if (!listenModel) throw missing(model.api);\n  return listenModel(model, audio, options);\n}\n\nexport function speak(\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options?: ProviderRequestOptions\n): SpeechEventStream {\n  const speakModel = speechApis.get(model.api)?.provider.speak;\n  if (!speakModel) throw missing(model.api);\n  return speakModel(model, input, options);\n}\n\nexport function connectRealtime(\n  model: RealtimeModel<string>,\n  context: RealtimeContext,\n  options?: RealtimeOptions\n): Promise<RealtimeSession> {\n  const connect = realtimeApis.get(model.api)?.provider.connect;\n  if (!connect) return Promise.reject(missing(model.api));\n  return connect(model, context, options);\n}\n"],
-  "mappings": "AA8BA,MAAM,oBAAoB,oBAAI,IAAgC;AAC9D,MAAM,aAAa,oBAAI,IAAyB;AAChD,MAAM,eAAe,oBAAI,IAA2B;AAEpD,SAAS,QAAQ,KAAoB;AACnC,SAAO,IAAI,MAAM,uCAAuC,GAAG,EAAE;AAC/D;AAEO,SAAS,iCAAiC,UAAgD;AAC/F,QAAM,QAAQ,EAAE,SAAS;AACzB,oBAAkB,IAAI,SAAS,KAAK,KAAK;AACzC,SAAO,MAAM;AACX,QAAI,kBAAkB,IAAI,SAAS,GAAG,MAAM,MAAO,mBAAkB,OAAO,SAAS,GAAG;AAAA,EAC1F;AACF;AAEO,SAAS,0BAA0B,UAAyC;AACjF,QAAM,QAAQ,EAAE,SAAS;AACzB,aAAW,IAAI,SAAS,KAAK,KAAK;AAClC,SAAO,MAAM;AACX,QAAI,WAAW,IAAI,SAAS,GAAG,MAAM,MAAO,YAAW,OAAO,SAAS,GAAG;AAAA,EAC5E;AACF;AAEO,SAAS,4BAA4B,UAA2C;AACrF,QAAM,QAAQ,EAAE,SAAS;AACzB,eAAa,IAAI,SAAS,KAAK,KAAK;AACpC,SAAO,MAAM;AACX,QAAI,aAAa,IAAI,SAAS,GAAG,MAAM,MAAO,cAAa,OAAO,SAAS,GAAG;AAAA,EAChF;AACF;AAEO,SAAS,WACd,OACA,OACA,SACqB;AACrB,QAAM,kBAAkB,kBAAkB,IAAI,MAAM,GAAG,GAAG,SAAS;AACnE,MAAI,CAAC,gBAAiB,QAAO,QAAQ,OAAO,QAAQ,MAAM,GAAG,CAAC;AAC9D,SAAO,gBAAgB,OAAO,OAAO,OAAO;AAC9C;AAEO,SAAS,OACd,OACA,OACA,SAC0B;AAC1B,QAAM,cAAc,kBAAkB,IAAI,MAAM,GAAG,GAAG,SAAS;AAC/D,MAAI,CAAC,YAAa,OAAM,QAAQ,MAAM,GAAG;AACzC,SAAO,YAAY,OAAO,OAAO,OAAO;AAC1C;AAEO,SAAS,MACd,OACA,OACA,SACmB;AACnB,QAAM,aAAa,WAAW,IAAI,MAAM,GAAG,GAAG,SAAS;AACvD,MAAI,CAAC,WAAY,OAAM,QAAQ,MAAM,GAAG;AACxC,SAAO,WAAW,OAAO,OAAO,OAAO;AACzC;AAEO,SAAS,gBACd,OACA,SACA,SAC0B;AAC1B,QAAM,UAAU,aAAa,IAAI,MAAM,GAAG,GAAG,SAAS;AACtD,MAAI,CAAC,QAAS,QAAO,QAAQ,OAAO,QAAQ,MAAM,GAAG,CAAC;AACtD,SAAO,QAAQ,OAAO,SAAS,OAAO;AACxC;",
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport type { AudioSource } from './audio.ts';\nimport { speakInSentences } from './chunk-speech.ts';\nimport type { RealtimeModel, SpeechModel, TranscriptionModel } from './models.ts';\nimport type {\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeOptions,\n  RealtimeSession,\n} from './realtime.ts';\nimport type { SpeechApiProvider, SpeechEventStream, SpeechInput } from './speech.ts';\nimport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEventStream,\n} from './transcription.ts';\nimport { listenWithVad } from './vad-listen.ts';\n\ninterface TranscriptionEntry {\n  provider: TranscriptionApiProvider;\n}\n\ninterface SpeechEntry {\n  provider: SpeechApiProvider;\n}\n\ninterface RealtimeEntry {\n  provider: RealtimeApiProvider;\n}\n\nconst transcriptionApis = new Map<string, TranscriptionEntry>();\nconst speechApis = new Map<string, SpeechEntry>();\nconst realtimeApis = new Map<string, RealtimeEntry>();\n\nfunction missing(api: string): Error {\n  return new Error(`No API provider registered for api: ${api}`);\n}\n\nexport function registerTranscriptionApiProvider(provider: TranscriptionApiProvider): () => void {\n  const entry = { provider };\n  transcriptionApis.set(provider.api, entry);\n  return () => {\n    if (transcriptionApis.get(provider.api) === entry) transcriptionApis.delete(provider.api);\n  };\n}\n\nexport function registerSpeechApiProvider(provider: SpeechApiProvider): () => void {\n  const entry = { provider };\n  speechApis.set(provider.api, entry);\n  return () => {\n    if (speechApis.get(provider.api) === entry) speechApis.delete(provider.api);\n  };\n}\n\nexport function registerRealtimeApiProvider(provider: RealtimeApiProvider): () => void {\n  const entry = { provider };\n  realtimeApis.set(provider.api, entry);\n  return () => {\n    if (realtimeApis.get(provider.api) === entry) realtimeApis.delete(provider.api);\n  };\n}\n\nexport function transcribe(\n  model: TranscriptionModel<string>,\n  input: AudioInput,\n  options?: ProviderRequestOptions\n): Promise<Transcript> {\n  const transcribeModel = transcriptionApis.get(model.api)?.provider.transcribe;\n  if (!transcribeModel) return Promise.reject(missing(model.api));\n  return transcribeModel(model, input, options);\n}\n\nexport function listen(\n  model: TranscriptionModel<string>,\n  audio: AudioSource,\n  options?: ListenOptions\n): TranscriptionEventStream {\n  const listenModel = transcriptionApis.get(model.api)?.provider.listen;\n  if (!listenModel) throw missing(model.api);\n  if (model.turnDetection === 'none') return listenWithVad(listenModel, model, audio, options);\n  return listenModel(model, audio, options);\n}\n\nexport function speak(\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options?: ProviderRequestOptions\n): SpeechEventStream {\n  const speakModel = speechApis.get(model.api)?.provider.speak;\n  if (!speakModel) throw missing(model.api);\n  if (!model.textStreaming) return speakInSentences(speakModel, model, input, options);\n  return speakModel(model, input, options);\n}\n\nexport function connectRealtime(\n  model: RealtimeModel<string>,\n  context: RealtimeContext,\n  options?: RealtimeOptions\n): Promise<RealtimeSession> {\n  const connect = realtimeApis.get(model.api)?.provider.connect;\n  if (!connect) return Promise.reject(missing(model.api));\n  return connect(model, context, options);\n}\n"],
+  "mappings": "AAEA,SAAS,wBAAwB;AAgBjC,SAAS,qBAAqB;AAc9B,MAAM,oBAAoB,oBAAI,IAAgC;AAC9D,MAAM,aAAa,oBAAI,IAAyB;AAChD,MAAM,eAAe,oBAAI,IAA2B;AAEpD,SAAS,QAAQ,KAAoB;AACnC,SAAO,IAAI,MAAM,uCAAuC,GAAG,EAAE;AAC/D;AAEO,SAAS,iCAAiC,UAAgD;AAC/F,QAAM,QAAQ,EAAE,SAAS;AACzB,oBAAkB,IAAI,SAAS,KAAK,KAAK;AACzC,SAAO,MAAM;AACX,QAAI,kBAAkB,IAAI,SAAS,GAAG,MAAM,MAAO,mBAAkB,OAAO,SAAS,GAAG;AAAA,EAC1F;AACF;AAEO,SAAS,0BAA0B,UAAyC;AACjF,QAAM,QAAQ,EAAE,SAAS;AACzB,aAAW,IAAI,SAAS,KAAK,KAAK;AAClC,SAAO,MAAM;AACX,QAAI,WAAW,IAAI,SAAS,GAAG,MAAM,MAAO,YAAW,OAAO,SAAS,GAAG;AAAA,EAC5E;AACF;AAEO,SAAS,4BAA4B,UAA2C;AACrF,QAAM,QAAQ,EAAE,SAAS;AACzB,eAAa,IAAI,SAAS,KAAK,KAAK;AACpC,SAAO,MAAM;AACX,QAAI,aAAa,IAAI,SAAS,GAAG,MAAM,MAAO,cAAa,OAAO,SAAS,GAAG;AAAA,EAChF;AACF;AAEO,SAAS,WACd,OACA,OACA,SACqB;AACrB,QAAM,kBAAkB,kBAAkB,IAAI,MAAM,GAAG,GAAG,SAAS;AACnE,MAAI,CAAC,gBAAiB,QAAO,QAAQ,OAAO,QAAQ,MAAM,GAAG,CAAC;AAC9D,SAAO,gBAAgB,OAAO,OAAO,OAAO;AAC9C;AAEO,SAAS,OACd,OACA,OACA,SAC0B;AAC1B,QAAM,cAAc,kBAAkB,IAAI,MAAM,GAAG,GAAG,SAAS;AAC/D,MAAI,CAAC,YAAa,OAAM,QAAQ,MAAM,GAAG;AACzC,MAAI,MAAM,kBAAkB,OAAQ,QAAO,cAAc,aAAa,OAAO,OAAO,OAAO;AAC3F,SAAO,YAAY,OAAO,OAAO,OAAO;AAC1C;AAEO,SAAS,MACd,OACA,OACA,SACmB;AACnB,QAAM,aAAa,WAAW,IAAI,MAAM,GAAG,GAAG,SAAS;AACvD,MAAI,CAAC,WAAY,OAAM,QAAQ,MAAM,GAAG;AACxC,MAAI,CAAC,MAAM,cAAe,QAAO,iBAAiB,YAAY,OAAO,OAAO,OAAO;AACnF,SAAO,WAAW,OAAO,OAAO,OAAO;AACzC;AAEO,SAAS,gBACd,OACA,SACA,SAC0B;AAC1B,QAAM,UAAU,aAAa,IAAI,MAAM,GAAG,GAAG,SAAS;AACtD,MAAI,CAAC,QAAS,QAAO,QAAQ,OAAO,QAAQ,MAAM,GAAG,CAAC;AACtD,SAAO,QAAQ,OAAO,SAAS,OAAO;AACxC;",
   "names": []
 }

package/dist/sentences.js

@@ -0,0 +1,88 @@
+const abbreviations = /* @__PURE__ */ new Set([
+  "mr",
+  "mrs",
+  "ms",
+  "dr",
+  "prof",
+  "sr",
+  "jr",
+  "st",
+  "vs",
+  "etc",
+  "eg",
+  "ie",
+  "fig",
+  "no"
+]);
+function wordBefore(text, index) {
+  let start = index;
+  while (start > 0 && /[A-Za-z]/.test(text[start - 1] ?? "")) start -= 1;
+  return text.slice(start, index).toLowerCase();
+}
+function boundary(text, index) {
+  const ch = text[index];
+  if (ch === "\n") return true;
+  if (ch !== "." && ch !== "!" && ch !== "?" && ch !== "\u2026") return false;
+  const previous = text[index - 1];
+  const next = text[index + 1];
+  if ((ch === "." || ch === "\u2026") && previous !== void 0 && next !== void 0 && previous >= "0" && previous <= "9" && next >= "0" && next <= "9") {
+    return false;
+  }
+  if (ch === "." && next === ".") return false;
+  if (ch === ".") {
+    const word = wordBefore(text, index);
+    if (abbreviations.has(word) || /^[A-Za-z]$/.test(word)) return false;
+  }
+  return true;
+}
+function splitSentences(buffer, flush) {
+  const sentences = [];
+  let start = 0;
+  let index = 0;
+  while (index < buffer.length) {
+    if (!boundary(buffer, index)) {
+      index += 1;
+      continue;
+    }
+    const ch = buffer[index];
+    const next = index + 1;
+    if (ch !== "\n" && next >= buffer.length && !flush) break;
+    if (ch !== "\n" && next < buffer.length && !/\s/.test(buffer[next] ?? "")) {
+      index += 1;
+      continue;
+    }
+    let end = next;
+    while (end < buffer.length && /\s/.test(buffer[end] ?? "")) end += 1;
+    const sentence = buffer.slice(start, end);
+    if (sentence.trim().length === 0) {
+      index = Math.max(end, index + 1);
+      continue;
+    }
+    sentences.push(sentence);
+    start = end;
+    index = end;
+  }
+  const rest = buffer.slice(start);
+  if (!flush) return { sentences, rest };
+  if (rest.length > 0) sentences.push(rest);
+  return { sentences, rest: "" };
+}
+async function* sentencesOf(text) {
+  if (typeof text === "string") {
+    for (const sentence of splitSentences(text, true).sentences) yield sentence;
+    return;
+  }
+  let buffer = "";
+  for await (const part of text) {
+    buffer += part;
+    const split = splitSentences(buffer, false);
+    buffer = split.rest;
+    for (const sentence of split.sentences) yield sentence;
+  }
+  for (const sentence of splitSentences(buffer, true).sentences) yield sentence;
+}
+export {
+  sentencesOf,
+  splitSentences
+};
+//# sourceMappingURL=sentences.js.map

package/dist/sentences.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../src/sentences.ts"],
+  "sourcesContent": ["const abbreviations = new Set([\n  'mr',\n  'mrs',\n  'ms',\n  'dr',\n  'prof',\n  'sr',\n  'jr',\n  'st',\n  'vs',\n  'etc',\n  'eg',\n  'ie',\n  'fig',\n  'no',\n]);\n\nfunction wordBefore(text: string, index: number): string {\n  let start = index;\n  while (start > 0 && /[A-Za-z]/.test(text[start - 1] ?? '')) start -= 1;\n  return text.slice(start, index).toLowerCase();\n}\n\nfunction boundary(text: string, index: number): boolean {\n  const ch = text[index];\n  if (ch === '\\n') return true;\n  if (ch !== '.' && ch !== '!' && ch !== '?' && ch !== '\u2026') return false;\n  const previous = text[index - 1];\n  const next = text[index + 1];\n  if (\n    (ch === '.' || ch === '\u2026') &&\n    previous !== undefined &&\n    next !== undefined &&\n    previous >= '0' &&\n    previous <= '9' &&\n    next >= '0' &&\n    next <= '9'\n  ) {\n    return false;\n  }\n  if (ch === '.' && next === '.') return false;\n  if (ch === '.') {\n    const word = wordBefore(text, index);\n    if (abbreviations.has(word) || /^[A-Za-z]$/.test(word)) return false;\n  }\n  return true;\n}\n\nexport function splitSentences(\n  buffer: string,\n  flush: boolean\n): { sentences: string[]; rest: string } {\n  const sentences: string[] = [];\n  let start = 0;\n  let index = 0;\n  while (index < buffer.length) {\n    if (!boundary(buffer, index)) {\n      index += 1;\n      continue;\n    }\n    const ch = buffer[index];\n    const next = index + 1;\n    if (ch !== '\\n' && next >= buffer.length && !flush) break;\n    if (ch !== '\\n' && next < buffer.length && !/\\s/.test(buffer[next] ?? '')) {\n      index += 1;\n      continue;\n    }\n    let end = next;\n    while (end < buffer.length && /\\s/.test(buffer[end] ?? '')) end += 1;\n    const sentence = buffer.slice(start, end);\n    if (sentence.trim().length === 0) {\n      index = Math.max(end, index + 1);\n      continue;\n    }\n    sentences.push(sentence);\n    start = end;\n    index = end;\n  }\n  const rest = buffer.slice(start);\n  if (!flush) return { sentences, rest };\n  if (rest.length > 0) sentences.push(rest);\n  return { sentences, rest: '' };\n}\n\nexport async function* sentencesOf(text: string | AsyncIterable<string>): AsyncGenerator<string> {\n  if (typeof text === 'string') {\n    for (const sentence of splitSentences(text, true).sentences) yield sentence;\n    return;\n  }\n  let buffer = '';\n  for await (const part of text) {\n    buffer += part;\n    const split = splitSentences(buffer, false);\n    buffer = split.rest;\n    for (const sentence of split.sentences) yield sentence;\n  }\n  for (const sentence of splitSentences(buffer, true).sentences) yield sentence;\n}\n"],
+  "mappings": "AAAA,MAAM,gBAAgB,oBAAI,IAAI;AAAA,EAC5B;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF,CAAC;AAED,SAAS,WAAW,MAAc,OAAuB;AACvD,MAAI,QAAQ;AACZ,SAAO,QAAQ,KAAK,WAAW,KAAK,KAAK,QAAQ,CAAC,KAAK,EAAE,EAAG,UAAS;AACrE,SAAO,KAAK,MAAM,OAAO,KAAK,EAAE,YAAY;AAC9C;AAEA,SAAS,SAAS,MAAc,OAAwB;AACtD,QAAM,KAAK,KAAK,KAAK;AACrB,MAAI,OAAO,KAAM,QAAO;AACxB,MAAI,OAAO,OAAO,OAAO,OAAO,OAAO,OAAO,OAAO,SAAK,QAAO;AACjE,QAAM,WAAW,KAAK,QAAQ,CAAC;AAC/B,QAAM,OAAO,KAAK,QAAQ,CAAC;AAC3B,OACG,OAAO,OAAO,OAAO,aACtB,aAAa,UACb,SAAS,UACT,YAAY,OACZ,YAAY,OACZ,QAAQ,OACR,QAAQ,KACR;AACA,WAAO;AAAA,EACT;AACA,MAAI,OAAO,OAAO,SAAS,IAAK,QAAO;AACvC,MAAI,OAAO,KAAK;AACd,UAAM,OAAO,WAAW,MAAM,KAAK;AACnC,QAAI,cAAc,IAAI,IAAI,KAAK,aAAa,KAAK,IAAI,EAAG,QAAO;AAAA,EACjE;AACA,SAAO;AACT;AAEO,SAAS,eACd,QACA,OACuC;AACvC,QAAM,YAAsB,CAAC;AAC7B,MAAI,QAAQ;AACZ,MAAI,QAAQ;AACZ,SAAO,QAAQ,OAAO,QAAQ;AAC5B,QAAI,CAAC,SAAS,QAAQ,KAAK,GAAG;AAC5B,eAAS;AACT;AAAA,IACF;AACA,UAAM,KAAK,OAAO,KAAK;AACvB,UAAM,OAAO,QAAQ;AACrB,QAAI,OAAO,QAAQ,QAAQ,OAAO,UAAU,CAAC,MAAO;AACpD,QAAI,OAAO,QAAQ,OAAO,OAAO,UAAU,CAAC,KAAK,KAAK,OAAO,IAAI,KAAK,EAAE,GAAG;AACzE,eAAS;AACT;AAAA,IACF;AACA,QAAI,MAAM;AACV,WAAO,MAAM,OAAO,UAAU,KAAK,KAAK,OAAO,GAAG,KAAK,EAAE,EAAG,QAAO;AACnE,UAAM,WAAW,OAAO,MAAM,OAAO,GAAG;AACxC,QAAI,SAAS,KAAK,EAAE,WAAW,GAAG;AAChC,cAAQ,KAAK,IAAI,KAAK,QAAQ,CAAC;AAC/B;AAAA,IACF;AACA,cAAU,KAAK,QAAQ;AACvB,YAAQ;AACR,YAAQ;AAAA,EACV;AACA,QAAM,OAAO,OAAO,MAAM,KAAK;AAC/B,MAAI,CAAC,MAAO,QAAO,EAAE,WAAW,KAAK;AACrC,MAAI,KAAK,SAAS,EAAG,WAAU,KAAK,IAAI;AACxC,SAAO,EAAE,WAAW,MAAM,GAAG;AAC/B;AAEA,gBAAuB,YAAY,MAA8D;AAC/F,MAAI,OAAO,SAAS,UAAU;AAC5B,eAAW,YAAY,eAAe,MAAM,IAAI,EAAE,UAAW,OAAM;AACnE;AAAA,EACF;AACA,MAAI,SAAS;AACb,mBAAiB,QAAQ,MAAM;AAC7B,cAAU;AACV,UAAM,QAAQ,eAAe,QAAQ,KAAK;AAC1C,aAAS,MAAM;AACf,eAAW,YAAY,MAAM,UAAW,OAAM;AAAA,EAChD;AACA,aAAW,YAAY,eAAe,QAAQ,IAAI,EAAE,UAAW,OAAM;AACvE;",
+  "names": []
+}

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/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": "1.4.0",
   "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",