@ai-ecoverse/pi-audio 1.12.0 → 1.13.0

package/README.md

@@ -23,6 +23,8 @@ pi-audio does not apply that patch. Hosts such as slicc-agent record monkeypatch
 
 `@ai-ecoverse/pi-audio/durable` appends a `voice.truncate` entry for a `VoiceCut`. The entry has no model messages. Its `edits` replace the target assistant message with the text heard so far and keep tool calls. The stored entry stays intact, and pi-durable derives the shorter message for the next model request. `@earendil-works/pi-durable` is an optional peer (`^1.1.0`); the root package does not import it. `VoiceCut.entryId` is the decimal string of pi-durable's numeric entry id.
 
+`@ai-ecoverse/pi-audio/elevenlabs` registers ElevenLabs speech adapters when `registerElevenLabs()` is called. `scribe_v2` posts multipart audio to `/speech-to-text`. `scribe_v2_realtime` streams base64 audio on `wss://…/speech-to-text/realtime` and commits turns with `turn_prediction` (`turnDetection: 'semantic'`). `eleven_flash_v2_5` and `eleven_multilingual_v2` post JSON to `/text-to-speech/{voice}/with-timestamps` for a string and stream text on `stream-input` for an async iterable, with character alignment. `eleven_v3` and `eleven_v4` stay on the HTTP path (`textStreaming: false`). Scribe prices are $0.22 per hour for batch and $0.39 per hour for realtime.
+
 `@ai-ecoverse/pi-audio/xai` registers xAI speech adapters when `registerXAI()` is called. `grok-voice-transcribe-2.0` posts multipart audio to `/stt` (fields before `file`) and streams raw binary frames on `wss://…/stt` after `transcript.created`, with Smart Turn (`turnDetection: 'semantic'`). The catalog id is shared, so `audioCost` uses the streaming rate of $0.20 per hour. `grok-voice-tts-1.0` posts JSON to `/tts` for a string and streams `text.delta` / `audio.delta` on `wss://…/tts` for an async iterable (`textStreaming: true`). Voices are eve, ara, rex, sal, and leo. Codecs are mp3, pcm, wav, mulaw, and alaw.
 
 `@ai-ecoverse/pi-audio/openai` registers OpenAI and OpenRouter speech adapters when `registerOpenAI()` is called. Batch transcription posts multipart audio to `/audio/transcriptions`. OpenRouter (`provider` `openrouter`, or a base URL on `openrouter.ai`) posts JSON `input_audio` instead. Speech posts JSON to `/audio/speech` and reads the raw audio response. OpenAI can return pcm, mp3, wav, opus, or flac. OpenRouter returns pcm or mp3. Realtime transcription uses a WebSocket at `/realtime?intent=transcription` for `gpt-realtime-whisper` and `gpt-live-transcribe`. The catalog key is type, provider, and id, so the batch model keeps the id `gpt-4o-transcribe`.

package/dist/elevenlabs/catalog.js

@@ -0,0 +1,136 @@
+import {
+  registerVoiceModel
+} from "../models.js";
+import { registerSpeechApiProvider, registerTranscriptionApiProvider } from "../registry.js";
+import { listenElevenLabs } from "./listen.js";
+import { speakElevenLabs } from "./speak.js";
+import { transcribeElevenLabs } from "./transcribe.js";
+const tokenCost = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };
+const eleven = "https://api.elevenlabs.io/v1";
+const pcm16 = (sampleRate) => ({
+  encoding: "pcm16",
+  sampleRate,
+  channels: 1
+});
+const law = (encoding) => ({
+  encoding,
+  sampleRate: 8e3,
+  channels: 1
+});
+const container = (encoding, sampleRate) => ({
+  encoding,
+  sampleRate,
+  channels: 1
+});
+const voices = [
+  { id: "21m00Tcm4TlvDq8ikWAM", name: "Rachel" },
+  { id: "pNInz6obpgDQGcFmaJgB", name: "Adam" },
+  { id: "EXAVITQu4vr4xnSDxMaL", name: "Bella" },
+  { id: "ErXwobaYiN019PkySvjV", name: "Antoni" },
+  { id: "TxGEqnHWrfWFTfGW9XjX", name: "Josh" }
+];
+const speechFormats = [
+  container("mp3", 44100),
+  container("mp3", 24e3),
+  container("mp3", 22050),
+  pcm16(16e3),
+  pcm16(24e3),
+  pcm16(8e3),
+  container("wav", 16e3),
+  law("mulaw"),
+  law("alaw"),
+  container("opus", 48e3)
+];
+function speech(id, textStreaming, perThousandChars) {
+  return {
+    id,
+    name: id,
+    api: "elevenlabs-tts",
+    provider: "elevenlabs",
+    baseUrl: eleven,
+    type: "speech",
+    textStreaming,
+    outputFormats: speechFormats,
+    voices,
+    cloning: true,
+    cost: tokenCost,
+    audioCost: { perThousandChars }
+  };
+}
+function elevenCatalog() {
+  const batch = {
+    id: "scribe_v2",
+    name: "scribe_v2",
+    api: "elevenlabs-scribe",
+    provider: "elevenlabs",
+    baseUrl: eleven,
+    type: "transcription",
+    streaming: false,
+    turnDetection: "none",
+    inputFormats: [
+      container("wav", 16e3),
+      container("mp3", 16e3),
+      container("flac", 16e3),
+      pcm16(16e3),
+      law("mulaw")
+    ],
+    diarization: true,
+    cost: tokenCost,
+    audioCost: { perInputMinute: 0.22 / 60 }
+  };
+  const realtime = {
+    id: "scribe_v2_realtime",
+    name: "scribe_v2_realtime",
+    api: "elevenlabs-scribe",
+    provider: "elevenlabs",
+    baseUrl: eleven,
+    type: "transcription",
+    streaming: true,
+    turnDetection: "semantic",
+    inputFormats: [
+      pcm16(8e3),
+      pcm16(16e3),
+      pcm16(22050),
+      pcm16(24e3),
+      pcm16(44100),
+      pcm16(48e3),
+      law("mulaw")
+    ],
+    diarization: true,
+    cost: tokenCost,
+    audioCost: { perInputMinute: 0.39 / 60 }
+  };
+  return [
+    batch,
+    realtime,
+    speech("eleven_flash_v2_5", true, 0.05),
+    speech("eleven_multilingual_v2", true, 0.1),
+    speech("eleven_v3", false, 0.1),
+    speech("eleven_v4", false, 0.1)
+  ];
+}
+function registerElevenLabs() {
+  const models = elevenCatalog();
+  const unregisterModels = models.map((model) => registerVoiceModel(model));
+  const unregisterScribe = registerTranscriptionApiProvider({
+    api: "elevenlabs-scribe",
+    transcribe: transcribeElevenLabs,
+    listen: listenElevenLabs
+  });
+  const unregisterSpeech = registerSpeechApiProvider({
+    api: "elevenlabs-tts",
+    speak: speakElevenLabs
+  });
+  let released = false;
+  return () => {
+    if (released) return;
+    released = true;
+    for (const unregister of unregisterModels) unregister();
+    unregisterScribe();
+    unregisterSpeech();
+  };
+}
+export {
+  registerElevenLabs
+};
+//# sourceMappingURL=catalog.js.map

package/dist/elevenlabs/catalog.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/elevenlabs/catalog.ts"],
+  "sourcesContent": ["import type { AudioFormat } from '../audio.ts';\nimport {\n  registerVoiceModel,\n  type SpeechModel,\n  type TranscriptionModel,\n  type VoiceInfo,\n} from '../models.ts';\nimport { registerSpeechApiProvider, registerTranscriptionApiProvider } from '../registry.ts';\nimport { listenElevenLabs } from './listen.ts';\nimport { speakElevenLabs } from './speak.ts';\nimport { transcribeElevenLabs } from './transcribe.ts';\n\nconst tokenCost = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };\nconst eleven = 'https://api.elevenlabs.io/v1';\n\nconst pcm16 = (sampleRate: number): AudioFormat => ({\n  encoding: 'pcm16',\n  sampleRate,\n  channels: 1,\n});\n\nconst law = (encoding: 'mulaw' | 'alaw'): AudioFormat => ({\n  encoding,\n  sampleRate: 8000,\n  channels: 1,\n});\n\nconst container = (encoding: 'wav' | 'mp3' | 'flac' | 'opus', sampleRate: number): AudioFormat => ({\n  encoding,\n  sampleRate,\n  channels: 1,\n});\n\nconst voices: VoiceInfo[] = [\n  { id: '21m00Tcm4TlvDq8ikWAM', name: 'Rachel' },\n  { id: 'pNInz6obpgDQGcFmaJgB', name: 'Adam' },\n  { id: 'EXAVITQu4vr4xnSDxMaL', name: 'Bella' },\n  { id: 'ErXwobaYiN019PkySvjV', name: 'Antoni' },\n  { id: 'TxGEqnHWrfWFTfGW9XjX', name: 'Josh' },\n];\n\nconst speechFormats: AudioFormat[] = [\n  container('mp3', 44100),\n  container('mp3', 24000),\n  container('mp3', 22050),\n  pcm16(16000),\n  pcm16(24000),\n  pcm16(8000),\n  container('wav', 16000),\n  law('mulaw'),\n  law('alaw'),\n  container('opus', 48000),\n];\n\nfunction speech(id: string, textStreaming: boolean, perThousandChars: number): SpeechModel<string> {\n  return {\n    id,\n    name: id,\n    api: 'elevenlabs-tts',\n    provider: 'elevenlabs',\n    baseUrl: eleven,\n    type: 'speech',\n    textStreaming,\n    outputFormats: speechFormats,\n    voices,\n    cloning: true,\n    cost: tokenCost,\n    audioCost: { perThousandChars },\n  };\n}\n\nfunction elevenCatalog(): Array<TranscriptionModel<string> | SpeechModel<string>> {\n  const batch: TranscriptionModel<string> = {\n    id: 'scribe_v2',\n    name: 'scribe_v2',\n    api: 'elevenlabs-scribe',\n    provider: 'elevenlabs',\n    baseUrl: eleven,\n    type: 'transcription',\n    streaming: false,\n    turnDetection: 'none',\n    inputFormats: [\n      container('wav', 16000),\n      container('mp3', 16000),\n      container('flac', 16000),\n      pcm16(16000),\n      law('mulaw'),\n    ],\n    diarization: true,\n    cost: tokenCost,\n    audioCost: { perInputMinute: 0.22 / 60 },\n  };\n  const realtime: TranscriptionModel<string> = {\n    id: 'scribe_v2_realtime',\n    name: 'scribe_v2_realtime',\n    api: 'elevenlabs-scribe',\n    provider: 'elevenlabs',\n    baseUrl: eleven,\n    type: 'transcription',\n    streaming: true,\n    turnDetection: 'semantic',\n    inputFormats: [\n      pcm16(8000),\n      pcm16(16000),\n      pcm16(22050),\n      pcm16(24000),\n      pcm16(44100),\n      pcm16(48000),\n      law('mulaw'),\n    ],\n    diarization: true,\n    cost: tokenCost,\n    audioCost: { perInputMinute: 0.39 / 60 },\n  };\n  return [\n    batch,\n    realtime,\n    speech('eleven_flash_v2_5', true, 0.05),\n    speech('eleven_multilingual_v2', true, 0.1),\n    speech('eleven_v3', false, 0.1),\n    speech('eleven_v4', false, 0.1),\n  ];\n}\n\nexport function registerElevenLabs(): () => void {\n  const models = elevenCatalog();\n  const unregisterModels = models.map((model) => registerVoiceModel(model));\n  const unregisterScribe = registerTranscriptionApiProvider({\n    api: 'elevenlabs-scribe',\n    transcribe: transcribeElevenLabs,\n    listen: listenElevenLabs,\n  });\n  const unregisterSpeech = registerSpeechApiProvider({\n    api: 'elevenlabs-tts',\n    speak: speakElevenLabs,\n  });\n  let released = false;\n  return () => {\n    if (released) return;\n    released = true;\n    for (const unregister of unregisterModels) unregister();\n    unregisterScribe();\n    unregisterSpeech();\n  };\n}\n"],
+  "mappings": "AACA;AAAA,EACE;AAAA,OAIK;AACP,SAAS,2BAA2B,wCAAwC;AAC5E,SAAS,wBAAwB;AACjC,SAAS,uBAAuB;AAChC,SAAS,4BAA4B;AAErC,MAAM,YAAY,EAAE,OAAO,GAAG,QAAQ,GAAG,WAAW,GAAG,YAAY,EAAE;AACrE,MAAM,SAAS;AAEf,MAAM,QAAQ,CAAC,gBAAqC;AAAA,EAClD,UAAU;AAAA,EACV;AAAA,EACA,UAAU;AACZ;AAEA,MAAM,MAAM,CAAC,cAA6C;AAAA,EACxD;AAAA,EACA,YAAY;AAAA,EACZ,UAAU;AACZ;AAEA,MAAM,YAAY,CAAC,UAA2C,gBAAqC;AAAA,EACjG;AAAA,EACA;AAAA,EACA,UAAU;AACZ;AAEA,MAAM,SAAsB;AAAA,EAC1B,EAAE,IAAI,wBAAwB,MAAM,SAAS;AAAA,EAC7C,EAAE,IAAI,wBAAwB,MAAM,OAAO;AAAA,EAC3C,EAAE,IAAI,wBAAwB,MAAM,QAAQ;AAAA,EAC5C,EAAE,IAAI,wBAAwB,MAAM,SAAS;AAAA,EAC7C,EAAE,IAAI,wBAAwB,MAAM,OAAO;AAC7C;AAEA,MAAM,gBAA+B;AAAA,EACnC,UAAU,OAAO,KAAK;AAAA,EACtB,UAAU,OAAO,IAAK;AAAA,EACtB,UAAU,OAAO,KAAK;AAAA,EACtB,MAAM,IAAK;AAAA,EACX,MAAM,IAAK;AAAA,EACX,MAAM,GAAI;AAAA,EACV,UAAU,OAAO,IAAK;AAAA,EACtB,IAAI,OAAO;AAAA,EACX,IAAI,MAAM;AAAA,EACV,UAAU,QAAQ,IAAK;AACzB;AAEA,SAAS,OAAO,IAAY,eAAwB,kBAA+C;AACjG,SAAO;AAAA,IACL;AAAA,IACA,MAAM;AAAA,IACN,KAAK;AAAA,IACL,UAAU;AAAA,IACV,SAAS;AAAA,IACT,MAAM;AAAA,IACN;AAAA,IACA,eAAe;AAAA,IACf;AAAA,IACA,SAAS;AAAA,IACT,MAAM;AAAA,IACN,WAAW,EAAE,iBAAiB;AAAA,EAChC;AACF;AAEA,SAAS,gBAAyE;AAChF,QAAM,QAAoC;AAAA,IACxC,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,KAAK;AAAA,IACL,UAAU;AAAA,IACV,SAAS;AAAA,IACT,MAAM;AAAA,IACN,WAAW;AAAA,IACX,eAAe;AAAA,IACf,cAAc;AAAA,MACZ,UAAU,OAAO,IAAK;AAAA,MACtB,UAAU,OAAO,IAAK;AAAA,MACtB,UAAU,QAAQ,IAAK;AAAA,MACvB,MAAM,IAAK;AAAA,MACX,IAAI,OAAO;AAAA,IACb;AAAA,IACA,aAAa;AAAA,IACb,MAAM;AAAA,IACN,WAAW,EAAE,gBAAgB,OAAO,GAAG;AAAA,EACzC;AACA,QAAM,WAAuC;AAAA,IAC3C,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,KAAK;AAAA,IACL,UAAU;AAAA,IACV,SAAS;AAAA,IACT,MAAM;AAAA,IACN,WAAW;AAAA,IACX,eAAe;AAAA,IACf,cAAc;AAAA,MACZ,MAAM,GAAI;AAAA,MACV,MAAM,IAAK;AAAA,MACX,MAAM,KAAK;AAAA,MACX,MAAM,IAAK;AAAA,MACX,MAAM,KAAK;AAAA,MACX,MAAM,IAAK;AAAA,MACX,IAAI,OAAO;AAAA,IACb;AAAA,IACA,aAAa;AAAA,IACb,MAAM;AAAA,IACN,WAAW,EAAE,gBAAgB,OAAO,GAAG;AAAA,EACzC;AACA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA,OAAO,qBAAqB,MAAM,IAAI;AAAA,IACtC,OAAO,0BAA0B,MAAM,GAAG;AAAA,IAC1C,OAAO,aAAa,OAAO,GAAG;AAAA,IAC9B,OAAO,aAAa,OAAO,GAAG;AAAA,EAChC;AACF;AAEO,SAAS,qBAAiC;AAC/C,QAAM,SAAS,cAAc;AAC7B,QAAM,mBAAmB,OAAO,IAAI,CAAC,UAAU,mBAAmB,KAAK,CAAC;AACxE,QAAM,mBAAmB,iCAAiC;AAAA,IACxD,KAAK;AAAA,IACL,YAAY;AAAA,IACZ,QAAQ;AAAA,EACV,CAAC;AACD,QAAM,mBAAmB,0BAA0B;AAAA,IACjD,KAAK;AAAA,IACL,OAAO;AAAA,EACT,CAAC;AACD,MAAI,WAAW;AACf,SAAO,MAAM;AACX,QAAI,SAAU;AACd,eAAW;AACX,eAAW,cAAc,iBAAkB,YAAW;AACtD,qBAAiB;AACjB,qBAAiB;AAAA,EACnB;AACF;",
+  "names": []
+}

package/dist/elevenlabs/index.js

@@ -0,0 +1,5 @@
+import { registerElevenLabs } from "./catalog.js";
+export {
+  registerElevenLabs
+};
+//# sourceMappingURL=index.js.map

package/dist/elevenlabs/index.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/elevenlabs/index.ts"],
+  "sourcesContent": ["export { registerElevenLabs } from './catalog.ts';\n"],
+  "mappings": "AAAA,SAAS,0BAA0B;",
+  "names": []
+}

package/dist/elevenlabs/listen.js

@@ -0,0 +1,268 @@
+import { chunkSeconds } from "../audio.js";
+import { encodeAudioBase64 } from "../convert.js";
+import { asVoiceError, isRecord, numberField, stringField } from "../openai/wire.js";
+import {
+  TranscriptionEventStream
+} from "../transcription.js";
+import { priceAudio } from "../usage.js";
+import {
+  awaitOpen,
+  connectSocket,
+  messageText,
+  sameFormat,
+  socketUrl,
+  sttAudioFormat,
+  xiHeaders
+} from "./wire.js";
+const failures = /* @__PURE__ */ new Set([
+  "error",
+  "auth_error",
+  "quota_exceeded",
+  "commit_throttled",
+  "unaccepted_terms",
+  "rate_limited",
+  "queue_overflow",
+  "resource_exhausted",
+  "session_time_limit_exceeded",
+  "input_error",
+  "invalid_request",
+  "chunk_size_exceeded",
+  "insufficient_audio_activity",
+  "transcriber_error"
+]);
+function terms(options) {
+  const vocabulary = options?.vocabulary;
+  if (!vocabulary) return [];
+  return vocabulary;
+}
+function commitStrategy(model) {
+  if (model.turnDetection === "semantic") return "turn_prediction";
+  if (model.turnDetection === "vad") return "vad";
+  return "manual";
+}
+function sttUrl(model, format, options) {
+  const url = socketUrl(model.baseUrl, "speech-to-text/realtime");
+  const params = new URLSearchParams();
+  params.set("model_id", model.id);
+  params.set("audio_format", sttAudioFormat(format));
+  params.set("commit_strategy", commitStrategy(model));
+  params.set("include_timestamps", "true");
+  if (options?.language) params.set("language_code", options.language);
+  for (const term of terms(options)) params.append("keyterms", term);
+  url.search = params.toString();
+  return url.toString();
+}
+function chosenFormat(model, chunk) {
+  const format = chunk === void 0 ? model.inputFormats[0] : chunk.format;
+  if (!format) throw new Error("No input format");
+  sttAudioFormat(format);
+  const known = model.inputFormats.some((item) => sameFormat(item, format));
+  if (!known) throw new Error(`Unsupported audio format ${format.encoding}`);
+  return format;
+}
+function textOr(event) {
+  const text = stringField(event, "text");
+  if (text === void 0) return "";
+  return text;
+}
+function secondsOf(word, key) {
+  const value = numberField(word, key);
+  if (value === void 0) return 0;
+  return value;
+}
+function wordFrom(word) {
+  if (stringField(word, "type") === "spacing") return void 0;
+  const speaker = stringField(word, "speaker_id");
+  const segment = {
+    text: textOr(word),
+    start: secondsOf(word, "start"),
+    end: secondsOf(word, "end")
+  };
+  if (speaker !== void 0) segment.speaker = speaker;
+  return segment;
+}
+function wordsOf(event, state) {
+  if (!Array.isArray(event.words)) return;
+  for (const word of event.words) {
+    if (!isRecord(word)) continue;
+    const segment = wordFrom(word);
+    if (segment) state.segments.push(segment);
+  }
+}
+function emitCommit(spoken, state, push) {
+  if (!state.started) {
+    state.started = true;
+    push({ type: "speech_start", at: 0 });
+  }
+  const segmentId = String(state.segment);
+  state.segment += 1;
+  state.turns.push(spoken);
+  push({ type: "final", segmentId, text: spoken });
+  state.started = false;
+  push({ type: "turn_end", text: spoken, at: 0 });
+}
+function applyListen(event, state, push) {
+  const type = stringField(event, "message_type");
+  if (type === "session_started") {
+    state.ready = true;
+    return;
+  }
+  if (type === "partial_transcript") {
+    if (!state.started) {
+      state.started = true;
+      push({ type: "speech_start", at: 0 });
+    }
+    push({ type: "partial", segmentId: String(state.segment), text: textOr(event) });
+    return;
+  }
+  if (type === "committed_transcript") {
+    const spoken = textOr(event);
+    state.lastCommit = spoken;
+    emitCommit(spoken, state, push);
+    return;
+  }
+  if (type === "committed_transcript_with_timestamps") {
+    const spoken = textOr(event);
+    wordsOf(event, state);
+    if (state.lastCommit !== spoken) emitCommit(spoken, state, push);
+    state.lastCommit = void 0;
+    return;
+  }
+  if (type !== void 0 && failures.has(type)) {
+    const message = stringField(event, "error");
+    state.failed = true;
+    push({
+      type: "error",
+      reason: "error",
+      error: message === void 0 ? "ElevenLabs transcription failed" : message
+    });
+  }
+}
+function finish(model, format, state, push) {
+  if (state.failed || state.completed) return;
+  state.completed = true;
+  const heard = chunkSeconds({ data: new Uint8Array(state.bytes), format, timestamp: 0 });
+  push({
+    type: "done",
+    transcript: {
+      text: state.turns.join(" "),
+      ...state.segments.length > 0 ? { segments: state.segments } : {},
+      usage: priceAudio(model.audioCost, { inputSeconds: heard })
+    }
+  });
+}
+function listenElevenLabs(model, audio, options) {
+  const stream = new TranscriptionEventStream();
+  void pumpListen(model, audio, options, stream);
+  return stream;
+}
+async function pumpListen(model, audio, options, stream) {
+  const state = {
+    turns: [],
+    segments: [],
+    segment: 0,
+    started: false,
+    bytes: 0,
+    opened: false,
+    ready: false,
+    clientDone: false,
+    failed: false,
+    completed: false,
+    lastCommit: void 0
+  };
+  let settled = false;
+  const push = (event) => {
+    if (settled) return;
+    if (event.type === "done" || event.type === "error") settled = true;
+    stream.push(event);
+  };
+  let socket;
+  const readyWait = Promise.withResolvers();
+  const doneWait = Promise.withResolvers();
+  const wake = (kind) => {
+    if (state.ready || state.failed) readyWait.resolve();
+    if (state.failed) doneWait.resolve();
+    if (!state.clientDone) return;
+    if (kind === "committed_transcript" || kind === "committed_transcript_with_timestamps") {
+      doneWait.resolve();
+    }
+    if (kind === "close") doneWait.resolve();
+  };
+  try {
+    if (options?.signal?.aborted) throw new DOMException("aborted", "AbortError");
+    const iterator = audio[Symbol.asyncIterator]();
+    const first = await iterator.next();
+    const format = chosenFormat(model, first.done ? void 0 : first.value);
+    socket = connectSocket(sttUrl(model, format, options), xiHeaders(model, options));
+    let incoming = Promise.resolve();
+    socket.addEventListener("message", (event) => {
+      incoming = incoming.then(() => onMessage(event.data, state, push, wake));
+    });
+    socket.addEventListener("close", () => {
+      if (state.opened && !state.failed && !state.completed && !state.clientDone) {
+        state.failed = true;
+        push({ type: "error", reason: "error", error: "ElevenLabs transcription socket closed" });
+      }
+      wake("close");
+    });
+    await awaitOpen(socket, "ElevenLabs transcription", (message) => {
+      state.failed = true;
+      push({ type: "error", reason: "error", error: message });
+      wake("fail");
+    });
+    if (state.failed) return;
+    state.opened = true;
+    if (!state.ready) await readyWait.promise;
+    if (state.failed) return;
+    if (!first.done) sendChunk(socket, first.value, format, state, false);
+    while (!state.failed) {
+      if (options?.signal?.aborted) throw new DOMException("aborted", "AbortError");
+      const next = await iterator.next();
+      if (next.done) break;
+      sendChunk(socket, next.value, format, state, false);
+    }
+    if (state.failed) return;
+    sendChunk(socket, void 0, format, state, true);
+    state.clientDone = true;
+    await doneWait.promise;
+    finish(model, format, state, push);
+  } catch (error) {
+    const voice = asVoiceError(error);
+    state.failed = true;
+    push({ type: "error", reason: voice.reason, error: voice.error });
+  } finally {
+    socket?.close();
+  }
+}
+function sendChunk(socket, chunk, format, state, commit) {
+  if (chunk && !sameFormat(chunk.format, format)) {
+    throw new Error(`Unsupported audio format ${chunk.format.encoding}`);
+  }
+  const data = chunk ? chunk.data : new Uint8Array();
+  state.bytes += data.byteLength;
+  socket.send(
+    JSON.stringify({
+      message_type: "input_audio_chunk",
+      audio_base_64: encodeAudioBase64(data),
+      commit,
+      sample_rate: format.sampleRate
+    })
+  );
+}
+async function onMessage(data, state, push, wake) {
+  try {
+    const parsed = JSON.parse(await messageText(data));
+    if (!isRecord(parsed)) throw new Error("ElevenLabs transcription event was not JSON");
+    applyListen(parsed, state, push);
+    wake(stringField(parsed, "message_type"));
+  } catch (error) {
+    const voice = asVoiceError(error);
+    state.failed = true;
+    push({ type: "error", reason: voice.reason, error: voice.error });
+    wake("fail");
+  }
+}
+export {
+  listenElevenLabs
+};
+//# sourceMappingURL=listen.js.map

package/dist/elevenlabs/listen.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/elevenlabs/listen.ts"],
+  "sourcesContent": ["import type { AudioChunk, AudioFormat } from '../audio.ts';\nimport { chunkSeconds } from '../audio.ts';\nimport { encodeAudioBase64 } from '../convert.ts';\nimport type { TranscriptionModel } from '../models.ts';\nimport { asVoiceError, isRecord, numberField, stringField } from '../openai/wire.ts';\nimport {\n  type ListenOptions,\n  type TranscriptionEvent,\n  TranscriptionEventStream,\n  type TranscriptSegment,\n} from '../transcription.ts';\nimport { priceAudio } from '../usage.ts';\nimport {\n  awaitOpen,\n  connectSocket,\n  messageText,\n  sameFormat,\n  socketUrl,\n  sttAudioFormat,\n  xiHeaders,\n} from './wire.ts';\n\nconst failures = new Set([\n  'error',\n  'auth_error',\n  'quota_exceeded',\n  'commit_throttled',\n  'unaccepted_terms',\n  'rate_limited',\n  'queue_overflow',\n  'resource_exhausted',\n  'session_time_limit_exceeded',\n  'input_error',\n  'invalid_request',\n  'chunk_size_exceeded',\n  'insufficient_audio_activity',\n  'transcriber_error',\n]);\n\ninterface ListenState {\n  turns: string[];\n  segments: TranscriptSegment[];\n  segment: number;\n  started: boolean;\n  bytes: number;\n  opened: boolean;\n  ready: boolean;\n  clientDone: boolean;\n  failed: boolean;\n  completed: boolean;\n  lastCommit: string | undefined;\n}\n\nfunction terms(options: ListenOptions | undefined): string[] {\n  const vocabulary = options?.vocabulary;\n  if (!vocabulary) return [];\n  return vocabulary;\n}\n\nfunction commitStrategy(model: TranscriptionModel<string>): string {\n  if (model.turnDetection === 'semantic') return 'turn_prediction';\n  if (model.turnDetection === 'vad') return 'vad';\n  return 'manual';\n}\n\nfunction sttUrl(\n  model: TranscriptionModel<string>,\n  format: AudioFormat,\n  options: ListenOptions | undefined\n): string {\n  const url = socketUrl(model.baseUrl, 'speech-to-text/realtime');\n  const params = new URLSearchParams();\n  params.set('model_id', model.id);\n  params.set('audio_format', sttAudioFormat(format));\n  params.set('commit_strategy', commitStrategy(model));\n  params.set('include_timestamps', 'true');\n  if (options?.language) params.set('language_code', options.language);\n  for (const term of terms(options)) params.append('keyterms', term);\n  url.search = params.toString();\n  return url.toString();\n}\n\nfunction chosenFormat(\n  model: TranscriptionModel<string>,\n  chunk: AudioChunk | undefined\n): AudioFormat {\n  const format = chunk === undefined ? model.inputFormats[0] : chunk.format;\n  if (!format) throw new Error('No input format');\n  sttAudioFormat(format);\n  const known = model.inputFormats.some((item) => sameFormat(item, format));\n  if (!known) throw new Error(`Unsupported audio format ${format.encoding}`);\n  return format;\n}\n\nfunction textOr(event: Record<string, unknown>): string {\n  const text = stringField(event, 'text');\n  if (text === undefined) return '';\n  return text;\n}\n\nfunction secondsOf(word: Record<string, unknown>, key: string): number {\n  const value = numberField(word, key);\n  if (value === undefined) return 0;\n  return value;\n}\n\nfunction wordFrom(word: Record<string, unknown>): TranscriptSegment | undefined {\n  if (stringField(word, 'type') === 'spacing') return undefined;\n  const speaker = stringField(word, 'speaker_id');\n  const segment: TranscriptSegment = {\n    text: textOr(word),\n    start: secondsOf(word, 'start'),\n    end: secondsOf(word, 'end'),\n  };\n  if (speaker !== undefined) segment.speaker = speaker;\n  return segment;\n}\n\nfunction wordsOf(event: Record<string, unknown>, state: ListenState) {\n  if (!Array.isArray(event.words)) return;\n  for (const word of event.words) {\n    if (!isRecord(word)) continue;\n    const segment = wordFrom(word);\n    if (segment) state.segments.push(segment);\n  }\n}\n\nfunction emitCommit(spoken: string, state: ListenState, push: (event: TranscriptionEvent) => void) {\n  if (!state.started) {\n    state.started = true;\n    push({ type: 'speech_start', at: 0 });\n  }\n  const segmentId = String(state.segment);\n  state.segment += 1;\n  state.turns.push(spoken);\n  push({ type: 'final', segmentId, text: spoken });\n  state.started = false;\n  push({ type: 'turn_end', text: spoken, at: 0 });\n}\n\nfunction applyListen(\n  event: Record<string, unknown>,\n  state: ListenState,\n  push: (event: TranscriptionEvent) => void\n) {\n  const type = stringField(event, 'message_type');\n  if (type === 'session_started') {\n    state.ready = true;\n    return;\n  }\n  if (type === 'partial_transcript') {\n    if (!state.started) {\n      state.started = true;\n      push({ type: 'speech_start', at: 0 });\n    }\n    push({ type: 'partial', segmentId: String(state.segment), text: textOr(event) });\n    return;\n  }\n  if (type === 'committed_transcript') {\n    const spoken = textOr(event);\n    state.lastCommit = spoken;\n    emitCommit(spoken, state, push);\n    return;\n  }\n  if (type === 'committed_transcript_with_timestamps') {\n    const spoken = textOr(event);\n    wordsOf(event, state);\n    if (state.lastCommit !== spoken) emitCommit(spoken, state, push);\n    state.lastCommit = undefined;\n    return;\n  }\n  if (type !== undefined && failures.has(type)) {\n    const message = stringField(event, 'error');\n    state.failed = true;\n    push({\n      type: 'error',\n      reason: 'error',\n      error: message === undefined ? 'ElevenLabs transcription failed' : message,\n    });\n  }\n}\n\nfunction finish(\n  model: TranscriptionModel<string>,\n  format: AudioFormat,\n  state: ListenState,\n  push: (event: TranscriptionEvent) => void\n) {\n  if (state.failed || state.completed) return;\n  state.completed = true;\n  const heard = chunkSeconds({ data: new Uint8Array(state.bytes), format, timestamp: 0 });\n  push({\n    type: 'done',\n    transcript: {\n      text: state.turns.join(' '),\n      ...(state.segments.length > 0 ? { segments: state.segments } : {}),\n      usage: priceAudio(model.audioCost, { inputSeconds: heard }),\n    },\n  });\n}\n\nexport function listenElevenLabs(\n  model: TranscriptionModel<string>,\n  audio: AsyncIterable<AudioChunk>,\n  options?: ListenOptions\n): TranscriptionEventStream {\n  const stream = new TranscriptionEventStream();\n  void pumpListen(model, audio, options, stream);\n  return stream;\n}\n\nasync function pumpListen(\n  model: TranscriptionModel<string>,\n  audio: AsyncIterable<AudioChunk>,\n  options: ListenOptions | undefined,\n  stream: TranscriptionEventStream\n) {\n  const state: ListenState = {\n    turns: [],\n    segments: [],\n    segment: 0,\n    started: false,\n    bytes: 0,\n    opened: false,\n    ready: false,\n    clientDone: false,\n    failed: false,\n    completed: false,\n    lastCommit: undefined,\n  };\n  let settled = false;\n  const push = (event: TranscriptionEvent) => {\n    if (settled) return;\n    if (event.type === 'done' || event.type === 'error') settled = true;\n    stream.push(event);\n  };\n  let socket: WebSocket | undefined;\n  const readyWait = Promise.withResolvers<void>();\n  const doneWait = Promise.withResolvers<void>();\n  const wake = (kind: string | undefined) => {\n    if (state.ready || state.failed) readyWait.resolve();\n    if (state.failed) doneWait.resolve();\n    if (!state.clientDone) return;\n    if (kind === 'committed_transcript' || kind === 'committed_transcript_with_timestamps') {\n      doneWait.resolve();\n    }\n    if (kind === 'close') doneWait.resolve();\n  };\n  try {\n    if (options?.signal?.aborted) throw new DOMException('aborted', 'AbortError');\n    const iterator = audio[Symbol.asyncIterator]();\n    const first = await iterator.next();\n    const format = chosenFormat(model, first.done ? undefined : first.value);\n    socket = connectSocket(sttUrl(model, format, options), xiHeaders(model, options));\n    let incoming = Promise.resolve();\n    socket.addEventListener('message', (event) => {\n      incoming = incoming.then(() => onMessage(event.data, state, push, wake));\n    });\n    socket.addEventListener('close', () => {\n      if (state.opened && !state.failed && !state.completed && !state.clientDone) {\n        state.failed = true;\n        push({ type: 'error', reason: 'error', error: 'ElevenLabs transcription socket closed' });\n      }\n      wake('close');\n    });\n    await awaitOpen(socket, 'ElevenLabs transcription', (message) => {\n      state.failed = true;\n      push({ type: 'error', reason: 'error', error: message });\n      wake('fail');\n    });\n    if (state.failed) return;\n    state.opened = true;\n    if (!state.ready) await readyWait.promise;\n    if (state.failed) return;\n    if (!first.done) sendChunk(socket, first.value, format, state, false);\n    while (!state.failed) {\n      if (options?.signal?.aborted) throw new DOMException('aborted', 'AbortError');\n      const next = await iterator.next();\n      if (next.done) break;\n      sendChunk(socket, next.value, format, state, false);\n    }\n    if (state.failed) return;\n    sendChunk(socket, undefined, format, state, true);\n    state.clientDone = true;\n    await doneWait.promise;\n    finish(model, format, state, push);\n  } catch (error) {\n    const voice = asVoiceError(error);\n    state.failed = true;\n    push({ type: 'error', reason: voice.reason, error: voice.error });\n  } finally {\n    socket?.close();\n  }\n}\n\nfunction sendChunk(\n  socket: WebSocket,\n  chunk: AudioChunk | undefined,\n  format: AudioFormat,\n  state: ListenState,\n  commit: boolean\n) {\n  if (chunk && !sameFormat(chunk.format, format)) {\n    throw new Error(`Unsupported audio format ${chunk.format.encoding}`);\n  }\n  const data = chunk ? chunk.data : new Uint8Array();\n  state.bytes += data.byteLength;\n  socket.send(\n    JSON.stringify({\n      message_type: 'input_audio_chunk',\n      audio_base_64: encodeAudioBase64(data),\n      commit,\n      sample_rate: format.sampleRate,\n    })\n  );\n}\n\nasync function onMessage(\n  data: unknown,\n  state: ListenState,\n  push: (event: TranscriptionEvent) => void,\n  wake: (kind: string | undefined) => void\n) {\n  try {\n    const parsed = JSON.parse(await messageText(data)) as unknown;\n    if (!isRecord(parsed)) throw new Error('ElevenLabs transcription event was not JSON');\n    applyListen(parsed, state, push);\n    wake(stringField(parsed, 'message_type'));\n  } catch (error) {\n    const voice = asVoiceError(error);\n    state.failed = true;\n    push({ type: 'error', reason: voice.reason, error: voice.error });\n    wake('fail');\n  }\n}\n"],
+  "mappings": "AACA,SAAS,oBAAoB;AAC7B,SAAS,yBAAyB;AAElC,SAAS,cAAc,UAAU,aAAa,mBAAmB;AACjE;AAAA,EAGE;AAAA,OAEK;AACP,SAAS,kBAAkB;AAC3B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,MAAM,WAAW,oBAAI,IAAI;AAAA,EACvB;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;AAgBD,SAAS,MAAM,SAA8C;AAC3D,QAAM,aAAa,SAAS;AAC5B,MAAI,CAAC,WAAY,QAAO,CAAC;AACzB,SAAO;AACT;AAEA,SAAS,eAAe,OAA2C;AACjE,MAAI,MAAM,kBAAkB,WAAY,QAAO;AAC/C,MAAI,MAAM,kBAAkB,MAAO,QAAO;AAC1C,SAAO;AACT;AAEA,SAAS,OACP,OACA,QACA,SACQ;AACR,QAAM,MAAM,UAAU,MAAM,SAAS,yBAAyB;AAC9D,QAAM,SAAS,IAAI,gBAAgB;AACnC,SAAO,IAAI,YAAY,MAAM,EAAE;AAC/B,SAAO,IAAI,gBAAgB,eAAe,MAAM,CAAC;AACjD,SAAO,IAAI,mBAAmB,eAAe,KAAK,CAAC;AACnD,SAAO,IAAI,sBAAsB,MAAM;AACvC,MAAI,SAAS,SAAU,QAAO,IAAI,iBAAiB,QAAQ,QAAQ;AACnE,aAAW,QAAQ,MAAM,OAAO,EAAG,QAAO,OAAO,YAAY,IAAI;AACjE,MAAI,SAAS,OAAO,SAAS;AAC7B,SAAO,IAAI,SAAS;AACtB;AAEA,SAAS,aACP,OACA,OACa;AACb,QAAM,SAAS,UAAU,SAAY,MAAM,aAAa,CAAC,IAAI,MAAM;AACnE,MAAI,CAAC,OAAQ,OAAM,IAAI,MAAM,iBAAiB;AAC9C,iBAAe,MAAM;AACrB,QAAM,QAAQ,MAAM,aAAa,KAAK,CAAC,SAAS,WAAW,MAAM,MAAM,CAAC;AACxE,MAAI,CAAC,MAAO,OAAM,IAAI,MAAM,4BAA4B,OAAO,QAAQ,EAAE;AACzE,SAAO;AACT;AAEA,SAAS,OAAO,OAAwC;AACtD,QAAM,OAAO,YAAY,OAAO,MAAM;AACtC,MAAI,SAAS,OAAW,QAAO;AAC/B,SAAO;AACT;AAEA,SAAS,UAAU,MAA+B,KAAqB;AACrE,QAAM,QAAQ,YAAY,MAAM,GAAG;AACnC,MAAI,UAAU,OAAW,QAAO;AAChC,SAAO;AACT;AAEA,SAAS,SAAS,MAA8D;AAC9E,MAAI,YAAY,MAAM,MAAM,MAAM,UAAW,QAAO;AACpD,QAAM,UAAU,YAAY,MAAM,YAAY;AAC9C,QAAM,UAA6B;AAAA,IACjC,MAAM,OAAO,IAAI;AAAA,IACjB,OAAO,UAAU,MAAM,OAAO;AAAA,IAC9B,KAAK,UAAU,MAAM,KAAK;AAAA,EAC5B;AACA,MAAI,YAAY,OAAW,SAAQ,UAAU;AAC7C,SAAO;AACT;AAEA,SAAS,QAAQ,OAAgC,OAAoB;AACnE,MAAI,CAAC,MAAM,QAAQ,MAAM,KAAK,EAAG;AACjC,aAAW,QAAQ,MAAM,OAAO;AAC9B,QAAI,CAAC,SAAS,IAAI,EAAG;AACrB,UAAM,UAAU,SAAS,IAAI;AAC7B,QAAI,QAAS,OAAM,SAAS,KAAK,OAAO;AAAA,EAC1C;AACF;AAEA,SAAS,WAAW,QAAgB,OAAoB,MAA2C;AACjG,MAAI,CAAC,MAAM,SAAS;AAClB,UAAM,UAAU;AAChB,SAAK,EAAE,MAAM,gBAAgB,IAAI,EAAE,CAAC;AAAA,EACtC;AACA,QAAM,YAAY,OAAO,MAAM,OAAO;AACtC,QAAM,WAAW;AACjB,QAAM,MAAM,KAAK,MAAM;AACvB,OAAK,EAAE,MAAM,SAAS,WAAW,MAAM,OAAO,CAAC;AAC/C,QAAM,UAAU;AAChB,OAAK,EAAE,MAAM,YAAY,MAAM,QAAQ,IAAI,EAAE,CAAC;AAChD;AAEA,SAAS,YACP,OACA,OACA,MACA;AACA,QAAM,OAAO,YAAY,OAAO,cAAc;AAC9C,MAAI,SAAS,mBAAmB;AAC9B,UAAM,QAAQ;AACd;AAAA,EACF;AACA,MAAI,SAAS,sBAAsB;AACjC,QAAI,CAAC,MAAM,SAAS;AAClB,YAAM,UAAU;AAChB,WAAK,EAAE,MAAM,gBAAgB,IAAI,EAAE,CAAC;AAAA,IACtC;AACA,SAAK,EAAE,MAAM,WAAW,WAAW,OAAO,MAAM,OAAO,GAAG,MAAM,OAAO,KAAK,EAAE,CAAC;AAC/E;AAAA,EACF;AACA,MAAI,SAAS,wBAAwB;AACnC,UAAM,SAAS,OAAO,KAAK;AAC3B,UAAM,aAAa;AACnB,eAAW,QAAQ,OAAO,IAAI;AAC9B;AAAA,EACF;AACA,MAAI,SAAS,wCAAwC;AACnD,UAAM,SAAS,OAAO,KAAK;AAC3B,YAAQ,OAAO,KAAK;AACpB,QAAI,MAAM,eAAe,OAAQ,YAAW,QAAQ,OAAO,IAAI;AAC/D,UAAM,aAAa;AACnB;AAAA,EACF;AACA,MAAI,SAAS,UAAa,SAAS,IAAI,IAAI,GAAG;AAC5C,UAAM,UAAU,YAAY,OAAO,OAAO;AAC1C,UAAM,SAAS;AACf,SAAK;AAAA,MACH,MAAM;AAAA,MACN,QAAQ;AAAA,MACR,OAAO,YAAY,SAAY,oCAAoC;AAAA,IACrE,CAAC;AAAA,EACH;AACF;AAEA,SAAS,OACP,OACA,QACA,OACA,MACA;AACA,MAAI,MAAM,UAAU,MAAM,UAAW;AACrC,QAAM,YAAY;AAClB,QAAM,QAAQ,aAAa,EAAE,MAAM,IAAI,WAAW,MAAM,KAAK,GAAG,QAAQ,WAAW,EAAE,CAAC;AACtF,OAAK;AAAA,IACH,MAAM;AAAA,IACN,YAAY;AAAA,MACV,MAAM,MAAM,MAAM,KAAK,GAAG;AAAA,MAC1B,GAAI,MAAM,SAAS,SAAS,IAAI,EAAE,UAAU,MAAM,SAAS,IAAI,CAAC;AAAA,MAChE,OAAO,WAAW,MAAM,WAAW,EAAE,cAAc,MAAM,CAAC;AAAA,IAC5D;AAAA,EACF,CAAC;AACH;AAEO,SAAS,iBACd,OACA,OACA,SAC0B;AAC1B,QAAM,SAAS,IAAI,yBAAyB;AAC5C,OAAK,WAAW,OAAO,OAAO,SAAS,MAAM;AAC7C,SAAO;AACT;AAEA,eAAe,WACb,OACA,OACA,SACA,QACA;AACA,QAAM,QAAqB;AAAA,IACzB,OAAO,CAAC;AAAA,IACR,UAAU,CAAC;AAAA,IACX,SAAS;AAAA,IACT,SAAS;AAAA,IACT,OAAO;AAAA,IACP,QAAQ;AAAA,IACR,OAAO;AAAA,IACP,YAAY;AAAA,IACZ,QAAQ;AAAA,IACR,WAAW;AAAA,IACX,YAAY;AAAA,EACd;AACA,MAAI,UAAU;AACd,QAAM,OAAO,CAAC,UAA8B;AAC1C,QAAI,QAAS;AACb,QAAI,MAAM,SAAS,UAAU,MAAM,SAAS,QAAS,WAAU;AAC/D,WAAO,KAAK,KAAK;AAAA,EACnB;AACA,MAAI;AACJ,QAAM,YAAY,QAAQ,cAAoB;AAC9C,QAAM,WAAW,QAAQ,cAAoB;AAC7C,QAAM,OAAO,CAAC,SAA6B;AACzC,QAAI,MAAM,SAAS,MAAM,OAAQ,WAAU,QAAQ;AACnD,QAAI,MAAM,OAAQ,UAAS,QAAQ;AACnC,QAAI,CAAC,MAAM,WAAY;AACvB,QAAI,SAAS,0BAA0B,SAAS,wCAAwC;AACtF,eAAS,QAAQ;AAAA,IACnB;AACA,QAAI,SAAS,QAAS,UAAS,QAAQ;AAAA,EACzC;AACA,MAAI;AACF,QAAI,SAAS,QAAQ,QAAS,OAAM,IAAI,aAAa,WAAW,YAAY;AAC5E,UAAM,WAAW,MAAM,OAAO,aAAa,EAAE;AAC7C,UAAM,QAAQ,MAAM,SAAS,KAAK;AAClC,UAAM,SAAS,aAAa,OAAO,MAAM,OAAO,SAAY,MAAM,KAAK;AACvE,aAAS,cAAc,OAAO,OAAO,QAAQ,OAAO,GAAG,UAAU,OAAO,OAAO,CAAC;AAChF,QAAI,WAAW,QAAQ,QAAQ;AAC/B,WAAO,iBAAiB,WAAW,CAAC,UAAU;AAC5C,iBAAW,SAAS,KAAK,MAAM,UAAU,MAAM,MAAM,OAAO,MAAM,IAAI,CAAC;AAAA,IACzE,CAAC;AACD,WAAO,iBAAiB,SAAS,MAAM;AACrC,UAAI,MAAM,UAAU,CAAC,MAAM,UAAU,CAAC,MAAM,aAAa,CAAC,MAAM,YAAY;AAC1E,cAAM,SAAS;AACf,aAAK,EAAE,MAAM,SAAS,QAAQ,SAAS,OAAO,yCAAyC,CAAC;AAAA,MAC1F;AACA,WAAK,OAAO;AAAA,IACd,CAAC;AACD,UAAM,UAAU,QAAQ,4BAA4B,CAAC,YAAY;AAC/D,YAAM,SAAS;AACf,WAAK,EAAE,MAAM,SAAS,QAAQ,SAAS,OAAO,QAAQ,CAAC;AACvD,WAAK,MAAM;AAAA,IACb,CAAC;AACD,QAAI,MAAM,OAAQ;AAClB,UAAM,SAAS;AACf,QAAI,CAAC,MAAM,MAAO,OAAM,UAAU;AAClC,QAAI,MAAM,OAAQ;AAClB,QAAI,CAAC,MAAM,KAAM,WAAU,QAAQ,MAAM,OAAO,QAAQ,OAAO,KAAK;AACpE,WAAO,CAAC,MAAM,QAAQ;AACpB,UAAI,SAAS,QAAQ,QAAS,OAAM,IAAI,aAAa,WAAW,YAAY;AAC5E,YAAM,OAAO,MAAM,SAAS,KAAK;AACjC,UAAI,KAAK,KAAM;AACf,gBAAU,QAAQ,KAAK,OAAO,QAAQ,OAAO,KAAK;AAAA,IACpD;AACA,QAAI,MAAM,OAAQ;AAClB,cAAU,QAAQ,QAAW,QAAQ,OAAO,IAAI;AAChD,UAAM,aAAa;AACnB,UAAM,SAAS;AACf,WAAO,OAAO,QAAQ,OAAO,IAAI;AAAA,EACnC,SAAS,OAAO;AACd,UAAM,QAAQ,aAAa,KAAK;AAChC,UAAM,SAAS;AACf,SAAK,EAAE,MAAM,SAAS,QAAQ,MAAM,QAAQ,OAAO,MAAM,MAAM,CAAC;AAAA,EAClE,UAAE;AACA,YAAQ,MAAM;AAAA,EAChB;AACF;AAEA,SAAS,UACP,QACA,OACA,QACA,OACA,QACA;AACA,MAAI,SAAS,CAAC,WAAW,MAAM,QAAQ,MAAM,GAAG;AAC9C,UAAM,IAAI,MAAM,4BAA4B,MAAM,OAAO,QAAQ,EAAE;AAAA,EACrE;AACA,QAAM,OAAO,QAAQ,MAAM,OAAO,IAAI,WAAW;AACjD,QAAM,SAAS,KAAK;AACpB,SAAO;AAAA,IACL,KAAK,UAAU;AAAA,MACb,cAAc;AAAA,MACd,eAAe,kBAAkB,IAAI;AAAA,MACrC;AAAA,MACA,aAAa,OAAO;AAAA,IACtB,CAAC;AAAA,EACH;AACF;AAEA,eAAe,UACb,MACA,OACA,MACA,MACA;AACA,MAAI;AACF,UAAM,SAAS,KAAK,MAAM,MAAM,YAAY,IAAI,CAAC;AACjD,QAAI,CAAC,SAAS,MAAM,EAAG,OAAM,IAAI,MAAM,6CAA6C;AACpF,gBAAY,QAAQ,OAAO,IAAI;AAC/B,SAAK,YAAY,QAAQ,cAAc,CAAC;AAAA,EAC1C,SAAS,OAAO;AACd,UAAM,QAAQ,aAAa,KAAK;AAChC,UAAM,SAAS;AACf,SAAK,EAAE,MAAM,SAAS,QAAQ,MAAM,QAAQ,OAAO,MAAM,MAAM,CAAC;AAChE,SAAK,MAAM;AAAA,EACb;AACF;",
+  "names": []
+}

package/dist/elevenlabs/speak.js

@@ -0,0 +1,251 @@
+import { chunkSeconds } from "../audio.js";
+import { decodeAudioBase64 } from "../convert.js";
+import { VoiceRequestError } from "../errors.js";
+import {
+  applyPayload,
+  asVoiceError,
+  endpoint,
+  isRecord,
+  readJson,
+  requestSignal,
+  stringField
+} from "../openai/wire.js";
+import { SpeechEventStream } from "../speech.js";
+import { priceAudio } from "../usage.js";
+import {
+  alignmentFrom,
+  awaitOpen,
+  connectSocket,
+  messageText,
+  outputFormatName,
+  sameFormat,
+  socketUrl,
+  xiHeaders
+} from "./wire.js";
+function languageCode(language) {
+  if (language === void 0) return void 0;
+  if (language.trim() === "") return void 0;
+  return language;
+}
+function resolveFormat(model, requested) {
+  const fallback = model.outputFormats[0];
+  if (!fallback) throw new Error("No output format");
+  if (!requested) return fallback;
+  const match = model.outputFormats.find((format) => sameFormat(format, requested));
+  if (!match) throw new Error(`Unsupported speech format ${requested.encoding}`);
+  return match;
+}
+function speechBody(model, text, language) {
+  const body = { text, model_id: model.id };
+  const code = languageCode(language);
+  if (code !== void 0 && model.id !== "eleven_multilingual_v2") body.language_code = code;
+  return body;
+}
+function speechUrl(model, voice, format) {
+  const url = socketUrl(model.baseUrl, `text-to-speech/${encodeURIComponent(voice)}/stream-input`);
+  const params = new URLSearchParams();
+  params.set("model_id", model.id);
+  params.set("output_format", outputFormatName(format));
+  params.set("sync_alignment", "true");
+  url.search = params.toString();
+  return url.toString();
+}
+async function readText(text, signal) {
+  if (typeof text === "string") return text;
+  let combined = "";
+  for await (const part of text) {
+    if (signal?.aborted) throw new DOMException("aborted", "AbortError");
+    combined += part;
+  }
+  return combined;
+}
+async function xiFetch(model, options, url, body) {
+  const headers = xiHeaders(model, options);
+  headers["Content-Type"] = "application/json";
+  const fetchImpl = options?.fetch ?? globalThis.fetch;
+  const response = await fetchImpl(url, {
+    method: "POST",
+    headers,
+    body,
+    signal: requestSignal(options)
+  });
+  await options?.onResponse?.({ status: response.status, headers: {} }, model);
+  if (!response.ok) {
+    const text = await response.text();
+    throw new Error(text === "" ? `HTTP ${response.status}` : text);
+  }
+  return response;
+}
+function speakElevenLabs(model, input, options) {
+  const stream = new SpeechEventStream();
+  void pumpSpeech(model, input, options, stream);
+  return stream;
+}
+async function pumpSpeech(model, input, options, stream) {
+  try {
+    if (options?.signal?.aborted) throw new DOMException("aborted", "AbortError");
+    const format = resolveFormat(model, input.format);
+    if (model.textStreaming && typeof input.text !== "string") {
+      await streamSpeech(model, input.text, input.voice, format, options, stream);
+      return;
+    }
+    const text = await readText(input.text, options?.signal);
+    await postSpeech(model, text, input.voice, input.language, format, options, stream);
+  } catch (error) {
+    const voice = asVoiceError(error);
+    stream.push({ type: "error", reason: voice.reason, error: voice.error });
+  }
+}
+async function postSpeech(model, text, voice, language, format, options, stream) {
+  const payload = speechBody(model, text, language);
+  const body = await applyPayload(payload, model, options);
+  if (!isRecord(body)) throw new Error("Speech payload was not JSON");
+  const query = `output_format=${outputFormatName(format)}`;
+  const path = `text-to-speech/${encodeURIComponent(voice)}/with-timestamps?${query}`;
+  const response = await xiFetch(
+    model,
+    options,
+    endpoint(model.baseUrl, path),
+    JSON.stringify(body)
+  );
+  const parsed = await readJson(response);
+  if (!isRecord(parsed)) throw new Error("Speech response had no audio");
+  const audio = stringField(parsed, "audio_base64");
+  if (audio === void 0) throw new Error("Speech response had no audio");
+  const data = decodeAudioBase64(audio);
+  const seconds = chunkSeconds({ data, format, timestamp: 0 });
+  const meters = { characters: text.length };
+  if (seconds > 0) meters.outputSeconds = seconds;
+  const alignment = alignmentFrom(parsed.alignment);
+  stream.push({ type: "start", format });
+  stream.push({
+    type: "audio",
+    chunk: { data, format, timestamp: 0 },
+    span: { itemId: model.id, start: 0, end: text.length },
+    ...alignment ? { alignment } : {}
+  });
+  stream.push({ type: "done", usage: priceAudio(model.audioCost, meters) });
+}
+async function streamSpeech(model, text, voice, format, options, stream) {
+  const state = {
+    opened: false,
+    started: false,
+    done: false,
+    failed: false,
+    outputSeconds: 0,
+    characters: 0
+  };
+  let settled = false;
+  const push = (event) => {
+    if (settled) return;
+    if (event.type === "done" || event.type === "error") settled = true;
+    stream.push(event);
+  };
+  const doneWait = Promise.withResolvers();
+  const wake = () => {
+    doneWait.resolve();
+  };
+  const socket = connectSocket(speechUrl(model, voice, format), xiHeaders(model, options));
+  let incoming = Promise.resolve();
+  socket.addEventListener("message", (event) => {
+    incoming = incoming.then(() => onMessage(event.data, model.id, format, state, push, wake));
+  });
+  socket.addEventListener("close", () => {
+    if (state.opened && !state.done && !state.failed) {
+      state.failed = true;
+      push({ type: "error", reason: "error", error: "ElevenLabs speech socket closed" });
+    }
+    wake();
+  });
+  try {
+    await awaitOpen(socket, "ElevenLabs speech", (message) => {
+      state.failed = true;
+      push({ type: "error", reason: "error", error: message });
+      wake();
+    });
+    if (state.failed) return;
+    state.opened = true;
+    state.characters = await writeText(socket, text, options?.signal);
+    if (state.failed) return;
+    socket.send(JSON.stringify({ text: " ", flush: true }));
+    socket.send(JSON.stringify({ text: "" }));
+    if (!state.done) await doneWait.promise;
+    if (state.failed) return;
+    const meters = { characters: state.characters };
+    if (state.outputSeconds > 0) meters.outputSeconds = state.outputSeconds;
+    push({ type: "done", usage: priceAudio(model.audioCost, meters) });
+  } catch (error) {
+    const voice2 = asVoiceError(error);
+    state.failed = true;
+    push({ type: "error", reason: voice2.reason, error: voice2.error });
+  } finally {
+    socket.close();
+  }
+}
+async function writeText(socket, text, signal) {
+  let combined = "";
+  socket.send(
+    JSON.stringify({
+      text: " ",
+      generation_config: { chunk_length_schedule: [50, 120, 160, 290] }
+    })
+  );
+  for await (const part of text) {
+    if (signal?.aborted) throw new VoiceRequestError("aborted", "aborted");
+    combined += part;
+    if (part === "") continue;
+    const delta = part.endsWith(" ") ? part : `${part} `;
+    socket.send(JSON.stringify({ text: delta }));
+  }
+  return combined.length;
+}
+async function onMessage(data, itemId, format, state, push, wake) {
+  try {
+    const parsed = JSON.parse(await messageText(data));
+    if (!isRecord(parsed)) throw new Error("ElevenLabs speech event was not JSON");
+    applySpeech(parsed, itemId, format, state, push);
+    if (state.done || state.failed) wake();
+  } catch (error) {
+    const voice = asVoiceError(error);
+    state.failed = true;
+    push({ type: "error", reason: voice.reason, error: voice.error });
+    wake();
+  }
+}
+function audioEvent(itemId, format, data, end, alignment) {
+  return {
+    type: "audio",
+    chunk: { data, format, timestamp: 0 },
+    span: { itemId, start: 0, end },
+    ...alignment ? { alignment } : {}
+  };
+}
+function applySpeech(event, itemId, format, state, push) {
+  if ("error" in event) {
+    const message = stringField(event, "error");
+    state.failed = true;
+    push({
+      type: "error",
+      reason: "error",
+      error: message === void 0 ? "ElevenLabs speech failed" : message
+    });
+    return;
+  }
+  if (event.isFinal === true) {
+    state.done = true;
+    return;
+  }
+  const audio = stringField(event, "audio");
+  if (audio === void 0 || audio === "") return;
+  const data = decodeAudioBase64(audio);
+  if (!state.started) {
+    state.started = true;
+    push({ type: "start", format });
+  }
+  state.outputSeconds += chunkSeconds({ data, format, timestamp: 0 });
+  push(audioEvent(itemId, format, data, state.characters, alignmentFrom(event.alignment)));
+}
+export {
+  speakElevenLabs
+};
+//# sourceMappingURL=speak.js.map

package/dist/elevenlabs/speak.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/elevenlabs/speak.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport type { AudioFormat, CharAlignment } from '../audio.ts';\nimport { chunkSeconds } from '../audio.ts';\nimport { decodeAudioBase64 } from '../convert.ts';\nimport { VoiceRequestError } from '../errors.ts';\nimport type { SpeechModel } from '../models.ts';\nimport {\n  applyPayload,\n  asVoiceError,\n  endpoint,\n  isRecord,\n  readJson,\n  requestSignal,\n  stringField,\n} from '../openai/wire.ts';\nimport { type SpeechEvent, SpeechEventStream, type SpeechInput } from '../speech.ts';\nimport { type AudioCounters, priceAudio } from '../usage.ts';\nimport {\n  alignmentFrom,\n  awaitOpen,\n  connectSocket,\n  messageText,\n  outputFormatName,\n  sameFormat,\n  socketUrl,\n  xiHeaders,\n} from './wire.ts';\n\ninterface SpeechState {\n  opened: boolean;\n  started: boolean;\n  done: boolean;\n  failed: boolean;\n  outputSeconds: number;\n  characters: number;\n}\n\nfunction languageCode(language: string | undefined): string | undefined {\n  if (language === undefined) return undefined;\n  if (language.trim() === '') return undefined;\n  return language;\n}\n\nfunction resolveFormat(\n  model: SpeechModel<string>,\n  requested: AudioFormat | undefined\n): AudioFormat {\n  const fallback = model.outputFormats[0];\n  if (!fallback) throw new Error('No output format');\n  if (!requested) return fallback;\n  const match = model.outputFormats.find((format) => sameFormat(format, requested));\n  if (!match) throw new Error(`Unsupported speech format ${requested.encoding}`);\n  return match;\n}\n\nfunction speechBody(\n  model: SpeechModel<string>,\n  text: string,\n  language: string | undefined\n): Record<string, unknown> {\n  const body: Record<string, unknown> = { text, model_id: model.id };\n  const code = languageCode(language);\n  if (code !== undefined && model.id !== 'eleven_multilingual_v2') body.language_code = code;\n  return body;\n}\n\nfunction speechUrl(model: SpeechModel<string>, voice: string, format: AudioFormat): string {\n  const url = socketUrl(model.baseUrl, `text-to-speech/${encodeURIComponent(voice)}/stream-input`);\n  const params = new URLSearchParams();\n  params.set('model_id', model.id);\n  params.set('output_format', outputFormatName(format));\n  params.set('sync_alignment', 'true');\n  url.search = params.toString();\n  return url.toString();\n}\n\nasync function readText(\n  text: string | AsyncIterable<string>,\n  signal: AbortSignal | undefined\n): Promise<string> {\n  if (typeof text === 'string') return text;\n  let combined = '';\n  for await (const part of text) {\n    if (signal?.aborted) throw new DOMException('aborted', 'AbortError');\n    combined += part;\n  }\n  return combined;\n}\n\nasync function xiFetch(\n  model: SpeechModel<string>,\n  options: ProviderRequestOptions | undefined,\n  url: string,\n  body: string\n): Promise<Response> {\n  const headers = xiHeaders(model, options);\n  headers['Content-Type'] = 'application/json';\n  const fetchImpl = options?.fetch ?? globalThis.fetch;\n  const response = await fetchImpl(url, {\n    method: 'POST',\n    headers,\n    body,\n    signal: requestSignal(options),\n  });\n  await options?.onResponse?.({ status: response.status, headers: {} }, model as never);\n  if (!response.ok) {\n    const text = await response.text();\n    throw new Error(text === '' ? `HTTP ${response.status}` : text);\n  }\n  return response;\n}\n\nexport function speakElevenLabs(\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options?: ProviderRequestOptions\n): SpeechEventStream {\n  const stream = new SpeechEventStream();\n  void pumpSpeech(model, input, options, stream);\n  return stream;\n}\n\nasync function pumpSpeech(\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options: ProviderRequestOptions | undefined,\n  stream: SpeechEventStream\n) {\n  try {\n    if (options?.signal?.aborted) throw new DOMException('aborted', 'AbortError');\n    const format = resolveFormat(model, input.format);\n    if (model.textStreaming && typeof input.text !== 'string') {\n      await streamSpeech(model, input.text, input.voice, format, options, stream);\n      return;\n    }\n    const text = await readText(input.text, options?.signal);\n    await postSpeech(model, text, input.voice, input.language, format, options, stream);\n  } catch (error) {\n    const voice = asVoiceError(error);\n    stream.push({ type: 'error', reason: voice.reason, error: voice.error });\n  }\n}\n\nasync function postSpeech(\n  model: SpeechModel<string>,\n  text: string,\n  voice: string,\n  language: string | undefined,\n  format: AudioFormat,\n  options: ProviderRequestOptions | undefined,\n  stream: SpeechEventStream\n) {\n  const payload = speechBody(model, text, language);\n  const body = await applyPayload(payload, model, options);\n  if (!isRecord(body)) throw new Error('Speech payload was not JSON');\n  const query = `output_format=${outputFormatName(format)}`;\n  const path = `text-to-speech/${encodeURIComponent(voice)}/with-timestamps?${query}`;\n  const response = await xiFetch(\n    model,\n    options,\n    endpoint(model.baseUrl, path),\n    JSON.stringify(body)\n  );\n  const parsed = await readJson(response);\n  if (!isRecord(parsed)) throw new Error('Speech response had no audio');\n  const audio = stringField(parsed, 'audio_base64');\n  if (audio === undefined) throw new Error('Speech response had no audio');\n  const data = decodeAudioBase64(audio);\n  const seconds = chunkSeconds({ data, format, timestamp: 0 });\n  const meters: AudioCounters = { characters: text.length };\n  if (seconds > 0) meters.outputSeconds = seconds;\n  const alignment = alignmentFrom(parsed.alignment);\n  stream.push({ type: 'start', format });\n  stream.push({\n    type: 'audio',\n    chunk: { data, format, timestamp: 0 },\n    span: { itemId: model.id, start: 0, end: text.length },\n    ...(alignment ? { alignment } : {}),\n  });\n  stream.push({ type: 'done', usage: priceAudio(model.audioCost, meters) });\n}\n\nasync function streamSpeech(\n  model: SpeechModel<string>,\n  text: AsyncIterable<string>,\n  voice: string,\n  format: AudioFormat,\n  options: ProviderRequestOptions | undefined,\n  stream: SpeechEventStream\n) {\n  const state: SpeechState = {\n    opened: false,\n    started: false,\n    done: false,\n    failed: false,\n    outputSeconds: 0,\n    characters: 0,\n  };\n  let settled = false;\n  const push = (event: SpeechEvent) => {\n    if (settled) return;\n    if (event.type === 'done' || event.type === 'error') settled = true;\n    stream.push(event);\n  };\n  const doneWait = Promise.withResolvers<void>();\n  const wake = () => {\n    doneWait.resolve();\n  };\n  const socket = connectSocket(speechUrl(model, voice, format), xiHeaders(model, options));\n  let incoming = Promise.resolve();\n  socket.addEventListener('message', (event) => {\n    incoming = incoming.then(() => onMessage(event.data, model.id, format, state, push, wake));\n  });\n  socket.addEventListener('close', () => {\n    if (state.opened && !state.done && !state.failed) {\n      state.failed = true;\n      push({ type: 'error', reason: 'error', error: 'ElevenLabs speech socket closed' });\n    }\n    wake();\n  });\n  try {\n    await awaitOpen(socket, 'ElevenLabs speech', (message) => {\n      state.failed = true;\n      push({ type: 'error', reason: 'error', error: message });\n      wake();\n    });\n    if (state.failed) return;\n    state.opened = true;\n    state.characters = await writeText(socket, text, options?.signal);\n    if (state.failed) return;\n    socket.send(JSON.stringify({ text: ' ', flush: true }));\n    socket.send(JSON.stringify({ text: '' }));\n    if (!state.done) await doneWait.promise;\n    if (state.failed) return;\n    const meters: AudioCounters = { characters: state.characters };\n    if (state.outputSeconds > 0) meters.outputSeconds = state.outputSeconds;\n    push({ type: 'done', usage: priceAudio(model.audioCost, meters) });\n  } catch (error) {\n    const voice = asVoiceError(error);\n    state.failed = true;\n    push({ type: 'error', reason: voice.reason, error: voice.error });\n  } finally {\n    socket.close();\n  }\n}\n\nasync function writeText(\n  socket: WebSocket,\n  text: AsyncIterable<string>,\n  signal: AbortSignal | undefined\n): Promise<number> {\n  let combined = '';\n  socket.send(\n    JSON.stringify({\n      text: ' ',\n      generation_config: { chunk_length_schedule: [50, 120, 160, 290] },\n    })\n  );\n  for await (const part of text) {\n    if (signal?.aborted) throw new VoiceRequestError('aborted', 'aborted');\n    combined += part;\n    if (part === '') continue;\n    const delta = part.endsWith(' ') ? part : `${part} `;\n    socket.send(JSON.stringify({ text: delta }));\n  }\n  return combined.length;\n}\n\nasync function onMessage(\n  data: unknown,\n  itemId: string,\n  format: AudioFormat,\n  state: SpeechState,\n  push: (event: SpeechEvent) => void,\n  wake: () => void\n) {\n  try {\n    const parsed = JSON.parse(await messageText(data)) as unknown;\n    if (!isRecord(parsed)) throw new Error('ElevenLabs speech event was not JSON');\n    applySpeech(parsed, itemId, format, state, push);\n    if (state.done || state.failed) wake();\n  } catch (error) {\n    const voice = asVoiceError(error);\n    state.failed = true;\n    push({ type: 'error', reason: voice.reason, error: voice.error });\n    wake();\n  }\n}\n\nfunction audioEvent(\n  itemId: string,\n  format: AudioFormat,\n  data: Uint8Array,\n  end: number,\n  alignment: CharAlignment | undefined\n): SpeechEvent {\n  return {\n    type: 'audio',\n    chunk: { data, format, timestamp: 0 },\n    span: { itemId, start: 0, end },\n    ...(alignment ? { alignment } : {}),\n  };\n}\n\nfunction applySpeech(\n  event: Record<string, unknown>,\n  itemId: string,\n  format: AudioFormat,\n  state: SpeechState,\n  push: (event: SpeechEvent) => void\n) {\n  if ('error' in event) {\n    const message = stringField(event, 'error');\n    state.failed = true;\n    push({\n      type: 'error',\n      reason: 'error',\n      error: message === undefined ? 'ElevenLabs speech failed' : message,\n    });\n    return;\n  }\n  if (event.isFinal === true) {\n    state.done = true;\n    return;\n  }\n  const audio = stringField(event, 'audio');\n  if (audio === undefined || audio === '') return;\n  const data = decodeAudioBase64(audio);\n  if (!state.started) {\n    state.started = true;\n    push({ type: 'start', format });\n  }\n  state.outputSeconds += chunkSeconds({ data, format, timestamp: 0 });\n  push(audioEvent(itemId, format, data, state.characters, alignmentFrom(event.alignment)));\n}\n"],
+  "mappings": "AAEA,SAAS,oBAAoB;AAC7B,SAAS,yBAAyB;AAClC,SAAS,yBAAyB;AAElC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAA2B,yBAA2C;AACtE,SAA6B,kBAAkB;AAC/C;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAWP,SAAS,aAAa,UAAkD;AACtE,MAAI,aAAa,OAAW,QAAO;AACnC,MAAI,SAAS,KAAK,MAAM,GAAI,QAAO;AACnC,SAAO;AACT;AAEA,SAAS,cACP,OACA,WACa;AACb,QAAM,WAAW,MAAM,cAAc,CAAC;AACtC,MAAI,CAAC,SAAU,OAAM,IAAI,MAAM,kBAAkB;AACjD,MAAI,CAAC,UAAW,QAAO;AACvB,QAAM,QAAQ,MAAM,cAAc,KAAK,CAAC,WAAW,WAAW,QAAQ,SAAS,CAAC;AAChF,MAAI,CAAC,MAAO,OAAM,IAAI,MAAM,6BAA6B,UAAU,QAAQ,EAAE;AAC7E,SAAO;AACT;AAEA,SAAS,WACP,OACA,MACA,UACyB;AACzB,QAAM,OAAgC,EAAE,MAAM,UAAU,MAAM,GAAG;AACjE,QAAM,OAAO,aAAa,QAAQ;AAClC,MAAI,SAAS,UAAa,MAAM,OAAO,yBAA0B,MAAK,gBAAgB;AACtF,SAAO;AACT;AAEA,SAAS,UAAU,OAA4B,OAAe,QAA6B;AACzF,QAAM,MAAM,UAAU,MAAM,SAAS,kBAAkB,mBAAmB,KAAK,CAAC,eAAe;AAC/F,QAAM,SAAS,IAAI,gBAAgB;AACnC,SAAO,IAAI,YAAY,MAAM,EAAE;AAC/B,SAAO,IAAI,iBAAiB,iBAAiB,MAAM,CAAC;AACpD,SAAO,IAAI,kBAAkB,MAAM;AACnC,MAAI,SAAS,OAAO,SAAS;AAC7B,SAAO,IAAI,SAAS;AACtB;AAEA,eAAe,SACb,MACA,QACiB;AACjB,MAAI,OAAO,SAAS,SAAU,QAAO;AACrC,MAAI,WAAW;AACf,mBAAiB,QAAQ,MAAM;AAC7B,QAAI,QAAQ,QAAS,OAAM,IAAI,aAAa,WAAW,YAAY;AACnE,gBAAY;AAAA,EACd;AACA,SAAO;AACT;AAEA,eAAe,QACb,OACA,SACA,KACA,MACmB;AACnB,QAAM,UAAU,UAAU,OAAO,OAAO;AACxC,UAAQ,cAAc,IAAI;AAC1B,QAAM,YAAY,SAAS,SAAS,WAAW;AAC/C,QAAM,WAAW,MAAM,UAAU,KAAK;AAAA,IACpC,QAAQ;AAAA,IACR;AAAA,IACA;AAAA,IACA,QAAQ,cAAc,OAAO;AAAA,EAC/B,CAAC;AACD,QAAM,SAAS,aAAa,EAAE,QAAQ,SAAS,QAAQ,SAAS,CAAC,EAAE,GAAG,KAAc;AACpF,MAAI,CAAC,SAAS,IAAI;AAChB,UAAM,OAAO,MAAM,SAAS,KAAK;AACjC,UAAM,IAAI,MAAM,SAAS,KAAK,QAAQ,SAAS,MAAM,KAAK,IAAI;AAAA,EAChE;AACA,SAAO;AACT;AAEO,SAAS,gBACd,OACA,OACA,SACmB;AACnB,QAAM,SAAS,IAAI,kBAAkB;AACrC,OAAK,WAAW,OAAO,OAAO,SAAS,MAAM;AAC7C,SAAO;AACT;AAEA,eAAe,WACb,OACA,OACA,SACA,QACA;AACA,MAAI;AACF,QAAI,SAAS,QAAQ,QAAS,OAAM,IAAI,aAAa,WAAW,YAAY;AAC5E,UAAM,SAAS,cAAc,OAAO,MAAM,MAAM;AAChD,QAAI,MAAM,iBAAiB,OAAO,MAAM,SAAS,UAAU;AACzD,YAAM,aAAa,OAAO,MAAM,MAAM,MAAM,OAAO,QAAQ,SAAS,MAAM;AAC1E;AAAA,IACF;AACA,UAAM,OAAO,MAAM,SAAS,MAAM,MAAM,SAAS,MAAM;AACvD,UAAM,WAAW,OAAO,MAAM,MAAM,OAAO,MAAM,UAAU,QAAQ,SAAS,MAAM;AAAA,EACpF,SAAS,OAAO;AACd,UAAM,QAAQ,aAAa,KAAK;AAChC,WAAO,KAAK,EAAE,MAAM,SAAS,QAAQ,MAAM,QAAQ,OAAO,MAAM,MAAM,CAAC;AAAA,EACzE;AACF;AAEA,eAAe,WACb,OACA,MACA,OACA,UACA,QACA,SACA,QACA;AACA,QAAM,UAAU,WAAW,OAAO,MAAM,QAAQ;AAChD,QAAM,OAAO,MAAM,aAAa,SAAS,OAAO,OAAO;AACvD,MAAI,CAAC,SAAS,IAAI,EAAG,OAAM,IAAI,MAAM,6BAA6B;AAClE,QAAM,QAAQ,iBAAiB,iBAAiB,MAAM,CAAC;AACvD,QAAM,OAAO,kBAAkB,mBAAmB,KAAK,CAAC,oBAAoB,KAAK;AACjF,QAAM,WAAW,MAAM;AAAA,IACrB;AAAA,IACA;AAAA,IACA,SAAS,MAAM,SAAS,IAAI;AAAA,IAC5B,KAAK,UAAU,IAAI;AAAA,EACrB;AACA,QAAM,SAAS,MAAM,SAAS,QAAQ;AACtC,MAAI,CAAC,SAAS,MAAM,EAAG,OAAM,IAAI,MAAM,8BAA8B;AACrE,QAAM,QAAQ,YAAY,QAAQ,cAAc;AAChD,MAAI,UAAU,OAAW,OAAM,IAAI,MAAM,8BAA8B;AACvE,QAAM,OAAO,kBAAkB,KAAK;AACpC,QAAM,UAAU,aAAa,EAAE,MAAM,QAAQ,WAAW,EAAE,CAAC;AAC3D,QAAM,SAAwB,EAAE,YAAY,KAAK,OAAO;AACxD,MAAI,UAAU,EAAG,QAAO,gBAAgB;AACxC,QAAM,YAAY,cAAc,OAAO,SAAS;AAChD,SAAO,KAAK,EAAE,MAAM,SAAS,OAAO,CAAC;AACrC,SAAO,KAAK;AAAA,IACV,MAAM;AAAA,IACN,OAAO,EAAE,MAAM,QAAQ,WAAW,EAAE;AAAA,IACpC,MAAM,EAAE,QAAQ,MAAM,IAAI,OAAO,GAAG,KAAK,KAAK,OAAO;AAAA,IACrD,GAAI,YAAY,EAAE,UAAU,IAAI,CAAC;AAAA,EACnC,CAAC;AACD,SAAO,KAAK,EAAE,MAAM,QAAQ,OAAO,WAAW,MAAM,WAAW,MAAM,EAAE,CAAC;AAC1E;AAEA,eAAe,aACb,OACA,MACA,OACA,QACA,SACA,QACA;AACA,QAAM,QAAqB;AAAA,IACzB,QAAQ;AAAA,IACR,SAAS;AAAA,IACT,MAAM;AAAA,IACN,QAAQ;AAAA,IACR,eAAe;AAAA,IACf,YAAY;AAAA,EACd;AACA,MAAI,UAAU;AACd,QAAM,OAAO,CAAC,UAAuB;AACnC,QAAI,QAAS;AACb,QAAI,MAAM,SAAS,UAAU,MAAM,SAAS,QAAS,WAAU;AAC/D,WAAO,KAAK,KAAK;AAAA,EACnB;AACA,QAAM,WAAW,QAAQ,cAAoB;AAC7C,QAAM,OAAO,MAAM;AACjB,aAAS,QAAQ;AAAA,EACnB;AACA,QAAM,SAAS,cAAc,UAAU,OAAO,OAAO,MAAM,GAAG,UAAU,OAAO,OAAO,CAAC;AACvF,MAAI,WAAW,QAAQ,QAAQ;AAC/B,SAAO,iBAAiB,WAAW,CAAC,UAAU;AAC5C,eAAW,SAAS,KAAK,MAAM,UAAU,MAAM,MAAM,MAAM,IAAI,QAAQ,OAAO,MAAM,IAAI,CAAC;AAAA,EAC3F,CAAC;AACD,SAAO,iBAAiB,SAAS,MAAM;AACrC,QAAI,MAAM,UAAU,CAAC,MAAM,QAAQ,CAAC,MAAM,QAAQ;AAChD,YAAM,SAAS;AACf,WAAK,EAAE,MAAM,SAAS,QAAQ,SAAS,OAAO,kCAAkC,CAAC;AAAA,IACnF;AACA,SAAK;AAAA,EACP,CAAC;AACD,MAAI;AACF,UAAM,UAAU,QAAQ,qBAAqB,CAAC,YAAY;AACxD,YAAM,SAAS;AACf,WAAK,EAAE,MAAM,SAAS,QAAQ,SAAS,OAAO,QAAQ,CAAC;AACvD,WAAK;AAAA,IACP,CAAC;AACD,QAAI,MAAM,OAAQ;AAClB,UAAM,SAAS;AACf,UAAM,aAAa,MAAM,UAAU,QAAQ,MAAM,SAAS,MAAM;AAChE,QAAI,MAAM,OAAQ;AAClB,WAAO,KAAK,KAAK,UAAU,EAAE,MAAM,KAAK,OAAO,KAAK,CAAC,CAAC;AACtD,WAAO,KAAK,KAAK,UAAU,EAAE,MAAM,GAAG,CAAC,CAAC;AACxC,QAAI,CAAC,MAAM,KAAM,OAAM,SAAS;AAChC,QAAI,MAAM,OAAQ;AAClB,UAAM,SAAwB,EAAE,YAAY,MAAM,WAAW;AAC7D,QAAI,MAAM,gBAAgB,EAAG,QAAO,gBAAgB,MAAM;AAC1D,SAAK,EAAE,MAAM,QAAQ,OAAO,WAAW,MAAM,WAAW,MAAM,EAAE,CAAC;AAAA,EACnE,SAAS,OAAO;AACd,UAAMA,SAAQ,aAAa,KAAK;AAChC,UAAM,SAAS;AACf,SAAK,EAAE,MAAM,SAAS,QAAQA,OAAM,QAAQ,OAAOA,OAAM,MAAM,CAAC;AAAA,EAClE,UAAE;AACA,WAAO,MAAM;AAAA,EACf;AACF;AAEA,eAAe,UACb,QACA,MACA,QACiB;AACjB,MAAI,WAAW;AACf,SAAO;AAAA,IACL,KAAK,UAAU;AAAA,MACb,MAAM;AAAA,MACN,mBAAmB,EAAE,uBAAuB,CAAC,IAAI,KAAK,KAAK,GAAG,EAAE;AAAA,IAClE,CAAC;AAAA,EACH;AACA,mBAAiB,QAAQ,MAAM;AAC7B,QAAI,QAAQ,QAAS,OAAM,IAAI,kBAAkB,WAAW,SAAS;AACrE,gBAAY;AACZ,QAAI,SAAS,GAAI;AACjB,UAAM,QAAQ,KAAK,SAAS,GAAG,IAAI,OAAO,GAAG,IAAI;AACjD,WAAO,KAAK,KAAK,UAAU,EAAE,MAAM,MAAM,CAAC,CAAC;AAAA,EAC7C;AACA,SAAO,SAAS;AAClB;AAEA,eAAe,UACb,MACA,QACA,QACA,OACA,MACA,MACA;AACA,MAAI;AACF,UAAM,SAAS,KAAK,MAAM,MAAM,YAAY,IAAI,CAAC;AACjD,QAAI,CAAC,SAAS,MAAM,EAAG,OAAM,IAAI,MAAM,sCAAsC;AAC7E,gBAAY,QAAQ,QAAQ,QAAQ,OAAO,IAAI;AAC/C,QAAI,MAAM,QAAQ,MAAM,OAAQ,MAAK;AAAA,EACvC,SAAS,OAAO;AACd,UAAM,QAAQ,aAAa,KAAK;AAChC,UAAM,SAAS;AACf,SAAK,EAAE,MAAM,SAAS,QAAQ,MAAM,QAAQ,OAAO,MAAM,MAAM,CAAC;AAChE,SAAK;AAAA,EACP;AACF;AAEA,SAAS,WACP,QACA,QACA,MACA,KACA,WACa;AACb,SAAO;AAAA,IACL,MAAM;AAAA,IACN,OAAO,EAAE,MAAM,QAAQ,WAAW,EAAE;AAAA,IACpC,MAAM,EAAE,QAAQ,OAAO,GAAG,IAAI;AAAA,IAC9B,GAAI,YAAY,EAAE,UAAU,IAAI,CAAC;AAAA,EACnC;AACF;AAEA,SAAS,YACP,OACA,QACA,QACA,OACA,MACA;AACA,MAAI,WAAW,OAAO;AACpB,UAAM,UAAU,YAAY,OAAO,OAAO;AAC1C,UAAM,SAAS;AACf,SAAK;AAAA,MACH,MAAM;AAAA,MACN,QAAQ;AAAA,MACR,OAAO,YAAY,SAAY,6BAA6B;AAAA,IAC9D,CAAC;AACD;AAAA,EACF;AACA,MAAI,MAAM,YAAY,MAAM;AAC1B,UAAM,OAAO;AACb;AAAA,EACF;AACA,QAAM,QAAQ,YAAY,OAAO,OAAO;AACxC,MAAI,UAAU,UAAa,UAAU,GAAI;AACzC,QAAM,OAAO,kBAAkB,KAAK;AACpC,MAAI,CAAC,MAAM,SAAS;AAClB,UAAM,UAAU;AAChB,SAAK,EAAE,MAAM,SAAS,OAAO,CAAC;AAAA,EAChC;AACA,QAAM,iBAAiB,aAAa,EAAE,MAAM,QAAQ,WAAW,EAAE,CAAC;AAClE,OAAK,WAAW,QAAQ,QAAQ,MAAM,MAAM,YAAY,cAAc,MAAM,SAAS,CAAC,CAAC;AACzF;",
+  "names": ["voice"]
+}

package/dist/elevenlabs/transcribe.js

@@ -0,0 +1,132 @@
+import { VoiceRequestError } from "../errors.js";
+import {
+  applyPayload,
+  asVoiceError,
+  endpoint,
+  isRecord,
+  numberField,
+  readAudio,
+  readJson,
+  requestSignal,
+  stringField
+} from "../openai/wire.js";
+import { priceAudio } from "../usage.js";
+import { audioFilename, encodeForm, formType, pcmS16, xiHeaders } from "./wire.js";
+function textOf(record, key) {
+  const text = stringField(record, key);
+  if (text === void 0) return "";
+  return text;
+}
+function secondsOf(record, key) {
+  const value = numberField(record, key);
+  if (value === void 0) return 0;
+  return value;
+}
+function wordFrom(word) {
+  const kind = stringField(word, "type");
+  if (kind === "spacing") return void 0;
+  const speaker = stringField(word, "speaker_id");
+  const segment = {
+    text: textOf(word, "text"),
+    start: secondsOf(word, "start"),
+    end: secondsOf(word, "end")
+  };
+  if (speaker !== void 0) segment.speaker = speaker;
+  return segment;
+}
+function wordsFrom(value) {
+  if (!Array.isArray(value)) return [];
+  const segments = [];
+  for (const word of value) {
+    if (!isRecord(word)) continue;
+    const segment = wordFrom(word);
+    if (segment) segments.push(segment);
+  }
+  return segments;
+}
+function transcriptFrom(body, model) {
+  if (!isRecord(body)) throw new Error("Transcription response had no text");
+  const direct = stringField(body, "text");
+  const channels = Array.isArray(body.transcripts) ? body.transcripts.filter(isRecord) : [];
+  const joined = channels.map((channel) => textOf(channel, "text")).join(" ");
+  const text = direct === void 0 ? joined : direct;
+  if (direct === void 0 && channels.length === 0)
+    throw new Error("Transcription response had no text");
+  const language = stringField(body, "language_code");
+  const words = direct === void 0 ? channels.flatMap((channel) => wordsFrom(channel.words)) : wordsFrom(body.words);
+  const duration = numberField(body, "audio_duration_secs");
+  return {
+    text,
+    ...language !== void 0 ? { language } : {},
+    ...words.length > 0 ? { segments: words } : {},
+    usage: priceAudio(model.audioCost, { inputSeconds: duration === void 0 ? 0 : duration })
+  };
+}
+function fileOf(body) {
+  const file = body.file;
+  if (!isRecord(file) || !(file.data instanceof Uint8Array) || typeof file.mimeType !== "string") {
+    throw new Error("Transcription payload was not a file");
+  }
+  const named = typeof file.filename === "string" ? file.filename : audioFilename(file.mimeType);
+  return { data: file.data, mimeType: file.mimeType, filename: named };
+}
+function fieldStrings(body) {
+  const fields = {};
+  for (const [key, value] of Object.entries(body)) {
+    if (key === "file") continue;
+    if (typeof value !== "string") continue;
+    fields[key] = value;
+  }
+  return fields;
+}
+async function xiFetch(model, options, url, body, contentType) {
+  const headers = xiHeaders(model, options);
+  headers["Content-Type"] = contentType;
+  const fetchImpl = options?.fetch ?? globalThis.fetch;
+  const response = await fetchImpl(url, {
+    method: "POST",
+    headers,
+    body,
+    signal: requestSignal(options)
+  });
+  await options?.onResponse?.({ status: response.status, headers: {} }, model);
+  if (!response.ok) {
+    const text = await response.text();
+    throw new Error(text === "" ? `HTTP ${response.status}` : text);
+  }
+  return response;
+}
+async function transcribeElevenLabs(model, input, options) {
+  try {
+    if (options?.signal?.aborted) throw new DOMException("aborted", "AbortError");
+    const fetchImpl = options?.fetch ?? globalThis.fetch;
+    const loaded = await readAudio(input, fetchImpl, requestSignal(options));
+    const filename = audioFilename(loaded.mimeType);
+    const payload = {
+      model_id: model.id,
+      timestamps_granularity: "word",
+      file: { data: loaded.data, mimeType: loaded.mimeType, filename }
+    };
+    if (model.diarization) payload.diarize = "true";
+    if (pcmS16(loaded.mimeType)) payload.file_format = "pcm_s16le_16";
+    const body = await applyPayload(payload, model, options);
+    if (!isRecord(body)) throw new Error("Transcription payload was not a file");
+    const file = fileOf(body);
+    const encoded = encodeForm(file, fieldStrings(body));
+    const response = await xiFetch(
+      model,
+      options,
+      endpoint(model.baseUrl, "speech-to-text"),
+      encoded,
+      formType()
+    );
+    return transcriptFrom(await readJson(response), model);
+  } catch (error) {
+    const voice = asVoiceError(error);
+    throw new VoiceRequestError(voice.reason, voice.error);
+  }
+}
+export {
+  transcribeElevenLabs
+};
+//# sourceMappingURL=transcribe.js.map

package/dist/elevenlabs/transcribe.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/elevenlabs/transcribe.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport { VoiceRequestError } from '../errors.ts';\nimport type { TranscriptionModel } from '../models.ts';\nimport {\n  applyPayload,\n  asVoiceError,\n  endpoint,\n  isRecord,\n  numberField,\n  readAudio,\n  readJson,\n  requestSignal,\n  stringField,\n} from '../openai/wire.ts';\nimport type { AudioInput, Transcript, TranscriptSegment } from '../transcription.ts';\nimport { priceAudio } from '../usage.ts';\nimport { audioFilename, type ElevenFile, encodeForm, formType, pcmS16, xiHeaders } from './wire.ts';\n\nfunction textOf(record: Record<string, unknown>, key: string): string {\n  const text = stringField(record, key);\n  if (text === undefined) return '';\n  return text;\n}\n\nfunction secondsOf(record: Record<string, unknown>, key: string): number {\n  const value = numberField(record, key);\n  if (value === undefined) return 0;\n  return value;\n}\n\nfunction wordFrom(word: Record<string, unknown>): TranscriptSegment | undefined {\n  const kind = stringField(word, 'type');\n  if (kind === 'spacing') return undefined;\n  const speaker = stringField(word, 'speaker_id');\n  const segment: TranscriptSegment = {\n    text: textOf(word, 'text'),\n    start: secondsOf(word, 'start'),\n    end: secondsOf(word, 'end'),\n  };\n  if (speaker !== undefined) segment.speaker = speaker;\n  return segment;\n}\n\nfunction wordsFrom(value: unknown): TranscriptSegment[] {\n  if (!Array.isArray(value)) return [];\n  const segments: TranscriptSegment[] = [];\n  for (const word of value) {\n    if (!isRecord(word)) continue;\n    const segment = wordFrom(word);\n    if (segment) segments.push(segment);\n  }\n  return segments;\n}\n\nfunction transcriptFrom(body: unknown, model: TranscriptionModel<string>): Transcript {\n  if (!isRecord(body)) throw new Error('Transcription response had no text');\n  const direct = stringField(body, 'text');\n  const channels = Array.isArray(body.transcripts) ? body.transcripts.filter(isRecord) : [];\n  const joined = channels.map((channel) => textOf(channel, 'text')).join(' ');\n  const text = direct === undefined ? joined : direct;\n  if (direct === undefined && channels.length === 0)\n    throw new Error('Transcription response had no text');\n  const language = stringField(body, 'language_code');\n  const words =\n    direct === undefined\n      ? channels.flatMap((channel) => wordsFrom(channel.words))\n      : wordsFrom(body.words);\n  const duration = numberField(body, 'audio_duration_secs');\n  return {\n    text,\n    ...(language !== undefined ? { language } : {}),\n    ...(words.length > 0 ? { segments: words } : {}),\n    usage: priceAudio(model.audioCost, { inputSeconds: duration === undefined ? 0 : duration }),\n  };\n}\n\nfunction fileOf(body: Record<string, unknown>): ElevenFile {\n  const file = body.file;\n  if (!isRecord(file) || !(file.data instanceof Uint8Array) || typeof file.mimeType !== 'string') {\n    throw new Error('Transcription payload was not a file');\n  }\n  const named = typeof file.filename === 'string' ? file.filename : audioFilename(file.mimeType);\n  return { data: file.data, mimeType: file.mimeType, filename: named };\n}\n\nfunction fieldStrings(body: Record<string, unknown>): Record<string, string> {\n  const fields: Record<string, string> = {};\n  for (const [key, value] of Object.entries(body)) {\n    if (key === 'file') continue;\n    if (typeof value !== 'string') continue;\n    fields[key] = value;\n  }\n  return fields;\n}\n\nasync function xiFetch(\n  model: TranscriptionModel<string>,\n  options: ProviderRequestOptions | undefined,\n  url: string,\n  body: BodyInit,\n  contentType: string\n): Promise<Response> {\n  const headers = xiHeaders(model, options);\n  headers['Content-Type'] = contentType;\n  const fetchImpl = options?.fetch ?? globalThis.fetch;\n  const response = await fetchImpl(url, {\n    method: 'POST',\n    headers,\n    body,\n    signal: requestSignal(options),\n  });\n  await options?.onResponse?.({ status: response.status, headers: {} }, model as never);\n  if (!response.ok) {\n    const text = await response.text();\n    throw new Error(text === '' ? `HTTP ${response.status}` : text);\n  }\n  return response;\n}\n\nexport async function transcribeElevenLabs(\n  model: TranscriptionModel<string>,\n  input: AudioInput,\n  options?: ProviderRequestOptions\n): Promise<Transcript> {\n  try {\n    if (options?.signal?.aborted) throw new DOMException('aborted', 'AbortError');\n    const fetchImpl = options?.fetch ?? globalThis.fetch;\n    const loaded = await readAudio(input, fetchImpl, requestSignal(options));\n    const filename = audioFilename(loaded.mimeType);\n    const payload: Record<string, unknown> = {\n      model_id: model.id,\n      timestamps_granularity: 'word',\n      file: { data: loaded.data, mimeType: loaded.mimeType, filename },\n    };\n    if (model.diarization) payload.diarize = 'true';\n    if (pcmS16(loaded.mimeType)) payload.file_format = 'pcm_s16le_16';\n    const body = await applyPayload(payload, model, options);\n    if (!isRecord(body)) throw new Error('Transcription payload was not a file');\n    const file = fileOf(body);\n    const encoded = encodeForm(file, fieldStrings(body));\n    const response = await xiFetch(\n      model,\n      options,\n      endpoint(model.baseUrl, 'speech-to-text'),\n      encoded,\n      formType()\n    );\n    return transcriptFrom(await readJson(response), model);\n  } catch (error) {\n    const voice = asVoiceError(error);\n    throw new VoiceRequestError(voice.reason, voice.error);\n  }\n}\n"],
+  "mappings": "AACA,SAAS,yBAAyB;AAElC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,kBAAkB;AAC3B,SAAS,eAAgC,YAAY,UAAU,QAAQ,iBAAiB;AAExF,SAAS,OAAO,QAAiC,KAAqB;AACpE,QAAM,OAAO,YAAY,QAAQ,GAAG;AACpC,MAAI,SAAS,OAAW,QAAO;AAC/B,SAAO;AACT;AAEA,SAAS,UAAU,QAAiC,KAAqB;AACvE,QAAM,QAAQ,YAAY,QAAQ,GAAG;AACrC,MAAI,UAAU,OAAW,QAAO;AAChC,SAAO;AACT;AAEA,SAAS,SAAS,MAA8D;AAC9E,QAAM,OAAO,YAAY,MAAM,MAAM;AACrC,MAAI,SAAS,UAAW,QAAO;AAC/B,QAAM,UAAU,YAAY,MAAM,YAAY;AAC9C,QAAM,UAA6B;AAAA,IACjC,MAAM,OAAO,MAAM,MAAM;AAAA,IACzB,OAAO,UAAU,MAAM,OAAO;AAAA,IAC9B,KAAK,UAAU,MAAM,KAAK;AAAA,EAC5B;AACA,MAAI,YAAY,OAAW,SAAQ,UAAU;AAC7C,SAAO;AACT;AAEA,SAAS,UAAU,OAAqC;AACtD,MAAI,CAAC,MAAM,QAAQ,KAAK,EAAG,QAAO,CAAC;AACnC,QAAM,WAAgC,CAAC;AACvC,aAAW,QAAQ,OAAO;AACxB,QAAI,CAAC,SAAS,IAAI,EAAG;AACrB,UAAM,UAAU,SAAS,IAAI;AAC7B,QAAI,QAAS,UAAS,KAAK,OAAO;AAAA,EACpC;AACA,SAAO;AACT;AAEA,SAAS,eAAe,MAAe,OAA+C;AACpF,MAAI,CAAC,SAAS,IAAI,EAAG,OAAM,IAAI,MAAM,oCAAoC;AACzE,QAAM,SAAS,YAAY,MAAM,MAAM;AACvC,QAAM,WAAW,MAAM,QAAQ,KAAK,WAAW,IAAI,KAAK,YAAY,OAAO,QAAQ,IAAI,CAAC;AACxF,QAAM,SAAS,SAAS,IAAI,CAAC,YAAY,OAAO,SAAS,MAAM,CAAC,EAAE,KAAK,GAAG;AAC1E,QAAM,OAAO,WAAW,SAAY,SAAS;AAC7C,MAAI,WAAW,UAAa,SAAS,WAAW;AAC9C,UAAM,IAAI,MAAM,oCAAoC;AACtD,QAAM,WAAW,YAAY,MAAM,eAAe;AAClD,QAAM,QACJ,WAAW,SACP,SAAS,QAAQ,CAAC,YAAY,UAAU,QAAQ,KAAK,CAAC,IACtD,UAAU,KAAK,KAAK;AAC1B,QAAM,WAAW,YAAY,MAAM,qBAAqB;AACxD,SAAO;AAAA,IACL;AAAA,IACA,GAAI,aAAa,SAAY,EAAE,SAAS,IAAI,CAAC;AAAA,IAC7C,GAAI,MAAM,SAAS,IAAI,EAAE,UAAU,MAAM,IAAI,CAAC;AAAA,IAC9C,OAAO,WAAW,MAAM,WAAW,EAAE,cAAc,aAAa,SAAY,IAAI,SAAS,CAAC;AAAA,EAC5F;AACF;AAEA,SAAS,OAAO,MAA2C;AACzD,QAAM,OAAO,KAAK;AAClB,MAAI,CAAC,SAAS,IAAI,KAAK,EAAE,KAAK,gBAAgB,eAAe,OAAO,KAAK,aAAa,UAAU;AAC9F,UAAM,IAAI,MAAM,sCAAsC;AAAA,EACxD;AACA,QAAM,QAAQ,OAAO,KAAK,aAAa,WAAW,KAAK,WAAW,cAAc,KAAK,QAAQ;AAC7F,SAAO,EAAE,MAAM,KAAK,MAAM,UAAU,KAAK,UAAU,UAAU,MAAM;AACrE;AAEA,SAAS,aAAa,MAAuD;AAC3E,QAAM,SAAiC,CAAC;AACxC,aAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,IAAI,GAAG;AAC/C,QAAI,QAAQ,OAAQ;AACpB,QAAI,OAAO,UAAU,SAAU;AAC/B,WAAO,GAAG,IAAI;AAAA,EAChB;AACA,SAAO;AACT;AAEA,eAAe,QACb,OACA,SACA,KACA,MACA,aACmB;AACnB,QAAM,UAAU,UAAU,OAAO,OAAO;AACxC,UAAQ,cAAc,IAAI;AAC1B,QAAM,YAAY,SAAS,SAAS,WAAW;AAC/C,QAAM,WAAW,MAAM,UAAU,KAAK;AAAA,IACpC,QAAQ;AAAA,IACR;AAAA,IACA;AAAA,IACA,QAAQ,cAAc,OAAO;AAAA,EAC/B,CAAC;AACD,QAAM,SAAS,aAAa,EAAE,QAAQ,SAAS,QAAQ,SAAS,CAAC,EAAE,GAAG,KAAc;AACpF,MAAI,CAAC,SAAS,IAAI;AAChB,UAAM,OAAO,MAAM,SAAS,KAAK;AACjC,UAAM,IAAI,MAAM,SAAS,KAAK,QAAQ,SAAS,MAAM,KAAK,IAAI;AAAA,EAChE;AACA,SAAO;AACT;AAEA,eAAsB,qBACpB,OACA,OACA,SACqB;AACrB,MAAI;AACF,QAAI,SAAS,QAAQ,QAAS,OAAM,IAAI,aAAa,WAAW,YAAY;AAC5E,UAAM,YAAY,SAAS,SAAS,WAAW;AAC/C,UAAM,SAAS,MAAM,UAAU,OAAO,WAAW,cAAc,OAAO,CAAC;AACvE,UAAM,WAAW,cAAc,OAAO,QAAQ;AAC9C,UAAM,UAAmC;AAAA,MACvC,UAAU,MAAM;AAAA,MAChB,wBAAwB;AAAA,MACxB,MAAM,EAAE,MAAM,OAAO,MAAM,UAAU,OAAO,UAAU,SAAS;AAAA,IACjE;AACA,QAAI,MAAM,YAAa,SAAQ,UAAU;AACzC,QAAI,OAAO,OAAO,QAAQ,EAAG,SAAQ,cAAc;AACnD,UAAM,OAAO,MAAM,aAAa,SAAS,OAAO,OAAO;AACvD,QAAI,CAAC,SAAS,IAAI,EAAG,OAAM,IAAI,MAAM,sCAAsC;AAC3E,UAAM,OAAO,OAAO,IAAI;AACxB,UAAM,UAAU,WAAW,MAAM,aAAa,IAAI,CAAC;AACnD,UAAM,WAAW,MAAM;AAAA,MACrB;AAAA,MACA;AAAA,MACA,SAAS,MAAM,SAAS,gBAAgB;AAAA,MACxC;AAAA,MACA,SAAS;AAAA,IACX;AACA,WAAO,eAAe,MAAM,SAAS,QAAQ,GAAG,KAAK;AAAA,EACvD,SAAS,OAAO;AACd,UAAM,QAAQ,aAAa,KAAK;AAChC,UAAM,IAAI,kBAAkB,MAAM,QAAQ,MAAM,KAAK;AAAA,EACvD;AACF;",
+  "names": []
+}

package/dist/elevenlabs/wire.js

@@ -0,0 +1,215 @@
+import { isRecord } from "../openai/wire.js";
+const boundary = "pi-audio-elevenlabs-boundary";
+const filenames = {
+  "audio/wav": "audio.wav",
+  "audio/wave": "audio.wav",
+  "audio/x-wav": "audio.wav",
+  "audio/mpeg": "audio.mp3",
+  "audio/mp3": "audio.mp3",
+  "audio/flac": "audio.flac",
+  "audio/ogg": "audio.ogg",
+  "audio/opus": "audio.opus",
+  "audio/pcm": "audio.pcm",
+  "audio/l16": "audio.pcm",
+  "audio/mulaw": "audio.mulaw",
+  "audio/basic": "audio.mulaw",
+  "audio/alaw": "audio.alaw",
+  "audio/mp4": "audio.mp4",
+  "audio/m4a": "audio.m4a",
+  "audio/aac": "audio.aac",
+  "audio/webm": "audio.webm"
+};
+const sttFormats = /* @__PURE__ */ new Set([
+  "pcm_8000",
+  "pcm_16000",
+  "pcm_22050",
+  "pcm_24000",
+  "pcm_44100",
+  "pcm_48000",
+  "ulaw_8000"
+]);
+function mediaType(mimeType) {
+  const type = mimeType.split(";")[0]?.trim().toLowerCase();
+  if (!type) throw new Error(`Unsupported audio type ${mimeType}`);
+  return type;
+}
+function audioFilename(mimeType) {
+  const name = filenames[mediaType(mimeType)];
+  if (!name) throw new Error(`Unsupported audio type ${mimeType}`);
+  return name;
+}
+function parameter(mimeType, name) {
+  for (const part of mimeType.split(";").slice(1)) {
+    const pieces = part.split("=");
+    const key = pieces[0]?.trim().toLowerCase();
+    const value = pieces[1]?.trim();
+    if (key === name && value) return value;
+  }
+  return void 0;
+}
+function pcmS16(mimeType) {
+  const type = mediaType(mimeType);
+  if (type !== "audio/pcm" && type !== "audio/l16") return false;
+  const rate = parameter(mimeType, "rate");
+  if (rate === void 0) return true;
+  return rate === "16000";
+}
+function encodeForm(file, fields) {
+  const chunks = [];
+  const encoder = new TextEncoder();
+  const write = (value) => {
+    chunks.push(encoder.encode(value));
+  };
+  for (const [name, value] of Object.entries(fields)) {
+    write(`--${boundary}\r
+Content-Disposition: form-data; name="${name}"\r
+\r
+${value}\r
+`);
+  }
+  const header = `--${boundary}\r
+Content-Disposition: form-data; name="file"; filename="${file.filename}"\r
+Content-Type: ${file.mimeType}\r
+\r
+`;
+  write(header);
+  chunks.push(file.data);
+  write(`\r
+--${boundary}--\r
+`);
+  const body = new Uint8Array(chunks.reduce((sum, chunk) => sum + chunk.length, 0));
+  let offset = 0;
+  for (const chunk of chunks) {
+    body.set(chunk, offset);
+    offset += chunk.length;
+  }
+  return body.buffer;
+}
+function formType() {
+  return `multipart/form-data; boundary=${boundary}`;
+}
+function headerValue(headers, name) {
+  if (!headers) return void 0;
+  const expected = name.toLowerCase();
+  for (const [key, value] of Object.entries(headers)) {
+    if (key.toLowerCase() === expected && value !== null && value.trim() !== "") return value;
+  }
+  return void 0;
+}
+function xiHeaders(model, options) {
+  const headers = { ...model.headers };
+  const extra = options?.headers;
+  if (extra) {
+    for (const [key, value] of Object.entries(extra)) {
+      if (value === null) delete headers[key];
+      else headers[key] = value;
+    }
+  }
+  if (headerValue(headers, "xi-api-key") === void 0) {
+    const apiKey = options?.apiKey;
+    if (!apiKey) throw new Error(`No API key for provider: ${model.provider}`);
+    headers["xi-api-key"] = apiKey;
+  }
+  return headers;
+}
+function socketUrl(baseUrl, path) {
+  const url = new URL(baseUrl);
+  url.protocol = url.protocol === "https:" ? "wss:" : "ws:";
+  url.pathname = `${url.pathname.replace(/\/$/, "")}/${path}`;
+  url.search = "";
+  return url;
+}
+function outputFormatName(format) {
+  if (format.encoding === "mp3" && format.sampleRate === 44100) return "mp3_44100_128";
+  if (format.encoding === "mp3" && format.sampleRate === 24e3) return "mp3_24000_48";
+  if (format.encoding === "mp3" && format.sampleRate === 22050) return "mp3_22050_32";
+  if (format.encoding === "pcm16") return `pcm_${format.sampleRate}`;
+  if (format.encoding === "wav") return `wav_${format.sampleRate}`;
+  if (format.encoding === "mulaw") return "ulaw_8000";
+  if (format.encoding === "alaw") return "alaw_8000";
+  if (format.encoding === "opus") return "opus_48000_128";
+  throw new Error(`Unsupported speech format ${format.encoding}`);
+}
+function sttAudioFormat(format) {
+  const name = format.encoding === "mulaw" ? "ulaw_8000" : `pcm_${format.sampleRate}`;
+  if (format.encoding !== "mulaw" && format.encoding !== "pcm16") {
+    throw new Error(`Unsupported audio format ${format.encoding}`);
+  }
+  if (!sttFormats.has(name)) throw new Error(`Unsupported audio format ${format.encoding}`);
+  return name;
+}
+function connectSocket(url, headers) {
+  const Socket = WebSocket;
+  return new Socket(url, { headers });
+}
+function awaitOpen(socket, label, fail) {
+  return new Promise((resolve) => {
+    let opened = false;
+    socket.addEventListener("open", () => {
+      opened = true;
+      resolve();
+    });
+    socket.addEventListener("error", () => {
+      fail(`${label} socket failed`);
+      if (!opened) resolve();
+    });
+    socket.addEventListener("close", () => {
+      if (!opened) {
+        fail(`${label} socket closed`);
+        resolve();
+      }
+    });
+  });
+}
+async function messageText(data) {
+  if (typeof data === "string") return data;
+  if (data instanceof ArrayBuffer) return new TextDecoder().decode(data);
+  if (data instanceof Uint8Array) return new TextDecoder().decode(data);
+  if (typeof Blob !== "undefined" && data instanceof Blob) return data.text();
+  throw new Error("Unsupported socket message");
+}
+function sameFormat(left, right) {
+  return left.encoding === right.encoding && left.sampleRate === right.sampleRate && left.channels === right.channels;
+}
+function numberList(value) {
+  if (!Array.isArray(value)) return void 0;
+  const numbers = [];
+  for (const item of value) {
+    if (typeof item === "number") numbers.push(item);
+  }
+  return numbers;
+}
+function durationMs(start, end) {
+  if (start === void 0 || end === void 0) return 0;
+  return (end - start) * 1e3;
+}
+function alignmentFrom(value) {
+  if (!isRecord(value)) return void 0;
+  const startMs = numberList(value.charStartTimesMs);
+  const durationMsList = numberList(value.charDurationsMs);
+  if (startMs && durationMsList) return { startMs, durationMs: durationMsList };
+  const starts = numberList(value.character_start_times_seconds);
+  const ends = numberList(value.character_end_times_seconds);
+  if (!starts || !ends) return void 0;
+  return {
+    startMs: starts.map((start) => start * 1e3),
+    durationMs: starts.map((start, index) => durationMs(start, ends[index]))
+  };
+}
+export {
+  alignmentFrom,
+  audioFilename,
+  awaitOpen,
+  connectSocket,
+  encodeForm,
+  formType,
+  mediaType,
+  messageText,
+  outputFormatName,
+  pcmS16,
+  sameFormat,
+  socketUrl,
+  sttAudioFormat,
+  xiHeaders
+};
+//# sourceMappingURL=wire.js.map

package/dist/elevenlabs/wire.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/elevenlabs/wire.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport type { AudioFormat, CharAlignment } from '../audio.ts';\nimport { isRecord } from '../openai/wire.ts';\n\nconst boundary = 'pi-audio-elevenlabs-boundary';\n\nconst filenames: Record<string, string> = {\n  'audio/wav': 'audio.wav',\n  'audio/wave': 'audio.wav',\n  'audio/x-wav': 'audio.wav',\n  'audio/mpeg': 'audio.mp3',\n  'audio/mp3': 'audio.mp3',\n  'audio/flac': 'audio.flac',\n  'audio/ogg': 'audio.ogg',\n  'audio/opus': 'audio.opus',\n  'audio/pcm': 'audio.pcm',\n  'audio/l16': 'audio.pcm',\n  'audio/mulaw': 'audio.mulaw',\n  'audio/basic': 'audio.mulaw',\n  'audio/alaw': 'audio.alaw',\n  'audio/mp4': 'audio.mp4',\n  'audio/m4a': 'audio.m4a',\n  'audio/aac': 'audio.aac',\n  'audio/webm': 'audio.webm',\n};\n\nconst sttFormats = new Set([\n  'pcm_8000',\n  'pcm_16000',\n  'pcm_22050',\n  'pcm_24000',\n  'pcm_44100',\n  'pcm_48000',\n  'ulaw_8000',\n]);\n\nexport interface ElevenFile {\n  data: Uint8Array;\n  mimeType: string;\n  filename: string;\n}\n\ninterface HeaderModel {\n  provider: string;\n  headers?: Record<string, string>;\n}\n\nexport function mediaType(mimeType: string): string {\n  const type = mimeType.split(';')[0]?.trim().toLowerCase();\n  if (!type) throw new Error(`Unsupported audio type ${mimeType}`);\n  return type;\n}\n\nexport function audioFilename(mimeType: string): string {\n  const name = filenames[mediaType(mimeType)];\n  if (!name) throw new Error(`Unsupported audio type ${mimeType}`);\n  return name;\n}\n\nfunction parameter(mimeType: string, name: string): string | undefined {\n  for (const part of mimeType.split(';').slice(1)) {\n    const pieces = part.split('=');\n    const key = pieces[0]?.trim().toLowerCase();\n    const value = pieces[1]?.trim();\n    if (key === name && value) return value;\n  }\n  return undefined;\n}\n\nexport function pcmS16(mimeType: string): boolean {\n  const type = mediaType(mimeType);\n  if (type !== 'audio/pcm' && type !== 'audio/l16') return false;\n  const rate = parameter(mimeType, 'rate');\n  if (rate === undefined) return true;\n  return rate === '16000';\n}\n\nexport function encodeForm(file: ElevenFile, fields: Record<string, string>): ArrayBuffer {\n  const chunks: Uint8Array[] = [];\n  const encoder = new TextEncoder();\n  const write = (value: string) => {\n    chunks.push(encoder.encode(value));\n  };\n  for (const [name, value] of Object.entries(fields)) {\n    write(`--${boundary}\\r\\nContent-Disposition: form-data; name=\"${name}\"\\r\\n\\r\\n${value}\\r\\n`);\n  }\n  const header =\n    `--${boundary}\\r\\nContent-Disposition: form-data; name=\"file\"; filename=\"${file.filename}\"\\r\\n` +\n    `Content-Type: ${file.mimeType}\\r\\n\\r\\n`;\n  write(header);\n  chunks.push(file.data);\n  write(`\\r\\n--${boundary}--\\r\\n`);\n  const body = new Uint8Array(chunks.reduce((sum, chunk) => sum + chunk.length, 0));\n  let offset = 0;\n  for (const chunk of chunks) {\n    body.set(chunk, offset);\n    offset += chunk.length;\n  }\n  return body.buffer;\n}\n\nexport function formType(): string {\n  return `multipart/form-data; boundary=${boundary}`;\n}\n\nfunction headerValue(\n  headers: Record<string, string | null> | undefined,\n  name: string\n): string | undefined {\n  if (!headers) return undefined;\n  const expected = name.toLowerCase();\n  for (const [key, value] of Object.entries(headers)) {\n    if (key.toLowerCase() === expected && value !== null && value.trim() !== '') return value;\n  }\n  return undefined;\n}\n\nexport function xiHeaders(\n  model: HeaderModel,\n  options: ProviderRequestOptions | undefined\n): Record<string, string> {\n  const headers: Record<string, string> = { ...model.headers };\n  const extra = options?.headers;\n  if (extra) {\n    for (const [key, value] of Object.entries(extra)) {\n      if (value === null) delete headers[key];\n      else headers[key] = value;\n    }\n  }\n  if (headerValue(headers, 'xi-api-key') === undefined) {\n    const apiKey = options?.apiKey;\n    if (!apiKey) throw new Error(`No API key for provider: ${model.provider}`);\n    headers['xi-api-key'] = apiKey;\n  }\n  return headers;\n}\n\nexport function socketUrl(baseUrl: string, path: string): URL {\n  const url = new URL(baseUrl);\n  url.protocol = url.protocol === 'https:' ? 'wss:' : 'ws:';\n  url.pathname = `${url.pathname.replace(/\\/$/, '')}/${path}`;\n  url.search = '';\n  return url;\n}\n\nexport function outputFormatName(format: AudioFormat): string {\n  if (format.encoding === 'mp3' && format.sampleRate === 44100) return 'mp3_44100_128';\n  if (format.encoding === 'mp3' && format.sampleRate === 24000) return 'mp3_24000_48';\n  if (format.encoding === 'mp3' && format.sampleRate === 22050) return 'mp3_22050_32';\n  if (format.encoding === 'pcm16') return `pcm_${format.sampleRate}`;\n  if (format.encoding === 'wav') return `wav_${format.sampleRate}`;\n  if (format.encoding === 'mulaw') return 'ulaw_8000';\n  if (format.encoding === 'alaw') return 'alaw_8000';\n  if (format.encoding === 'opus') return 'opus_48000_128';\n  throw new Error(`Unsupported speech format ${format.encoding}`);\n}\n\nexport function sttAudioFormat(format: AudioFormat): string {\n  const name = format.encoding === 'mulaw' ? 'ulaw_8000' : `pcm_${format.sampleRate}`;\n  if (format.encoding !== 'mulaw' && format.encoding !== 'pcm16') {\n    throw new Error(`Unsupported audio format ${format.encoding}`);\n  }\n  if (!sttFormats.has(name)) throw new Error(`Unsupported audio format ${format.encoding}`);\n  return name;\n}\n\nexport function connectSocket(url: string, headers: Record<string, string>): WebSocket {\n  const Socket = WebSocket as unknown as new (\n    address: string,\n    init: { headers: Record<string, string> }\n  ) => WebSocket;\n  return new Socket(url, { headers });\n}\n\nexport function awaitOpen(\n  socket: WebSocket,\n  label: string,\n  fail: (message: string) => void\n): Promise<void> {\n  return new Promise((resolve) => {\n    let opened = false;\n    socket.addEventListener('open', () => {\n      opened = true;\n      resolve();\n    });\n    socket.addEventListener('error', () => {\n      fail(`${label} socket failed`);\n      if (!opened) resolve();\n    });\n    socket.addEventListener('close', () => {\n      if (!opened) {\n        fail(`${label} socket closed`);\n        resolve();\n      }\n    });\n  });\n}\n\nexport async function messageText(data: unknown): Promise<string> {\n  if (typeof data === 'string') return data;\n  if (data instanceof ArrayBuffer) return new TextDecoder().decode(data);\n  if (data instanceof Uint8Array) return new TextDecoder().decode(data);\n  if (typeof Blob !== 'undefined' && data instanceof Blob) return data.text();\n  throw new Error('Unsupported socket message');\n}\n\nexport function sameFormat(left: AudioFormat, right: AudioFormat): boolean {\n  return (\n    left.encoding === right.encoding &&\n    left.sampleRate === right.sampleRate &&\n    left.channels === right.channels\n  );\n}\n\nfunction numberList(value: unknown): number[] | undefined {\n  if (!Array.isArray(value)) return undefined;\n  const numbers: number[] = [];\n  for (const item of value) {\n    if (typeof item === 'number') numbers.push(item);\n  }\n  return numbers;\n}\n\nfunction durationMs(start: number | undefined, end: number | undefined): number {\n  if (start === undefined || end === undefined) return 0;\n  return (end - start) * 1000;\n}\n\nexport function alignmentFrom(value: unknown): CharAlignment | undefined {\n  if (!isRecord(value)) return undefined;\n  const startMs = numberList(value.charStartTimesMs);\n  const durationMsList = numberList(value.charDurationsMs);\n  if (startMs && durationMsList) return { startMs, durationMs: durationMsList };\n  const starts = numberList(value.character_start_times_seconds);\n  const ends = numberList(value.character_end_times_seconds);\n  if (!starts || !ends) return undefined;\n  return {\n    startMs: starts.map((start) => start * 1000),\n    durationMs: starts.map((start, index) => durationMs(start, ends[index])),\n  };\n}\n"],
+  "mappings": "AAEA,SAAS,gBAAgB;AAEzB,MAAM,WAAW;AAEjB,MAAM,YAAoC;AAAA,EACxC,aAAa;AAAA,EACb,cAAc;AAAA,EACd,eAAe;AAAA,EACf,cAAc;AAAA,EACd,aAAa;AAAA,EACb,cAAc;AAAA,EACd,aAAa;AAAA,EACb,cAAc;AAAA,EACd,aAAa;AAAA,EACb,aAAa;AAAA,EACb,eAAe;AAAA,EACf,eAAe;AAAA,EACf,cAAc;AAAA,EACd,aAAa;AAAA,EACb,aAAa;AAAA,EACb,aAAa;AAAA,EACb,cAAc;AAChB;AAEA,MAAM,aAAa,oBAAI,IAAI;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF,CAAC;AAaM,SAAS,UAAU,UAA0B;AAClD,QAAM,OAAO,SAAS,MAAM,GAAG,EAAE,CAAC,GAAG,KAAK,EAAE,YAAY;AACxD,MAAI,CAAC,KAAM,OAAM,IAAI,MAAM,0BAA0B,QAAQ,EAAE;AAC/D,SAAO;AACT;AAEO,SAAS,cAAc,UAA0B;AACtD,QAAM,OAAO,UAAU,UAAU,QAAQ,CAAC;AAC1C,MAAI,CAAC,KAAM,OAAM,IAAI,MAAM,0BAA0B,QAAQ,EAAE;AAC/D,SAAO;AACT;AAEA,SAAS,UAAU,UAAkB,MAAkC;AACrE,aAAW,QAAQ,SAAS,MAAM,GAAG,EAAE,MAAM,CAAC,GAAG;AAC/C,UAAM,SAAS,KAAK,MAAM,GAAG;AAC7B,UAAM,MAAM,OAAO,CAAC,GAAG,KAAK,EAAE,YAAY;AAC1C,UAAM,QAAQ,OAAO,CAAC,GAAG,KAAK;AAC9B,QAAI,QAAQ,QAAQ,MAAO,QAAO;AAAA,EACpC;AACA,SAAO;AACT;AAEO,SAAS,OAAO,UAA2B;AAChD,QAAM,OAAO,UAAU,QAAQ;AAC/B,MAAI,SAAS,eAAe,SAAS,YAAa,QAAO;AACzD,QAAM,OAAO,UAAU,UAAU,MAAM;AACvC,MAAI,SAAS,OAAW,QAAO;AAC/B,SAAO,SAAS;AAClB;AAEO,SAAS,WAAW,MAAkB,QAA6C;AACxF,QAAM,SAAuB,CAAC;AAC9B,QAAM,UAAU,IAAI,YAAY;AAChC,QAAM,QAAQ,CAAC,UAAkB;AAC/B,WAAO,KAAK,QAAQ,OAAO,KAAK,CAAC;AAAA,EACnC;AACA,aAAW,CAAC,MAAM,KAAK,KAAK,OAAO,QAAQ,MAAM,GAAG;AAClD,UAAM,KAAK,QAAQ;AAAA,wCAA6C,IAAI;AAAA;AAAA,EAAY,KAAK;AAAA,CAAM;AAAA,EAC7F;AACA,QAAM,SACJ,KAAK,QAAQ;AAAA,yDAA8D,KAAK,QAAQ;AAAA,gBACvE,KAAK,QAAQ;AAAA;AAAA;AAChC,QAAM,MAAM;AACZ,SAAO,KAAK,KAAK,IAAI;AACrB,QAAM;AAAA,IAAS,QAAQ;AAAA,CAAQ;AAC/B,QAAM,OAAO,IAAI,WAAW,OAAO,OAAO,CAAC,KAAK,UAAU,MAAM,MAAM,QAAQ,CAAC,CAAC;AAChF,MAAI,SAAS;AACb,aAAW,SAAS,QAAQ;AAC1B,SAAK,IAAI,OAAO,MAAM;AACtB,cAAU,MAAM;AAAA,EAClB;AACA,SAAO,KAAK;AACd;AAEO,SAAS,WAAmB;AACjC,SAAO,iCAAiC,QAAQ;AAClD;AAEA,SAAS,YACP,SACA,MACoB;AACpB,MAAI,CAAC,QAAS,QAAO;AACrB,QAAM,WAAW,KAAK,YAAY;AAClC,aAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,OAAO,GAAG;AAClD,QAAI,IAAI,YAAY,MAAM,YAAY,UAAU,QAAQ,MAAM,KAAK,MAAM,GAAI,QAAO;AAAA,EACtF;AACA,SAAO;AACT;AAEO,SAAS,UACd,OACA,SACwB;AACxB,QAAM,UAAkC,EAAE,GAAG,MAAM,QAAQ;AAC3D,QAAM,QAAQ,SAAS;AACvB,MAAI,OAAO;AACT,eAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,KAAK,GAAG;AAChD,UAAI,UAAU,KAAM,QAAO,QAAQ,GAAG;AAAA,UACjC,SAAQ,GAAG,IAAI;AAAA,IACtB;AAAA,EACF;AACA,MAAI,YAAY,SAAS,YAAY,MAAM,QAAW;AACpD,UAAM,SAAS,SAAS;AACxB,QAAI,CAAC,OAAQ,OAAM,IAAI,MAAM,4BAA4B,MAAM,QAAQ,EAAE;AACzE,YAAQ,YAAY,IAAI;AAAA,EAC1B;AACA,SAAO;AACT;AAEO,SAAS,UAAU,SAAiB,MAAmB;AAC5D,QAAM,MAAM,IAAI,IAAI,OAAO;AAC3B,MAAI,WAAW,IAAI,aAAa,WAAW,SAAS;AACpD,MAAI,WAAW,GAAG,IAAI,SAAS,QAAQ,OAAO,EAAE,CAAC,IAAI,IAAI;AACzD,MAAI,SAAS;AACb,SAAO;AACT;AAEO,SAAS,iBAAiB,QAA6B;AAC5D,MAAI,OAAO,aAAa,SAAS,OAAO,eAAe,MAAO,QAAO;AACrE,MAAI,OAAO,aAAa,SAAS,OAAO,eAAe,KAAO,QAAO;AACrE,MAAI,OAAO,aAAa,SAAS,OAAO,eAAe,MAAO,QAAO;AACrE,MAAI,OAAO,aAAa,QAAS,QAAO,OAAO,OAAO,UAAU;AAChE,MAAI,OAAO,aAAa,MAAO,QAAO,OAAO,OAAO,UAAU;AAC9D,MAAI,OAAO,aAAa,QAAS,QAAO;AACxC,MAAI,OAAO,aAAa,OAAQ,QAAO;AACvC,MAAI,OAAO,aAAa,OAAQ,QAAO;AACvC,QAAM,IAAI,MAAM,6BAA6B,OAAO,QAAQ,EAAE;AAChE;AAEO,SAAS,eAAe,QAA6B;AAC1D,QAAM,OAAO,OAAO,aAAa,UAAU,cAAc,OAAO,OAAO,UAAU;AACjF,MAAI,OAAO,aAAa,WAAW,OAAO,aAAa,SAAS;AAC9D,UAAM,IAAI,MAAM,4BAA4B,OAAO,QAAQ,EAAE;AAAA,EAC/D;AACA,MAAI,CAAC,WAAW,IAAI,IAAI,EAAG,OAAM,IAAI,MAAM,4BAA4B,OAAO,QAAQ,EAAE;AACxF,SAAO;AACT;AAEO,SAAS,cAAc,KAAa,SAA4C;AACrF,QAAM,SAAS;AAIf,SAAO,IAAI,OAAO,KAAK,EAAE,QAAQ,CAAC;AACpC;AAEO,SAAS,UACd,QACA,OACA,MACe;AACf,SAAO,IAAI,QAAQ,CAAC,YAAY;AAC9B,QAAI,SAAS;AACb,WAAO,iBAAiB,QAAQ,MAAM;AACpC,eAAS;AACT,cAAQ;AAAA,IACV,CAAC;AACD,WAAO,iBAAiB,SAAS,MAAM;AACrC,WAAK,GAAG,KAAK,gBAAgB;AAC7B,UAAI,CAAC,OAAQ,SAAQ;AAAA,IACvB,CAAC;AACD,WAAO,iBAAiB,SAAS,MAAM;AACrC,UAAI,CAAC,QAAQ;AACX,aAAK,GAAG,KAAK,gBAAgB;AAC7B,gBAAQ;AAAA,MACV;AAAA,IACF,CAAC;AAAA,EACH,CAAC;AACH;AAEA,eAAsB,YAAY,MAAgC;AAChE,MAAI,OAAO,SAAS,SAAU,QAAO;AACrC,MAAI,gBAAgB,YAAa,QAAO,IAAI,YAAY,EAAE,OAAO,IAAI;AACrE,MAAI,gBAAgB,WAAY,QAAO,IAAI,YAAY,EAAE,OAAO,IAAI;AACpE,MAAI,OAAO,SAAS,eAAe,gBAAgB,KAAM,QAAO,KAAK,KAAK;AAC1E,QAAM,IAAI,MAAM,4BAA4B;AAC9C;AAEO,SAAS,WAAW,MAAmB,OAA6B;AACzE,SACE,KAAK,aAAa,MAAM,YACxB,KAAK,eAAe,MAAM,cAC1B,KAAK,aAAa,MAAM;AAE5B;AAEA,SAAS,WAAW,OAAsC;AACxD,MAAI,CAAC,MAAM,QAAQ,KAAK,EAAG,QAAO;AAClC,QAAM,UAAoB,CAAC;AAC3B,aAAW,QAAQ,OAAO;AACxB,QAAI,OAAO,SAAS,SAAU,SAAQ,KAAK,IAAI;AAAA,EACjD;AACA,SAAO;AACT;AAEA,SAAS,WAAW,OAA2B,KAAiC;AAC9E,MAAI,UAAU,UAAa,QAAQ,OAAW,QAAO;AACrD,UAAQ,MAAM,SAAS;AACzB;AAEO,SAAS,cAAc,OAA2C;AACvE,MAAI,CAAC,SAAS,KAAK,EAAG,QAAO;AAC7B,QAAM,UAAU,WAAW,MAAM,gBAAgB;AACjD,QAAM,iBAAiB,WAAW,MAAM,eAAe;AACvD,MAAI,WAAW,eAAgB,QAAO,EAAE,SAAS,YAAY,eAAe;AAC5E,QAAM,SAAS,WAAW,MAAM,6BAA6B;AAC7D,QAAM,OAAO,WAAW,MAAM,2BAA2B;AACzD,MAAI,CAAC,UAAU,CAAC,KAAM,QAAO;AAC7B,SAAO;AAAA,IACL,SAAS,OAAO,IAAI,CAAC,UAAU,QAAQ,GAAI;AAAA,IAC3C,YAAY,OAAO,IAAI,CAAC,OAAO,UAAU,WAAW,OAAO,KAAK,KAAK,CAAC,CAAC;AAAA,EACzE;AACF;",
+  "names": []
+}

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

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

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

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

package/dist/types/elevenlabs/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 listenElevenLabs(model: TranscriptionModel<string>, audio: AsyncIterable<AudioChunk>, options?: ListenOptions): TranscriptionEventStream;

package/dist/types/elevenlabs/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 speakElevenLabs(model: SpeechModel<string>, input: SpeechInput, options?: ProviderRequestOptions): SpeechEventStream;

package/dist/types/elevenlabs/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 transcribeElevenLabs(model: TranscriptionModel<string>, input: AudioInput, options?: ProviderRequestOptions): Promise<Transcript>;

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

@@ -0,0 +1,26 @@
+import type { ProviderRequestOptions } from '@earendil-works/pi-ai';
+import type { AudioFormat, CharAlignment } from '../audio.ts';
+export interface ElevenFile {
+    data: Uint8Array;
+    mimeType: string;
+    filename: string;
+}
+interface HeaderModel {
+    provider: string;
+    headers?: Record<string, string>;
+}
+export declare function mediaType(mimeType: string): string;
+export declare function audioFilename(mimeType: string): string;
+export declare function pcmS16(mimeType: string): boolean;
+export declare function encodeForm(file: ElevenFile, fields: Record<string, string>): ArrayBuffer;
+export declare function formType(): string;
+export declare function xiHeaders(model: HeaderModel, options: ProviderRequestOptions | undefined): Record<string, string>;
+export declare function socketUrl(baseUrl: string, path: string): URL;
+export declare function outputFormatName(format: AudioFormat): string;
+export declare function sttAudioFormat(format: AudioFormat): string;
+export declare function connectSocket(url: string, headers: Record<string, string>): WebSocket;
+export declare function awaitOpen(socket: WebSocket, label: string, fail: (message: string) => void): Promise<void>;
+export declare function messageText(data: unknown): Promise<string>;
+export declare function sameFormat(left: AudioFormat, right: AudioFormat): boolean;
+export declare function alignmentFrom(value: unknown): CharAlignment | undefined;
+export {};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-ecoverse/pi-audio",
-  "version": "1.12.0",
+  "version": "1.13.0",
   "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",
@@ -46,6 +46,10 @@
       "types": "./dist/types/xai/index.d.ts",
       "default": "./dist/xai/index.js"
     },
+    "./elevenlabs": {
+      "types": "./dist/types/elevenlabs/index.d.ts",
+      "default": "./dist/elevenlabs/index.js"
+    },
     "./durable": {
       "types": "./dist/types/durable/index.d.ts",
       "default": "./dist/durable/index.js"