@ai-ecoverse/pi-audio 1.9.0 → 1.11.0

package/dist/xai/speak.js

@@ -0,0 +1,208 @@
+import { chunkSeconds } from "../audio.js";
+import { decodeAudioBase64 } from "../convert.js";
+import { VoiceRequestError } from "../errors.js";
+import {
+  applyPayload,
+  asVoiceError,
+  endpoint,
+  isRecord,
+  requestHeaders,
+  stringField,
+  voiceFetch
+} from "../openai/wire.js";
+import { SpeechEventStream } from "../speech.js";
+import { priceAudio } from "../usage.js";
+import { awaitOpen, codecName, connectSocket, messageText, sameFormat, socketUrl } from "./wire.js";
+function languageOf(language) {
+  if (language === void 0 || language.trim() === "") return "auto";
+  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 outputFormat(format) {
+  const body = {
+    codec: codecName(format.encoding),
+    sample_rate: format.sampleRate
+  };
+  if (format.encoding === "mp3") body.bit_rate = 128e3;
+  return body;
+}
+function ttsUrl(baseUrl, voice, language, format) {
+  const url = socketUrl(baseUrl, "tts");
+  const params = new URLSearchParams();
+  params.set("language", language);
+  params.set("voice", voice);
+  params.set("codec", codecName(format.encoding));
+  params.set("sample_rate", String(format.sampleRate));
+  if (format.encoding === "mp3") params.set("bit_rate", "128000");
+  url.search = params.toString();
+  return url.toString();
+}
+function speakXAI(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);
+    const language = languageOf(input.language);
+    if (typeof input.text === "string") {
+      await postSpeech(model, input.text, input.voice, language, format, options, stream);
+      return;
+    }
+    await streamSpeech(model, input.text, input.voice, 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 = {
+    text,
+    voice_id: voice,
+    language,
+    output_format: outputFormat(format)
+  };
+  const body = await applyPayload(payload, model, options);
+  if (!isRecord(body)) throw new Error("Speech payload was not JSON");
+  const response = await voiceFetch(
+    model,
+    options,
+    endpoint(model.baseUrl, "tts"),
+    JSON.stringify(body),
+    "application/json"
+  );
+  const data = new Uint8Array(await response.arrayBuffer());
+  const seconds = chunkSeconds({ data, format, timestamp: 0 });
+  const meters = { characters: text.length };
+  if (seconds > 0) meters.outputSeconds = seconds;
+  stream.push({ type: "start", format });
+  stream.push({
+    type: "audio",
+    chunk: { data, format, timestamp: 0 },
+    span: { itemId: model.id, start: 0, end: text.length }
+  });
+  stream.push({ type: "done", usage: priceAudio(model.audioCost, meters) });
+}
+async function streamSpeech(model, text, voice, language, 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 headers = requestHeaders(model, options);
+  const socket = connectSocket(ttsUrl(model.baseUrl, voice, language, format), headers);
+  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: "xAI speech socket closed" });
+    }
+    wake();
+  });
+  try {
+    await awaitOpen(socket, "xAI 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({ type: "text.done" }));
+    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 = "";
+  for await (const part of text) {
+    if (signal?.aborted) throw new VoiceRequestError("aborted", "aborted");
+    combined += part;
+    if (part !== "") socket.send(JSON.stringify({ type: "text.delta", delta: part }));
+  }
+  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("xAI 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 applySpeech(event, itemId, format, state, push) {
+  const type = stringField(event, "type");
+  if (type === "audio.delta") {
+    const delta = stringField(event, "delta");
+    const payload = delta === void 0 ? "" : delta;
+    const data = decodeAudioBase64(payload);
+    if (!state.started) {
+      state.started = true;
+      push({ type: "start", format });
+    }
+    state.outputSeconds += chunkSeconds({ data, format, timestamp: 0 });
+    push({
+      type: "audio",
+      chunk: { data, format, timestamp: 0 },
+      span: { itemId, start: 0, end: state.characters }
+    });
+    return;
+  }
+  if (type === "audio.done") {
+    state.done = true;
+    return;
+  }
+  if (type === "error") {
+    const message = stringField(event, "message");
+    state.failed = true;
+    push({
+      type: "error",
+      reason: "error",
+      error: message === void 0 ? "xAI speech failed" : message
+    });
+  }
+}
+export {
+  speakXAI
+};
+//# sourceMappingURL=speak.js.map

package/dist/xai/speak.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/xai/speak.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport type { AudioFormat } 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  requestHeaders,\n  stringField,\n  voiceFetch,\n} from '../openai/wire.ts';\nimport { type SpeechEvent, SpeechEventStream, type SpeechInput } from '../speech.ts';\nimport { type AudioCounters, priceAudio } from '../usage.ts';\nimport { awaitOpen, codecName, connectSocket, messageText, sameFormat, socketUrl } 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 languageOf(language: string | undefined): string {\n  if (language === undefined || language.trim() === '') return 'auto';\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 outputFormat(format: AudioFormat): Record<string, number | string> {\n  const body: Record<string, number | string> = {\n    codec: codecName(format.encoding),\n    sample_rate: format.sampleRate,\n  };\n  if (format.encoding === 'mp3') body.bit_rate = 128000;\n  return body;\n}\n\nfunction ttsUrl(baseUrl: string, voice: string, language: string, format: AudioFormat): string {\n  const url = socketUrl(baseUrl, 'tts');\n  const params = new URLSearchParams();\n  params.set('language', language);\n  params.set('voice', voice);\n  params.set('codec', codecName(format.encoding));\n  params.set('sample_rate', String(format.sampleRate));\n  if (format.encoding === 'mp3') params.set('bit_rate', '128000');\n  url.search = params.toString();\n  return url.toString();\n}\n\nexport function speakXAI(\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    const language = languageOf(input.language);\n    if (typeof input.text === 'string') {\n      await postSpeech(model, input.text, input.voice, language, format, options, stream);\n      return;\n    }\n    await streamSpeech(model, input.text, input.voice, 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,\n  format: AudioFormat,\n  options: ProviderRequestOptions | undefined,\n  stream: SpeechEventStream\n) {\n  const payload = {\n    text,\n    voice_id: voice,\n    language,\n    output_format: outputFormat(format),\n  };\n  const body = await applyPayload(payload, model, options);\n  if (!isRecord(body)) throw new Error('Speech payload was not JSON');\n  const response = await voiceFetch(\n    model,\n    options,\n    endpoint(model.baseUrl, 'tts'),\n    JSON.stringify(body),\n    'application/json'\n  );\n  const data = new Uint8Array(await response.arrayBuffer());\n  const seconds = chunkSeconds({ data, format, timestamp: 0 });\n  const meters: AudioCounters = { characters: text.length };\n  if (seconds > 0) meters.outputSeconds = seconds;\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  });\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  language: 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 headers = requestHeaders(model, options);\n  const socket = connectSocket(ttsUrl(model.baseUrl, voice, language, format), headers);\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: 'xAI speech socket closed' });\n    }\n    wake();\n  });\n  try {\n    await awaitOpen(socket, 'xAI 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({ type: 'text.done' }));\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  for await (const part of text) {\n    if (signal?.aborted) throw new VoiceRequestError('aborted', 'aborted');\n    combined += part;\n    if (part !== '') socket.send(JSON.stringify({ type: 'text.delta', delta: part }));\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('xAI 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 applySpeech(\n  event: Record<string, unknown>,\n  itemId: string,\n  format: AudioFormat,\n  state: SpeechState,\n  push: (event: SpeechEvent) => void\n) {\n  const type = stringField(event, 'type');\n  if (type === 'audio.delta') {\n    const delta = stringField(event, 'delta');\n    const payload = delta === undefined ? '' : delta;\n    const data = decodeAudioBase64(payload);\n    if (!state.started) {\n      state.started = true;\n      push({ type: 'start', format });\n    }\n    state.outputSeconds += chunkSeconds({ data, format, timestamp: 0 });\n    push({\n      type: 'audio',\n      chunk: { data, format, timestamp: 0 },\n      span: { itemId, start: 0, end: state.characters },\n    });\n    return;\n  }\n  if (type === 'audio.done') {\n    state.done = true;\n    return;\n  }\n  if (type === 'error') {\n    const message = stringField(event, 'message');\n    state.failed = true;\n    push({\n      type: 'error',\n      reason: 'error',\n      error: message === undefined ? 'xAI speech failed' : message,\n    });\n  }\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,SAAS,WAAW,WAAW,eAAe,aAAa,YAAY,iBAAiB;AAWxF,SAAS,WAAW,UAAsC;AACxD,MAAI,aAAa,UAAa,SAAS,KAAK,MAAM,GAAI,QAAO;AAC7D,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,aAAa,QAAsD;AAC1E,QAAM,OAAwC;AAAA,IAC5C,OAAO,UAAU,OAAO,QAAQ;AAAA,IAChC,aAAa,OAAO;AAAA,EACtB;AACA,MAAI,OAAO,aAAa,MAAO,MAAK,WAAW;AAC/C,SAAO;AACT;AAEA,SAAS,OAAO,SAAiB,OAAe,UAAkB,QAA6B;AAC7F,QAAM,MAAM,UAAU,SAAS,KAAK;AACpC,QAAM,SAAS,IAAI,gBAAgB;AACnC,SAAO,IAAI,YAAY,QAAQ;AAC/B,SAAO,IAAI,SAAS,KAAK;AACzB,SAAO,IAAI,SAAS,UAAU,OAAO,QAAQ,CAAC;AAC9C,SAAO,IAAI,eAAe,OAAO,OAAO,UAAU,CAAC;AACnD,MAAI,OAAO,aAAa,MAAO,QAAO,IAAI,YAAY,QAAQ;AAC9D,MAAI,SAAS,OAAO,SAAS;AAC7B,SAAO,IAAI,SAAS;AACtB;AAEO,SAAS,SACd,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,UAAM,WAAW,WAAW,MAAM,QAAQ;AAC1C,QAAI,OAAO,MAAM,SAAS,UAAU;AAClC,YAAM,WAAW,OAAO,MAAM,MAAM,MAAM,OAAO,UAAU,QAAQ,SAAS,MAAM;AAClF;AAAA,IACF;AACA,UAAM,aAAa,OAAO,MAAM,MAAM,MAAM,OAAO,UAAU,QAAQ,SAAS,MAAM;AAAA,EACtF,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;AAAA,IACd;AAAA,IACA,UAAU;AAAA,IACV;AAAA,IACA,eAAe,aAAa,MAAM;AAAA,EACpC;AACA,QAAM,OAAO,MAAM,aAAa,SAAS,OAAO,OAAO;AACvD,MAAI,CAAC,SAAS,IAAI,EAAG,OAAM,IAAI,MAAM,6BAA6B;AAClE,QAAM,WAAW,MAAM;AAAA,IACrB;AAAA,IACA;AAAA,IACA,SAAS,MAAM,SAAS,KAAK;AAAA,IAC7B,KAAK,UAAU,IAAI;AAAA,IACnB;AAAA,EACF;AACA,QAAM,OAAO,IAAI,WAAW,MAAM,SAAS,YAAY,CAAC;AACxD,QAAM,UAAU,aAAa,EAAE,MAAM,QAAQ,WAAW,EAAE,CAAC;AAC3D,QAAM,SAAwB,EAAE,YAAY,KAAK,OAAO;AACxD,MAAI,UAAU,EAAG,QAAO,gBAAgB;AACxC,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,EACvD,CAAC;AACD,SAAO,KAAK,EAAE,MAAM,QAAQ,OAAO,WAAW,MAAM,WAAW,MAAM,EAAE,CAAC;AAC1E;AAEA,eAAe,aACb,OACA,MACA,OACA,UACA,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,UAAU,eAAe,OAAO,OAAO;AAC7C,QAAM,SAAS,cAAc,OAAO,MAAM,SAAS,OAAO,UAAU,MAAM,GAAG,OAAO;AACpF,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,2BAA2B,CAAC;AAAA,IAC5E;AACA,SAAK;AAAA,EACP,CAAC;AACD,MAAI;AACF,UAAM,UAAU,QAAQ,cAAc,CAAC,YAAY;AACjD,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,YAAY,CAAC,CAAC;AACjD,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,mBAAiB,QAAQ,MAAM;AAC7B,QAAI,QAAQ,QAAS,OAAM,IAAI,kBAAkB,WAAW,SAAS;AACrE,gBAAY;AACZ,QAAI,SAAS,GAAI,QAAO,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,OAAO,KAAK,CAAC,CAAC;AAAA,EAClF;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,+BAA+B;AACtE,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,YACP,OACA,QACA,QACA,OACA,MACA;AACA,QAAM,OAAO,YAAY,OAAO,MAAM;AACtC,MAAI,SAAS,eAAe;AAC1B,UAAM,QAAQ,YAAY,OAAO,OAAO;AACxC,UAAM,UAAU,UAAU,SAAY,KAAK;AAC3C,UAAM,OAAO,kBAAkB,OAAO;AACtC,QAAI,CAAC,MAAM,SAAS;AAClB,YAAM,UAAU;AAChB,WAAK,EAAE,MAAM,SAAS,OAAO,CAAC;AAAA,IAChC;AACA,UAAM,iBAAiB,aAAa,EAAE,MAAM,QAAQ,WAAW,EAAE,CAAC;AAClE,SAAK;AAAA,MACH,MAAM;AAAA,MACN,OAAO,EAAE,MAAM,QAAQ,WAAW,EAAE;AAAA,MACpC,MAAM,EAAE,QAAQ,OAAO,GAAG,KAAK,MAAM,WAAW;AAAA,IAClD,CAAC;AACD;AAAA,EACF;AACA,MAAI,SAAS,cAAc;AACzB,UAAM,OAAO;AACb;AAAA,EACF;AACA,MAAI,SAAS,SAAS;AACpB,UAAM,UAAU,YAAY,OAAO,SAAS;AAC5C,UAAM,SAAS;AACf,SAAK;AAAA,MACH,MAAM;AAAA,MACN,QAAQ;AAAA,MACR,OAAO,YAAY,SAAY,sBAAsB;AAAA,IACvD,CAAC;AAAA,EACH;AACF;",
+  "names": ["voice"]
+}

package/dist/xai/transcribe.js

@@ -0,0 +1,113 @@
+import { VoiceRequestError } from "../errors.js";
+import {
+  applyPayload,
+  asVoiceError,
+  endpoint,
+  isRecord,
+  numberField,
+  readAudio,
+  readJson,
+  requestSignal,
+  stringField,
+  voiceFetch
+} from "../openai/wire.js";
+import { priceAudio } from "../usage.js";
+import { audioFilename, encodeForm, formType, rawFields } from "./wire.js";
+function durationOf(record) {
+  const duration = numberField(record, "duration");
+  if (duration === void 0) return 0;
+  return duration;
+}
+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 speakerOf(record) {
+  const value = record.speaker;
+  if (typeof value === "number") return String(value);
+  if (typeof value === "string") return value;
+  return void 0;
+}
+function wordFrom(word) {
+  const speaker = speakerOf(word);
+  const segment = {
+    text: textOf(word, "text"),
+    start: secondsOf(word, "start"),
+    end: secondsOf(word, "end")
+  };
+  if (speaker !== void 0) segment.speaker = speaker;
+  return segment;
+}
+function transcriptFrom(body, model) {
+  if (!isRecord(body)) throw new Error("Transcription response had no text");
+  const text = stringField(body, "text");
+  if (text === void 0) throw new Error("Transcription response had no text");
+  const language = stringField(body, "language");
+  const words = Array.isArray(body.words) ? body.words.filter(isRecord).map(wordFrom) : void 0;
+  return {
+    text,
+    ...language !== void 0 ? { language } : {},
+    ...words !== void 0 ? { segments: words } : {},
+    usage: priceAudio(model.audioCost, { inputSeconds: durationOf(body) })
+  };
+}
+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 transcribeXAI(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: model.id,
+      file: { data: loaded.data, mimeType: loaded.mimeType, filename }
+    };
+    if (model.diarization) payload.diarize = "true";
+    const raw = rawFields(loaded.mimeType, model);
+    if (raw) {
+      payload.audio_format = raw.audio_format;
+      payload.sample_rate = raw.sample_rate;
+    }
+    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 voiceFetch(
+      model,
+      options,
+      endpoint(model.baseUrl, "stt"),
+      encoded,
+      formType()
+    );
+    return transcriptFrom(await readJson(response), model);
+  } catch (error) {
+    const voice = asVoiceError(error);
+    throw new VoiceRequestError(voice.reason, voice.error);
+  }
+}
+export {
+  transcribeXAI
+};
+//# sourceMappingURL=transcribe.js.map

package/dist/xai/transcribe.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/xai/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  voiceFetch,\n} from '../openai/wire.ts';\nimport type { AudioInput, Transcript, TranscriptSegment } from '../transcription.ts';\nimport { priceAudio } from '../usage.ts';\nimport { audioFilename, encodeForm, formType, rawFields, type XaiFile } from './wire.ts';\n\nfunction durationOf(record: Record<string, unknown>): number {\n  const duration = numberField(record, 'duration');\n  if (duration === undefined) return 0;\n  return duration;\n}\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 speakerOf(record: Record<string, unknown>): string | undefined {\n  const value = record.speaker;\n  if (typeof value === 'number') return String(value);\n  if (typeof value === 'string') return value;\n  return undefined;\n}\n\nfunction wordFrom(word: Record<string, unknown>): TranscriptSegment {\n  const speaker = speakerOf(word);\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 transcriptFrom(body: unknown, model: TranscriptionModel<string>): Transcript {\n  if (!isRecord(body)) throw new Error('Transcription response had no text');\n  const text = stringField(body, 'text');\n  if (text === undefined) throw new Error('Transcription response had no text');\n  const language = stringField(body, 'language');\n  const words = Array.isArray(body.words) ? body.words.filter(isRecord).map(wordFrom) : undefined;\n  return {\n    text,\n    ...(language !== undefined ? { language } : {}),\n    ...(words !== undefined ? { segments: words } : {}),\n    usage: priceAudio(model.audioCost, { inputSeconds: durationOf(body) }),\n  };\n}\n\nfunction fileOf(body: Record<string, unknown>): XaiFile {\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\nexport async function transcribeXAI(\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: model.id,\n      file: { data: loaded.data, mimeType: loaded.mimeType, filename },\n    };\n    if (model.diarization) payload.diarize = 'true';\n    const raw = rawFields(loaded.mimeType, model);\n    if (raw) {\n      payload.audio_format = raw.audio_format;\n      payload.sample_rate = raw.sample_rate;\n    }\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 voiceFetch(\n      model,\n      options,\n      endpoint(model.baseUrl, 'stt'),\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,EACA;AAAA,OACK;AAEP,SAAS,kBAAkB;AAC3B,SAAS,eAAe,YAAY,UAAU,iBAA+B;AAE7E,SAAS,WAAW,QAAyC;AAC3D,QAAM,WAAW,YAAY,QAAQ,UAAU;AAC/C,MAAI,aAAa,OAAW,QAAO;AACnC,SAAO;AACT;AAEA,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,UAAU,QAAqD;AACtE,QAAM,QAAQ,OAAO;AACrB,MAAI,OAAO,UAAU,SAAU,QAAO,OAAO,KAAK;AAClD,MAAI,OAAO,UAAU,SAAU,QAAO;AACtC,SAAO;AACT;AAEA,SAAS,SAAS,MAAkD;AAClE,QAAM,UAAU,UAAU,IAAI;AAC9B,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,eAAe,MAAe,OAA+C;AACpF,MAAI,CAAC,SAAS,IAAI,EAAG,OAAM,IAAI,MAAM,oCAAoC;AACzE,QAAM,OAAO,YAAY,MAAM,MAAM;AACrC,MAAI,SAAS,OAAW,OAAM,IAAI,MAAM,oCAAoC;AAC5E,QAAM,WAAW,YAAY,MAAM,UAAU;AAC7C,QAAM,QAAQ,MAAM,QAAQ,KAAK,KAAK,IAAI,KAAK,MAAM,OAAO,QAAQ,EAAE,IAAI,QAAQ,IAAI;AACtF,SAAO;AAAA,IACL;AAAA,IACA,GAAI,aAAa,SAAY,EAAE,SAAS,IAAI,CAAC;AAAA,IAC7C,GAAI,UAAU,SAAY,EAAE,UAAU,MAAM,IAAI,CAAC;AAAA,IACjD,OAAO,WAAW,MAAM,WAAW,EAAE,cAAc,WAAW,IAAI,EAAE,CAAC;AAAA,EACvE;AACF;AAEA,SAAS,OAAO,MAAwC;AACtD,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,eAAsB,cACpB,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,OAAO,MAAM;AAAA,MACb,MAAM,EAAE,MAAM,OAAO,MAAM,UAAU,OAAO,UAAU,SAAS;AAAA,IACjE;AACA,QAAI,MAAM,YAAa,SAAQ,UAAU;AACzC,UAAM,MAAM,UAAU,OAAO,UAAU,KAAK;AAC5C,QAAI,KAAK;AACP,cAAQ,eAAe,IAAI;AAC3B,cAAQ,cAAc,IAAI;AAAA,IAC5B;AACA,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,KAAK;AAAA,MAC7B;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/xai/wire.js

@@ -0,0 +1,171 @@
+const boundary = "pi-audio-xai-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"
+};
+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 ${mimeType}`);
+  return name;
+}
+function declaredRate(model, encoding) {
+  const match = model.inputFormats.find((format) => format.encoding === encoding);
+  if (!match) return "16000";
+  return String(match.sampleRate);
+}
+function parameter(mimeType, name) {
+  for (const part of mimeType.split(";").slice(1)) {
+    const pieces = part.split("=");
+    const key = pieces[0]?.trim().toLowerCase();
+    const value = pieces[1]?.trim();
+    if (key === name && value) return value;
+  }
+  return void 0;
+}
+function rawFields(mimeType, model) {
+  const type = mediaType(mimeType);
+  const rate = parameter(mimeType, "rate");
+  if (type === "audio/pcm" || type === "audio/l16") {
+    const sampleRate = rate === void 0 ? declaredRate(model, "pcm16") : rate;
+    return { audio_format: "pcm", sample_rate: sampleRate };
+  }
+  if (type === "audio/mulaw" || type === "audio/basic") {
+    const sampleRate = rate === void 0 ? declaredRate(model, "mulaw") : rate;
+    return { audio_format: "mulaw", sample_rate: sampleRate };
+  }
+  if (type === "audio/alaw") {
+    const sampleRate = rate === void 0 ? declaredRate(model, "alaw") : rate;
+    return { audio_format: "alaw", sample_rate: sampleRate };
+  }
+  return void 0;
+}
+function encodeForm(file, fields) {
+  const chunks = [];
+  const encoder = new TextEncoder();
+  const write = (value) => {
+    chunks.push(encoder.encode(value));
+  };
+  for (const [name, value] of Object.entries(fields)) {
+    write(`--${boundary}\r
+Content-Disposition: form-data; name="${name}"\r
+\r
+${value}\r
+`);
+  }
+  write(
+    `--${boundary}\r
+Content-Disposition: form-data; name="file"; filename="${file.filename}"\r
+Content-Type: ${file.mimeType}\r
+\r
+`
+  );
+  chunks.push(file.data);
+  write(`\r
+--${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=${boundary}`;
+}
+function socketUrl(baseUrl, path) {
+  const url = new URL(baseUrl);
+  url.protocol = url.protocol === "https:" ? "wss:" : "ws:";
+  url.pathname = `${url.pathname.replace(/\/$/, "")}/${path}`;
+  url.search = "";
+  return url;
+}
+function codecName(encoding) {
+  if (encoding === "pcm16") return "pcm";
+  if (encoding === "mp3") return "mp3";
+  if (encoding === "wav") return "wav";
+  if (encoding === "mulaw") return "mulaw";
+  if (encoding === "alaw") return "alaw";
+  throw new Error(`xAI speech cannot return ${encoding}`);
+}
+function encodingName(encoding) {
+  if (encoding === "pcm16") return "pcm";
+  if (encoding === "mulaw") return "mulaw";
+  if (encoding === "alaw") return "alaw";
+  throw new Error(`Unsupported audio format ${encoding}`);
+}
+function connectSocket(url, headers) {
+  const Socket = WebSocket;
+  return new Socket(url, { headers });
+}
+function awaitOpen(socket, label, fail) {
+  return new Promise((resolve) => {
+    let opened = false;
+    socket.addEventListener("open", () => {
+      opened = true;
+      resolve();
+    });
+    socket.addEventListener("error", () => {
+      fail(`${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 copiedBytes(data) {
+  const copy = new ArrayBuffer(data.byteLength);
+  new Uint8Array(copy).set(data);
+  return copy;
+}
+export {
+  audioFilename,
+  awaitOpen,
+  codecName,
+  connectSocket,
+  copiedBytes,
+  encodeForm,
+  encodingName,
+  formType,
+  mediaType,
+  messageText,
+  rawFields,
+  sameFormat,
+  socketUrl
+};
+//# sourceMappingURL=wire.js.map

package/dist/xai/wire.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/xai/wire.ts"],
+  "sourcesContent": ["import type { AudioFormat } from '../audio.ts';\nimport type { TranscriptionModel } from '../models.ts';\n\nconst boundary = 'pi-audio-xai-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\nexport interface XaiFile {\n  data: Uint8Array;\n  mimeType: string;\n  filename: 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 ${mimeType}`);\n  return name;\n}\n\nfunction declaredRate(\n  model: TranscriptionModel<string>,\n  encoding: AudioFormat['encoding']\n): string {\n  const match = model.inputFormats.find((format) => format.encoding === encoding);\n  if (!match) return '16000';\n  return String(match.sampleRate);\n}\n\nfunction parameter(mimeType: string, name: string): string | undefined {\n  for (const part of mimeType.split(';').slice(1)) {\n    const pieces = part.split('=');\n    const key = pieces[0]?.trim().toLowerCase();\n    const value = pieces[1]?.trim();\n    if (key === name && value) return value;\n  }\n  return undefined;\n}\n\nexport function rawFields(\n  mimeType: string,\n  model: TranscriptionModel<string>\n): Record<string, string> | undefined {\n  const type = mediaType(mimeType);\n  const rate = parameter(mimeType, 'rate');\n  if (type === 'audio/pcm' || type === 'audio/l16') {\n    const sampleRate = rate === undefined ? declaredRate(model, 'pcm16') : rate;\n    return { audio_format: 'pcm', sample_rate: sampleRate };\n  }\n  if (type === 'audio/mulaw' || type === 'audio/basic') {\n    const sampleRate = rate === undefined ? declaredRate(model, 'mulaw') : rate;\n    return { audio_format: 'mulaw', sample_rate: sampleRate };\n  }\n  if (type === 'audio/alaw') {\n    const sampleRate = rate === undefined ? declaredRate(model, 'alaw') : rate;\n    return { audio_format: 'alaw', sample_rate: sampleRate };\n  }\n  return undefined;\n}\n\nexport function encodeForm(file: XaiFile, fields: Record<string, string>): ArrayBuffer {\n  const chunks: Uint8Array[] = [];\n  const encoder = new TextEncoder();\n  const write = (value: string) => {\n    chunks.push(encoder.encode(value));\n  };\n  for (const [name, value] of Object.entries(fields)) {\n    write(`--${boundary}\\r\\nContent-Disposition: form-data; name=\"${name}\"\\r\\n\\r\\n${value}\\r\\n`);\n  }\n  write(\n    `--${boundary}\\r\\nContent-Disposition: form-data; name=\"file\"; filename=\"${file.filename}\"\\r\\nContent-Type: ${file.mimeType}\\r\\n\\r\\n`\n  );\n  chunks.push(file.data);\n  write(`\\r\\n--${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=${boundary}`;\n}\n\nexport function socketUrl(baseUrl: string, path: string): URL {\n  const url = new URL(baseUrl);\n  url.protocol = url.protocol === 'https:' ? 'wss:' : 'ws:';\n  url.pathname = `${url.pathname.replace(/\\/$/, '')}/${path}`;\n  url.search = '';\n  return url;\n}\n\nexport function codecName(encoding: AudioFormat['encoding']): string {\n  if (encoding === 'pcm16') return 'pcm';\n  if (encoding === 'mp3') return 'mp3';\n  if (encoding === 'wav') return 'wav';\n  if (encoding === 'mulaw') return 'mulaw';\n  if (encoding === 'alaw') return 'alaw';\n  throw new Error(`xAI speech cannot return ${encoding}`);\n}\n\nexport function encodingName(encoding: AudioFormat['encoding']): string {\n  if (encoding === 'pcm16') return 'pcm';\n  if (encoding === 'mulaw') return 'mulaw';\n  if (encoding === 'alaw') return 'alaw';\n  throw new Error(`Unsupported audio format ${encoding}`);\n}\n\nexport function connectSocket(url: string, headers: Record<string, string>): WebSocket {\n  const Socket = WebSocket as unknown as new (\n    address: string,\n    init: { headers: Record<string, string> }\n  ) => WebSocket;\n  return new Socket(url, { headers });\n}\n\nexport function awaitOpen(\n  socket: WebSocket,\n  label: string,\n  fail: (message: string) => void\n): Promise<void> {\n  return new Promise((resolve) => {\n    let opened = false;\n    socket.addEventListener('open', () => {\n      opened = true;\n      resolve();\n    });\n    socket.addEventListener('error', () => {\n      fail(`${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\nexport function copiedBytes(data: Uint8Array): ArrayBuffer {\n  const copy = new ArrayBuffer(data.byteLength);\n  new Uint8Array(copy).set(data);\n  return copy;\n}\n"],
+  "mappings": "AAGA,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;AAQO,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,aACP,OACA,UACQ;AACR,QAAM,QAAQ,MAAM,aAAa,KAAK,CAAC,WAAW,OAAO,aAAa,QAAQ;AAC9E,MAAI,CAAC,MAAO,QAAO;AACnB,SAAO,OAAO,MAAM,UAAU;AAChC;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,UACd,UACA,OACoC;AACpC,QAAM,OAAO,UAAU,QAAQ;AAC/B,QAAM,OAAO,UAAU,UAAU,MAAM;AACvC,MAAI,SAAS,eAAe,SAAS,aAAa;AAChD,UAAM,aAAa,SAAS,SAAY,aAAa,OAAO,OAAO,IAAI;AACvE,WAAO,EAAE,cAAc,OAAO,aAAa,WAAW;AAAA,EACxD;AACA,MAAI,SAAS,iBAAiB,SAAS,eAAe;AACpD,UAAM,aAAa,SAAS,SAAY,aAAa,OAAO,OAAO,IAAI;AACvE,WAAO,EAAE,cAAc,SAAS,aAAa,WAAW;AAAA,EAC1D;AACA,MAAI,SAAS,cAAc;AACzB,UAAM,aAAa,SAAS,SAAY,aAAa,OAAO,MAAM,IAAI;AACtE,WAAO,EAAE,cAAc,QAAQ,aAAa,WAAW;AAAA,EACzD;AACA,SAAO;AACT;AAEO,SAAS,WAAW,MAAe,QAA6C;AACrF,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;AAAA,IACE,KAAK,QAAQ;AAAA,yDAA8D,KAAK,QAAQ;AAAA,gBAAsB,KAAK,QAAQ;AAAA;AAAA;AAAA,EAC7H;AACA,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;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,UAAU,UAA2C;AACnE,MAAI,aAAa,QAAS,QAAO;AACjC,MAAI,aAAa,MAAO,QAAO;AAC/B,MAAI,aAAa,MAAO,QAAO;AAC/B,MAAI,aAAa,QAAS,QAAO;AACjC,MAAI,aAAa,OAAQ,QAAO;AAChC,QAAM,IAAI,MAAM,4BAA4B,QAAQ,EAAE;AACxD;AAEO,SAAS,aAAa,UAA2C;AACtE,MAAI,aAAa,QAAS,QAAO;AACjC,MAAI,aAAa,QAAS,QAAO;AACjC,MAAI,aAAa,OAAQ,QAAO;AAChC,QAAM,IAAI,MAAM,4BAA4B,QAAQ,EAAE;AACxD;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;AAEO,SAAS,YAAY,MAA+B;AACzD,QAAM,OAAO,IAAI,YAAY,KAAK,UAAU;AAC5C,MAAI,WAAW,IAAI,EAAE,IAAI,IAAI;AAC7B,SAAO;AACT;",
+  "names": []
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-ecoverse/pi-audio",
-  "version": "1.9.0",
+  "version": "1.11.0",
   "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",
@@ -45,6 +45,14 @@
     "./xai": {
       "types": "./dist/types/xai/index.d.ts",
       "default": "./dist/xai/index.js"
+    },
+    "./durable": {
+      "types": "./dist/types/durable/index.d.ts",
+      "default": "./dist/durable/index.js"
+    },
+    "./gemini-live": {
+      "types": "./dist/types/gemini-live/index.d.ts",
+      "default": "./dist/gemini-live/index.js"
     }
   },
   "files": [
@@ -68,6 +76,7 @@
     "@ai-ecoverse/slicc-shared-web": "1.21.1",
     "@biomejs/biome": "2.5.15",
     "@earendil-works/pi-ai": "1.1.0",
+    "@earendil-works/pi-durable": "1.1.0",
     "@types/node": "24.19.1",
     "esbuild": "0.28.2",
     "lefthook": "2.2.1",
@@ -76,6 +85,12 @@
     "typescript": "7.0.2"
   },
   "peerDependencies": {
-    "@earendil-works/pi-ai": "1.1.0"
+    "@earendil-works/pi-ai": "1.1.0",
+    "@earendil-works/pi-durable": "^1.1.0"
+  },
+  "peerDependenciesMeta": {
+    "@earendil-works/pi-durable": {
+      "optional": true
+    }
   }
 }