@ai-ecoverse/pi-audio 1.10.0 → 1.12.0

package/dist/xai/listen.js

@@ -0,0 +1,211 @@
+import { chunkSeconds } from "../audio.js";
+import {
+  asVoiceError,
+  isRecord,
+  numberField,
+  requestHeaders,
+  stringField
+} from "../openai/wire.js";
+import {
+  TranscriptionEventStream
+} from "../transcription.js";
+import { priceAudio } from "../usage.js";
+import {
+  awaitOpen,
+  connectSocket,
+  copiedBytes,
+  encodingName,
+  messageText,
+  sameFormat,
+  socketUrl
+} from "./wire.js";
+function terms(options) {
+  const vocabulary = options?.vocabulary;
+  if (!vocabulary) return [];
+  return vocabulary;
+}
+function sttUrl(model, format, options) {
+  const url = socketUrl(model.baseUrl, "stt");
+  const params = new URLSearchParams();
+  params.set("model", model.id);
+  params.set("sample_rate", String(format.sampleRate));
+  params.set("encoding", encodingName(format.encoding));
+  params.set("interim_results", "true");
+  if (model.turnDetection === "semantic") params.set("smart_turn", "0.5");
+  if (model.diarization) params.set("diarize", "true");
+  if (options?.language) params.set("language", options.language);
+  for (const term of terms(options)) params.append("keyterm", 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");
+  encodingName(format.encoding);
+  const known = model.inputFormats.some((item) => sameFormat(item, format));
+  if (!known) throw new Error(`Unsupported audio format ${format.encoding}`);
+  return format;
+}
+function textOr(event, key) {
+  const text = stringField(event, key);
+  if (text === void 0) return "";
+  return text;
+}
+function applyListen(event, state, push) {
+  const type = stringField(event, "type");
+  if (type === "transcript.created") {
+    state.ready = true;
+    return;
+  }
+  if (type === "transcript.partial") {
+    const spoken = textOr(event, "text");
+    if (!state.started) {
+      state.started = true;
+      const start = numberField(event, "start");
+      push({ type: "speech_start", at: start === void 0 ? 0 : start * 1e3 });
+    }
+    if (event.is_final === true) {
+      const segmentId = String(state.segment);
+      state.segment += 1;
+      state.turns.push(spoken);
+      push({ type: "final", segmentId, text: spoken });
+      if (event.speech_final === true) {
+        state.started = false;
+        push({ type: "turn_end", text: spoken, at: 0 });
+      }
+      return;
+    }
+    push({ type: "partial", segmentId: String(state.segment), text: spoken });
+    return;
+  }
+  if (type === "transcript.done") {
+    const text = stringField(event, "text");
+    state.doneText = text === void 0 ? state.turns.join(" ") : text;
+    const duration = numberField(event, "duration");
+    if (duration !== void 0) state.duration = duration;
+    return;
+  }
+  if (type === "error") {
+    const message = stringField(event, "message");
+    state.failed = true;
+    push({
+      type: "error",
+      reason: "error",
+      error: message === void 0 ? "xAI 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 });
+  const seconds = state.duration === void 0 ? heard : state.duration;
+  const text = state.doneText === void 0 ? state.turns.join(" ") : state.doneText;
+  push({
+    type: "done",
+    transcript: { text, usage: priceAudio(model.audioCost, { inputSeconds: seconds }) }
+  });
+}
+function listenXAI(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: [],
+    segment: 0,
+    started: false,
+    bytes: 0,
+    opened: false,
+    ready: false,
+    clientDone: false,
+    failed: false,
+    completed: false,
+    duration: void 0,
+    doneText: 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 = () => {
+    if (state.ready || state.failed) readyWait.resolve();
+    if (state.failed || state.doneText !== void 0) 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), requestHeaders(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: "xAI transcription socket closed" });
+      }
+      wake();
+      if (state.clientDone) doneWait.resolve();
+    });
+    await awaitOpen(socket, "xAI transcription", (message) => {
+      state.failed = true;
+      push({ type: "error", reason: "error", error: message });
+      wake();
+    });
+    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);
+    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);
+    }
+    if (state.failed) return;
+    state.clientDone = true;
+    socket.send(JSON.stringify({ type: "audio.done" }));
+    if (state.doneText === void 0) 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) {
+  if (!sameFormat(chunk.format, format)) {
+    throw new Error(`Unsupported audio format ${chunk.format.encoding}`);
+  }
+  state.bytes += chunk.data.byteLength;
+  socket.send(copiedBytes(chunk.data));
+}
+async function onMessage(data, state, push, wake) {
+  try {
+    const parsed = JSON.parse(await messageText(data));
+    if (!isRecord(parsed)) throw new Error("xAI transcription event was not JSON");
+    applyListen(parsed, state, push);
+    wake();
+  } catch (error) {
+    const voice = asVoiceError(error);
+    state.failed = true;
+    push({ type: "error", reason: voice.reason, error: voice.error });
+    wake();
+  }
+}
+export {
+  listenXAI
+};
+//# sourceMappingURL=listen.js.map

package/dist/xai/listen.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/xai/listen.ts"],
+  "sourcesContent": ["import type { AudioChunk, AudioFormat } from '../audio.ts';\nimport { chunkSeconds } from '../audio.ts';\nimport type { TranscriptionModel } from '../models.ts';\nimport {\n  asVoiceError,\n  isRecord,\n  numberField,\n  requestHeaders,\n  stringField,\n} from '../openai/wire.ts';\nimport {\n  type ListenOptions,\n  type TranscriptionEvent,\n  TranscriptionEventStream,\n} from '../transcription.ts';\nimport { priceAudio } from '../usage.ts';\nimport {\n  awaitOpen,\n  connectSocket,\n  copiedBytes,\n  encodingName,\n  messageText,\n  sameFormat,\n  socketUrl,\n} from './wire.ts';\n\ninterface ListenState {\n  turns: string[];\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  duration: number | undefined;\n  doneText: 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 sttUrl(\n  model: TranscriptionModel<string>,\n  format: AudioFormat,\n  options: ListenOptions | undefined\n): string {\n  const url = socketUrl(model.baseUrl, 'stt');\n  const params = new URLSearchParams();\n  params.set('model', model.id);\n  params.set('sample_rate', String(format.sampleRate));\n  params.set('encoding', encodingName(format.encoding));\n  params.set('interim_results', 'true');\n  if (model.turnDetection === 'semantic') params.set('smart_turn', '0.5');\n  if (model.diarization) params.set('diarize', 'true');\n  if (options?.language) params.set('language', options.language);\n  for (const term of terms(options)) params.append('keyterm', 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  encodingName(format.encoding);\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>, key: string): string {\n  const text = stringField(event, key);\n  if (text === undefined) return '';\n  return text;\n}\n\nfunction applyListen(\n  event: Record<string, unknown>,\n  state: ListenState,\n  push: (event: TranscriptionEvent) => void\n) {\n  const type = stringField(event, 'type');\n  if (type === 'transcript.created') {\n    state.ready = true;\n    return;\n  }\n  if (type === 'transcript.partial') {\n    const spoken = textOr(event, 'text');\n    if (!state.started) {\n      state.started = true;\n      const start = numberField(event, 'start');\n      push({ type: 'speech_start', at: start === undefined ? 0 : start * 1000 });\n    }\n    if (event.is_final === true) {\n      const segmentId = String(state.segment);\n      state.segment += 1;\n      state.turns.push(spoken);\n      push({ type: 'final', segmentId, text: spoken });\n      if (event.speech_final === true) {\n        state.started = false;\n        push({ type: 'turn_end', text: spoken, at: 0 });\n      }\n      return;\n    }\n    push({ type: 'partial', segmentId: String(state.segment), text: spoken });\n    return;\n  }\n  if (type === 'transcript.done') {\n    const text = stringField(event, 'text');\n    state.doneText = text === undefined ? state.turns.join(' ') : text;\n    const duration = numberField(event, 'duration');\n    if (duration !== undefined) state.duration = duration;\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 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  const seconds = state.duration === undefined ? heard : state.duration;\n  const text = state.doneText === undefined ? state.turns.join(' ') : state.doneText;\n  push({\n    type: 'done',\n    transcript: { text, usage: priceAudio(model.audioCost, { inputSeconds: seconds }) },\n  });\n}\n\nexport function listenXAI(\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    segment: 0,\n    started: false,\n    bytes: 0,\n    opened: false,\n    ready: false,\n    clientDone: false,\n    failed: false,\n    completed: false,\n    duration: undefined,\n    doneText: 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 = () => {\n    if (state.ready || state.failed) readyWait.resolve();\n    if (state.failed || state.doneText !== undefined) 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), requestHeaders(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: 'xAI transcription socket closed' });\n      }\n      wake();\n      if (state.clientDone) doneWait.resolve();\n    });\n    await awaitOpen(socket, 'xAI transcription', (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    if (!state.ready) await readyWait.promise;\n    if (state.failed) return;\n    if (!first.done) sendChunk(socket, first.value, format, state);\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);\n    }\n    if (state.failed) return;\n    state.clientDone = true;\n    socket.send(JSON.stringify({ type: 'audio.done' }));\n    if (state.doneText === undefined) 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(socket: WebSocket, chunk: AudioChunk, format: AudioFormat, state: ListenState) {\n  if (!sameFormat(chunk.format, format)) {\n    throw new Error(`Unsupported audio format ${chunk.format.encoding}`);\n  }\n  state.bytes += chunk.data.byteLength;\n  socket.send(copiedBytes(chunk.data));\n}\n\nasync function onMessage(\n  data: unknown,\n  state: ListenState,\n  push: (event: TranscriptionEvent) => 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 transcription event was not JSON');\n    applyListen(parsed, state, push);\n    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"],
+  "mappings": "AACA,SAAS,oBAAoB;AAE7B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP;AAAA,EAGE;AAAA,OACK;AACP,SAAS,kBAAkB;AAC3B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAgBP,SAAS,MAAM,SAA8C;AAC3D,QAAM,aAAa,SAAS;AAC5B,MAAI,CAAC,WAAY,QAAO,CAAC;AACzB,SAAO;AACT;AAEA,SAAS,OACP,OACA,QACA,SACQ;AACR,QAAM,MAAM,UAAU,MAAM,SAAS,KAAK;AAC1C,QAAM,SAAS,IAAI,gBAAgB;AACnC,SAAO,IAAI,SAAS,MAAM,EAAE;AAC5B,SAAO,IAAI,eAAe,OAAO,OAAO,UAAU,CAAC;AACnD,SAAO,IAAI,YAAY,aAAa,OAAO,QAAQ,CAAC;AACpD,SAAO,IAAI,mBAAmB,MAAM;AACpC,MAAI,MAAM,kBAAkB,WAAY,QAAO,IAAI,cAAc,KAAK;AACtE,MAAI,MAAM,YAAa,QAAO,IAAI,WAAW,MAAM;AACnD,MAAI,SAAS,SAAU,QAAO,IAAI,YAAY,QAAQ,QAAQ;AAC9D,aAAW,QAAQ,MAAM,OAAO,EAAG,QAAO,OAAO,WAAW,IAAI;AAChE,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,eAAa,OAAO,QAAQ;AAC5B,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,OAAgC,KAAqB;AACnE,QAAM,OAAO,YAAY,OAAO,GAAG;AACnC,MAAI,SAAS,OAAW,QAAO;AAC/B,SAAO;AACT;AAEA,SAAS,YACP,OACA,OACA,MACA;AACA,QAAM,OAAO,YAAY,OAAO,MAAM;AACtC,MAAI,SAAS,sBAAsB;AACjC,UAAM,QAAQ;AACd;AAAA,EACF;AACA,MAAI,SAAS,sBAAsB;AACjC,UAAM,SAAS,OAAO,OAAO,MAAM;AACnC,QAAI,CAAC,MAAM,SAAS;AAClB,YAAM,UAAU;AAChB,YAAM,QAAQ,YAAY,OAAO,OAAO;AACxC,WAAK,EAAE,MAAM,gBAAgB,IAAI,UAAU,SAAY,IAAI,QAAQ,IAAK,CAAC;AAAA,IAC3E;AACA,QAAI,MAAM,aAAa,MAAM;AAC3B,YAAM,YAAY,OAAO,MAAM,OAAO;AACtC,YAAM,WAAW;AACjB,YAAM,MAAM,KAAK,MAAM;AACvB,WAAK,EAAE,MAAM,SAAS,WAAW,MAAM,OAAO,CAAC;AAC/C,UAAI,MAAM,iBAAiB,MAAM;AAC/B,cAAM,UAAU;AAChB,aAAK,EAAE,MAAM,YAAY,MAAM,QAAQ,IAAI,EAAE,CAAC;AAAA,MAChD;AACA;AAAA,IACF;AACA,SAAK,EAAE,MAAM,WAAW,WAAW,OAAO,MAAM,OAAO,GAAG,MAAM,OAAO,CAAC;AACxE;AAAA,EACF;AACA,MAAI,SAAS,mBAAmB;AAC9B,UAAM,OAAO,YAAY,OAAO,MAAM;AACtC,UAAM,WAAW,SAAS,SAAY,MAAM,MAAM,KAAK,GAAG,IAAI;AAC9D,UAAM,WAAW,YAAY,OAAO,UAAU;AAC9C,QAAI,aAAa,OAAW,OAAM,WAAW;AAC7C;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,6BAA6B;AAAA,IAC9D,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,QAAM,UAAU,MAAM,aAAa,SAAY,QAAQ,MAAM;AAC7D,QAAM,OAAO,MAAM,aAAa,SAAY,MAAM,MAAM,KAAK,GAAG,IAAI,MAAM;AAC1E,OAAK;AAAA,IACH,MAAM;AAAA,IACN,YAAY,EAAE,MAAM,OAAO,WAAW,MAAM,WAAW,EAAE,cAAc,QAAQ,CAAC,EAAE;AAAA,EACpF,CAAC;AACH;AAEO,SAAS,UACd,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,SAAS;AAAA,IACT,SAAS;AAAA,IACT,OAAO;AAAA,IACP,QAAQ;AAAA,IACR,OAAO;AAAA,IACP,YAAY;AAAA,IACZ,QAAQ;AAAA,IACR,WAAW;AAAA,IACX,UAAU;AAAA,IACV,UAAU;AAAA,EACZ;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,MAAM;AACjB,QAAI,MAAM,SAAS,MAAM,OAAQ,WAAU,QAAQ;AACnD,QAAI,MAAM,UAAU,MAAM,aAAa,OAAW,UAAS,QAAQ;AAAA,EACrE;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,eAAe,OAAO,OAAO,CAAC;AACrF,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,kCAAkC,CAAC;AAAA,MACnF;AACA,WAAK;AACL,UAAI,MAAM,WAAY,UAAS,QAAQ;AAAA,IACzC,CAAC;AACD,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,QAAI,CAAC,MAAM,MAAO,OAAM,UAAU;AAClC,QAAI,MAAM,OAAQ;AAClB,QAAI,CAAC,MAAM,KAAM,WAAU,QAAQ,MAAM,OAAO,QAAQ,KAAK;AAC7D,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,KAAK;AAAA,IAC7C;AACA,QAAI,MAAM,OAAQ;AAClB,UAAM,aAAa;AACnB,WAAO,KAAK,KAAK,UAAU,EAAE,MAAM,aAAa,CAAC,CAAC;AAClD,QAAI,MAAM,aAAa,OAAW,OAAM,SAAS;AACjD,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,UAAU,QAAmB,OAAmB,QAAqB,OAAoB;AAChG,MAAI,CAAC,WAAW,MAAM,QAAQ,MAAM,GAAG;AACrC,UAAM,IAAI,MAAM,4BAA4B,MAAM,OAAO,QAAQ,EAAE;AAAA,EACrE;AACA,QAAM,SAAS,MAAM,KAAK;AAC1B,SAAO,KAAK,YAAY,MAAM,IAAI,CAAC;AACrC;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,sCAAsC;AAC7E,gBAAY,QAAQ,OAAO,IAAI;AAC/B,SAAK;AAAA,EACP,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;",
+  "names": []
+}

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": []
+}