@ai-ecoverse/pi-audio 1.18.1 → 1.20.0

package/README.md

@@ -23,6 +23,10 @@ pi-audio does not apply that patch. Hosts such as slicc-agent record monkeypatch
 
 `@ai-ecoverse/pi-audio/durable` appends a `voice.truncate` entry for a `VoiceCut`. The entry has no model messages. Its `edits` replace the target assistant message with the text heard so far and keep tool calls. The stored entry stays intact, and pi-durable derives the shorter message for the next model request. `@earendil-works/pi-durable` is an optional peer (`^1.1.0`); the root package does not import it. `VoiceCut.entryId` is the decimal string of pi-durable's numeric entry id.
 
+`@ai-ecoverse/pi-audio/google` registers Gemini speech adapters when `registerGoogle()` is called. `gemini-3.8-flash-tts` and `gemini-3.8-flash-lite-tts` post JSON to `models/{id}:generateContent` with `responseModalities: ["AUDIO"]` and a prebuilt `voiceName`. Unary audio is WAV at 24 kHz unless the format asks for `AUDIO_L16`, `AUDIO_MULAW`, or `AUDIO_ALAW`. `textStreaming` is false. Output audio is priced from 25 tokens per second: $0.0135 per minute for Flash and $0.009 per minute for Flash-Lite through December 31, 2026. Text input is $0.50 per million tokens and is not on `audioCost`. Auth is `x-goog-api-key`.
+
+`@ai-ecoverse/pi-audio/fish` registers Fish Audio speech adapters when `registerFish()` is called. `transcribe-1-pro` posts multipart field `audio` to `/asr` with header `model: transcribe-1-pro` ($0.36 per hour). `ignore_timestamps` is false, so word `segments` and `speaker_turns` come back. There is no realtime ASR. `s2.1-pro` posts JSON to `/tts` for a string and streams MessagePack `text` events on `wss://…/tts/live` for an async iterable (`textStreaming: true`). `s2.1-pro-free` is the same path at $0. Paid speech is $15 per million UTF-8 bytes, stored as $0.015 per thousand characters. Voices are caller `reference_id`s (`cloning: true`).
+
 `@ai-ecoverse/pi-audio/mistral` registers Mistral speech adapters when `registerMistral()` is called. `voxtral-mini-latest` posts multipart audio to `/audio/transcriptions` ($0.003 per minute, diarization on). `voxtral-mini-transcribe-realtime-2602` streams base64 PCM on `wss://…/audio/transcriptions/realtime` after `session.created` ($0.006 per minute). Realtime has no server endpointing, so `turnDetection` is `none` and the registry VAD wrapper applies. `voxtral-mini-tts-2603` posts JSON to `/audio/speech` and reads SSE `speech.audio.delta` float32 PCM at 24 kHz ($0.016 per thousand characters). Hosted voice ids such as `en_paul_neutral` differ from the open-weight presets. `textStreaming` is false.
 
 `@ai-ecoverse/pi-audio/elevenlabs` registers ElevenLabs speech adapters when `registerElevenLabs()` is called. `scribe_v2` posts multipart audio to `/speech-to-text`. `scribe_v2_realtime` streams base64 audio on `wss://…/speech-to-text/realtime` and commits turns with `turn_prediction` (`turnDetection: 'semantic'`). `eleven_flash_v2_5` and `eleven_multilingual_v2` post JSON to `/text-to-speech/{voice}/with-timestamps` for a string and stream text on `stream-input` for an async iterable, with character alignment. `eleven_v3` and `eleven_v4` stay on the HTTP path (`textStreaming: false`). Scribe prices are $0.22 per hour for batch and $0.39 per hour for realtime.

package/dist/fish/catalog.js

@@ -0,0 +1,78 @@
+import { registerVoiceModel } from "../models.js";
+import { registerSpeechApiProvider, registerTranscriptionApiProvider } from "../registry.js";
+import { speakFish } from "./speak.js";
+import { transcribeFish } from "./transcribe.js";
+const tokenCost = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };
+const fish = "https://api.fish.audio/v1";
+const container = (encoding, sampleRate) => ({
+  encoding,
+  sampleRate,
+  channels: 1
+});
+const outputFormats = [
+  container("pcm16", 44100),
+  container("wav", 44100),
+  container("mp3", 44100),
+  container("opus", 48e3)
+];
+function speechModel(id, perThousandChars) {
+  return {
+    id,
+    name: id,
+    api: "fish-tts",
+    provider: "fish",
+    baseUrl: fish,
+    type: "speech",
+    textStreaming: true,
+    outputFormats,
+    cloning: true,
+    cost: tokenCost,
+    audioCost: { perThousandChars }
+  };
+}
+function fishCatalog() {
+  const batch = {
+    id: "transcribe-1-pro",
+    name: "transcribe-1-pro",
+    api: "fish-asr",
+    provider: "fish",
+    baseUrl: fish,
+    type: "transcription",
+    streaming: false,
+    turnDetection: "none",
+    inputFormats: [
+      container("wav", 16e3),
+      container("mp3", 16e3),
+      container("flac", 16e3),
+      container("opus", 16e3)
+    ],
+    diarization: true,
+    cost: tokenCost,
+    audioCost: { perInputMinute: 0.36 / 60 }
+  };
+  return [batch, speechModel("s2.1-pro", 0.015), speechModel("s2.1-pro-free", 0)];
+}
+function registerFish() {
+  const models = fishCatalog();
+  const unregisterModels = models.map((model) => registerVoiceModel(model));
+  const unregisterTranscription = registerTranscriptionApiProvider({
+    api: "fish-asr",
+    transcribe: transcribeFish
+  });
+  const unregisterSpeech = registerSpeechApiProvider({
+    api: "fish-tts",
+    speak: speakFish
+  });
+  let released = false;
+  return () => {
+    if (released) return;
+    released = true;
+    for (const unregister of unregisterModels) unregister();
+    unregisterTranscription();
+    unregisterSpeech();
+  };
+}
+export {
+  registerFish
+};
+//# sourceMappingURL=catalog.js.map

package/dist/fish/catalog.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/fish/catalog.ts"],
+  "sourcesContent": ["import type { AudioFormat } from '../audio.ts';\nimport { registerVoiceModel, type SpeechModel, type TranscriptionModel } from '../models.ts';\nimport { registerSpeechApiProvider, registerTranscriptionApiProvider } from '../registry.ts';\nimport { speakFish } from './speak.ts';\nimport { transcribeFish } from './transcribe.ts';\n\nconst tokenCost = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };\nconst fish = 'https://api.fish.audio/v1';\n\nconst container = (\n  encoding: 'wav' | 'mp3' | 'flac' | 'opus' | 'pcm16',\n  sampleRate: number\n): AudioFormat => ({\n  encoding,\n  sampleRate,\n  channels: 1,\n});\n\nconst outputFormats = [\n  container('pcm16', 44100),\n  container('wav', 44100),\n  container('mp3', 44100),\n  container('opus', 48000),\n];\n\nfunction speechModel(id: string, perThousandChars: number): SpeechModel<string> {\n  return {\n    id,\n    name: id,\n    api: 'fish-tts',\n    provider: 'fish',\n    baseUrl: fish,\n    type: 'speech',\n    textStreaming: true,\n    outputFormats,\n    cloning: true,\n    cost: tokenCost,\n    audioCost: { perThousandChars },\n  };\n}\n\nfunction fishCatalog(): Array<TranscriptionModel<string> | SpeechModel<string>> {\n  const batch: TranscriptionModel<string> = {\n    id: 'transcribe-1-pro',\n    name: 'transcribe-1-pro',\n    api: 'fish-asr',\n    provider: 'fish',\n    baseUrl: fish,\n    type: 'transcription',\n    streaming: false,\n    turnDetection: 'none',\n    inputFormats: [\n      container('wav', 16000),\n      container('mp3', 16000),\n      container('flac', 16000),\n      container('opus', 16000),\n    ],\n    diarization: true,\n    cost: tokenCost,\n    audioCost: { perInputMinute: 0.36 / 60 },\n  };\n  return [batch, speechModel('s2.1-pro', 0.015), speechModel('s2.1-pro-free', 0)];\n}\n\nexport function registerFish(): () => void {\n  const models = fishCatalog();\n  const unregisterModels = models.map((model) => registerVoiceModel(model));\n  const unregisterTranscription = registerTranscriptionApiProvider({\n    api: 'fish-asr',\n    transcribe: transcribeFish,\n  });\n  const unregisterSpeech = registerSpeechApiProvider({\n    api: 'fish-tts',\n    speak: speakFish,\n  });\n  let released = false;\n  return () => {\n    if (released) return;\n    released = true;\n    for (const unregister of unregisterModels) unregister();\n    unregisterTranscription();\n    unregisterSpeech();\n  };\n}\n"],
+  "mappings": "AACA,SAAS,0BAAqE;AAC9E,SAAS,2BAA2B,wCAAwC;AAC5E,SAAS,iBAAiB;AAC1B,SAAS,sBAAsB;AAE/B,MAAM,YAAY,EAAE,OAAO,GAAG,QAAQ,GAAG,WAAW,GAAG,YAAY,EAAE;AACrE,MAAM,OAAO;AAEb,MAAM,YAAY,CAChB,UACA,gBACiB;AAAA,EACjB;AAAA,EACA;AAAA,EACA,UAAU;AACZ;AAEA,MAAM,gBAAgB;AAAA,EACpB,UAAU,SAAS,KAAK;AAAA,EACxB,UAAU,OAAO,KAAK;AAAA,EACtB,UAAU,OAAO,KAAK;AAAA,EACtB,UAAU,QAAQ,IAAK;AACzB;AAEA,SAAS,YAAY,IAAY,kBAA+C;AAC9E,SAAO;AAAA,IACL;AAAA,IACA,MAAM;AAAA,IACN,KAAK;AAAA,IACL,UAAU;AAAA,IACV,SAAS;AAAA,IACT,MAAM;AAAA,IACN,eAAe;AAAA,IACf;AAAA,IACA,SAAS;AAAA,IACT,MAAM;AAAA,IACN,WAAW,EAAE,iBAAiB;AAAA,EAChC;AACF;AAEA,SAAS,cAAuE;AAC9E,QAAM,QAAoC;AAAA,IACxC,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,KAAK;AAAA,IACL,UAAU;AAAA,IACV,SAAS;AAAA,IACT,MAAM;AAAA,IACN,WAAW;AAAA,IACX,eAAe;AAAA,IACf,cAAc;AAAA,MACZ,UAAU,OAAO,IAAK;AAAA,MACtB,UAAU,OAAO,IAAK;AAAA,MACtB,UAAU,QAAQ,IAAK;AAAA,MACvB,UAAU,QAAQ,IAAK;AAAA,IACzB;AAAA,IACA,aAAa;AAAA,IACb,MAAM;AAAA,IACN,WAAW,EAAE,gBAAgB,OAAO,GAAG;AAAA,EACzC;AACA,SAAO,CAAC,OAAO,YAAY,YAAY,KAAK,GAAG,YAAY,iBAAiB,CAAC,CAAC;AAChF;AAEO,SAAS,eAA2B;AACzC,QAAM,SAAS,YAAY;AAC3B,QAAM,mBAAmB,OAAO,IAAI,CAAC,UAAU,mBAAmB,KAAK,CAAC;AACxE,QAAM,0BAA0B,iCAAiC;AAAA,IAC/D,KAAK;AAAA,IACL,YAAY;AAAA,EACd,CAAC;AACD,QAAM,mBAAmB,0BAA0B;AAAA,IACjD,KAAK;AAAA,IACL,OAAO;AAAA,EACT,CAAC;AACD,MAAI,WAAW;AACf,SAAO,MAAM;AACX,QAAI,SAAU;AACd,eAAW;AACX,eAAW,cAAc,iBAAkB,YAAW;AACtD,4BAAwB;AACxB,qBAAiB;AAAA,EACnB;AACF;",
+  "names": []
+}

package/dist/fish/index.js

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

package/dist/fish/index.js.map

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

package/dist/fish/speak.js

@@ -0,0 +1,269 @@
+import { chunkSeconds } from "../audio.js";
+import { VoiceRequestError } from "../errors.js";
+import {
+  applyPayload,
+  asVoiceError,
+  endpoint,
+  isRecord,
+  requestSignal,
+  stringField
+} from "../openai/wire.js";
+import { SpeechEventStream } from "../speech.js";
+import { priceAudio } from "../usage.js";
+import {
+  awaitOpen,
+  connectSocket,
+  fishHeaders,
+  messageBytes,
+  pack,
+  sameFormat,
+  socketUrl,
+  speechFormatName,
+  unpack
+} from "./wire.js";
+function languageCode(language) {
+  if (language === void 0) return void 0;
+  if (language.trim() === "") return void 0;
+  return language;
+}
+function resolveFormat(model, requested) {
+  const fallback = model.outputFormats[0];
+  if (!fallback) throw new Error("No output format");
+  if (!requested) return fallback;
+  const match = model.outputFormats.find((format) => sameFormat(format, requested));
+  if (!match) throw new Error(`Unsupported speech format ${requested.encoding}`);
+  return match;
+}
+function speechBody(text, voice, language, format) {
+  const body = {
+    text,
+    reference_id: voice,
+    format: speechFormatName(format),
+    sample_rate: format.sampleRate,
+    latency: "normal"
+  };
+  const code = languageCode(language);
+  if (code !== void 0) body.language = code;
+  return body;
+}
+function startFrame(voice, language, format) {
+  const request = {
+    text: "",
+    format: speechFormatName(format),
+    reference_id: voice,
+    sample_rate: format.sampleRate,
+    latency: "balanced"
+  };
+  const code = languageCode(language);
+  if (code !== void 0) request.language = code;
+  return { event: "start", request };
+}
+async function readText(text, signal) {
+  if (typeof text === "string") return text;
+  let combined = "";
+  for await (const part of text) {
+    if (signal?.aborted) throw new DOMException("aborted", "AbortError");
+    combined += part;
+  }
+  return combined;
+}
+function errorText(text, status) {
+  if (text === "") return `HTTP ${status}`;
+  let parsed;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    return text;
+  }
+  if (!isRecord(parsed) || Array.isArray(parsed)) return text;
+  const message = stringField(parsed, "message");
+  if (message === void 0) return text;
+  return message;
+}
+async function fishFetch(model, options, url, body) {
+  const headers = fishHeaders(model, options);
+  headers["Content-Type"] = "application/json";
+  const fetchImpl = options?.fetch ?? globalThis.fetch;
+  const response = await fetchImpl(url, {
+    method: "POST",
+    headers,
+    body,
+    signal: requestSignal(options)
+  });
+  await options?.onResponse?.({ status: response.status, headers: {} }, model);
+  if (!response.ok) {
+    const text = await response.text();
+    throw new Error(errorText(text, response.status));
+  }
+  return response;
+}
+function sendFrame(socket, value) {
+  const packed = pack(value);
+  const copy = new ArrayBuffer(packed.byteLength);
+  new Uint8Array(copy).set(packed);
+  socket.send(copy);
+}
+function speakFish(model, input, options) {
+  const stream = new SpeechEventStream();
+  void pumpSpeech(model, input, options, stream);
+  return stream;
+}
+async function pumpSpeech(model, input, options, stream) {
+  try {
+    if (options?.signal?.aborted) throw new DOMException("aborted", "AbortError");
+    const format = resolveFormat(model, input.format);
+    if (model.textStreaming && typeof input.text !== "string") {
+      await streamSpeech(model, input.text, input.voice, input.language, format, options, stream);
+      return;
+    }
+    const text = await readText(input.text, options?.signal);
+    await postSpeech(model, text, input.voice, input.language, format, options, stream);
+  } catch (error) {
+    const voice = asVoiceError(error);
+    stream.push({ type: "error", reason: voice.reason, error: voice.error });
+  }
+}
+async function postSpeech(model, text, voice, language, format, options, stream) {
+  const payload = speechBody(text, voice, language, format);
+  const body = await applyPayload(payload, model, options);
+  if (!isRecord(body) || Array.isArray(body)) throw new Error("Speech payload was not JSON");
+  const response = await fishFetch(
+    model,
+    options,
+    endpoint(model.baseUrl, "tts"),
+    JSON.stringify(body)
+  );
+  const data = new Uint8Array(await response.arrayBuffer());
+  if (data.byteLength === 0) throw new Error("Speech response had no audio");
+  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,
+    heard: 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 = (kind) => {
+    if (state.failed) doneWait.resolve();
+    if (kind === "finish" || kind === "close") doneWait.resolve();
+  };
+  const url = socketUrl(model.baseUrl, "tts/live");
+  const socket = connectSocket(url.toString(), fishHeaders(model, options));
+  let incoming = Promise.resolve();
+  socket.addEventListener("message", (event) => {
+    incoming = incoming.then(() => onMessage(event.data, model.id, format, state, push, wake));
+  });
+  socket.addEventListener("close", () => {
+    if (state.opened && !state.done && !state.failed) {
+      state.failed = true;
+      push({ type: "error", reason: "error", error: "Fish speech socket closed" });
+    }
+    wake("close");
+  });
+  try {
+    await awaitOpen(socket, "Fish speech", (message) => {
+      state.failed = true;
+      push({ type: "error", reason: "error", error: message });
+      wake("close");
+    });
+    if (state.failed) return;
+    state.opened = true;
+    sendFrame(socket, startFrame(voice, language, format));
+    state.characters = await writeText(socket, text, options?.signal);
+    if (state.failed) return;
+    sendFrame(socket, { event: "flush" });
+    sendFrame(socket, { event: "stop" });
+    if (!state.done) await doneWait.promise;
+    if (state.failed) return;
+    if (!state.heard) throw new Error("Speech response had no audio");
+    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 === "") continue;
+    sendFrame(socket, { event: "text", text: part });
+  }
+  return combined.length;
+}
+async function onMessage(data, itemId, format, state, push, wake) {
+  try {
+    const parsed = unpack(await messageBytes(data));
+    if (!isRecord(parsed) || Array.isArray(parsed) || parsed instanceof Uint8Array) {
+      throw new Error("Fish speech event was not a map");
+    }
+    const kind = stringField(parsed, "event");
+    applySpeech(parsed, kind, itemId, format, state, push);
+    wake(kind);
+  } catch (error) {
+    const voice = asVoiceError(error);
+    state.failed = true;
+    push({ type: "error", reason: voice.reason, error: voice.error });
+    wake("close");
+  }
+}
+function applySpeech(event, kind, itemId, format, state, push) {
+  if (kind === "error") {
+    const message = stringField(event, "message");
+    state.failed = true;
+    push({
+      type: "error",
+      reason: "error",
+      error: message === void 0 ? "Fish speech failed" : message
+    });
+    return;
+  }
+  if (kind === "finish") {
+    state.done = true;
+    return;
+  }
+  if (kind !== "audio") return;
+  const audio = event.audio;
+  if (!(audio instanceof Uint8Array) || audio.byteLength === 0) return;
+  if (!state.started) {
+    state.started = true;
+    push({ type: "start", format });
+  }
+  state.heard = true;
+  state.outputSeconds += chunkSeconds({ data: audio, format, timestamp: 0 });
+  push({
+    type: "audio",
+    chunk: { data: audio, format, timestamp: 0 },
+    span: { itemId, start: 0, end: state.characters }
+  });
+}
+export {
+  speakFish
+};
+//# sourceMappingURL=speak.js.map

package/dist/fish/speak.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/fish/speak.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport type { AudioFormat } from '../audio.ts';\nimport { chunkSeconds } from '../audio.ts';\nimport { VoiceRequestError } from '../errors.ts';\nimport type { SpeechModel } from '../models.ts';\nimport {\n  applyPayload,\n  asVoiceError,\n  endpoint,\n  isRecord,\n  requestSignal,\n  stringField,\n} from '../openai/wire.ts';\nimport { type SpeechEvent, SpeechEventStream, type SpeechInput } from '../speech.ts';\nimport { type AudioCounters, priceAudio } from '../usage.ts';\nimport {\n  awaitOpen,\n  connectSocket,\n  fishHeaders,\n  messageBytes,\n  type PackValue,\n  pack,\n  sameFormat,\n  socketUrl,\n  speechFormatName,\n  unpack,\n} from './wire.ts';\n\ninterface SpeechState {\n  opened: boolean;\n  started: boolean;\n  done: boolean;\n  failed: boolean;\n  heard: boolean;\n  outputSeconds: number;\n  characters: number;\n}\n\nfunction languageCode(language: string | undefined): string | undefined {\n  if (language === undefined) return undefined;\n  if (language.trim() === '') return undefined;\n  return language;\n}\n\nfunction resolveFormat(\n  model: SpeechModel<string>,\n  requested: AudioFormat | undefined\n): AudioFormat {\n  const fallback = model.outputFormats[0];\n  if (!fallback) throw new Error('No output format');\n  if (!requested) return fallback;\n  const match = model.outputFormats.find((format) => sameFormat(format, requested));\n  if (!match) throw new Error(`Unsupported speech format ${requested.encoding}`);\n  return match;\n}\n\nfunction speechBody(\n  text: string,\n  voice: string,\n  language: string | undefined,\n  format: AudioFormat\n): Record<string, unknown> {\n  const body: Record<string, unknown> = {\n    text,\n    reference_id: voice,\n    format: speechFormatName(format),\n    sample_rate: format.sampleRate,\n    latency: 'normal',\n  };\n  const code = languageCode(language);\n  if (code !== undefined) body.language = code;\n  return body;\n}\n\nfunction startFrame(voice: string, language: string | undefined, format: AudioFormat): PackValue {\n  const request: { [key: string]: PackValue } = {\n    text: '',\n    format: speechFormatName(format),\n    reference_id: voice,\n    sample_rate: format.sampleRate,\n    latency: 'balanced',\n  };\n  const code = languageCode(language);\n  if (code !== undefined) request.language = code;\n  return { event: 'start', request };\n}\n\nasync function readText(\n  text: string | AsyncIterable<string>,\n  signal: AbortSignal | undefined\n): Promise<string> {\n  if (typeof text === 'string') return text;\n  let combined = '';\n  for await (const part of text) {\n    if (signal?.aborted) throw new DOMException('aborted', 'AbortError');\n    combined += part;\n  }\n  return combined;\n}\n\nfunction errorText(text: string, status: number): string {\n  if (text === '') return `HTTP ${status}`;\n  let parsed: unknown;\n  try {\n    parsed = JSON.parse(text);\n  } catch {\n    return text;\n  }\n  if (!isRecord(parsed) || Array.isArray(parsed)) return text;\n  const message = stringField(parsed, 'message');\n  if (message === undefined) return text;\n  return message;\n}\n\nasync function fishFetch(\n  model: SpeechModel<string>,\n  options: ProviderRequestOptions | undefined,\n  url: string,\n  body: string\n): Promise<Response> {\n  const headers = fishHeaders(model, options);\n  headers['Content-Type'] = 'application/json';\n  const fetchImpl = options?.fetch ?? globalThis.fetch;\n  const response = await fetchImpl(url, {\n    method: 'POST',\n    headers,\n    body,\n    signal: requestSignal(options),\n  });\n  await options?.onResponse?.({ status: response.status, headers: {} }, model as never);\n  if (!response.ok) {\n    const text = await response.text();\n    throw new Error(errorText(text, response.status));\n  }\n  return response;\n}\n\nfunction sendFrame(socket: WebSocket, value: PackValue): void {\n  const packed = pack(value);\n  const copy = new ArrayBuffer(packed.byteLength);\n  new Uint8Array(copy).set(packed);\n  socket.send(copy);\n}\n\nexport function speakFish(\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options?: ProviderRequestOptions\n): SpeechEventStream {\n  const stream = new SpeechEventStream();\n  void pumpSpeech(model, input, options, stream);\n  return stream;\n}\n\nasync function pumpSpeech(\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options: ProviderRequestOptions | undefined,\n  stream: SpeechEventStream\n) {\n  try {\n    if (options?.signal?.aborted) throw new DOMException('aborted', 'AbortError');\n    const format = resolveFormat(model, input.format);\n    if (model.textStreaming && typeof input.text !== 'string') {\n      await streamSpeech(model, input.text, input.voice, input.language, format, options, stream);\n      return;\n    }\n    const text = await readText(input.text, options?.signal);\n    await postSpeech(model, text, input.voice, input.language, format, options, stream);\n  } catch (error) {\n    const voice = asVoiceError(error);\n    stream.push({ type: 'error', reason: voice.reason, error: voice.error });\n  }\n}\n\nasync function postSpeech(\n  model: SpeechModel<string>,\n  text: string,\n  voice: string,\n  language: string | undefined,\n  format: AudioFormat,\n  options: ProviderRequestOptions | undefined,\n  stream: SpeechEventStream\n) {\n  const payload = speechBody(text, voice, language, format);\n  const body = await applyPayload(payload, model, options);\n  if (!isRecord(body) || Array.isArray(body)) throw new Error('Speech payload was not JSON');\n  const response = await fishFetch(\n    model,\n    options,\n    endpoint(model.baseUrl, 'tts'),\n    JSON.stringify(body)\n  );\n  const data = new Uint8Array(await response.arrayBuffer());\n  if (data.byteLength === 0) throw new Error('Speech response had no audio');\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 | undefined,\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    heard: 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 = (kind: string | undefined) => {\n    if (state.failed) doneWait.resolve();\n    if (kind === 'finish' || kind === 'close') doneWait.resolve();\n  };\n  const url = socketUrl(model.baseUrl, 'tts/live');\n  const socket = connectSocket(url.toString(), fishHeaders(model, options));\n  let incoming = Promise.resolve();\n  socket.addEventListener('message', (event) => {\n    incoming = incoming.then(() => onMessage(event.data, model.id, format, state, push, wake));\n  });\n  socket.addEventListener('close', () => {\n    if (state.opened && !state.done && !state.failed) {\n      state.failed = true;\n      push({ type: 'error', reason: 'error', error: 'Fish speech socket closed' });\n    }\n    wake('close');\n  });\n  try {\n    await awaitOpen(socket, 'Fish speech', (message) => {\n      state.failed = true;\n      push({ type: 'error', reason: 'error', error: message });\n      wake('close');\n    });\n    if (state.failed) return;\n    state.opened = true;\n    sendFrame(socket, startFrame(voice, language, format));\n    state.characters = await writeText(socket, text, options?.signal);\n    if (state.failed) return;\n    sendFrame(socket, { event: 'flush' });\n    sendFrame(socket, { event: 'stop' });\n    if (!state.done) await doneWait.promise;\n    if (state.failed) return;\n    if (!state.heard) throw new Error('Speech response had no audio');\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 === '') continue;\n    sendFrame(socket, { event: 'text', text: 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: (kind: string | undefined) => void\n) {\n  try {\n    const parsed = unpack(await messageBytes(data));\n    if (!isRecord(parsed) || Array.isArray(parsed) || parsed instanceof Uint8Array) {\n      throw new Error('Fish speech event was not a map');\n    }\n    const kind = stringField(parsed, 'event');\n    applySpeech(parsed, kind, itemId, format, state, push);\n    wake(kind);\n  } catch (error) {\n    const voice = asVoiceError(error);\n    state.failed = true;\n    push({ type: 'error', reason: voice.reason, error: voice.error });\n    wake('close');\n  }\n}\n\nfunction applySpeech(\n  event: Record<string, unknown>,\n  kind: string | undefined,\n  itemId: string,\n  format: AudioFormat,\n  state: SpeechState,\n  push: (event: SpeechEvent) => void\n) {\n  if (kind === 'error') {\n    const message = stringField(event, 'message');\n    state.failed = true;\n    push({\n      type: 'error',\n      reason: 'error',\n      error: message === undefined ? 'Fish speech failed' : message,\n    });\n    return;\n  }\n  if (kind === 'finish') {\n    state.done = true;\n    return;\n  }\n  if (kind !== 'audio') return;\n  const audio = event.audio;\n  if (!(audio instanceof Uint8Array) || audio.byteLength === 0) return;\n  if (!state.started) {\n    state.started = true;\n    push({ type: 'start', format });\n  }\n  state.heard = true;\n  state.outputSeconds += chunkSeconds({ data: audio, format, timestamp: 0 });\n  push({\n    type: 'audio',\n    chunk: { data: audio, format, timestamp: 0 },\n    span: { itemId, start: 0, end: state.characters },\n  });\n}\n"],
+  "mappings": "AAEA,SAAS,oBAAoB;AAC7B,SAAS,yBAAyB;AAElC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAA2B,yBAA2C;AACtE,SAA6B,kBAAkB;AAC/C;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAYP,SAAS,aAAa,UAAkD;AACtE,MAAI,aAAa,OAAW,QAAO;AACnC,MAAI,SAAS,KAAK,MAAM,GAAI,QAAO;AACnC,SAAO;AACT;AAEA,SAAS,cACP,OACA,WACa;AACb,QAAM,WAAW,MAAM,cAAc,CAAC;AACtC,MAAI,CAAC,SAAU,OAAM,IAAI,MAAM,kBAAkB;AACjD,MAAI,CAAC,UAAW,QAAO;AACvB,QAAM,QAAQ,MAAM,cAAc,KAAK,CAAC,WAAW,WAAW,QAAQ,SAAS,CAAC;AAChF,MAAI,CAAC,MAAO,OAAM,IAAI,MAAM,6BAA6B,UAAU,QAAQ,EAAE;AAC7E,SAAO;AACT;AAEA,SAAS,WACP,MACA,OACA,UACA,QACyB;AACzB,QAAM,OAAgC;AAAA,IACpC;AAAA,IACA,cAAc;AAAA,IACd,QAAQ,iBAAiB,MAAM;AAAA,IAC/B,aAAa,OAAO;AAAA,IACpB,SAAS;AAAA,EACX;AACA,QAAM,OAAO,aAAa,QAAQ;AAClC,MAAI,SAAS,OAAW,MAAK,WAAW;AACxC,SAAO;AACT;AAEA,SAAS,WAAW,OAAe,UAA8B,QAAgC;AAC/F,QAAM,UAAwC;AAAA,IAC5C,MAAM;AAAA,IACN,QAAQ,iBAAiB,MAAM;AAAA,IAC/B,cAAc;AAAA,IACd,aAAa,OAAO;AAAA,IACpB,SAAS;AAAA,EACX;AACA,QAAM,OAAO,aAAa,QAAQ;AAClC,MAAI,SAAS,OAAW,SAAQ,WAAW;AAC3C,SAAO,EAAE,OAAO,SAAS,QAAQ;AACnC;AAEA,eAAe,SACb,MACA,QACiB;AACjB,MAAI,OAAO,SAAS,SAAU,QAAO;AACrC,MAAI,WAAW;AACf,mBAAiB,QAAQ,MAAM;AAC7B,QAAI,QAAQ,QAAS,OAAM,IAAI,aAAa,WAAW,YAAY;AACnE,gBAAY;AAAA,EACd;AACA,SAAO;AACT;AAEA,SAAS,UAAU,MAAc,QAAwB;AACvD,MAAI,SAAS,GAAI,QAAO,QAAQ,MAAM;AACtC,MAAI;AACJ,MAAI;AACF,aAAS,KAAK,MAAM,IAAI;AAAA,EAC1B,QAAQ;AACN,WAAO;AAAA,EACT;AACA,MAAI,CAAC,SAAS,MAAM,KAAK,MAAM,QAAQ,MAAM,EAAG,QAAO;AACvD,QAAM,UAAU,YAAY,QAAQ,SAAS;AAC7C,MAAI,YAAY,OAAW,QAAO;AAClC,SAAO;AACT;AAEA,eAAe,UACb,OACA,SACA,KACA,MACmB;AACnB,QAAM,UAAU,YAAY,OAAO,OAAO;AAC1C,UAAQ,cAAc,IAAI;AAC1B,QAAM,YAAY,SAAS,SAAS,WAAW;AAC/C,QAAM,WAAW,MAAM,UAAU,KAAK;AAAA,IACpC,QAAQ;AAAA,IACR;AAAA,IACA;AAAA,IACA,QAAQ,cAAc,OAAO;AAAA,EAC/B,CAAC;AACD,QAAM,SAAS,aAAa,EAAE,QAAQ,SAAS,QAAQ,SAAS,CAAC,EAAE,GAAG,KAAc;AACpF,MAAI,CAAC,SAAS,IAAI;AAChB,UAAM,OAAO,MAAM,SAAS,KAAK;AACjC,UAAM,IAAI,MAAM,UAAU,MAAM,SAAS,MAAM,CAAC;AAAA,EAClD;AACA,SAAO;AACT;AAEA,SAAS,UAAU,QAAmB,OAAwB;AAC5D,QAAM,SAAS,KAAK,KAAK;AACzB,QAAM,OAAO,IAAI,YAAY,OAAO,UAAU;AAC9C,MAAI,WAAW,IAAI,EAAE,IAAI,MAAM;AAC/B,SAAO,KAAK,IAAI;AAClB;AAEO,SAAS,UACd,OACA,OACA,SACmB;AACnB,QAAM,SAAS,IAAI,kBAAkB;AACrC,OAAK,WAAW,OAAO,OAAO,SAAS,MAAM;AAC7C,SAAO;AACT;AAEA,eAAe,WACb,OACA,OACA,SACA,QACA;AACA,MAAI;AACF,QAAI,SAAS,QAAQ,QAAS,OAAM,IAAI,aAAa,WAAW,YAAY;AAC5E,UAAM,SAAS,cAAc,OAAO,MAAM,MAAM;AAChD,QAAI,MAAM,iBAAiB,OAAO,MAAM,SAAS,UAAU;AACzD,YAAM,aAAa,OAAO,MAAM,MAAM,MAAM,OAAO,MAAM,UAAU,QAAQ,SAAS,MAAM;AAC1F;AAAA,IACF;AACA,UAAM,OAAO,MAAM,SAAS,MAAM,MAAM,SAAS,MAAM;AACvD,UAAM,WAAW,OAAO,MAAM,MAAM,OAAO,MAAM,UAAU,QAAQ,SAAS,MAAM;AAAA,EACpF,SAAS,OAAO;AACd,UAAM,QAAQ,aAAa,KAAK;AAChC,WAAO,KAAK,EAAE,MAAM,SAAS,QAAQ,MAAM,QAAQ,OAAO,MAAM,MAAM,CAAC;AAAA,EACzE;AACF;AAEA,eAAe,WACb,OACA,MACA,OACA,UACA,QACA,SACA,QACA;AACA,QAAM,UAAU,WAAW,MAAM,OAAO,UAAU,MAAM;AACxD,QAAM,OAAO,MAAM,aAAa,SAAS,OAAO,OAAO;AACvD,MAAI,CAAC,SAAS,IAAI,KAAK,MAAM,QAAQ,IAAI,EAAG,OAAM,IAAI,MAAM,6BAA6B;AACzF,QAAM,WAAW,MAAM;AAAA,IACrB;AAAA,IACA;AAAA,IACA,SAAS,MAAM,SAAS,KAAK;AAAA,IAC7B,KAAK,UAAU,IAAI;AAAA,EACrB;AACA,QAAM,OAAO,IAAI,WAAW,MAAM,SAAS,YAAY,CAAC;AACxD,MAAI,KAAK,eAAe,EAAG,OAAM,IAAI,MAAM,8BAA8B;AACzE,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,OAAO;AAAA,IACP,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,CAAC,SAA6B;AACzC,QAAI,MAAM,OAAQ,UAAS,QAAQ;AACnC,QAAI,SAAS,YAAY,SAAS,QAAS,UAAS,QAAQ;AAAA,EAC9D;AACA,QAAM,MAAM,UAAU,MAAM,SAAS,UAAU;AAC/C,QAAM,SAAS,cAAc,IAAI,SAAS,GAAG,YAAY,OAAO,OAAO,CAAC;AACxE,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,4BAA4B,CAAC;AAAA,IAC7E;AACA,SAAK,OAAO;AAAA,EACd,CAAC;AACD,MAAI;AACF,UAAM,UAAU,QAAQ,eAAe,CAAC,YAAY;AAClD,YAAM,SAAS;AACf,WAAK,EAAE,MAAM,SAAS,QAAQ,SAAS,OAAO,QAAQ,CAAC;AACvD,WAAK,OAAO;AAAA,IACd,CAAC;AACD,QAAI,MAAM,OAAQ;AAClB,UAAM,SAAS;AACf,cAAU,QAAQ,WAAW,OAAO,UAAU,MAAM,CAAC;AACrD,UAAM,aAAa,MAAM,UAAU,QAAQ,MAAM,SAAS,MAAM;AAChE,QAAI,MAAM,OAAQ;AAClB,cAAU,QAAQ,EAAE,OAAO,QAAQ,CAAC;AACpC,cAAU,QAAQ,EAAE,OAAO,OAAO,CAAC;AACnC,QAAI,CAAC,MAAM,KAAM,OAAM,SAAS;AAChC,QAAI,MAAM,OAAQ;AAClB,QAAI,CAAC,MAAM,MAAO,OAAM,IAAI,MAAM,8BAA8B;AAChE,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;AACjB,cAAU,QAAQ,EAAE,OAAO,QAAQ,MAAM,KAAK,CAAC;AAAA,EACjD;AACA,SAAO,SAAS;AAClB;AAEA,eAAe,UACb,MACA,QACA,QACA,OACA,MACA,MACA;AACA,MAAI;AACF,UAAM,SAAS,OAAO,MAAM,aAAa,IAAI,CAAC;AAC9C,QAAI,CAAC,SAAS,MAAM,KAAK,MAAM,QAAQ,MAAM,KAAK,kBAAkB,YAAY;AAC9E,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AACA,UAAM,OAAO,YAAY,QAAQ,OAAO;AACxC,gBAAY,QAAQ,MAAM,QAAQ,QAAQ,OAAO,IAAI;AACrD,SAAK,IAAI;AAAA,EACX,SAAS,OAAO;AACd,UAAM,QAAQ,aAAa,KAAK;AAChC,UAAM,SAAS;AACf,SAAK,EAAE,MAAM,SAAS,QAAQ,MAAM,QAAQ,OAAO,MAAM,MAAM,CAAC;AAChE,SAAK,OAAO;AAAA,EACd;AACF;AAEA,SAAS,YACP,OACA,MACA,QACA,QACA,OACA,MACA;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,uBAAuB;AAAA,IACxD,CAAC;AACD;AAAA,EACF;AACA,MAAI,SAAS,UAAU;AACrB,UAAM,OAAO;AACb;AAAA,EACF;AACA,MAAI,SAAS,QAAS;AACtB,QAAM,QAAQ,MAAM;AACpB,MAAI,EAAE,iBAAiB,eAAe,MAAM,eAAe,EAAG;AAC9D,MAAI,CAAC,MAAM,SAAS;AAClB,UAAM,UAAU;AAChB,SAAK,EAAE,MAAM,SAAS,OAAO,CAAC;AAAA,EAChC;AACA,QAAM,QAAQ;AACd,QAAM,iBAAiB,aAAa,EAAE,MAAM,OAAO,QAAQ,WAAW,EAAE,CAAC;AACzE,OAAK;AAAA,IACH,MAAM;AAAA,IACN,OAAO,EAAE,MAAM,OAAO,QAAQ,WAAW,EAAE;AAAA,IAC3C,MAAM,EAAE,QAAQ,OAAO,GAAG,KAAK,MAAM,WAAW;AAAA,EAClD,CAAC;AACH;",
+  "names": ["voice"]
+}

package/dist/fish/transcribe.js

@@ -0,0 +1,158 @@
+import { VoiceRequestError } from "../errors.js";
+import {
+  applyPayload,
+  asVoiceError,
+  endpoint,
+  isRecord,
+  numberField,
+  readAudio,
+  readJson,
+  requestSignal,
+  stringField
+} from "../openai/wire.js";
+import { priceAudio } from "../usage.js";
+import { audioFilename, encodeForm, fishHeaders, formType } from "./wire.js";
+function textOf(record, key) {
+  const text = stringField(record, key);
+  if (text === void 0) return "";
+  return text;
+}
+function secondsOf(record, key) {
+  const value = numberField(record, key);
+  if (value === void 0) return 0;
+  return value;
+}
+function durationOf(body) {
+  const duration = numberField(body, "duration");
+  if (duration === void 0) return 0;
+  return duration;
+}
+function languageOf(body) {
+  const code = stringField(body, "language_code");
+  if (code !== void 0) return code;
+  return stringField(body, "language");
+}
+function segmentFrom(word, withSpeaker) {
+  const text = textOf(word, "text");
+  if (text === "") return void 0;
+  const segment = {
+    text,
+    start: secondsOf(word, "start"),
+    end: secondsOf(word, "end")
+  };
+  if (!withSpeaker) return segment;
+  const speaker = stringField(word, "speaker");
+  if (speaker !== void 0) segment.speaker = speaker;
+  return segment;
+}
+function segmentsFrom(value, withSpeaker) {
+  if (!Array.isArray(value)) return [];
+  const segments = [];
+  for (const word of value) {
+    if (!isRecord(word) || Array.isArray(word)) continue;
+    const segment = segmentFrom(word, withSpeaker);
+    if (segment) segments.push(segment);
+  }
+  return segments;
+}
+function timedFrom(body) {
+  const turns = segmentsFrom(body.speaker_turns, true);
+  if (turns.length > 0) return turns;
+  return segmentsFrom(body.segments, false);
+}
+function transcriptFrom(body, model) {
+  if (!isRecord(body) || Array.isArray(body)) throw new Error("Transcription response had no text");
+  const direct = stringField(body, "text");
+  if (direct === void 0) throw new Error("Transcription response had no text");
+  const timed = timedFrom(body);
+  const language = languageOf(body);
+  return {
+    text: direct,
+    ...language !== void 0 ? { language } : {},
+    ...timed.length > 0 ? { segments: timed } : {},
+    usage: priceAudio(model.audioCost, { inputSeconds: durationOf(body) })
+  };
+}
+function fileOf(body) {
+  const file = body.audio;
+  if (!isRecord(file) || Array.isArray(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 === "audio") continue;
+    if (typeof value !== "string") continue;
+    fields[key] = value;
+  }
+  return fields;
+}
+function errorText(text, status) {
+  if (text === "") return `HTTP ${status}`;
+  let parsed;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    return text;
+  }
+  if (!isRecord(parsed) || Array.isArray(parsed)) return text;
+  const message = stringField(parsed, "message");
+  if (message === void 0) return text;
+  return message;
+}
+async function fishFetch(model, options, url, body, contentType) {
+  const headers = fishHeaders(model, options);
+  headers["Content-Type"] = contentType;
+  const fetchImpl = options?.fetch ?? globalThis.fetch;
+  const response = await fetchImpl(url, {
+    method: "POST",
+    headers,
+    body,
+    signal: requestSignal(options)
+  });
+  await options?.onResponse?.({ status: response.status, headers: {} }, model);
+  if (!response.ok) {
+    const text = await response.text();
+    throw new Error(errorText(text, response.status));
+  }
+  return response;
+}
+async function transcribeFish(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 payload = {
+      ignore_timestamps: "false",
+      diarize: model.diarization ? "true" : "false",
+      audio: {
+        data: loaded.data,
+        mimeType: loaded.mimeType,
+        filename: audioFilename(loaded.mimeType)
+      }
+    };
+    const body = await applyPayload(payload, model, options);
+    if (!isRecord(body) || Array.isArray(body))
+      throw new Error("Transcription payload was not a file");
+    const file = fileOf(body);
+    const encoded = encodeForm(file, fieldStrings(body));
+    const response = await fishFetch(
+      model,
+      options,
+      endpoint(model.baseUrl, "asr"),
+      encoded,
+      formType()
+    );
+    return transcriptFrom(await readJson(response), model);
+  } catch (error) {
+    const voice = asVoiceError(error);
+    throw new VoiceRequestError(voice.reason, voice.error);
+  }
+}
+export {
+  transcribeFish
+};
+//# sourceMappingURL=transcribe.js.map

package/dist/fish/transcribe.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/fish/transcribe.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport { VoiceRequestError } from '../errors.ts';\nimport type { TranscriptionModel } from '../models.ts';\nimport {\n  applyPayload,\n  asVoiceError,\n  endpoint,\n  isRecord,\n  numberField,\n  readAudio,\n  readJson,\n  requestSignal,\n  stringField,\n} from '../openai/wire.ts';\nimport type { AudioInput, Transcript, TranscriptSegment } from '../transcription.ts';\nimport { priceAudio } from '../usage.ts';\nimport { audioFilename, encodeForm, type FishFile, fishHeaders, formType } from './wire.ts';\n\nfunction textOf(record: Record<string, unknown>, key: string): string {\n  const text = stringField(record, key);\n  if (text === undefined) return '';\n  return text;\n}\n\nfunction secondsOf(record: Record<string, unknown>, key: string): number {\n  const value = numberField(record, key);\n  if (value === undefined) return 0;\n  return value;\n}\n\nfunction durationOf(body: Record<string, unknown>): number {\n  const duration = numberField(body, 'duration');\n  if (duration === undefined) return 0;\n  return duration;\n}\n\nfunction languageOf(body: Record<string, unknown>): string | undefined {\n  const code = stringField(body, 'language_code');\n  if (code !== undefined) return code;\n  return stringField(body, 'language');\n}\n\nfunction segmentFrom(\n  word: Record<string, unknown>,\n  withSpeaker: boolean\n): TranscriptSegment | undefined {\n  const text = textOf(word, 'text');\n  if (text === '') return undefined;\n  const segment: TranscriptSegment = {\n    text,\n    start: secondsOf(word, 'start'),\n    end: secondsOf(word, 'end'),\n  };\n  if (!withSpeaker) return segment;\n  const speaker = stringField(word, 'speaker');\n  if (speaker !== undefined) segment.speaker = speaker;\n  return segment;\n}\n\nfunction segmentsFrom(value: unknown, withSpeaker: boolean): TranscriptSegment[] {\n  if (!Array.isArray(value)) return [];\n  const segments: TranscriptSegment[] = [];\n  for (const word of value) {\n    if (!isRecord(word) || Array.isArray(word)) continue;\n    const segment = segmentFrom(word, withSpeaker);\n    if (segment) segments.push(segment);\n  }\n  return segments;\n}\n\nfunction timedFrom(body: Record<string, unknown>): TranscriptSegment[] {\n  const turns = segmentsFrom(body.speaker_turns, true);\n  if (turns.length > 0) return turns;\n  return segmentsFrom(body.segments, false);\n}\n\nfunction transcriptFrom(body: unknown, model: TranscriptionModel<string>): Transcript {\n  if (!isRecord(body) || Array.isArray(body)) throw new Error('Transcription response had no text');\n  const direct = stringField(body, 'text');\n  if (direct === undefined) throw new Error('Transcription response had no text');\n  const timed = timedFrom(body);\n  const language = languageOf(body);\n  return {\n    text: direct,\n    ...(language !== undefined ? { language } : {}),\n    ...(timed.length > 0 ? { segments: timed } : {}),\n    usage: priceAudio(model.audioCost, { inputSeconds: durationOf(body) }),\n  };\n}\n\nfunction fileOf(body: Record<string, unknown>): FishFile {\n  const file = body.audio;\n  if (\n    !isRecord(file) ||\n    Array.isArray(file) ||\n    !(file.data instanceof Uint8Array) ||\n    typeof file.mimeType !== 'string'\n  ) {\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 === 'audio') continue;\n    if (typeof value !== 'string') continue;\n    fields[key] = value;\n  }\n  return fields;\n}\n\nfunction errorText(text: string, status: number): string {\n  if (text === '') return `HTTP ${status}`;\n  let parsed: unknown;\n  try {\n    parsed = JSON.parse(text);\n  } catch {\n    return text;\n  }\n  if (!isRecord(parsed) || Array.isArray(parsed)) return text;\n  const message = stringField(parsed, 'message');\n  if (message === undefined) return text;\n  return message;\n}\n\nasync function fishFetch(\n  model: TranscriptionModel<string>,\n  options: ProviderRequestOptions | undefined,\n  url: string,\n  body: BodyInit,\n  contentType: string\n): Promise<Response> {\n  const headers = fishHeaders(model, options);\n  headers['Content-Type'] = contentType;\n  const fetchImpl = options?.fetch ?? globalThis.fetch;\n  const response = await fetchImpl(url, {\n    method: 'POST',\n    headers,\n    body,\n    signal: requestSignal(options),\n  });\n  await options?.onResponse?.({ status: response.status, headers: {} }, model as never);\n  if (!response.ok) {\n    const text = await response.text();\n    throw new Error(errorText(text, response.status));\n  }\n  return response;\n}\n\nexport async function transcribeFish(\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 payload: Record<string, unknown> = {\n      ignore_timestamps: 'false',\n      diarize: model.diarization ? 'true' : 'false',\n      audio: {\n        data: loaded.data,\n        mimeType: loaded.mimeType,\n        filename: audioFilename(loaded.mimeType),\n      },\n    };\n    const body = await applyPayload(payload, model, options);\n    if (!isRecord(body) || Array.isArray(body))\n      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 fishFetch(\n      model,\n      options,\n      endpoint(model.baseUrl, 'asr'),\n      encoded,\n      formType()\n    );\n    return transcriptFrom(await readJson(response), model);\n  } catch (error) {\n    const voice = asVoiceError(error);\n    throw new VoiceRequestError(voice.reason, voice.error);\n  }\n}\n"],
+  "mappings": "AACA,SAAS,yBAAyB;AAElC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,kBAAkB;AAC3B,SAAS,eAAe,YAA2B,aAAa,gBAAgB;AAEhF,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,WAAW,MAAuC;AACzD,QAAM,WAAW,YAAY,MAAM,UAAU;AAC7C,MAAI,aAAa,OAAW,QAAO;AACnC,SAAO;AACT;AAEA,SAAS,WAAW,MAAmD;AACrE,QAAM,OAAO,YAAY,MAAM,eAAe;AAC9C,MAAI,SAAS,OAAW,QAAO;AAC/B,SAAO,YAAY,MAAM,UAAU;AACrC;AAEA,SAAS,YACP,MACA,aAC+B;AAC/B,QAAM,OAAO,OAAO,MAAM,MAAM;AAChC,MAAI,SAAS,GAAI,QAAO;AACxB,QAAM,UAA6B;AAAA,IACjC;AAAA,IACA,OAAO,UAAU,MAAM,OAAO;AAAA,IAC9B,KAAK,UAAU,MAAM,KAAK;AAAA,EAC5B;AACA,MAAI,CAAC,YAAa,QAAO;AACzB,QAAM,UAAU,YAAY,MAAM,SAAS;AAC3C,MAAI,YAAY,OAAW,SAAQ,UAAU;AAC7C,SAAO;AACT;AAEA,SAAS,aAAa,OAAgB,aAA2C;AAC/E,MAAI,CAAC,MAAM,QAAQ,KAAK,EAAG,QAAO,CAAC;AACnC,QAAM,WAAgC,CAAC;AACvC,aAAW,QAAQ,OAAO;AACxB,QAAI,CAAC,SAAS,IAAI,KAAK,MAAM,QAAQ,IAAI,EAAG;AAC5C,UAAM,UAAU,YAAY,MAAM,WAAW;AAC7C,QAAI,QAAS,UAAS,KAAK,OAAO;AAAA,EACpC;AACA,SAAO;AACT;AAEA,SAAS,UAAU,MAAoD;AACrE,QAAM,QAAQ,aAAa,KAAK,eAAe,IAAI;AACnD,MAAI,MAAM,SAAS,EAAG,QAAO;AAC7B,SAAO,aAAa,KAAK,UAAU,KAAK;AAC1C;AAEA,SAAS,eAAe,MAAe,OAA+C;AACpF,MAAI,CAAC,SAAS,IAAI,KAAK,MAAM,QAAQ,IAAI,EAAG,OAAM,IAAI,MAAM,oCAAoC;AAChG,QAAM,SAAS,YAAY,MAAM,MAAM;AACvC,MAAI,WAAW,OAAW,OAAM,IAAI,MAAM,oCAAoC;AAC9E,QAAM,QAAQ,UAAU,IAAI;AAC5B,QAAM,WAAW,WAAW,IAAI;AAChC,SAAO;AAAA,IACL,MAAM;AAAA,IACN,GAAI,aAAa,SAAY,EAAE,SAAS,IAAI,CAAC;AAAA,IAC7C,GAAI,MAAM,SAAS,IAAI,EAAE,UAAU,MAAM,IAAI,CAAC;AAAA,IAC9C,OAAO,WAAW,MAAM,WAAW,EAAE,cAAc,WAAW,IAAI,EAAE,CAAC;AAAA,EACvE;AACF;AAEA,SAAS,OAAO,MAAyC;AACvD,QAAM,OAAO,KAAK;AAClB,MACE,CAAC,SAAS,IAAI,KACd,MAAM,QAAQ,IAAI,KAClB,EAAE,KAAK,gBAAgB,eACvB,OAAO,KAAK,aAAa,UACzB;AACA,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,QAAS;AACrB,QAAI,OAAO,UAAU,SAAU;AAC/B,WAAO,GAAG,IAAI;AAAA,EAChB;AACA,SAAO;AACT;AAEA,SAAS,UAAU,MAAc,QAAwB;AACvD,MAAI,SAAS,GAAI,QAAO,QAAQ,MAAM;AACtC,MAAI;AACJ,MAAI;AACF,aAAS,KAAK,MAAM,IAAI;AAAA,EAC1B,QAAQ;AACN,WAAO;AAAA,EACT;AACA,MAAI,CAAC,SAAS,MAAM,KAAK,MAAM,QAAQ,MAAM,EAAG,QAAO;AACvD,QAAM,UAAU,YAAY,QAAQ,SAAS;AAC7C,MAAI,YAAY,OAAW,QAAO;AAClC,SAAO;AACT;AAEA,eAAe,UACb,OACA,SACA,KACA,MACA,aACmB;AACnB,QAAM,UAAU,YAAY,OAAO,OAAO;AAC1C,UAAQ,cAAc,IAAI;AAC1B,QAAM,YAAY,SAAS,SAAS,WAAW;AAC/C,QAAM,WAAW,MAAM,UAAU,KAAK;AAAA,IACpC,QAAQ;AAAA,IACR;AAAA,IACA;AAAA,IACA,QAAQ,cAAc,OAAO;AAAA,EAC/B,CAAC;AACD,QAAM,SAAS,aAAa,EAAE,QAAQ,SAAS,QAAQ,SAAS,CAAC,EAAE,GAAG,KAAc;AACpF,MAAI,CAAC,SAAS,IAAI;AAChB,UAAM,OAAO,MAAM,SAAS,KAAK;AACjC,UAAM,IAAI,MAAM,UAAU,MAAM,SAAS,MAAM,CAAC;AAAA,EAClD;AACA,SAAO;AACT;AAEA,eAAsB,eACpB,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,UAAmC;AAAA,MACvC,mBAAmB;AAAA,MACnB,SAAS,MAAM,cAAc,SAAS;AAAA,MACtC,OAAO;AAAA,QACL,MAAM,OAAO;AAAA,QACb,UAAU,OAAO;AAAA,QACjB,UAAU,cAAc,OAAO,QAAQ;AAAA,MACzC;AAAA,IACF;AACA,UAAM,OAAO,MAAM,aAAa,SAAS,OAAO,OAAO;AACvD,QAAI,CAAC,SAAS,IAAI,KAAK,MAAM,QAAQ,IAAI;AACvC,YAAM,IAAI,MAAM,sCAAsC;AACxD,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": []
+}