@ai-ecoverse/pi-audio 1.19.0 → 1.20.0

package/README.md

@@ -23,6 +23,8 @@ pi-audio does not apply that patch. Hosts such as slicc-agent record monkeypatch
 
 `@ai-ecoverse/pi-audio/durable` appends a `voice.truncate` entry for a `VoiceCut`. The entry has no model messages. Its `edits` replace the target assistant message with the text heard so far and keep tool calls. The stored entry stays intact, and pi-durable derives the shorter message for the next model request. `@earendil-works/pi-durable` is an optional peer (`^1.1.0`); the root package does not import it. `VoiceCut.entryId` is the decimal string of pi-durable's numeric entry id.
 
+`@ai-ecoverse/pi-audio/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.

package/dist/google/catalog.js

@@ -0,0 +1,61 @@
+import { registerVoiceModel } from "../models.js";
+import { registerSpeechApiProvider } from "../registry.js";
+import { speakGoogle } from "./speak.js";
+const tokenCost = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };
+const google = "https://generativelanguage.googleapis.com/v1beta";
+const format = (encoding, sampleRate) => ({
+  encoding,
+  sampleRate,
+  channels: 1
+});
+const outputFormats = [
+  format("wav", 24e3),
+  format("pcm16", 24e3),
+  format("mulaw", 8e3),
+  format("alaw", 8e3)
+];
+const voices = [
+  { id: "Kore", name: "Kore" },
+  { id: "Puck", name: "Puck" },
+  { id: "Charon", name: "Charon" },
+  { id: "Fenrir", name: "Fenrir" },
+  { id: "Aoede", name: "Aoede" }
+];
+function speechModel(id, perOutputMinute) {
+  return {
+    id,
+    name: id,
+    api: "google-tts",
+    provider: "google",
+    baseUrl: google,
+    type: "speech",
+    textStreaming: false,
+    outputFormats,
+    voices,
+    cloning: true,
+    cost: tokenCost,
+    audioCost: { perOutputMinute }
+  };
+}
+function registerGoogle() {
+  const models = [
+    speechModel("gemini-3.8-flash-tts", 25 * 9 * 60 / 1e6),
+    speechModel("gemini-3.8-flash-lite-tts", 25 * 6 * 60 / 1e6)
+  ];
+  const unregisterModels = models.map((model) => registerVoiceModel(model));
+  const unregisterSpeech = registerSpeechApiProvider({
+    api: "google-tts",
+    speak: speakGoogle
+  });
+  let released = false;
+  return () => {
+    if (released) return;
+    released = true;
+    for (const unregister of unregisterModels) unregister();
+    unregisterSpeech();
+  };
+}
+export {
+  registerGoogle
+};
+//# sourceMappingURL=catalog.js.map

package/dist/google/catalog.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/google/catalog.ts"],
+  "sourcesContent": ["import type { AudioFormat } from '../audio.ts';\nimport { registerVoiceModel, type SpeechModel, type VoiceInfo } from '../models.ts';\nimport { registerSpeechApiProvider } from '../registry.ts';\nimport { speakGoogle } from './speak.ts';\n\nconst tokenCost = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };\nconst google = 'https://generativelanguage.googleapis.com/v1beta';\n\nconst format = (encoding: 'wav' | 'pcm16' | 'mulaw' | 'alaw', sampleRate: number): AudioFormat => ({\n  encoding,\n  sampleRate,\n  channels: 1,\n});\n\nconst outputFormats = [\n  format('wav', 24000),\n  format('pcm16', 24000),\n  format('mulaw', 8000),\n  format('alaw', 8000),\n];\n\nconst voices: VoiceInfo[] = [\n  { id: 'Kore', name: 'Kore' },\n  { id: 'Puck', name: 'Puck' },\n  { id: 'Charon', name: 'Charon' },\n  { id: 'Fenrir', name: 'Fenrir' },\n  { id: 'Aoede', name: 'Aoede' },\n];\n\nfunction speechModel(id: string, perOutputMinute: number): SpeechModel<string> {\n  return {\n    id,\n    name: id,\n    api: 'google-tts',\n    provider: 'google',\n    baseUrl: google,\n    type: 'speech',\n    textStreaming: false,\n    outputFormats,\n    voices,\n    cloning: true,\n    cost: tokenCost,\n    audioCost: { perOutputMinute },\n  };\n}\n\nexport function registerGoogle(): () => void {\n  const models = [\n    speechModel('gemini-3.8-flash-tts', (25 * 9 * 60) / 1_000_000),\n    speechModel('gemini-3.8-flash-lite-tts', (25 * 6 * 60) / 1_000_000),\n  ];\n  const unregisterModels = models.map((model) => registerVoiceModel(model));\n  const unregisterSpeech = registerSpeechApiProvider({\n    api: 'google-tts',\n    speak: speakGoogle,\n  });\n  let released = false;\n  return () => {\n    if (released) return;\n    released = true;\n    for (const unregister of unregisterModels) unregister();\n    unregisterSpeech();\n  };\n}\n"],
+  "mappings": "AACA,SAAS,0BAA4D;AACrE,SAAS,iCAAiC;AAC1C,SAAS,mBAAmB;AAE5B,MAAM,YAAY,EAAE,OAAO,GAAG,QAAQ,GAAG,WAAW,GAAG,YAAY,EAAE;AACrE,MAAM,SAAS;AAEf,MAAM,SAAS,CAAC,UAA8C,gBAAqC;AAAA,EACjG;AAAA,EACA;AAAA,EACA,UAAU;AACZ;AAEA,MAAM,gBAAgB;AAAA,EACpB,OAAO,OAAO,IAAK;AAAA,EACnB,OAAO,SAAS,IAAK;AAAA,EACrB,OAAO,SAAS,GAAI;AAAA,EACpB,OAAO,QAAQ,GAAI;AACrB;AAEA,MAAM,SAAsB;AAAA,EAC1B,EAAE,IAAI,QAAQ,MAAM,OAAO;AAAA,EAC3B,EAAE,IAAI,QAAQ,MAAM,OAAO;AAAA,EAC3B,EAAE,IAAI,UAAU,MAAM,SAAS;AAAA,EAC/B,EAAE,IAAI,UAAU,MAAM,SAAS;AAAA,EAC/B,EAAE,IAAI,SAAS,MAAM,QAAQ;AAC/B;AAEA,SAAS,YAAY,IAAY,iBAA8C;AAC7E,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;AAAA,IACA,SAAS;AAAA,IACT,MAAM;AAAA,IACN,WAAW,EAAE,gBAAgB;AAAA,EAC/B;AACF;AAEO,SAAS,iBAA6B;AAC3C,QAAM,SAAS;AAAA,IACb,YAAY,wBAAyB,KAAK,IAAI,KAAM,GAAS;AAAA,IAC7D,YAAY,6BAA8B,KAAK,IAAI,KAAM,GAAS;AAAA,EACpE;AACA,QAAM,mBAAmB,OAAO,IAAI,CAAC,UAAU,mBAAmB,KAAK,CAAC;AACxE,QAAM,mBAAmB,0BAA0B;AAAA,IACjD,KAAK;AAAA,IACL,OAAO;AAAA,EACT,CAAC;AACD,MAAI,WAAW;AACf,SAAO,MAAM;AACX,QAAI,SAAU;AACd,eAAW;AACX,eAAW,cAAc,iBAAkB,YAAW;AACtD,qBAAiB;AAAA,EACnB;AACF;",
+  "names": []
+}

package/dist/google/index.js

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

package/dist/google/index.js.map

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

package/dist/google/speak.js

@@ -0,0 +1,160 @@
+import { chunkSeconds } from "../audio.js";
+import { decodeAudioBase64 } from "../convert.js";
+import {
+  applyPayload,
+  asVoiceError,
+  endpoint,
+  isRecord,
+  readJson,
+  requestSignal,
+  stringField
+} from "../openai/wire.js";
+import { SpeechEventStream } from "../speech.js";
+import { priceAudio } from "../usage.js";
+import { googleHeaders, sameFormat, speechMime } from "./wire.js";
+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, format) {
+  return {
+    contents: [{ role: "user", parts: [{ text }] }],
+    generationConfig: {
+      responseModalities: ["AUDIO"],
+      responseFormat: { audio: { mimeType: speechMime(format) } },
+      speechConfig: {
+        voiceConfig: { prebuiltVoiceConfig: { voiceName: voice } }
+      }
+    }
+  };
+}
+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 direct = stringField(parsed, "message");
+  if (direct !== void 0) return direct;
+  const error = parsed.error;
+  if (typeof error === "string") return error;
+  if (isRecord(error) && !Array.isArray(error)) {
+    const nested = stringField(error, "message");
+    if (nested !== void 0) return nested;
+  }
+  return text;
+}
+function inlineData(part) {
+  const camel = part.inlineData;
+  const snake = part.inline_data;
+  const inline = isRecord(camel) && !Array.isArray(camel) ? camel : snake;
+  if (!isRecord(inline) || Array.isArray(inline)) return void 0;
+  return stringField(inline, "data");
+}
+function audioParts(body) {
+  const candidates = body.candidates;
+  if (!Array.isArray(candidates)) return [];
+  const first = candidates[0];
+  if (!isRecord(first) || Array.isArray(first)) return [];
+  const content = first.content;
+  if (!isRecord(content) || Array.isArray(content)) return [];
+  const parts = content.parts;
+  if (!Array.isArray(parts)) return [];
+  const chunks = [];
+  for (const part of parts) {
+    if (!isRecord(part) || Array.isArray(part)) continue;
+    const data = inlineData(part);
+    if (data === void 0 || data === "") continue;
+    chunks.push(data);
+  }
+  return chunks;
+}
+function concat(chunks) {
+  const length = chunks.reduce((sum, chunk) => sum + chunk.byteLength, 0);
+  const out = new Uint8Array(length);
+  let offset = 0;
+  for (const chunk of chunks) {
+    out.set(chunk, offset);
+    offset += chunk.byteLength;
+  }
+  return out;
+}
+async function googleFetch(model, options, url, body) {
+  const headers = googleHeaders(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 speakGoogle(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 text = await readText(input.text, options?.signal);
+    await postSpeech(model, text, input.voice, 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, format, options, stream) {
+  const payload = speechBody(text, voice, 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 googleFetch(
+    model,
+    options,
+    endpoint(model.baseUrl, `models/${encodeURIComponent(model.id)}:generateContent`),
+    JSON.stringify(body)
+  );
+  const parsed = await readJson(response);
+  if (!isRecord(parsed) || Array.isArray(parsed)) throw new Error("Speech response had no audio");
+  const encoded = audioParts(parsed);
+  if (encoded.length === 0) throw new Error("Speech response had no audio");
+  const data = concat(encoded.map((item) => decodeAudioBase64(item)));
+  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) });
+}
+export {
+  speakGoogle
+};
+//# sourceMappingURL=speak.js.map

package/dist/google/speak.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/google/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 type { SpeechModel } from '../models.ts';\nimport {\n  applyPayload,\n  asVoiceError,\n  endpoint,\n  isRecord,\n  readJson,\n  requestSignal,\n  stringField,\n} from '../openai/wire.ts';\nimport { SpeechEventStream, type SpeechInput } from '../speech.ts';\nimport { type AudioCounters, priceAudio } from '../usage.ts';\nimport { googleHeaders, sameFormat, speechMime } from './wire.ts';\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(text: string, voice: string, format: AudioFormat): Record<string, unknown> {\n  return {\n    contents: [{ role: 'user', parts: [{ text }] }],\n    generationConfig: {\n      responseModalities: ['AUDIO'],\n      responseFormat: { audio: { mimeType: speechMime(format) } },\n      speechConfig: {\n        voiceConfig: { prebuiltVoiceConfig: { voiceName: voice } },\n      },\n    },\n  };\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 direct = stringField(parsed, 'message');\n  if (direct !== undefined) return direct;\n  const error = parsed.error;\n  if (typeof error === 'string') return error;\n  if (isRecord(error) && !Array.isArray(error)) {\n    const nested = stringField(error, 'message');\n    if (nested !== undefined) return nested;\n  }\n  return text;\n}\n\nfunction inlineData(part: Record<string, unknown>): string | undefined {\n  const camel = part.inlineData;\n  const snake = part.inline_data;\n  const inline = isRecord(camel) && !Array.isArray(camel) ? camel : snake;\n  if (!isRecord(inline) || Array.isArray(inline)) return undefined;\n  return stringField(inline, 'data');\n}\n\nfunction audioParts(body: Record<string, unknown>): string[] {\n  const candidates = body.candidates;\n  if (!Array.isArray(candidates)) return [];\n  const first = candidates[0];\n  if (!isRecord(first) || Array.isArray(first)) return [];\n  const content = first.content;\n  if (!isRecord(content) || Array.isArray(content)) return [];\n  const parts = content.parts;\n  if (!Array.isArray(parts)) return [];\n  const chunks: string[] = [];\n  for (const part of parts) {\n    if (!isRecord(part) || Array.isArray(part)) continue;\n    const data = inlineData(part);\n    if (data === undefined || data === '') continue;\n    chunks.push(data);\n  }\n  return chunks;\n}\n\nfunction concat(chunks: Uint8Array[]): Uint8Array {\n  const length = chunks.reduce((sum, chunk) => sum + chunk.byteLength, 0);\n  const out = new Uint8Array(length);\n  let offset = 0;\n  for (const chunk of chunks) {\n    out.set(chunk, offset);\n    offset += chunk.byteLength;\n  }\n  return out;\n}\n\nasync function googleFetch(\n  model: SpeechModel<string>,\n  options: ProviderRequestOptions | undefined,\n  url: string,\n  body: string\n): Promise<Response> {\n  const headers = googleHeaders(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\nexport function speakGoogle(\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 text = await readText(input.text, options?.signal);\n    await postSpeech(model, text, input.voice, 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  format: AudioFormat,\n  options: ProviderRequestOptions | undefined,\n  stream: SpeechEventStream\n) {\n  const payload = speechBody(text, voice, 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 googleFetch(\n    model,\n    options,\n    endpoint(model.baseUrl, `models/${encodeURIComponent(model.id)}:generateContent`),\n    JSON.stringify(body)\n  );\n  const parsed = await readJson(response);\n  if (!isRecord(parsed) || Array.isArray(parsed)) throw new Error('Speech response had no audio');\n  const encoded = audioParts(parsed);\n  if (encoded.length === 0) throw new Error('Speech response had no audio');\n  const data = concat(encoded.map((item) => decodeAudioBase64(item)));\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"],
+  "mappings": "AAEA,SAAS,oBAAoB;AAC7B,SAAS,yBAAyB;AAElC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAA2C;AACpD,SAA6B,kBAAkB;AAC/C,SAAS,eAAe,YAAY,kBAAkB;AAEtD,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,WAAW,MAAc,OAAe,QAA8C;AAC7F,SAAO;AAAA,IACL,UAAU,CAAC,EAAE,MAAM,QAAQ,OAAO,CAAC,EAAE,KAAK,CAAC,EAAE,CAAC;AAAA,IAC9C,kBAAkB;AAAA,MAChB,oBAAoB,CAAC,OAAO;AAAA,MAC5B,gBAAgB,EAAE,OAAO,EAAE,UAAU,WAAW,MAAM,EAAE,EAAE;AAAA,MAC1D,cAAc;AAAA,QACZ,aAAa,EAAE,qBAAqB,EAAE,WAAW,MAAM,EAAE;AAAA,MAC3D;AAAA,IACF;AAAA,EACF;AACF;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,SAAS,YAAY,QAAQ,SAAS;AAC5C,MAAI,WAAW,OAAW,QAAO;AACjC,QAAM,QAAQ,OAAO;AACrB,MAAI,OAAO,UAAU,SAAU,QAAO;AACtC,MAAI,SAAS,KAAK,KAAK,CAAC,MAAM,QAAQ,KAAK,GAAG;AAC5C,UAAM,SAAS,YAAY,OAAO,SAAS;AAC3C,QAAI,WAAW,OAAW,QAAO;AAAA,EACnC;AACA,SAAO;AACT;AAEA,SAAS,WAAW,MAAmD;AACrE,QAAM,QAAQ,KAAK;AACnB,QAAM,QAAQ,KAAK;AACnB,QAAM,SAAS,SAAS,KAAK,KAAK,CAAC,MAAM,QAAQ,KAAK,IAAI,QAAQ;AAClE,MAAI,CAAC,SAAS,MAAM,KAAK,MAAM,QAAQ,MAAM,EAAG,QAAO;AACvD,SAAO,YAAY,QAAQ,MAAM;AACnC;AAEA,SAAS,WAAW,MAAyC;AAC3D,QAAM,aAAa,KAAK;AACxB,MAAI,CAAC,MAAM,QAAQ,UAAU,EAAG,QAAO,CAAC;AACxC,QAAM,QAAQ,WAAW,CAAC;AAC1B,MAAI,CAAC,SAAS,KAAK,KAAK,MAAM,QAAQ,KAAK,EAAG,QAAO,CAAC;AACtD,QAAM,UAAU,MAAM;AACtB,MAAI,CAAC,SAAS,OAAO,KAAK,MAAM,QAAQ,OAAO,EAAG,QAAO,CAAC;AAC1D,QAAM,QAAQ,QAAQ;AACtB,MAAI,CAAC,MAAM,QAAQ,KAAK,EAAG,QAAO,CAAC;AACnC,QAAM,SAAmB,CAAC;AAC1B,aAAW,QAAQ,OAAO;AACxB,QAAI,CAAC,SAAS,IAAI,KAAK,MAAM,QAAQ,IAAI,EAAG;AAC5C,UAAM,OAAO,WAAW,IAAI;AAC5B,QAAI,SAAS,UAAa,SAAS,GAAI;AACvC,WAAO,KAAK,IAAI;AAAA,EAClB;AACA,SAAO;AACT;AAEA,SAAS,OAAO,QAAkC;AAChD,QAAM,SAAS,OAAO,OAAO,CAAC,KAAK,UAAU,MAAM,MAAM,YAAY,CAAC;AACtE,QAAM,MAAM,IAAI,WAAW,MAAM;AACjC,MAAI,SAAS;AACb,aAAW,SAAS,QAAQ;AAC1B,QAAI,IAAI,OAAO,MAAM;AACrB,cAAU,MAAM;AAAA,EAClB;AACA,SAAO;AACT;AAEA,eAAe,YACb,OACA,SACA,KACA,MACmB;AACnB,QAAM,UAAU,cAAc,OAAO,OAAO;AAC5C,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;AAEO,SAAS,YACd,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,OAAO,MAAM,SAAS,MAAM,MAAM,SAAS,MAAM;AACvD,UAAM,WAAW,OAAO,MAAM,MAAM,OAAO,QAAQ,SAAS,MAAM;AAAA,EACpE,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,QACA,SACA,QACA;AACA,QAAM,UAAU,WAAW,MAAM,OAAO,MAAM;AAC9C,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,UAAU,mBAAmB,MAAM,EAAE,CAAC,kBAAkB;AAAA,IAChF,KAAK,UAAU,IAAI;AAAA,EACrB;AACA,QAAM,SAAS,MAAM,SAAS,QAAQ;AACtC,MAAI,CAAC,SAAS,MAAM,KAAK,MAAM,QAAQ,MAAM,EAAG,OAAM,IAAI,MAAM,8BAA8B;AAC9F,QAAM,UAAU,WAAW,MAAM;AACjC,MAAI,QAAQ,WAAW,EAAG,OAAM,IAAI,MAAM,8BAA8B;AACxE,QAAM,OAAO,OAAO,QAAQ,IAAI,CAAC,SAAS,kBAAkB,IAAI,CAAC,CAAC;AAClE,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;",
+  "names": []
+}

package/dist/google/wire.js

@@ -0,0 +1,43 @@
+const mimes = {
+  wav: "AUDIO_WAV",
+  pcm16: "AUDIO_L16",
+  mulaw: "AUDIO_MULAW",
+  alaw: "AUDIO_ALAW"
+};
+function speechMime(format) {
+  const name = mimes[format.encoding];
+  if (!name) throw new Error(`Unsupported speech format ${format.encoding}`);
+  return name;
+}
+function sameFormat(left, right) {
+  return left.encoding === right.encoding && left.sampleRate === right.sampleRate && left.channels === right.channels;
+}
+function headerValue(headers, name) {
+  const expected = name.toLowerCase();
+  for (const [key, value] of Object.entries(headers)) {
+    if (key.toLowerCase() === expected && value.trim() !== "") return value;
+  }
+  return void 0;
+}
+function googleHeaders(model, options) {
+  const headers = { ...model.headers };
+  const extra = options?.headers;
+  if (extra) {
+    for (const [key, value] of Object.entries(extra)) {
+      if (value === null) delete headers[key];
+      else headers[key] = value;
+    }
+  }
+  if (headerValue(headers, "x-goog-api-key") === void 0) {
+    const apiKey = options?.apiKey;
+    if (!apiKey) throw new Error(`No API key for provider: ${model.provider}`);
+    headers["x-goog-api-key"] = apiKey;
+  }
+  return headers;
+}
+export {
+  googleHeaders,
+  sameFormat,
+  speechMime
+};
+//# sourceMappingURL=wire.js.map

package/dist/google/wire.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/google/wire.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport type { AudioFormat } from '../audio.ts';\n\nconst mimes: Record<string, string> = {\n  wav: 'AUDIO_WAV',\n  pcm16: 'AUDIO_L16',\n  mulaw: 'AUDIO_MULAW',\n  alaw: 'AUDIO_ALAW',\n};\n\nexport function speechMime(format: AudioFormat): string {\n  const name = mimes[format.encoding];\n  if (!name) throw new Error(`Unsupported speech format ${format.encoding}`);\n  return name;\n}\n\nexport function sameFormat(left: AudioFormat, right: AudioFormat): boolean {\n  return (\n    left.encoding === right.encoding &&\n    left.sampleRate === right.sampleRate &&\n    left.channels === right.channels\n  );\n}\n\nfunction headerValue(headers: Record<string, string>, name: string): string | undefined {\n  const expected = name.toLowerCase();\n  for (const [key, value] of Object.entries(headers)) {\n    if (key.toLowerCase() === expected && value.trim() !== '') return value;\n  }\n  return undefined;\n}\n\nexport function googleHeaders(\n  model: { provider: string; headers?: Record<string, string> },\n  options: ProviderRequestOptions | undefined\n): Record<string, string> {\n  const headers: Record<string, string> = { ...model.headers };\n  const extra = options?.headers;\n  if (extra) {\n    for (const [key, value] of Object.entries(extra)) {\n      if (value === null) delete headers[key];\n      else headers[key] = value;\n    }\n  }\n  if (headerValue(headers, 'x-goog-api-key') === undefined) {\n    const apiKey = options?.apiKey;\n    if (!apiKey) throw new Error(`No API key for provider: ${model.provider}`);\n    headers['x-goog-api-key'] = apiKey;\n  }\n  return headers;\n}\n"],
+  "mappings": "AAGA,MAAM,QAAgC;AAAA,EACpC,KAAK;AAAA,EACL,OAAO;AAAA,EACP,OAAO;AAAA,EACP,MAAM;AACR;AAEO,SAAS,WAAW,QAA6B;AACtD,QAAM,OAAO,MAAM,OAAO,QAAQ;AAClC,MAAI,CAAC,KAAM,OAAM,IAAI,MAAM,6BAA6B,OAAO,QAAQ,EAAE;AACzE,SAAO;AACT;AAEO,SAAS,WAAW,MAAmB,OAA6B;AACzE,SACE,KAAK,aAAa,MAAM,YACxB,KAAK,eAAe,MAAM,cAC1B,KAAK,aAAa,MAAM;AAE5B;AAEA,SAAS,YAAY,SAAiC,MAAkC;AACtF,QAAM,WAAW,KAAK,YAAY;AAClC,aAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,OAAO,GAAG;AAClD,QAAI,IAAI,YAAY,MAAM,YAAY,MAAM,KAAK,MAAM,GAAI,QAAO;AAAA,EACpE;AACA,SAAO;AACT;AAEO,SAAS,cACd,OACA,SACwB;AACxB,QAAM,UAAkC,EAAE,GAAG,MAAM,QAAQ;AAC3D,QAAM,QAAQ,SAAS;AACvB,MAAI,OAAO;AACT,eAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,KAAK,GAAG;AAChD,UAAI,UAAU,KAAM,QAAO,QAAQ,GAAG;AAAA,UACjC,SAAQ,GAAG,IAAI;AAAA,IACtB;AAAA,EACF;AACA,MAAI,YAAY,SAAS,gBAAgB,MAAM,QAAW;AACxD,UAAM,SAAS,SAAS;AACxB,QAAI,CAAC,OAAQ,OAAM,IAAI,MAAM,4BAA4B,MAAM,QAAQ,EAAE;AACzE,YAAQ,gBAAgB,IAAI;AAAA,EAC9B;AACA,SAAO;AACT;",
+  "names": []
+}

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

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

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

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

package/dist/types/google/speak.d.ts

@@ -0,0 +1,4 @@
+import type { ProviderRequestOptions } from '@earendil-works/pi-ai';
+import type { SpeechModel } from '../models.ts';
+import { SpeechEventStream, type SpeechInput } from '../speech.ts';
+export declare function speakGoogle(model: SpeechModel<string>, input: SpeechInput, options?: ProviderRequestOptions): SpeechEventStream;

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

@@ -0,0 +1,8 @@
+import type { ProviderRequestOptions } from '@earendil-works/pi-ai';
+import type { AudioFormat } from '../audio.ts';
+export declare function speechMime(format: AudioFormat): string;
+export declare function sameFormat(left: AudioFormat, right: AudioFormat): boolean;
+export declare function googleHeaders(model: {
+    provider: string;
+    headers?: Record<string, string>;
+}, options: ProviderRequestOptions | undefined): Record<string, string>;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-ecoverse/pi-audio",
-  "version": "1.19.0",
+  "version": "1.20.0",
   "description": "Voice bindings for pi: speech-to-text, text-to-speech and realtime speech for pi-coding-agent plugins and slicc-agent",
   "type": "module",
   "license": "Apache-2.0",
@@ -58,6 +58,10 @@
       "types": "./dist/types/fish/index.d.ts",
       "default": "./dist/fish/index.js"
     },
+    "./google": {
+      "types": "./dist/types/google/index.d.ts",
+      "default": "./dist/google/index.js"
+    },
     "./durable": {
       "types": "./dist/types/durable/index.d.ts",
       "default": "./dist/durable/index.js"