@ai-ecoverse/pi-audio 1.4.0 → 1.6.0

package/dist/openai/wire.js

@@ -0,0 +1,201 @@
+import { encodeAudioBase64 } from "../convert.js";
+import { VoiceRequestError } from "../errors.js";
+const boundary = "pi-audio-boundary";
+const mimeFiles = {
+  "audio/wav": { format: "wav", filename: "audio.wav" },
+  "audio/wave": { format: "wav", filename: "audio.wav" },
+  "audio/x-wav": { format: "wav", filename: "audio.wav" },
+  "audio/mpeg": { format: "mp3", filename: "audio.mp3" },
+  "audio/mp3": { format: "mp3", filename: "audio.mp3" },
+  "audio/flac": { format: "flac", filename: "audio.flac" },
+  "audio/ogg": { format: "ogg", filename: "audio.ogg" },
+  "audio/opus": { format: "ogg", filename: "audio.ogg" },
+  "audio/webm": { format: "webm", filename: "audio.webm" },
+  "audio/mp4": { format: "m4a", filename: "audio.m4a" },
+  "audio/m4a": { format: "m4a", filename: "audio.m4a" },
+  "audio/aac": { format: "aac", filename: "audio.aac" },
+  "audio/pcm": { format: "pcm", filename: "audio.pcm" }
+};
+function isOpenRouter(model) {
+  return model.provider === "openrouter" || model.baseUrl.includes("openrouter.ai");
+}
+function endpoint(baseUrl, path) {
+  return `${baseUrl.replace(/\/$/, "")}/${path}`;
+}
+function realtimeUrl(baseUrl) {
+  const url = new URL(baseUrl);
+  url.protocol = url.protocol === "https:" ? "wss:" : "ws:";
+  url.pathname = `${url.pathname.replace(/\/$/, "")}/realtime`;
+  url.search = "intent=transcription";
+  return url.toString();
+}
+function audioFile(mimeType) {
+  const found = mimeFiles[mimeType.toLowerCase()];
+  if (!found) throw new Error(`Unsupported audio type ${mimeType}`);
+  return found;
+}
+function isRecord(value) {
+  return typeof value === "object" && value !== null;
+}
+function stringField(record, key) {
+  const value = record[key];
+  return typeof value === "string" ? value : void 0;
+}
+function numberField(record, key) {
+  const value = record[key];
+  return typeof value === "number" ? value : void 0;
+}
+function asVoiceError(error) {
+  if (error instanceof VoiceRequestError) return { reason: error.reason, error: error.message };
+  if (error instanceof Error && error.name === "AbortError")
+    return { reason: "aborted", error: "aborted" };
+  if (error instanceof Error) return { reason: "error", error: error.message };
+  return { reason: "error", error: String(error) };
+}
+function headerValue(headers, name) {
+  if (!headers) return void 0;
+  const expected = name.toLowerCase();
+  for (const [key, value] of Object.entries(headers)) {
+    if (key.toLowerCase() === expected && value !== null && value.trim() !== "") return value;
+  }
+  return void 0;
+}
+function hasAuth(headers) {
+  return headerValue(headers, "authorization") !== void 0 || headerValue(headers, "cf-aig-authorization") !== void 0;
+}
+function requestHeaders(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 (!hasAuth(headers)) {
+    const apiKey = options?.apiKey;
+    if (!apiKey) throw new Error(`No API key for provider: ${model.provider}`);
+    headers.Authorization = `Bearer ${apiKey}`;
+  }
+  return headers;
+}
+function requestSignal(options) {
+  const signals = [];
+  if (options?.signal) signals.push(options.signal);
+  const timeout = options?.timeoutMs;
+  if (timeout !== void 0 && timeout > 0) signals.push(AbortSignal.timeout(timeout));
+  if (signals.length === 0) return void 0;
+  if (signals.length === 1) return signals[0];
+  return AbortSignal.any(signals);
+}
+function encodeMultipart(file, fields) {
+  const chunks = [];
+  const encoder = new TextEncoder();
+  const write = (value) => chunks.push(encoder.encode(value));
+  write(
+    `--${boundary}\r
+Content-Disposition: form-data; name="file"; filename="${file.filename}"\r
+Content-Type: ${file.mimeType}\r
+\r
+`
+  );
+  chunks.push(file.data);
+  write("\r\n");
+  for (const [name, value] of Object.entries(fields)) {
+    write(`--${boundary}\r
+Content-Disposition: form-data; name="${name}"\r
+\r
+${value}\r
+`);
+  }
+  write(`--${boundary}--\r
+`);
+  const body = new Uint8Array(chunks.reduce((sum, chunk) => sum + chunk.length, 0));
+  let offset = 0;
+  for (const chunk of chunks) {
+    body.set(chunk, offset);
+    offset += chunk.length;
+  }
+  return body;
+}
+function multipartType() {
+  return `multipart/form-data; boundary=${boundary}`;
+}
+async function readAudio(input, fetchImpl, signal) {
+  if (!("url" in input)) return { data: input.data, mimeType: input.mimeType };
+  const response = await fetchImpl(input.url, { signal });
+  if (!response.ok) throw new Error(`Audio URL failed: HTTP ${response.status}`);
+  const mime = response.headers.get("content-type");
+  const mimeType = mime === null || mime.trim() === "" ? "audio/wav" : mime.split(";")[0].trim();
+  return { data: new Uint8Array(await response.arrayBuffer()), mimeType };
+}
+function audioPayload(file, openRouter, modelId) {
+  if (openRouter) {
+    return {
+      model: modelId,
+      input_audio: { data: encodeAudioBase64(file.data), format: file.format }
+    };
+  }
+  return { model: modelId, file };
+}
+async function applyPayload(payload, model, options) {
+  const hook = options?.onPayload;
+  const replaced = await hook?.(payload, model);
+  return replaced === void 0 ? payload : replaced;
+}
+function headerRecord(headers) {
+  const record = {};
+  headers.forEach((value, key) => {
+    record[key] = value;
+  });
+  return record;
+}
+async function voiceFetch(model, options, url, body, contentType) {
+  const headers = requestHeaders(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: headerRecord(response.headers) },
+    model
+  );
+  if (!response.ok) {
+    const text = await response.text();
+    throw new Error(text === "" ? `HTTP ${response.status}` : text);
+  }
+  return response;
+}
+async function readJson(response) {
+  const text = await response.text();
+  try {
+    return JSON.parse(text);
+  } catch {
+    throw new Error("Transcription response was not JSON");
+  }
+}
+export {
+  applyPayload,
+  asVoiceError,
+  audioFile,
+  audioPayload,
+  encodeMultipart,
+  endpoint,
+  hasAuth,
+  isOpenRouter,
+  isRecord,
+  multipartType,
+  numberField,
+  readAudio,
+  readJson,
+  realtimeUrl,
+  requestHeaders,
+  requestSignal,
+  stringField,
+  voiceFetch
+};
+//# sourceMappingURL=wire.js.map

package/dist/openai/wire.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/openai/wire.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport { encodeAudioBase64 } from '../convert.ts';\nimport { VoiceRequestError } from '../errors.ts';\n\nconst boundary = 'pi-audio-boundary';\n\nconst mimeFiles: Record<string, { format: string; filename: string }> = {\n  'audio/wav': { format: 'wav', filename: 'audio.wav' },\n  'audio/wave': { format: 'wav', filename: 'audio.wav' },\n  'audio/x-wav': { format: 'wav', filename: 'audio.wav' },\n  'audio/mpeg': { format: 'mp3', filename: 'audio.mp3' },\n  'audio/mp3': { format: 'mp3', filename: 'audio.mp3' },\n  'audio/flac': { format: 'flac', filename: 'audio.flac' },\n  'audio/ogg': { format: 'ogg', filename: 'audio.ogg' },\n  'audio/opus': { format: 'ogg', filename: 'audio.ogg' },\n  'audio/webm': { format: 'webm', filename: 'audio.webm' },\n  'audio/mp4': { format: 'm4a', filename: 'audio.m4a' },\n  'audio/m4a': { format: 'm4a', filename: 'audio.m4a' },\n  'audio/aac': { format: 'aac', filename: 'audio.aac' },\n  'audio/pcm': { format: 'pcm', filename: 'audio.pcm' },\n};\n\nexport interface AudioFile {\n  data: Uint8Array;\n  mimeType: string;\n  filename: string;\n  format: string;\n}\n\nexport interface JsonBody {\n  [key: string]: unknown;\n}\n\ninterface HeaderModel {\n  provider: string;\n  headers?: Record<string, string>;\n}\n\nexport function isOpenRouter(model: { provider: string; baseUrl: string }): boolean {\n  return model.provider === 'openrouter' || model.baseUrl.includes('openrouter.ai');\n}\n\nexport function endpoint(baseUrl: string, path: string): string {\n  return `${baseUrl.replace(/\\/$/, '')}/${path}`;\n}\n\nexport function realtimeUrl(baseUrl: string): string {\n  const url = new URL(baseUrl);\n  url.protocol = url.protocol === 'https:' ? 'wss:' : 'ws:';\n  url.pathname = `${url.pathname.replace(/\\/$/, '')}/realtime`;\n  url.search = 'intent=transcription';\n  return url.toString();\n}\n\nexport function audioFile(mimeType: string): { format: string; filename: string } {\n  const found = mimeFiles[mimeType.toLowerCase()];\n  if (!found) throw new Error(`Unsupported audio type ${mimeType}`);\n  return found;\n}\n\nexport function isRecord(value: unknown): value is Record<string, unknown> {\n  return typeof value === 'object' && value !== null;\n}\n\nexport function stringField(record: Record<string, unknown>, key: string): string | undefined {\n  const value = record[key];\n  return typeof value === 'string' ? value : undefined;\n}\n\nexport function numberField(record: Record<string, unknown>, key: string): number | undefined {\n  const value = record[key];\n  return typeof value === 'number' ? value : undefined;\n}\n\nexport function asVoiceError(error: unknown): { reason: 'aborted' | 'error'; error: string } {\n  if (error instanceof VoiceRequestError) return { reason: error.reason, error: error.message };\n  if (error instanceof Error && error.name === 'AbortError')\n    return { reason: 'aborted', error: 'aborted' };\n  if (error instanceof Error) return { reason: 'error', error: error.message };\n  return { reason: 'error', error: String(error) };\n}\n\nfunction headerValue(\n  headers: Record<string, string | null> | undefined,\n  name: string\n): string | undefined {\n  if (!headers) return undefined;\n  const expected = name.toLowerCase();\n  for (const [key, value] of Object.entries(headers)) {\n    if (key.toLowerCase() === expected && value !== null && value.trim() !== '') return value;\n  }\n  return undefined;\n}\n\nexport function hasAuth(headers: Record<string, string | null> | undefined): boolean {\n  return (\n    headerValue(headers, 'authorization') !== undefined ||\n    headerValue(headers, 'cf-aig-authorization') !== undefined\n  );\n}\n\nexport function requestHeaders(\n  model: HeaderModel,\n  options: ProviderRequestOptions | undefined\n): Record<string, string> {\n  const headers: Record<string, string> = { ...model.headers };\n  const extra = options?.headers;\n  if (extra) {\n    for (const [key, value] of Object.entries(extra)) {\n      if (value === null) delete headers[key];\n      else headers[key] = value;\n    }\n  }\n  if (!hasAuth(headers)) {\n    const apiKey = options?.apiKey;\n    if (!apiKey) throw new Error(`No API key for provider: ${model.provider}`);\n    headers.Authorization = `Bearer ${apiKey}`;\n  }\n  return headers;\n}\n\nexport function requestSignal(\n  options: ProviderRequestOptions | undefined\n): AbortSignal | undefined {\n  const signals: AbortSignal[] = [];\n  if (options?.signal) signals.push(options.signal);\n  const timeout = options?.timeoutMs;\n  if (timeout !== undefined && timeout > 0) signals.push(AbortSignal.timeout(timeout));\n  if (signals.length === 0) return undefined;\n  if (signals.length === 1) return signals[0];\n  return AbortSignal.any(signals);\n}\n\nexport function encodeMultipart(file: AudioFile, fields: Record<string, string>): Uint8Array {\n  const chunks: Uint8Array[] = [];\n  const encoder = new TextEncoder();\n  const write = (value: string) => chunks.push(encoder.encode(value));\n  write(\n    `--${boundary}\\r\\nContent-Disposition: form-data; name=\"file\"; filename=\"${file.filename}\"\\r\\nContent-Type: ${file.mimeType}\\r\\n\\r\\n`\n  );\n  chunks.push(file.data);\n  write('\\r\\n');\n  for (const [name, value] of Object.entries(fields)) {\n    write(`--${boundary}\\r\\nContent-Disposition: form-data; name=\"${name}\"\\r\\n\\r\\n${value}\\r\\n`);\n  }\n  write(`--${boundary}--\\r\\n`);\n  const body = new Uint8Array(chunks.reduce((sum, chunk) => sum + chunk.length, 0));\n  let offset = 0;\n  for (const chunk of chunks) {\n    body.set(chunk, offset);\n    offset += chunk.length;\n  }\n  return body;\n}\n\nexport function multipartType(): string {\n  return `multipart/form-data; boundary=${boundary}`;\n}\n\nexport async function readAudio(\n  input: { data: Uint8Array; mimeType: string } | { url: string },\n  fetchImpl: typeof fetch,\n  signal: AbortSignal | undefined\n): Promise<{ data: Uint8Array; mimeType: string }> {\n  if (!('url' in input)) return { data: input.data, mimeType: input.mimeType };\n  const response = await fetchImpl(input.url, { signal });\n  if (!response.ok) throw new Error(`Audio URL failed: HTTP ${response.status}`);\n  const mime = response.headers.get('content-type');\n  const mimeType = mime === null || mime.trim() === '' ? 'audio/wav' : mime.split(';')[0].trim();\n  return { data: new Uint8Array(await response.arrayBuffer()), mimeType };\n}\n\nexport function audioPayload(file: AudioFile, openRouter: boolean, modelId: string): JsonBody {\n  if (openRouter) {\n    return {\n      model: modelId,\n      input_audio: { data: encodeAudioBase64(file.data), format: file.format },\n    };\n  }\n  return { model: modelId, file };\n}\n\nexport async function applyPayload(\n  payload: unknown,\n  model: unknown,\n  options: ProviderRequestOptions | undefined\n): Promise<unknown> {\n  const hook = options?.onPayload as\n    | ((body: unknown, voiceModel: unknown) => unknown | undefined | Promise<unknown | undefined>)\n    | undefined;\n  const replaced = await hook?.(payload, model);\n  return replaced === undefined ? payload : replaced;\n}\n\nfunction headerRecord(headers: Headers): Record<string, string> {\n  const record: Record<string, string> = {};\n  headers.forEach((value, key) => {\n    record[key] = value;\n  });\n  return record;\n}\n\nexport async function voiceFetch(\n  model: HeaderModel & { id: string },\n  options: ProviderRequestOptions | undefined,\n  url: string,\n  body: BodyInit,\n  contentType: string\n): Promise<Response> {\n  const headers = requestHeaders(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?.(\n    { status: response.status, headers: headerRecord(response.headers) },\n    model as never\n  );\n  if (!response.ok) {\n    const text = await response.text();\n    throw new Error(text === '' ? `HTTP ${response.status}` : text);\n  }\n  return response;\n}\n\nexport async function readJson(response: Response): Promise<unknown> {\n  const text = await response.text();\n  try {\n    return JSON.parse(text);\n  } catch {\n    throw new Error('Transcription response was not JSON');\n  }\n}\n"],
+  "mappings": "AACA,SAAS,yBAAyB;AAClC,SAAS,yBAAyB;AAElC,MAAM,WAAW;AAEjB,MAAM,YAAkE;AAAA,EACtE,aAAa,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACpD,cAAc,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACrD,eAAe,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACtD,cAAc,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACrD,aAAa,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACpD,cAAc,EAAE,QAAQ,QAAQ,UAAU,aAAa;AAAA,EACvD,aAAa,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACpD,cAAc,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACrD,cAAc,EAAE,QAAQ,QAAQ,UAAU,aAAa;AAAA,EACvD,aAAa,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACpD,aAAa,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACpD,aAAa,EAAE,QAAQ,OAAO,UAAU,YAAY;AAAA,EACpD,aAAa,EAAE,QAAQ,OAAO,UAAU,YAAY;AACtD;AAkBO,SAAS,aAAa,OAAuD;AAClF,SAAO,MAAM,aAAa,gBAAgB,MAAM,QAAQ,SAAS,eAAe;AAClF;AAEO,SAAS,SAAS,SAAiB,MAAsB;AAC9D,SAAO,GAAG,QAAQ,QAAQ,OAAO,EAAE,CAAC,IAAI,IAAI;AAC9C;AAEO,SAAS,YAAY,SAAyB;AACnD,QAAM,MAAM,IAAI,IAAI,OAAO;AAC3B,MAAI,WAAW,IAAI,aAAa,WAAW,SAAS;AACpD,MAAI,WAAW,GAAG,IAAI,SAAS,QAAQ,OAAO,EAAE,CAAC;AACjD,MAAI,SAAS;AACb,SAAO,IAAI,SAAS;AACtB;AAEO,SAAS,UAAU,UAAwD;AAChF,QAAM,QAAQ,UAAU,SAAS,YAAY,CAAC;AAC9C,MAAI,CAAC,MAAO,OAAM,IAAI,MAAM,0BAA0B,QAAQ,EAAE;AAChE,SAAO;AACT;AAEO,SAAS,SAAS,OAAkD;AACzE,SAAO,OAAO,UAAU,YAAY,UAAU;AAChD;AAEO,SAAS,YAAY,QAAiC,KAAiC;AAC5F,QAAM,QAAQ,OAAO,GAAG;AACxB,SAAO,OAAO,UAAU,WAAW,QAAQ;AAC7C;AAEO,SAAS,YAAY,QAAiC,KAAiC;AAC5F,QAAM,QAAQ,OAAO,GAAG;AACxB,SAAO,OAAO,UAAU,WAAW,QAAQ;AAC7C;AAEO,SAAS,aAAa,OAAgE;AAC3F,MAAI,iBAAiB,kBAAmB,QAAO,EAAE,QAAQ,MAAM,QAAQ,OAAO,MAAM,QAAQ;AAC5F,MAAI,iBAAiB,SAAS,MAAM,SAAS;AAC3C,WAAO,EAAE,QAAQ,WAAW,OAAO,UAAU;AAC/C,MAAI,iBAAiB,MAAO,QAAO,EAAE,QAAQ,SAAS,OAAO,MAAM,QAAQ;AAC3E,SAAO,EAAE,QAAQ,SAAS,OAAO,OAAO,KAAK,EAAE;AACjD;AAEA,SAAS,YACP,SACA,MACoB;AACpB,MAAI,CAAC,QAAS,QAAO;AACrB,QAAM,WAAW,KAAK,YAAY;AAClC,aAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,OAAO,GAAG;AAClD,QAAI,IAAI,YAAY,MAAM,YAAY,UAAU,QAAQ,MAAM,KAAK,MAAM,GAAI,QAAO;AAAA,EACtF;AACA,SAAO;AACT;AAEO,SAAS,QAAQ,SAA6D;AACnF,SACE,YAAY,SAAS,eAAe,MAAM,UAC1C,YAAY,SAAS,sBAAsB,MAAM;AAErD;AAEO,SAAS,eACd,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,CAAC,QAAQ,OAAO,GAAG;AACrB,UAAM,SAAS,SAAS;AACxB,QAAI,CAAC,OAAQ,OAAM,IAAI,MAAM,4BAA4B,MAAM,QAAQ,EAAE;AACzE,YAAQ,gBAAgB,UAAU,MAAM;AAAA,EAC1C;AACA,SAAO;AACT;AAEO,SAAS,cACd,SACyB;AACzB,QAAM,UAAyB,CAAC;AAChC,MAAI,SAAS,OAAQ,SAAQ,KAAK,QAAQ,MAAM;AAChD,QAAM,UAAU,SAAS;AACzB,MAAI,YAAY,UAAa,UAAU,EAAG,SAAQ,KAAK,YAAY,QAAQ,OAAO,CAAC;AACnF,MAAI,QAAQ,WAAW,EAAG,QAAO;AACjC,MAAI,QAAQ,WAAW,EAAG,QAAO,QAAQ,CAAC;AAC1C,SAAO,YAAY,IAAI,OAAO;AAChC;AAEO,SAAS,gBAAgB,MAAiB,QAA4C;AAC3F,QAAM,SAAuB,CAAC;AAC9B,QAAM,UAAU,IAAI,YAAY;AAChC,QAAM,QAAQ,CAAC,UAAkB,OAAO,KAAK,QAAQ,OAAO,KAAK,CAAC;AAClE;AAAA,IACE,KAAK,QAAQ;AAAA,yDAA8D,KAAK,QAAQ;AAAA,gBAAsB,KAAK,QAAQ;AAAA;AAAA;AAAA,EAC7H;AACA,SAAO,KAAK,KAAK,IAAI;AACrB,QAAM,MAAM;AACZ,aAAW,CAAC,MAAM,KAAK,KAAK,OAAO,QAAQ,MAAM,GAAG;AAClD,UAAM,KAAK,QAAQ;AAAA,wCAA6C,IAAI;AAAA;AAAA,EAAY,KAAK;AAAA,CAAM;AAAA,EAC7F;AACA,QAAM,KAAK,QAAQ;AAAA,CAAQ;AAC3B,QAAM,OAAO,IAAI,WAAW,OAAO,OAAO,CAAC,KAAK,UAAU,MAAM,MAAM,QAAQ,CAAC,CAAC;AAChF,MAAI,SAAS;AACb,aAAW,SAAS,QAAQ;AAC1B,SAAK,IAAI,OAAO,MAAM;AACtB,cAAU,MAAM;AAAA,EAClB;AACA,SAAO;AACT;AAEO,SAAS,gBAAwB;AACtC,SAAO,iCAAiC,QAAQ;AAClD;AAEA,eAAsB,UACpB,OACA,WACA,QACiD;AACjD,MAAI,EAAE,SAAS,OAAQ,QAAO,EAAE,MAAM,MAAM,MAAM,UAAU,MAAM,SAAS;AAC3E,QAAM,WAAW,MAAM,UAAU,MAAM,KAAK,EAAE,OAAO,CAAC;AACtD,MAAI,CAAC,SAAS,GAAI,OAAM,IAAI,MAAM,0BAA0B,SAAS,MAAM,EAAE;AAC7E,QAAM,OAAO,SAAS,QAAQ,IAAI,cAAc;AAChD,QAAM,WAAW,SAAS,QAAQ,KAAK,KAAK,MAAM,KAAK,cAAc,KAAK,MAAM,GAAG,EAAE,CAAC,EAAE,KAAK;AAC7F,SAAO,EAAE,MAAM,IAAI,WAAW,MAAM,SAAS,YAAY,CAAC,GAAG,SAAS;AACxE;AAEO,SAAS,aAAa,MAAiB,YAAqB,SAA2B;AAC5F,MAAI,YAAY;AACd,WAAO;AAAA,MACL,OAAO;AAAA,MACP,aAAa,EAAE,MAAM,kBAAkB,KAAK,IAAI,GAAG,QAAQ,KAAK,OAAO;AAAA,IACzE;AAAA,EACF;AACA,SAAO,EAAE,OAAO,SAAS,KAAK;AAChC;AAEA,eAAsB,aACpB,SACA,OACA,SACkB;AAClB,QAAM,OAAO,SAAS;AAGtB,QAAM,WAAW,MAAM,OAAO,SAAS,KAAK;AAC5C,SAAO,aAAa,SAAY,UAAU;AAC5C;AAEA,SAAS,aAAa,SAA0C;AAC9D,QAAM,SAAiC,CAAC;AACxC,UAAQ,QAAQ,CAAC,OAAO,QAAQ;AAC9B,WAAO,GAAG,IAAI;AAAA,EAChB,CAAC;AACD,SAAO;AACT;AAEA,eAAsB,WACpB,OACA,SACA,KACA,MACA,aACmB;AACnB,QAAM,UAAU,eAAe,OAAO,OAAO;AAC7C,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,IACb,EAAE,QAAQ,SAAS,QAAQ,SAAS,aAAa,SAAS,OAAO,EAAE;AAAA,IACnE;AAAA,EACF;AACA,MAAI,CAAC,SAAS,IAAI;AAChB,UAAM,OAAO,MAAM,SAAS,KAAK;AACjC,UAAM,IAAI,MAAM,SAAS,KAAK,QAAQ,SAAS,MAAM,KAAK,IAAI;AAAA,EAChE;AACA,SAAO;AACT;AAEA,eAAsB,SAAS,UAAsC;AACnE,QAAM,OAAO,MAAM,SAAS,KAAK;AACjC,MAAI;AACF,WAAO,KAAK,MAAM,IAAI;AAAA,EACxB,QAAQ;AACN,UAAM,IAAI,MAAM,qCAAqC;AAAA,EACvD;AACF;",
+  "names": []
+}

package/dist/realtime.js

@@ -1 +1,94 @@
+import {
+  alignmentSeconds,
+  chunkSeconds
+} from "./audio.js";
+import { VoiceRequestError } from "./errors.js";
+import { priceAudio } from "./usage.js";
+function planDelivery(delivery, requested) {
+  const support = delivery[requested];
+  if (support !== "native" && support !== "approximate") {
+    throw new VoiceRequestError("error", `unsupported delivery: ${requested}`);
+  }
+  return {
+    delivery: requested,
+    support,
+    waitForIdle: requested === "when_idle" && support === "approximate"
+  };
+}
+function planTruncation(truncation, heard) {
+  if (truncation === "client" && heard !== void 0) {
+    return { item: { itemId: heard.itemId, audioEndMs: heard.heardMs } };
+  }
+  return {};
+}
+function realtimeSocket(url, options, protocols = []) {
+  const gateway = options?.gateway;
+  if (gateway === void 0) return { url, protocols: [...protocols] };
+  return {
+    url: gateway.url,
+    protocols: [`cf-aig-authorization.${gateway.token}`, ...protocols]
+  };
+}
+class RealtimeLedger {
+  #inputSeconds = 0;
+  #outputSeconds = 0;
+  #textMessages = 0;
+  #busy = false;
+  #held = [];
+  get inputSeconds() {
+    return this.#inputSeconds;
+  }
+  get outputSeconds() {
+    return this.#outputSeconds;
+  }
+  get textMessages() {
+    return this.#textMessages;
+  }
+  get busy() {
+    return this.#busy;
+  }
+  addInput(chunk) {
+    this.#inputSeconds += chunkSeconds(chunk);
+  }
+  addOutput(chunk, alignment) {
+    const seconds = chunkSeconds(chunk);
+    this.#outputSeconds += seconds > 0 ? seconds : alignmentSeconds(alignment);
+  }
+  addText() {
+    this.#textMessages += 1;
+  }
+  observe(event) {
+    if (event.type === "output_audio") this.addOutput(event.chunk, event.alignment);
+    if (event.type === "output_audio" || event.type === "output_transcript") this.#busy = true;
+    if (event.type === "response_done") this.#release();
+  }
+  submit(delivery, requested, run) {
+    const plan = planDelivery(delivery, requested);
+    if (plan.waitForIdle && this.#busy) this.#held.push(run);
+    else run();
+    return plan;
+  }
+  settle() {
+    this.#release();
+  }
+  usage(rates) {
+    return priceAudio(rates, {
+      inputSeconds: this.#inputSeconds,
+      outputSeconds: this.#outputSeconds,
+      textMessages: this.#textMessages
+    });
+  }
+  #release() {
+    this.#busy = false;
+    const held = this.#held;
+    this.#held = [];
+    for (const run of held) run();
+  }
+}
+export {
+  RealtimeLedger,
+  planDelivery,
+  planTruncation,
+  realtimeSocket
+};
 //# sourceMappingURL=realtime.js.map

package/dist/realtime.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
-  "sources": [],
-  "sourcesContent": [],
-  "mappings": "",
+  "sources": ["../src/realtime.ts"],
+  "sourcesContent": ["import type {\n  Message,\n  ProviderRequestOptions,\n  Tool,\n  ToolCall,\n  ToolResultMessage,\n  Usage,\n} from '@earendil-works/pi-ai';\nimport {\n  type AudioChunk,\n  alignmentSeconds,\n  type CharAlignment,\n  chunkSeconds,\n  type TextSpan,\n} from './audio.ts';\nimport { VoiceRequestError } from './errors.ts';\nimport type { Delivery, DeliverySupport, RealtimeModel } from './models.ts';\nimport { priceAudio, type RealtimeAudioCost } from './usage.ts';\n\nexport interface RealtimeContext {\n  instructions: string;\n  tools: Tool[];\n  voice: string;\n  history?: Message[];\n}\n\nexport interface RealtimeToken {\n  value: string;\n  expiresAt: number;\n  model: string;\n}\n\nexport interface RealtimeOptions extends ProviderRequestOptions {\n  token?: RealtimeToken;\n  transport?: 'websocket' | 'webrtc';\n  gateway?: { url: string; token: string };\n  resumeHandle?: string;\n}\n\nexport type RealtimeEvent =\n  | { type: 'session_start'; sessionId: string }\n  | { type: 'input_speech_start' }\n  | { type: 'input_transcript'; itemId: string; text: string; final: boolean }\n  | {\n      type: 'output_audio';\n      itemId: string;\n      chunk: AudioChunk;\n      span?: TextSpan;\n      alignment?: CharAlignment;\n    }\n  | { type: 'output_transcript'; itemId: string; text: string; final: boolean }\n  | { type: 'tool_call'; call: ToolCall; blocking: boolean }\n  | { type: 'tool_call_cancelled'; callId: string }\n  | { type: 'response_done'; itemId: string; interrupted: boolean; usage: Usage }\n  | { type: 'go_away'; timeLeftMs: number }\n  | { type: 'resumption'; handle: string }\n  | { type: 'error'; error: string; fatal: boolean };\n\nexport interface RealtimeSession extends AsyncIterable<RealtimeEvent> {\n  readonly model: RealtimeModel<string>;\n  appendAudio(chunk: AudioChunk): void;\n  appendVideoFrame?(jpeg: Uint8Array): void;\n  sendText(text: string): void;\n  inject(text: string, delivery: Delivery): void;\n  respond(callId: string, result: ToolResultMessage, delivery?: Delivery): void;\n  interrupt(heard?: { itemId: string; heardMs: number }): void;\n  update(patch: Partial<Pick<RealtimeContext, 'instructions' | 'tools' | 'voice'>>): void;\n  close(): Promise<{ usage: Usage }>;\n}\n\nexport interface RealtimeApiProvider<TApi extends string = string> {\n  api: TApi;\n  connect: (\n    model: RealtimeModel<TApi>,\n    context: RealtimeContext,\n    options?: RealtimeOptions\n  ) => Promise<RealtimeSession>;\n}\n\nexport interface DeliveryPlan {\n  delivery: Delivery;\n  support: DeliverySupport;\n  waitForIdle: boolean;\n}\n\nexport interface TruncationPlan {\n  item?: { itemId: string; audioEndMs: number };\n}\n\nexport interface RealtimeSocket {\n  url: string;\n  protocols: string[];\n}\n\nexport function planDelivery(\n  delivery: Record<Delivery, DeliverySupport>,\n  requested: Delivery\n): DeliveryPlan {\n  const support = delivery[requested];\n  if (support !== 'native' && support !== 'approximate') {\n    throw new VoiceRequestError('error', `unsupported delivery: ${requested}`);\n  }\n  return {\n    delivery: requested,\n    support,\n    waitForIdle: requested === 'when_idle' && support === 'approximate',\n  };\n}\n\nexport function planTruncation(\n  truncation: RealtimeModel<string>['truncation'],\n  heard?: { itemId: string; heardMs: number }\n): TruncationPlan {\n  if (truncation === 'client' && heard !== undefined) {\n    return { item: { itemId: heard.itemId, audioEndMs: heard.heardMs } };\n  }\n  return {};\n}\n\nexport function realtimeSocket(\n  url: string,\n  options?: Pick<RealtimeOptions, 'gateway'>,\n  protocols: readonly string[] = []\n): RealtimeSocket {\n  const gateway = options?.gateway;\n  if (gateway === undefined) return { url, protocols: [...protocols] };\n  return {\n    url: gateway.url,\n    protocols: [`cf-aig-authorization.${gateway.token}`, ...protocols],\n  };\n}\n\nexport class RealtimeLedger {\n  #inputSeconds = 0;\n  #outputSeconds = 0;\n  #textMessages = 0;\n  #busy = false;\n  #held: Array<() => void> = [];\n\n  get inputSeconds(): number {\n    return this.#inputSeconds;\n  }\n\n  get outputSeconds(): number {\n    return this.#outputSeconds;\n  }\n\n  get textMessages(): number {\n    return this.#textMessages;\n  }\n\n  get busy(): boolean {\n    return this.#busy;\n  }\n\n  addInput(chunk: AudioChunk): void {\n    this.#inputSeconds += chunkSeconds(chunk);\n  }\n\n  addOutput(chunk: AudioChunk, alignment?: CharAlignment): void {\n    const seconds = chunkSeconds(chunk);\n    this.#outputSeconds += seconds > 0 ? seconds : alignmentSeconds(alignment);\n  }\n\n  addText(): void {\n    this.#textMessages += 1;\n  }\n\n  observe(event: RealtimeEvent): void {\n    if (event.type === 'output_audio') this.addOutput(event.chunk, event.alignment);\n    if (event.type === 'output_audio' || event.type === 'output_transcript') this.#busy = true;\n    if (event.type === 'response_done') this.#release();\n  }\n\n  submit(\n    delivery: Record<Delivery, DeliverySupport>,\n    requested: Delivery,\n    run: () => void\n  ): DeliveryPlan {\n    const plan = planDelivery(delivery, requested);\n    if (plan.waitForIdle && this.#busy) this.#held.push(run);\n    else run();\n    return plan;\n  }\n\n  settle(): void {\n    this.#release();\n  }\n\n  usage(rates: RealtimeAudioCost): Usage {\n    return priceAudio(rates, {\n      inputSeconds: this.#inputSeconds,\n      outputSeconds: this.#outputSeconds,\n      textMessages: this.#textMessages,\n    });\n  }\n\n  #release(): void {\n    this.#busy = false;\n    const held = this.#held;\n    this.#held = [];\n    for (const run of held) run();\n  }\n}\n"],
+  "mappings": "AAQA;AAAA,EAEE;AAAA,EAEA;AAAA,OAEK;AACP,SAAS,yBAAyB;AAElC,SAAS,kBAA0C;AA6E5C,SAAS,aACd,UACA,WACc;AACd,QAAM,UAAU,SAAS,SAAS;AAClC,MAAI,YAAY,YAAY,YAAY,eAAe;AACrD,UAAM,IAAI,kBAAkB,SAAS,yBAAyB,SAAS,EAAE;AAAA,EAC3E;AACA,SAAO;AAAA,IACL,UAAU;AAAA,IACV;AAAA,IACA,aAAa,cAAc,eAAe,YAAY;AAAA,EACxD;AACF;AAEO,SAAS,eACd,YACA,OACgB;AAChB,MAAI,eAAe,YAAY,UAAU,QAAW;AAClD,WAAO,EAAE,MAAM,EAAE,QAAQ,MAAM,QAAQ,YAAY,MAAM,QAAQ,EAAE;AAAA,EACrE;AACA,SAAO,CAAC;AACV;AAEO,SAAS,eACd,KACA,SACA,YAA+B,CAAC,GAChB;AAChB,QAAM,UAAU,SAAS;AACzB,MAAI,YAAY,OAAW,QAAO,EAAE,KAAK,WAAW,CAAC,GAAG,SAAS,EAAE;AACnE,SAAO;AAAA,IACL,KAAK,QAAQ;AAAA,IACb,WAAW,CAAC,wBAAwB,QAAQ,KAAK,IAAI,GAAG,SAAS;AAAA,EACnE;AACF;AAEO,MAAM,eAAe;AAAA,EAC1B,gBAAgB;AAAA,EAChB,iBAAiB;AAAA,EACjB,gBAAgB;AAAA,EAChB,QAAQ;AAAA,EACR,QAA2B,CAAC;AAAA,EAE5B,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,gBAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAgB;AAClB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,SAAS,OAAyB;AAChC,SAAK,iBAAiB,aAAa,KAAK;AAAA,EAC1C;AAAA,EAEA,UAAU,OAAmB,WAAiC;AAC5D,UAAM,UAAU,aAAa,KAAK;AAClC,SAAK,kBAAkB,UAAU,IAAI,UAAU,iBAAiB,SAAS;AAAA,EAC3E;AAAA,EAEA,UAAgB;AACd,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEA,QAAQ,OAA4B;AAClC,QAAI,MAAM,SAAS,eAAgB,MAAK,UAAU,MAAM,OAAO,MAAM,SAAS;AAC9E,QAAI,MAAM,SAAS,kBAAkB,MAAM,SAAS,oBAAqB,MAAK,QAAQ;AACtF,QAAI,MAAM,SAAS,gBAAiB,MAAK,SAAS;AAAA,EACpD;AAAA,EAEA,OACE,UACA,WACA,KACc;AACd,UAAM,OAAO,aAAa,UAAU,SAAS;AAC7C,QAAI,KAAK,eAAe,KAAK,MAAO,MAAK,MAAM,KAAK,GAAG;AAAA,QAClD,KAAI;AACT,WAAO;AAAA,EACT;AAAA,EAEA,SAAe;AACb,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,OAAiC;AACrC,WAAO,WAAW,OAAO;AAAA,MACvB,cAAc,KAAK;AAAA,MACnB,eAAe,KAAK;AAAA,MACpB,cAAc,KAAK;AAAA,IACrB,CAAC;AAAA,EACH;AAAA,EAEA,WAAiB;AACf,SAAK,QAAQ;AACb,UAAM,OAAO,KAAK;AAClB,SAAK,QAAQ,CAAC;AACd,eAAW,OAAO,KAAM,KAAI;AAAA,EAC9B;AACF;",
   "names": []
 }

package/dist/registry.js

@@ -1,4 +1,5 @@
 import { speakInSentences } from "./chunk-speech.js";
+import { VoiceRequestError } from "./errors.js";
 import { listenWithVad } from "./vad-listen.js";
 const transcriptionApis = /* @__PURE__ */ new Map();
 const speechApis = /* @__PURE__ */ new Map();
@@ -45,6 +46,14 @@ function speak(model, input, options) {
   return speakModel(model, input, options);
 }
 function connectRealtime(model, context, options) {
+  if (options?.transport !== void 0 && !model.transports.includes(options.transport)) {
+    return Promise.reject(
+      new VoiceRequestError("error", `unsupported transport: ${options.transport}`)
+    );
+  }
+  if (options?.resumeHandle !== void 0 && !model.resumable) {
+    return Promise.reject(new VoiceRequestError("error", "session is not resumable"));
+  }
   const connect = realtimeApis.get(model.api)?.provider.connect;
   if (!connect) return Promise.reject(missing(model.api));
   return connect(model, context, options);

package/dist/registry.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["../src/registry.ts"],
-  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport type { AudioSource } from './audio.ts';\nimport { speakInSentences } from './chunk-speech.ts';\nimport type { RealtimeModel, SpeechModel, TranscriptionModel } from './models.ts';\nimport type {\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeOptions,\n  RealtimeSession,\n} from './realtime.ts';\nimport type { SpeechApiProvider, SpeechEventStream, SpeechInput } from './speech.ts';\nimport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEventStream,\n} from './transcription.ts';\nimport { listenWithVad } from './vad-listen.ts';\n\ninterface TranscriptionEntry {\n  provider: TranscriptionApiProvider;\n}\n\ninterface SpeechEntry {\n  provider: SpeechApiProvider;\n}\n\ninterface RealtimeEntry {\n  provider: RealtimeApiProvider;\n}\n\nconst transcriptionApis = new Map<string, TranscriptionEntry>();\nconst speechApis = new Map<string, SpeechEntry>();\nconst realtimeApis = new Map<string, RealtimeEntry>();\n\nfunction missing(api: string): Error {\n  return new Error(`No API provider registered for api: ${api}`);\n}\n\nexport function registerTranscriptionApiProvider(provider: TranscriptionApiProvider): () => void {\n  const entry = { provider };\n  transcriptionApis.set(provider.api, entry);\n  return () => {\n    if (transcriptionApis.get(provider.api) === entry) transcriptionApis.delete(provider.api);\n  };\n}\n\nexport function registerSpeechApiProvider(provider: SpeechApiProvider): () => void {\n  const entry = { provider };\n  speechApis.set(provider.api, entry);\n  return () => {\n    if (speechApis.get(provider.api) === entry) speechApis.delete(provider.api);\n  };\n}\n\nexport function registerRealtimeApiProvider(provider: RealtimeApiProvider): () => void {\n  const entry = { provider };\n  realtimeApis.set(provider.api, entry);\n  return () => {\n    if (realtimeApis.get(provider.api) === entry) realtimeApis.delete(provider.api);\n  };\n}\n\nexport function transcribe(\n  model: TranscriptionModel<string>,\n  input: AudioInput,\n  options?: ProviderRequestOptions\n): Promise<Transcript> {\n  const transcribeModel = transcriptionApis.get(model.api)?.provider.transcribe;\n  if (!transcribeModel) return Promise.reject(missing(model.api));\n  return transcribeModel(model, input, options);\n}\n\nexport function listen(\n  model: TranscriptionModel<string>,\n  audio: AudioSource,\n  options?: ListenOptions\n): TranscriptionEventStream {\n  const listenModel = transcriptionApis.get(model.api)?.provider.listen;\n  if (!listenModel) throw missing(model.api);\n  if (model.turnDetection === 'none') return listenWithVad(listenModel, model, audio, options);\n  return listenModel(model, audio, options);\n}\n\nexport function speak(\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options?: ProviderRequestOptions\n): SpeechEventStream {\n  const speakModel = speechApis.get(model.api)?.provider.speak;\n  if (!speakModel) throw missing(model.api);\n  if (!model.textStreaming) return speakInSentences(speakModel, model, input, options);\n  return speakModel(model, input, options);\n}\n\nexport function connectRealtime(\n  model: RealtimeModel<string>,\n  context: RealtimeContext,\n  options?: RealtimeOptions\n): Promise<RealtimeSession> {\n  const connect = realtimeApis.get(model.api)?.provider.connect;\n  if (!connect) return Promise.reject(missing(model.api));\n  return connect(model, context, options);\n}\n"],
-  "mappings": "AAEA,SAAS,wBAAwB;AAgBjC,SAAS,qBAAqB;AAc9B,MAAM,oBAAoB,oBAAI,IAAgC;AAC9D,MAAM,aAAa,oBAAI,IAAyB;AAChD,MAAM,eAAe,oBAAI,IAA2B;AAEpD,SAAS,QAAQ,KAAoB;AACnC,SAAO,IAAI,MAAM,uCAAuC,GAAG,EAAE;AAC/D;AAEO,SAAS,iCAAiC,UAAgD;AAC/F,QAAM,QAAQ,EAAE,SAAS;AACzB,oBAAkB,IAAI,SAAS,KAAK,KAAK;AACzC,SAAO,MAAM;AACX,QAAI,kBAAkB,IAAI,SAAS,GAAG,MAAM,MAAO,mBAAkB,OAAO,SAAS,GAAG;AAAA,EAC1F;AACF;AAEO,SAAS,0BAA0B,UAAyC;AACjF,QAAM,QAAQ,EAAE,SAAS;AACzB,aAAW,IAAI,SAAS,KAAK,KAAK;AAClC,SAAO,MAAM;AACX,QAAI,WAAW,IAAI,SAAS,GAAG,MAAM,MAAO,YAAW,OAAO,SAAS,GAAG;AAAA,EAC5E;AACF;AAEO,SAAS,4BAA4B,UAA2C;AACrF,QAAM,QAAQ,EAAE,SAAS;AACzB,eAAa,IAAI,SAAS,KAAK,KAAK;AACpC,SAAO,MAAM;AACX,QAAI,aAAa,IAAI,SAAS,GAAG,MAAM,MAAO,cAAa,OAAO,SAAS,GAAG;AAAA,EAChF;AACF;AAEO,SAAS,WACd,OACA,OACA,SACqB;AACrB,QAAM,kBAAkB,kBAAkB,IAAI,MAAM,GAAG,GAAG,SAAS;AACnE,MAAI,CAAC,gBAAiB,QAAO,QAAQ,OAAO,QAAQ,MAAM,GAAG,CAAC;AAC9D,SAAO,gBAAgB,OAAO,OAAO,OAAO;AAC9C;AAEO,SAAS,OACd,OACA,OACA,SAC0B;AAC1B,QAAM,cAAc,kBAAkB,IAAI,MAAM,GAAG,GAAG,SAAS;AAC/D,MAAI,CAAC,YAAa,OAAM,QAAQ,MAAM,GAAG;AACzC,MAAI,MAAM,kBAAkB,OAAQ,QAAO,cAAc,aAAa,OAAO,OAAO,OAAO;AAC3F,SAAO,YAAY,OAAO,OAAO,OAAO;AAC1C;AAEO,SAAS,MACd,OACA,OACA,SACmB;AACnB,QAAM,aAAa,WAAW,IAAI,MAAM,GAAG,GAAG,SAAS;AACvD,MAAI,CAAC,WAAY,OAAM,QAAQ,MAAM,GAAG;AACxC,MAAI,CAAC,MAAM,cAAe,QAAO,iBAAiB,YAAY,OAAO,OAAO,OAAO;AACnF,SAAO,WAAW,OAAO,OAAO,OAAO;AACzC;AAEO,SAAS,gBACd,OACA,SACA,SAC0B;AAC1B,QAAM,UAAU,aAAa,IAAI,MAAM,GAAG,GAAG,SAAS;AACtD,MAAI,CAAC,QAAS,QAAO,QAAQ,OAAO,QAAQ,MAAM,GAAG,CAAC;AACtD,SAAO,QAAQ,OAAO,SAAS,OAAO;AACxC;",
+  "sourcesContent": ["import type { ProviderRequestOptions } from '@earendil-works/pi-ai';\nimport type { AudioSource } from './audio.ts';\nimport { speakInSentences } from './chunk-speech.ts';\nimport { VoiceRequestError } from './errors.ts';\nimport type { RealtimeModel, SpeechModel, TranscriptionModel } from './models.ts';\nimport type {\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeOptions,\n  RealtimeSession,\n} from './realtime.ts';\nimport type { SpeechApiProvider, SpeechEventStream, SpeechInput } from './speech.ts';\nimport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEventStream,\n} from './transcription.ts';\nimport { listenWithVad } from './vad-listen.ts';\n\ninterface TranscriptionEntry {\n  provider: TranscriptionApiProvider;\n}\n\ninterface SpeechEntry {\n  provider: SpeechApiProvider;\n}\n\ninterface RealtimeEntry {\n  provider: RealtimeApiProvider;\n}\n\nconst transcriptionApis = new Map<string, TranscriptionEntry>();\nconst speechApis = new Map<string, SpeechEntry>();\nconst realtimeApis = new Map<string, RealtimeEntry>();\n\nfunction missing(api: string): Error {\n  return new Error(`No API provider registered for api: ${api}`);\n}\n\nexport function registerTranscriptionApiProvider(provider: TranscriptionApiProvider): () => void {\n  const entry = { provider };\n  transcriptionApis.set(provider.api, entry);\n  return () => {\n    if (transcriptionApis.get(provider.api) === entry) transcriptionApis.delete(provider.api);\n  };\n}\n\nexport function registerSpeechApiProvider(provider: SpeechApiProvider): () => void {\n  const entry = { provider };\n  speechApis.set(provider.api, entry);\n  return () => {\n    if (speechApis.get(provider.api) === entry) speechApis.delete(provider.api);\n  };\n}\n\nexport function registerRealtimeApiProvider(provider: RealtimeApiProvider): () => void {\n  const entry = { provider };\n  realtimeApis.set(provider.api, entry);\n  return () => {\n    if (realtimeApis.get(provider.api) === entry) realtimeApis.delete(provider.api);\n  };\n}\n\nexport function transcribe(\n  model: TranscriptionModel<string>,\n  input: AudioInput,\n  options?: ProviderRequestOptions\n): Promise<Transcript> {\n  const transcribeModel = transcriptionApis.get(model.api)?.provider.transcribe;\n  if (!transcribeModel) return Promise.reject(missing(model.api));\n  return transcribeModel(model, input, options);\n}\n\nexport function listen(\n  model: TranscriptionModel<string>,\n  audio: AudioSource,\n  options?: ListenOptions\n): TranscriptionEventStream {\n  const listenModel = transcriptionApis.get(model.api)?.provider.listen;\n  if (!listenModel) throw missing(model.api);\n  if (model.turnDetection === 'none') return listenWithVad(listenModel, model, audio, options);\n  return listenModel(model, audio, options);\n}\n\nexport function speak(\n  model: SpeechModel<string>,\n  input: SpeechInput,\n  options?: ProviderRequestOptions\n): SpeechEventStream {\n  const speakModel = speechApis.get(model.api)?.provider.speak;\n  if (!speakModel) throw missing(model.api);\n  if (!model.textStreaming) return speakInSentences(speakModel, model, input, options);\n  return speakModel(model, input, options);\n}\n\nexport function connectRealtime(\n  model: RealtimeModel<string>,\n  context: RealtimeContext,\n  options?: RealtimeOptions\n): Promise<RealtimeSession> {\n  if (options?.transport !== undefined && !model.transports.includes(options.transport)) {\n    return Promise.reject(\n      new VoiceRequestError('error', `unsupported transport: ${options.transport}`)\n    );\n  }\n  if (options?.resumeHandle !== undefined && !model.resumable) {\n    return Promise.reject(new VoiceRequestError('error', 'session is not resumable'));\n  }\n  const connect = realtimeApis.get(model.api)?.provider.connect;\n  if (!connect) return Promise.reject(missing(model.api));\n  return connect(model, context, options);\n}\n"],
+  "mappings": "AAEA,SAAS,wBAAwB;AACjC,SAAS,yBAAyB;AAgBlC,SAAS,qBAAqB;AAc9B,MAAM,oBAAoB,oBAAI,IAAgC;AAC9D,MAAM,aAAa,oBAAI,IAAyB;AAChD,MAAM,eAAe,oBAAI,IAA2B;AAEpD,SAAS,QAAQ,KAAoB;AACnC,SAAO,IAAI,MAAM,uCAAuC,GAAG,EAAE;AAC/D;AAEO,SAAS,iCAAiC,UAAgD;AAC/F,QAAM,QAAQ,EAAE,SAAS;AACzB,oBAAkB,IAAI,SAAS,KAAK,KAAK;AACzC,SAAO,MAAM;AACX,QAAI,kBAAkB,IAAI,SAAS,GAAG,MAAM,MAAO,mBAAkB,OAAO,SAAS,GAAG;AAAA,EAC1F;AACF;AAEO,SAAS,0BAA0B,UAAyC;AACjF,QAAM,QAAQ,EAAE,SAAS;AACzB,aAAW,IAAI,SAAS,KAAK,KAAK;AAClC,SAAO,MAAM;AACX,QAAI,WAAW,IAAI,SAAS,GAAG,MAAM,MAAO,YAAW,OAAO,SAAS,GAAG;AAAA,EAC5E;AACF;AAEO,SAAS,4BAA4B,UAA2C;AACrF,QAAM,QAAQ,EAAE,SAAS;AACzB,eAAa,IAAI,SAAS,KAAK,KAAK;AACpC,SAAO,MAAM;AACX,QAAI,aAAa,IAAI,SAAS,GAAG,MAAM,MAAO,cAAa,OAAO,SAAS,GAAG;AAAA,EAChF;AACF;AAEO,SAAS,WACd,OACA,OACA,SACqB;AACrB,QAAM,kBAAkB,kBAAkB,IAAI,MAAM,GAAG,GAAG,SAAS;AACnE,MAAI,CAAC,gBAAiB,QAAO,QAAQ,OAAO,QAAQ,MAAM,GAAG,CAAC;AAC9D,SAAO,gBAAgB,OAAO,OAAO,OAAO;AAC9C;AAEO,SAAS,OACd,OACA,OACA,SAC0B;AAC1B,QAAM,cAAc,kBAAkB,IAAI,MAAM,GAAG,GAAG,SAAS;AAC/D,MAAI,CAAC,YAAa,OAAM,QAAQ,MAAM,GAAG;AACzC,MAAI,MAAM,kBAAkB,OAAQ,QAAO,cAAc,aAAa,OAAO,OAAO,OAAO;AAC3F,SAAO,YAAY,OAAO,OAAO,OAAO;AAC1C;AAEO,SAAS,MACd,OACA,OACA,SACmB;AACnB,QAAM,aAAa,WAAW,IAAI,MAAM,GAAG,GAAG,SAAS;AACvD,MAAI,CAAC,WAAY,OAAM,QAAQ,MAAM,GAAG;AACxC,MAAI,CAAC,MAAM,cAAe,QAAO,iBAAiB,YAAY,OAAO,OAAO,OAAO;AACnF,SAAO,WAAW,OAAO,OAAO,OAAO;AACzC;AAEO,SAAS,gBACd,OACA,SACA,SAC0B;AAC1B,MAAI,SAAS,cAAc,UAAa,CAAC,MAAM,WAAW,SAAS,QAAQ,SAAS,GAAG;AACrF,WAAO,QAAQ;AAAA,MACb,IAAI,kBAAkB,SAAS,0BAA0B,QAAQ,SAAS,EAAE;AAAA,IAC9E;AAAA,EACF;AACA,MAAI,SAAS,iBAAiB,UAAa,CAAC,MAAM,WAAW;AAC3D,WAAO,QAAQ,OAAO,IAAI,kBAAkB,SAAS,0BAA0B,CAAC;AAAA,EAClF;AACA,QAAM,UAAU,aAAa,IAAI,MAAM,GAAG,GAAG,SAAS;AACtD,MAAI,CAAC,QAAS,QAAO,QAAQ,OAAO,QAAQ,MAAM,GAAG,CAAC;AACtD,SAAO,QAAQ,OAAO,SAAS,OAAO;AACxC;",
   "names": []
 }

package/dist/types/audio.d.ts

@@ -34,3 +34,5 @@ export interface AudioSink {
     readonly playedMs: number;
     onPlayback(listener: (tag: AudioTag, offsetMs: number) => void): () => void;
 }
+export declare function chunkSeconds(chunk: AudioChunk): number;
+export declare function alignmentSeconds(alignment: CharAlignment | undefined): number;

package/dist/types/faux/realtime.d.ts

@@ -1,7 +1,7 @@
 import type { AudioChunk } from '../audio.ts';
 import { type Delivery, type DeliverySupport, type RealtimeModel } from '../models.ts';
-import type { RealtimeEvent, RealtimeSession } from '../realtime.ts';
-import { type RealtimeAudioCost } from '../usage.ts';
+import { type RealtimeEvent, type RealtimeSession, type RealtimeSocket, type TruncationPlan } from '../realtime.ts';
+import type { RealtimeAudioCost } from '../usage.ts';
 export type FauxRealtimeCall = {
     method: 'appendAudio';
     chunk: AudioChunk;
@@ -15,17 +15,20 @@ export type FauxRealtimeCall = {
     method: 'inject';
     text: string;
     delivery: Delivery;
+    support: DeliverySupport;
 } | {
     method: 'respond';
     callId: string;
     result: Parameters<RealtimeSession['respond']>[1];
     delivery?: Delivery;
+    support?: DeliverySupport;
 } | {
     method: 'interrupt';
     heard?: {
         itemId: string;
         heardMs: number;
     };
+    plan: TruncationPlan;
 } | {
     method: 'update';
     patch: Parameters<RealtimeSession['update']>[0];
@@ -48,6 +51,7 @@ export interface FauxRealtimeOptions extends FauxRealtimeFlags {
 export interface FauxRealtime {
     model: RealtimeModel<string>;
     calls: FauxRealtimeCall[];
+    readonly socket: RealtimeSocket | undefined;
     emit(event: RealtimeEvent): void;
     setFlags(flags: FauxRealtimeFlags): void;
     unregister(): void;

package/dist/types/index.d.ts

@@ -5,7 +5,8 @@ export { convertAudio, convertAudioSource, decodeAudioBase64, encodeAudioBase64,
 export { VoiceRequestError } from './errors.ts';
 export type { Delivery, DeliverySupport, KnownRealtimeApi, KnownSpeechApi, KnownTranscriptionApi, RealtimeApi, RealtimeModel, SpeechApi, SpeechModel, TranscriptionApi, TranscriptionModel, VoiceInfo, VoiceModel, VoiceModelType, } from './models.ts';
 export { getVoiceModel, listVoiceModels, registerVoiceModel, } from './models.ts';
-export type { RealtimeApiProvider, RealtimeContext, RealtimeEvent, RealtimeOptions, RealtimeSession, RealtimeToken, } from './realtime.ts';
+export type { DeliveryPlan, RealtimeApiProvider, RealtimeContext, RealtimeEvent, RealtimeOptions, RealtimeSession, RealtimeSocket, RealtimeToken, TruncationPlan, } from './realtime.ts';
+export { planDelivery, planTruncation, RealtimeLedger, realtimeSocket } from './realtime.ts';
 export { connectRealtime, listen, registerRealtimeApiProvider, registerSpeechApiProvider, registerTranscriptionApiProvider, speak, transcribe, } from './registry.ts';
 export type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';
 export { SpeechEventStream } from './speech.ts';

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

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

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

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

package/dist/types/openai/listen.d.ts

@@ -0,0 +1,4 @@
+import type { AudioChunk } from '../audio.ts';
+import type { TranscriptionModel } from '../models.ts';
+import { type ListenOptions, TranscriptionEventStream } from '../transcription.ts';
+export declare function listenOpenAI(model: TranscriptionModel<string>, audio: AsyncIterable<AudioChunk>, options?: ListenOptions): TranscriptionEventStream;

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

package/dist/types/openai/transcribe.d.ts

@@ -0,0 +1,4 @@
+import type { ProviderRequestOptions } from '@earendil-works/pi-ai';
+import type { TranscriptionModel } from '../models.ts';
+import type { AudioInput, Transcript } from '../transcription.ts';
+export declare function transcribeOpenAI(model: TranscriptionModel<string>, input: AudioInput, options?: ProviderRequestOptions): Promise<Transcript>;

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

@@ -0,0 +1,52 @@
+import type { ProviderRequestOptions } from '@earendil-works/pi-ai';
+export interface AudioFile {
+    data: Uint8Array;
+    mimeType: string;
+    filename: string;
+    format: string;
+}
+export interface JsonBody {
+    [key: string]: unknown;
+}
+interface HeaderModel {
+    provider: string;
+    headers?: Record<string, string>;
+}
+export declare function isOpenRouter(model: {
+    provider: string;
+    baseUrl: string;
+}): boolean;
+export declare function endpoint(baseUrl: string, path: string): string;
+export declare function realtimeUrl(baseUrl: string): string;
+export declare function audioFile(mimeType: string): {
+    format: string;
+    filename: string;
+};
+export declare function isRecord(value: unknown): value is Record<string, unknown>;
+export declare function stringField(record: Record<string, unknown>, key: string): string | undefined;
+export declare function numberField(record: Record<string, unknown>, key: string): number | undefined;
+export declare function asVoiceError(error: unknown): {
+    reason: 'aborted' | 'error';
+    error: string;
+};
+export declare function hasAuth(headers: Record<string, string | null> | undefined): boolean;
+export declare function requestHeaders(model: HeaderModel, options: ProviderRequestOptions | undefined): Record<string, string>;
+export declare function requestSignal(options: ProviderRequestOptions | undefined): AbortSignal | undefined;
+export declare function encodeMultipart(file: AudioFile, fields: Record<string, string>): Uint8Array;
+export declare function multipartType(): string;
+export declare function readAudio(input: {
+    data: Uint8Array;
+    mimeType: string;
+} | {
+    url: string;
+}, fetchImpl: typeof fetch, signal: AbortSignal | undefined): Promise<{
+    data: Uint8Array;
+    mimeType: string;
+}>;
+export declare function audioPayload(file: AudioFile, openRouter: boolean, modelId: string): JsonBody;
+export declare function applyPayload(payload: unknown, model: unknown, options: ProviderRequestOptions | undefined): Promise<unknown>;
+export declare function voiceFetch(model: HeaderModel & {
+    id: string;
+}, options: ProviderRequestOptions | undefined, url: string, body: BodyInit, contentType: string): Promise<Response>;
+export declare function readJson(response: Response): Promise<unknown>;
+export {};

package/dist/types/realtime.d.ts

@@ -1,6 +1,7 @@
 import type { Message, ProviderRequestOptions, Tool, ToolCall, ToolResultMessage, Usage } from '@earendil-works/pi-ai';
-import type { AudioChunk, CharAlignment, TextSpan } from './audio.ts';
-import type { Delivery, RealtimeModel } from './models.ts';
+import { type AudioChunk, type CharAlignment, type TextSpan } from './audio.ts';
+import type { Delivery, DeliverySupport, RealtimeModel } from './models.ts';
+import { type RealtimeAudioCost } from './usage.ts';
 export interface RealtimeContext {
     instructions: string;
     tools: Tool[];
@@ -85,3 +86,38 @@ export interface RealtimeApiProvider<TApi extends string = string> {
     api: TApi;
     connect: (model: RealtimeModel<TApi>, context: RealtimeContext, options?: RealtimeOptions) => Promise<RealtimeSession>;
 }
+export interface DeliveryPlan {
+    delivery: Delivery;
+    support: DeliverySupport;
+    waitForIdle: boolean;
+}
+export interface TruncationPlan {
+    item?: {
+        itemId: string;
+        audioEndMs: number;
+    };
+}
+export interface RealtimeSocket {
+    url: string;
+    protocols: string[];
+}
+export declare function planDelivery(delivery: Record<Delivery, DeliverySupport>, requested: Delivery): DeliveryPlan;
+export declare function planTruncation(truncation: RealtimeModel<string>['truncation'], heard?: {
+    itemId: string;
+    heardMs: number;
+}): TruncationPlan;
+export declare function realtimeSocket(url: string, options?: Pick<RealtimeOptions, 'gateway'>, protocols?: readonly string[]): RealtimeSocket;
+export declare class RealtimeLedger {
+    #private;
+    get inputSeconds(): number;
+    get outputSeconds(): number;
+    get textMessages(): number;
+    get busy(): boolean;
+    addInput(chunk: AudioChunk): void;
+    addOutput(chunk: AudioChunk, alignment?: CharAlignment): void;
+    addText(): void;
+    observe(event: RealtimeEvent): void;
+    submit(delivery: Record<Delivery, DeliverySupport>, requested: Delivery, run: () => void): DeliveryPlan;
+    settle(): void;
+    usage(rates: RealtimeAudioCost): Usage;
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-ecoverse/pi-audio",
-  "version": "1.4.0",
+  "version": "1.6.0",
   "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",
@@ -33,6 +33,10 @@
     "./worklet": {
       "types": "./dist/types/worklet/convert.d.ts",
       "default": "./dist/worklet/convert.js"
+    },
+    "./openai": {
+      "types": "./dist/types/openai/index.d.ts",
+      "default": "./dist/openai/index.js"
     }
   },
   "files": [