@ai-ecoverse/pi-audio 1.10.0 → 1.12.0

package/dist/index.js

@@ -24,6 +24,7 @@ import {
 } from "./registry.js";
 import { SpeechEventStream } from "./speech.js";
 import { askAgent, bindThinker, talkerTools } from "./talker.js";
+import { mintRealtimeToken } from "./tokens.js";
 import { TranscriptionEventStream } from "./transcription.js";
 import { voiceTurns } from "./turns.js";
 import { addAudioCounters, priceAudio } from "./usage.js";
@@ -46,6 +47,7 @@ export {
   getVoiceModel,
   listVoiceModels,
   listen,
+  mintRealtimeToken,
   planDelivery,
   planTruncation,
   priceAudio,

package/dist/index.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["../src/index.ts"],
-  "sourcesContent": ["export type {\n  AudioChunk,\n  AudioEncoding,\n  AudioFormat,\n  AudioSink,\n  AudioSource,\n  AudioTag,\n  CharAlignment,\n  TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport {\n  convertAudio,\n  convertAudioSource,\n  decodeAudioBase64,\n  encodeAudioBase64,\n} from './convert.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n  Delivery,\n  DeliverySupport,\n  KnownRealtimeApi,\n  KnownSpeechApi,\n  KnownTranscriptionApi,\n  RealtimeApi,\n  RealtimeModel,\n  SpeechApi,\n  SpeechModel,\n  TranscriptionApi,\n  TranscriptionModel,\n  VoiceInfo,\n  VoiceModel,\n  VoiceModelType,\n} from './models.ts';\nexport {\n  getVoiceModel,\n  listVoiceModels,\n  registerVoiceModel,\n} from './models.ts';\nexport type {\n  DeliveryPlan,\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeEvent,\n  RealtimeOptions,\n  RealtimeSession,\n  RealtimeSocket,\n  RealtimeToken,\n  TruncationPlan,\n} from './realtime.ts';\nexport { planDelivery, planTruncation, RealtimeLedger, realtimeSocket } from './realtime.ts';\nexport {\n  connectRealtime,\n  listen,\n  registerRealtimeApiProvider,\n  registerSpeechApiProvider,\n  registerTranscriptionApiProvider,\n  speak,\n  transcribe,\n} from './registry.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type { VoiceAgentEvent, VoiceMessageChange } from './spoken.ts';\nexport type {\n  BindThinkerOptions,\n  BoundThinker,\n  Thinker,\n  ThinkerCaption,\n  ThinkerPhase,\n  ThinkerUpdate,\n} from './talker.ts';\nexport { askAgent, bindThinker, talkerTools } from './talker.ts';\nexport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEvent,\n  TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n  BargeIn,\n  CaptionEvent,\n  EchoRecommendation,\n  VoiceConversation,\n  VoiceCut,\n  VoiceModels,\n  VoicePhase,\n  VoiceSession,\n  VoiceTurnMeta,\n  VoiceTurnOptions,\n} from './turns.ts';\nexport { voiceTurns } from './turns.ts';\nexport type {\n  AudioCost,\n  AudioCounters,\n  AudioUsage,\n  RealtimeAudioCost,\n  SpeechAudioCost,\n  TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\nexport type { VadHook, VadUpdate } from './vad.ts';\nexport { energyVad } from './vad.ts';\n"],
-  "mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAC3B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAYP,SAAS,cAAc,gBAAgB,gBAAgB,sBAAsB;AAC7E;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,yBAAyB;AAUlC,SAAS,UAAU,aAAa,mBAAmB;AASnD,SAAS,gCAAgC;AAazC,SAAS,kBAAkB;AAS3B,SAAS,kBAAkB,kBAAkB;AAE7C,SAAS,iBAAiB;",
+  "sourcesContent": ["export type {\n  AudioChunk,\n  AudioEncoding,\n  AudioFormat,\n  AudioSink,\n  AudioSource,\n  AudioTag,\n  CharAlignment,\n  TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport {\n  convertAudio,\n  convertAudioSource,\n  decodeAudioBase64,\n  encodeAudioBase64,\n} from './convert.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n  Delivery,\n  DeliverySupport,\n  KnownRealtimeApi,\n  KnownSpeechApi,\n  KnownTranscriptionApi,\n  RealtimeApi,\n  RealtimeModel,\n  SpeechApi,\n  SpeechModel,\n  TranscriptionApi,\n  TranscriptionModel,\n  VoiceInfo,\n  VoiceModel,\n  VoiceModelType,\n} from './models.ts';\nexport {\n  getVoiceModel,\n  listVoiceModels,\n  registerVoiceModel,\n} from './models.ts';\nexport type {\n  DeliveryPlan,\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeEvent,\n  RealtimeOptions,\n  RealtimeSession,\n  RealtimeSocket,\n  RealtimeToken,\n  TruncationPlan,\n} from './realtime.ts';\nexport { planDelivery, planTruncation, RealtimeLedger, realtimeSocket } from './realtime.ts';\nexport {\n  connectRealtime,\n  listen,\n  registerRealtimeApiProvider,\n  registerSpeechApiProvider,\n  registerTranscriptionApiProvider,\n  speak,\n  transcribe,\n} from './registry.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type { VoiceAgentEvent, VoiceMessageChange } from './spoken.ts';\nexport type {\n  BindThinkerOptions,\n  BoundThinker,\n  Thinker,\n  ThinkerCaption,\n  ThinkerPhase,\n  ThinkerUpdate,\n} from './talker.ts';\nexport { askAgent, bindThinker, talkerTools } from './talker.ts';\nexport type { RealtimeConstraints } from './tokens.ts';\nexport { mintRealtimeToken } from './tokens.ts';\nexport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEvent,\n  TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n  BargeIn,\n  CaptionEvent,\n  EchoRecommendation,\n  VoiceConversation,\n  VoiceCut,\n  VoiceModels,\n  VoicePhase,\n  VoiceSession,\n  VoiceTurnMeta,\n  VoiceTurnOptions,\n} from './turns.ts';\nexport { voiceTurns } from './turns.ts';\nexport type {\n  AudioCost,\n  AudioCounters,\n  AudioUsage,\n  RealtimeAudioCost,\n  SpeechAudioCost,\n  TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\nexport type { VadHook, VadUpdate } from './vad.ts';\nexport { energyVad } from './vad.ts';\n"],
+  "mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAC3B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAYP,SAAS,cAAc,gBAAgB,gBAAgB,sBAAsB;AAC7E;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,yBAAyB;AAUlC,SAAS,UAAU,aAAa,mBAAmB;AAEnD,SAAS,yBAAyB;AASlC,SAAS,gCAAgC;AAazC,SAAS,kBAAkB;AAS3B,SAAS,kBAAkB,kBAAkB;AAE7C,SAAS,iBAAiB;",
   "names": []
 }

package/dist/tokens.js

@@ -0,0 +1,200 @@
+import { VoiceRequestError } from "./errors.js";
+const openaiSecrets = "https://api.openai.com/v1/realtime/client_secrets";
+const xaiSecrets = "https://api.x.ai/v1/realtime/client_secrets";
+const geminiTokens = "https://generativelanguage.googleapis.com/v1beta/auth_tokens";
+const elevenSigned = "https://api.elevenlabs.io/v1/convai/conversation/get-signed-url";
+const elevenWebRtc = "https://api.elevenlabs.io/v1/convai/conversation/token";
+const fishSessions = "https://api.fish.audio/v1/agent/sessions";
+const fifteenMinutes = 15 * 60 * 1e3;
+const thirtyMinutes = 30 * 60 * 1e3;
+const oneMinute = 60 * 1e3;
+const xaiLifetimeSeconds = 300;
+function apiKey(options) {
+  const key = options.apiKey;
+  if (typeof key === "string" && key.length > 0) return key;
+  throw new VoiceRequestError("error", "missing realtime credential");
+}
+function mergedHeaders(options, defaults) {
+  const headers = { ...defaults };
+  const extra = options.headers;
+  if (extra === void 0) return headers;
+  for (const [key, value] of Object.entries(extra)) {
+    if (value === null) delete headers[key];
+    else headers[key] = value;
+  }
+  return headers;
+}
+function omitUndefined(source) {
+  const out = {};
+  for (const [key, value] of Object.entries(source)) {
+    if (value !== void 0) out[key] = value;
+  }
+  return out;
+}
+function functionTool(tool) {
+  return {
+    type: "function",
+    name: tool.name,
+    description: tool.description,
+    parameters: tool.parameters
+  };
+}
+function declaration(tool) {
+  return {
+    name: tool.name,
+    description: tool.description,
+    parameters: tool.parameters
+  };
+}
+function requiredAgent(constraints) {
+  const id = constraints.agentId;
+  if (typeof id === "string" && id.length > 0) return id;
+  throw new VoiceRequestError("error", "missing agent id");
+}
+function expiresAt(value, fallback) {
+  if (typeof value === "number" && Number.isFinite(value)) {
+    return value < 1e12 ? value * 1e3 : value;
+  }
+  if (typeof value === "string") {
+    const parsed = Date.parse(value);
+    if (Number.isFinite(parsed)) return parsed;
+  }
+  return fallback;
+}
+function fromField(field) {
+  return (json, fallback) => expiresAt(json[field], fallback);
+}
+function openaiBody(model, constraints) {
+  const session = omitUndefined({
+    type: "realtime",
+    model: model.id,
+    instructions: constraints.instructions,
+    tools: constraints.tools?.map(functionTool)
+  });
+  if (constraints.voice !== void 0) session.audio = { output: { voice: constraints.voice } };
+  return { session };
+}
+function geminiBody(model, constraints, now) {
+  const instructions = constraints.instructions;
+  const voice = constraints.voice;
+  const tools = constraints.tools;
+  const config = omitUndefined({
+    responseModalities: ["AUDIO"],
+    sessionResumption: {},
+    systemInstruction: instructions === void 0 ? void 0 : { parts: [{ text: instructions }] },
+    speechConfig: voice === void 0 ? void 0 : { voiceConfig: { prebuiltVoiceConfig: { voiceName: voice } } },
+    tools: tools === void 0 ? void 0 : [{ functionDeclarations: tools.map(declaration) }]
+  });
+  return {
+    uses: constraints.uses ?? 1,
+    expireTime: new Date(now + thirtyMinutes).toISOString(),
+    newSessionExpireTime: new Date(now + oneMinute).toISOString(),
+    liveConnectConstraints: { model: `models/${model.id}`, config }
+  };
+}
+function fishBody(constraints) {
+  const overrides = omitUndefined({
+    system_prompt: constraints.instructions,
+    voice_id: constraints.voice
+  });
+  return omitUndefined({
+    agent_id: requiredAgent(constraints),
+    overrides: Object.keys(overrides).length > 0 ? overrides : void 0
+  });
+}
+function webrtcOnly(model) {
+  return model.transports.length === 1 && model.transports[0] === "webrtc";
+}
+function elevenUrl(model, constraints) {
+  const url = new URL(webrtcOnly(model) ? elevenWebRtc : elevenSigned);
+  url.searchParams.set("agent_id", requiredAgent(constraints));
+  return url.toString();
+}
+function mintCall(model, constraints, options, key, now) {
+  const jsonHeaders = { "Content-Type": "application/json" };
+  switch (model.api) {
+    case "openai-realtime":
+      return {
+        url: openaiSecrets,
+        method: "POST",
+        headers: mergedHeaders(options, { ...jsonHeaders, Authorization: `Bearer ${key}` }),
+        body: JSON.stringify(openaiBody(model, constraints)),
+        field: "value",
+        expires: fromField("expires_at"),
+        fallbackExpires: now
+      };
+    case "xai-realtime":
+      return {
+        url: xaiSecrets,
+        method: "POST",
+        headers: mergedHeaders(options, { ...jsonHeaders, Authorization: `Bearer ${key}` }),
+        body: JSON.stringify({ expires_after: { seconds: xaiLifetimeSeconds } }),
+        field: "value",
+        expires: fromField("expires_at"),
+        fallbackExpires: now + xaiLifetimeSeconds * 1e3
+      };
+    case "gemini-live":
+      return {
+        url: geminiTokens,
+        method: "POST",
+        headers: mergedHeaders(options, { ...jsonHeaders, "x-goog-api-key": key }),
+        body: JSON.stringify(geminiBody(model, constraints, now)),
+        field: "name",
+        expires: fromField("expireTime"),
+        fallbackExpires: now + thirtyMinutes
+      };
+    case "elevenlabs-agents":
+      return {
+        url: elevenUrl(model, constraints),
+        method: "GET",
+        headers: mergedHeaders(options, { "xi-api-key": key }),
+        field: webrtcOnly(model) ? "token" : "signed_url",
+        expires: fromField("expires_at"),
+        fallbackExpires: now + fifteenMinutes
+      };
+    case "fish-agents":
+      return {
+        url: fishSessions,
+        method: "POST",
+        headers: mergedHeaders(options, { ...jsonHeaders, Authorization: `Bearer ${key}` }),
+        body: JSON.stringify(fishBody(constraints)),
+        field: "token",
+        expires: fromField("expires_at"),
+        fallbackExpires: now
+      };
+    default:
+      throw new VoiceRequestError("error", "unsupported realtime api");
+  }
+}
+function readValue(json, field) {
+  if (typeof json !== "object" || json === null) {
+    throw new VoiceRequestError("error", "mint failed (empty token)");
+  }
+  const value = json[field];
+  if (typeof value === "string" && value.length > 0) return value;
+  throw new VoiceRequestError("error", "mint failed (empty token)");
+}
+async function mintRealtimeToken(model, constraints, options) {
+  const settings = options ?? {};
+  const key = apiKey(settings);
+  const now = Date.now();
+  const call = mintCall(model, constraints, settings, key, now);
+  const fetch = settings.fetch ?? globalThis.fetch;
+  const response = await fetch(call.url, {
+    method: call.method,
+    headers: call.headers,
+    body: call.body,
+    signal: settings.signal
+  });
+  if (!response.ok) throw new VoiceRequestError("error", `mint failed (${response.status})`);
+  const json = await response.json();
+  return {
+    value: readValue(json, call.field),
+    expiresAt: call.expires(json, call.fallbackExpires),
+    model: model.id
+  };
+}
+export {
+  mintRealtimeToken
+};
+//# sourceMappingURL=tokens.js.map

package/dist/tokens.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../src/tokens.ts"],
+  "sourcesContent": ["import type { ProviderRequestOptions, Tool } from '@earendil-works/pi-ai';\nimport { VoiceRequestError } from './errors.ts';\nimport type { RealtimeModel } from './models.ts';\nimport type { RealtimeToken } from './realtime.ts';\n\nexport interface RealtimeConstraints {\n  instructions?: string;\n  tools?: Tool[];\n  voice?: string;\n  uses?: number;\n  agentId?: string;\n}\n\nconst openaiSecrets = 'https://api.openai.com/v1/realtime/client_secrets';\nconst xaiSecrets = 'https://api.x.ai/v1/realtime/client_secrets';\nconst geminiTokens = 'https://generativelanguage.googleapis.com/v1beta/auth_tokens';\nconst elevenSigned = 'https://api.elevenlabs.io/v1/convai/conversation/get-signed-url';\nconst elevenWebRtc = 'https://api.elevenlabs.io/v1/convai/conversation/token';\nconst fishSessions = 'https://api.fish.audio/v1/agent/sessions';\nconst fifteenMinutes = 15 * 60 * 1000;\nconst thirtyMinutes = 30 * 60 * 1000;\nconst oneMinute = 60 * 1000;\nconst xaiLifetimeSeconds = 300;\n\ntype Json = Record<string, unknown>;\n\ninterface MintCall {\n  url: string;\n  method: 'GET' | 'POST';\n  headers: Record<string, string>;\n  body?: string;\n  field: string;\n  expires: (json: Json, fallback: number) => number;\n  fallbackExpires: number;\n}\n\nfunction apiKey(options: ProviderRequestOptions): string {\n  const key = options.apiKey;\n  if (typeof key === 'string' && key.length > 0) return key;\n  throw new VoiceRequestError('error', 'missing realtime credential');\n}\n\nfunction mergedHeaders(\n  options: ProviderRequestOptions,\n  defaults: Record<string, string>\n): Record<string, string> {\n  const headers = { ...defaults };\n  const extra = options.headers;\n  if (extra === undefined) return headers;\n  for (const [key, value] of Object.entries(extra)) {\n    if (value === null) delete headers[key];\n    else headers[key] = value;\n  }\n  return headers;\n}\n\nfunction omitUndefined(source: Json): Json {\n  const out: Json = {};\n  for (const [key, value] of Object.entries(source)) {\n    if (value !== undefined) out[key] = value;\n  }\n  return out;\n}\n\nfunction functionTool(tool: Tool): Json {\n  return {\n    type: 'function',\n    name: tool.name,\n    description: tool.description,\n    parameters: tool.parameters as unknown,\n  };\n}\n\nfunction declaration(tool: Tool): Json {\n  return {\n    name: tool.name,\n    description: tool.description,\n    parameters: tool.parameters as unknown,\n  };\n}\n\nfunction requiredAgent(constraints: RealtimeConstraints): string {\n  const id = constraints.agentId;\n  if (typeof id === 'string' && id.length > 0) return id;\n  throw new VoiceRequestError('error', 'missing agent id');\n}\n\nfunction expiresAt(value: unknown, fallback: number): number {\n  if (typeof value === 'number' && Number.isFinite(value)) {\n    return value < 1e12 ? value * 1000 : value;\n  }\n  if (typeof value === 'string') {\n    const parsed = Date.parse(value);\n    if (Number.isFinite(parsed)) return parsed;\n  }\n  return fallback;\n}\n\nfunction fromField(field: string): (json: Json, fallback: number) => number {\n  return (json, fallback) => expiresAt(json[field], fallback);\n}\n\nfunction openaiBody(model: RealtimeModel<string>, constraints: RealtimeConstraints): Json {\n  const session = omitUndefined({\n    type: 'realtime',\n    model: model.id,\n    instructions: constraints.instructions,\n    tools: constraints.tools?.map(functionTool),\n  });\n  if (constraints.voice !== undefined) session.audio = { output: { voice: constraints.voice } };\n  return { session };\n}\n\nfunction geminiBody(\n  model: RealtimeModel<string>,\n  constraints: RealtimeConstraints,\n  now: number\n): Json {\n  const instructions = constraints.instructions;\n  const voice = constraints.voice;\n  const tools = constraints.tools;\n  const config = omitUndefined({\n    responseModalities: ['AUDIO'],\n    sessionResumption: {},\n    systemInstruction: instructions === undefined ? undefined : { parts: [{ text: instructions }] },\n    speechConfig:\n      voice === undefined\n        ? undefined\n        : { voiceConfig: { prebuiltVoiceConfig: { voiceName: voice } } },\n    tools: tools === undefined ? undefined : [{ functionDeclarations: tools.map(declaration) }],\n  });\n  return {\n    uses: constraints.uses ?? 1,\n    expireTime: new Date(now + thirtyMinutes).toISOString(),\n    newSessionExpireTime: new Date(now + oneMinute).toISOString(),\n    liveConnectConstraints: { model: `models/${model.id}`, config },\n  };\n}\n\nfunction fishBody(constraints: RealtimeConstraints): Json {\n  const overrides = omitUndefined({\n    system_prompt: constraints.instructions,\n    voice_id: constraints.voice,\n  });\n  return omitUndefined({\n    agent_id: requiredAgent(constraints),\n    overrides: Object.keys(overrides).length > 0 ? overrides : undefined,\n  });\n}\n\nfunction webrtcOnly(model: RealtimeModel<string>): boolean {\n  return model.transports.length === 1 && model.transports[0] === 'webrtc';\n}\n\nfunction elevenUrl(model: RealtimeModel<string>, constraints: RealtimeConstraints): string {\n  const url = new URL(webrtcOnly(model) ? elevenWebRtc : elevenSigned);\n  url.searchParams.set('agent_id', requiredAgent(constraints));\n  return url.toString();\n}\n\nfunction mintCall(\n  model: RealtimeModel<string>,\n  constraints: RealtimeConstraints,\n  options: ProviderRequestOptions,\n  key: string,\n  now: number\n): MintCall {\n  const jsonHeaders = { 'Content-Type': 'application/json' };\n  switch (model.api) {\n    case 'openai-realtime':\n      return {\n        url: openaiSecrets,\n        method: 'POST',\n        headers: mergedHeaders(options, { ...jsonHeaders, Authorization: `Bearer ${key}` }),\n        body: JSON.stringify(openaiBody(model, constraints)),\n        field: 'value',\n        expires: fromField('expires_at'),\n        fallbackExpires: now,\n      };\n    case 'xai-realtime':\n      return {\n        url: xaiSecrets,\n        method: 'POST',\n        headers: mergedHeaders(options, { ...jsonHeaders, Authorization: `Bearer ${key}` }),\n        body: JSON.stringify({ expires_after: { seconds: xaiLifetimeSeconds } }),\n        field: 'value',\n        expires: fromField('expires_at'),\n        fallbackExpires: now + xaiLifetimeSeconds * 1000,\n      };\n    case 'gemini-live':\n      return {\n        url: geminiTokens,\n        method: 'POST',\n        headers: mergedHeaders(options, { ...jsonHeaders, 'x-goog-api-key': key }),\n        body: JSON.stringify(geminiBody(model, constraints, now)),\n        field: 'name',\n        expires: fromField('expireTime'),\n        fallbackExpires: now + thirtyMinutes,\n      };\n    case 'elevenlabs-agents':\n      return {\n        url: elevenUrl(model, constraints),\n        method: 'GET',\n        headers: mergedHeaders(options, { 'xi-api-key': key }),\n        field: webrtcOnly(model) ? 'token' : 'signed_url',\n        expires: fromField('expires_at'),\n        fallbackExpires: now + fifteenMinutes,\n      };\n    case 'fish-agents':\n      return {\n        url: fishSessions,\n        method: 'POST',\n        headers: mergedHeaders(options, { ...jsonHeaders, Authorization: `Bearer ${key}` }),\n        body: JSON.stringify(fishBody(constraints)),\n        field: 'token',\n        expires: fromField('expires_at'),\n        fallbackExpires: now,\n      };\n    default:\n      throw new VoiceRequestError('error', 'unsupported realtime api');\n  }\n}\n\nfunction readValue(json: unknown, field: string): string {\n  if (typeof json !== 'object' || json === null) {\n    throw new VoiceRequestError('error', 'mint failed (empty token)');\n  }\n  const value = (json as Json)[field];\n  if (typeof value === 'string' && value.length > 0) return value;\n  throw new VoiceRequestError('error', 'mint failed (empty token)');\n}\n\nexport async function mintRealtimeToken(\n  model: RealtimeModel<string>,\n  constraints: RealtimeConstraints,\n  options?: ProviderRequestOptions\n): Promise<RealtimeToken> {\n  const settings = options ?? {};\n  const key = apiKey(settings);\n  const now = Date.now();\n  const call = mintCall(model, constraints, settings, key, now);\n  const fetch = settings.fetch ?? globalThis.fetch;\n  const response = await fetch(call.url, {\n    method: call.method,\n    headers: call.headers,\n    body: call.body,\n    signal: settings.signal,\n  });\n  if (!response.ok) throw new VoiceRequestError('error', `mint failed (${response.status})`);\n  const json: unknown = await response.json();\n  return {\n    value: readValue(json, call.field),\n    expiresAt: call.expires(json as Json, call.fallbackExpires),\n    model: model.id,\n  };\n}\n"],
+  "mappings": "AACA,SAAS,yBAAyB;AAYlC,MAAM,gBAAgB;AACtB,MAAM,aAAa;AACnB,MAAM,eAAe;AACrB,MAAM,eAAe;AACrB,MAAM,eAAe;AACrB,MAAM,eAAe;AACrB,MAAM,iBAAiB,KAAK,KAAK;AACjC,MAAM,gBAAgB,KAAK,KAAK;AAChC,MAAM,YAAY,KAAK;AACvB,MAAM,qBAAqB;AAc3B,SAAS,OAAO,SAAyC;AACvD,QAAM,MAAM,QAAQ;AACpB,MAAI,OAAO,QAAQ,YAAY,IAAI,SAAS,EAAG,QAAO;AACtD,QAAM,IAAI,kBAAkB,SAAS,6BAA6B;AACpE;AAEA,SAAS,cACP,SACA,UACwB;AACxB,QAAM,UAAU,EAAE,GAAG,SAAS;AAC9B,QAAM,QAAQ,QAAQ;AACtB,MAAI,UAAU,OAAW,QAAO;AAChC,aAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,KAAK,GAAG;AAChD,QAAI,UAAU,KAAM,QAAO,QAAQ,GAAG;AAAA,QACjC,SAAQ,GAAG,IAAI;AAAA,EACtB;AACA,SAAO;AACT;AAEA,SAAS,cAAc,QAAoB;AACzC,QAAM,MAAY,CAAC;AACnB,aAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,MAAM,GAAG;AACjD,QAAI,UAAU,OAAW,KAAI,GAAG,IAAI;AAAA,EACtC;AACA,SAAO;AACT;AAEA,SAAS,aAAa,MAAkB;AACtC,SAAO;AAAA,IACL,MAAM;AAAA,IACN,MAAM,KAAK;AAAA,IACX,aAAa,KAAK;AAAA,IAClB,YAAY,KAAK;AAAA,EACnB;AACF;AAEA,SAAS,YAAY,MAAkB;AACrC,SAAO;AAAA,IACL,MAAM,KAAK;AAAA,IACX,aAAa,KAAK;AAAA,IAClB,YAAY,KAAK;AAAA,EACnB;AACF;AAEA,SAAS,cAAc,aAA0C;AAC/D,QAAM,KAAK,YAAY;AACvB,MAAI,OAAO,OAAO,YAAY,GAAG,SAAS,EAAG,QAAO;AACpD,QAAM,IAAI,kBAAkB,SAAS,kBAAkB;AACzD;AAEA,SAAS,UAAU,OAAgB,UAA0B;AAC3D,MAAI,OAAO,UAAU,YAAY,OAAO,SAAS,KAAK,GAAG;AACvD,WAAO,QAAQ,OAAO,QAAQ,MAAO;AAAA,EACvC;AACA,MAAI,OAAO,UAAU,UAAU;AAC7B,UAAM,SAAS,KAAK,MAAM,KAAK;AAC/B,QAAI,OAAO,SAAS,MAAM,EAAG,QAAO;AAAA,EACtC;AACA,SAAO;AACT;AAEA,SAAS,UAAU,OAAyD;AAC1E,SAAO,CAAC,MAAM,aAAa,UAAU,KAAK,KAAK,GAAG,QAAQ;AAC5D;AAEA,SAAS,WAAW,OAA8B,aAAwC;AACxF,QAAM,UAAU,cAAc;AAAA,IAC5B,MAAM;AAAA,IACN,OAAO,MAAM;AAAA,IACb,cAAc,YAAY;AAAA,IAC1B,OAAO,YAAY,OAAO,IAAI,YAAY;AAAA,EAC5C,CAAC;AACD,MAAI,YAAY,UAAU,OAAW,SAAQ,QAAQ,EAAE,QAAQ,EAAE,OAAO,YAAY,MAAM,EAAE;AAC5F,SAAO,EAAE,QAAQ;AACnB;AAEA,SAAS,WACP,OACA,aACA,KACM;AACN,QAAM,eAAe,YAAY;AACjC,QAAM,QAAQ,YAAY;AAC1B,QAAM,QAAQ,YAAY;AAC1B,QAAM,SAAS,cAAc;AAAA,IAC3B,oBAAoB,CAAC,OAAO;AAAA,IAC5B,mBAAmB,CAAC;AAAA,IACpB,mBAAmB,iBAAiB,SAAY,SAAY,EAAE,OAAO,CAAC,EAAE,MAAM,aAAa,CAAC,EAAE;AAAA,IAC9F,cACE,UAAU,SACN,SACA,EAAE,aAAa,EAAE,qBAAqB,EAAE,WAAW,MAAM,EAAE,EAAE;AAAA,IACnE,OAAO,UAAU,SAAY,SAAY,CAAC,EAAE,sBAAsB,MAAM,IAAI,WAAW,EAAE,CAAC;AAAA,EAC5F,CAAC;AACD,SAAO;AAAA,IACL,MAAM,YAAY,QAAQ;AAAA,IAC1B,YAAY,IAAI,KAAK,MAAM,aAAa,EAAE,YAAY;AAAA,IACtD,sBAAsB,IAAI,KAAK,MAAM,SAAS,EAAE,YAAY;AAAA,IAC5D,wBAAwB,EAAE,OAAO,UAAU,MAAM,EAAE,IAAI,OAAO;AAAA,EAChE;AACF;AAEA,SAAS,SAAS,aAAwC;AACxD,QAAM,YAAY,cAAc;AAAA,IAC9B,eAAe,YAAY;AAAA,IAC3B,UAAU,YAAY;AAAA,EACxB,CAAC;AACD,SAAO,cAAc;AAAA,IACnB,UAAU,cAAc,WAAW;AAAA,IACnC,WAAW,OAAO,KAAK,SAAS,EAAE,SAAS,IAAI,YAAY;AAAA,EAC7D,CAAC;AACH;AAEA,SAAS,WAAW,OAAuC;AACzD,SAAO,MAAM,WAAW,WAAW,KAAK,MAAM,WAAW,CAAC,MAAM;AAClE;AAEA,SAAS,UAAU,OAA8B,aAA0C;AACzF,QAAM,MAAM,IAAI,IAAI,WAAW,KAAK,IAAI,eAAe,YAAY;AACnE,MAAI,aAAa,IAAI,YAAY,cAAc,WAAW,CAAC;AAC3D,SAAO,IAAI,SAAS;AACtB;AAEA,SAAS,SACP,OACA,aACA,SACA,KACA,KACU;AACV,QAAM,cAAc,EAAE,gBAAgB,mBAAmB;AACzD,UAAQ,MAAM,KAAK;AAAA,IACjB,KAAK;AACH,aAAO;AAAA,QACL,KAAK;AAAA,QACL,QAAQ;AAAA,QACR,SAAS,cAAc,SAAS,EAAE,GAAG,aAAa,eAAe,UAAU,GAAG,GAAG,CAAC;AAAA,QAClF,MAAM,KAAK,UAAU,WAAW,OAAO,WAAW,CAAC;AAAA,QACnD,OAAO;AAAA,QACP,SAAS,UAAU,YAAY;AAAA,QAC/B,iBAAiB;AAAA,MACnB;AAAA,IACF,KAAK;AACH,aAAO;AAAA,QACL,KAAK;AAAA,QACL,QAAQ;AAAA,QACR,SAAS,cAAc,SAAS,EAAE,GAAG,aAAa,eAAe,UAAU,GAAG,GAAG,CAAC;AAAA,QAClF,MAAM,KAAK,UAAU,EAAE,eAAe,EAAE,SAAS,mBAAmB,EAAE,CAAC;AAAA,QACvE,OAAO;AAAA,QACP,SAAS,UAAU,YAAY;AAAA,QAC/B,iBAAiB,MAAM,qBAAqB;AAAA,MAC9C;AAAA,IACF,KAAK;AACH,aAAO;AAAA,QACL,KAAK;AAAA,QACL,QAAQ;AAAA,QACR,SAAS,cAAc,SAAS,EAAE,GAAG,aAAa,kBAAkB,IAAI,CAAC;AAAA,QACzE,MAAM,KAAK,UAAU,WAAW,OAAO,aAAa,GAAG,CAAC;AAAA,QACxD,OAAO;AAAA,QACP,SAAS,UAAU,YAAY;AAAA,QAC/B,iBAAiB,MAAM;AAAA,MACzB;AAAA,IACF,KAAK;AACH,aAAO;AAAA,QACL,KAAK,UAAU,OAAO,WAAW;AAAA,QACjC,QAAQ;AAAA,QACR,SAAS,cAAc,SAAS,EAAE,cAAc,IAAI,CAAC;AAAA,QACrD,OAAO,WAAW,KAAK,IAAI,UAAU;AAAA,QACrC,SAAS,UAAU,YAAY;AAAA,QAC/B,iBAAiB,MAAM;AAAA,MACzB;AAAA,IACF,KAAK;AACH,aAAO;AAAA,QACL,KAAK;AAAA,QACL,QAAQ;AAAA,QACR,SAAS,cAAc,SAAS,EAAE,GAAG,aAAa,eAAe,UAAU,GAAG,GAAG,CAAC;AAAA,QAClF,MAAM,KAAK,UAAU,SAAS,WAAW,CAAC;AAAA,QAC1C,OAAO;AAAA,QACP,SAAS,UAAU,YAAY;AAAA,QAC/B,iBAAiB;AAAA,MACnB;AAAA,IACF;AACE,YAAM,IAAI,kBAAkB,SAAS,0BAA0B;AAAA,EACnE;AACF;AAEA,SAAS,UAAU,MAAe,OAAuB;AACvD,MAAI,OAAO,SAAS,YAAY,SAAS,MAAM;AAC7C,UAAM,IAAI,kBAAkB,SAAS,2BAA2B;AAAA,EAClE;AACA,QAAM,QAAS,KAAc,KAAK;AAClC,MAAI,OAAO,UAAU,YAAY,MAAM,SAAS,EAAG,QAAO;AAC1D,QAAM,IAAI,kBAAkB,SAAS,2BAA2B;AAClE;AAEA,eAAsB,kBACpB,OACA,aACA,SACwB;AACxB,QAAM,WAAW,WAAW,CAAC;AAC7B,QAAM,MAAM,OAAO,QAAQ;AAC3B,QAAM,MAAM,KAAK,IAAI;AACrB,QAAM,OAAO,SAAS,OAAO,aAAa,UAAU,KAAK,GAAG;AAC5D,QAAM,QAAQ,SAAS,SAAS,WAAW;AAC3C,QAAM,WAAW,MAAM,MAAM,KAAK,KAAK;AAAA,IACrC,QAAQ,KAAK;AAAA,IACb,SAAS,KAAK;AAAA,IACd,MAAM,KAAK;AAAA,IACX,QAAQ,SAAS;AAAA,EACnB,CAAC;AACD,MAAI,CAAC,SAAS,GAAI,OAAM,IAAI,kBAAkB,SAAS,gBAAgB,SAAS,MAAM,GAAG;AACzF,QAAM,OAAgB,MAAM,SAAS,KAAK;AAC1C,SAAO;AAAA,IACL,OAAO,UAAU,MAAM,KAAK,KAAK;AAAA,IACjC,WAAW,KAAK,QAAQ,MAAc,KAAK,eAAe;AAAA,IAC1D,OAAO,MAAM;AAAA,EACf;AACF;",
+  "names": []
+}

package/dist/types/gemini-live/index.d.ts

@@ -0,0 +1,7 @@
+import type { RealtimeModel } from '../models.ts';
+import { type GeminiOptions } from './session.ts';
+export declare const gemini38Live: RealtimeModel<"gemini-live">;
+export declare const gemini25NativeAudio: RealtimeModel<"gemini-live">;
+export declare const gemini31FlashLive: RealtimeModel<"gemini-live">;
+export declare function registerGeminiLive(options?: GeminiOptions): () => void;
+export type { GeminiOptions, VoiceWireFactory } from './session.ts';

package/dist/types/gemini-live/session.d.ts

@@ -0,0 +1,12 @@
+import type { AudioFormat } from '../audio.ts';
+import type { RealtimeModel } from '../models.ts';
+import { type VoiceSocketConstructor, type VoiceWire, type VoiceWireTarget } from '../openai-realtime/socket.ts';
+import type { RealtimeContext, RealtimeOptions, RealtimeSession } from '../realtime.ts';
+export type VoiceWireFactory = (target: VoiceWireTarget) => VoiceWire;
+export interface GeminiOptions {
+    wire?: VoiceWireFactory;
+    socket?: VoiceSocketConstructor;
+    vocabulary?: string[];
+}
+export declare function connectGeminiSession(model: RealtimeModel<string>, context: RealtimeContext, options: RealtimeOptions | undefined, gemini: GeminiOptions): Promise<RealtimeSession>;
+export declare function pcmFormat(sampleRate: number): AudioFormat;

package/dist/types/index.d.ts

@@ -13,6 +13,8 @@ export { SpeechEventStream } from './speech.ts';
 export type { VoiceAgentEvent, VoiceMessageChange } from './spoken.ts';
 export type { BindThinkerOptions, BoundThinker, Thinker, ThinkerCaption, ThinkerPhase, ThinkerUpdate, } from './talker.ts';
 export { askAgent, bindThinker, talkerTools } from './talker.ts';
+export type { RealtimeConstraints } from './tokens.ts';
+export { mintRealtimeToken } from './tokens.ts';
 export type { AudioInput, ListenOptions, Transcript, TranscriptionApiProvider, TranscriptionEvent, TranscriptSegment, } from './transcription.ts';
 export { TranscriptionEventStream } from './transcription.ts';
 export type { BargeIn, CaptionEvent, EchoRecommendation, VoiceConversation, VoiceCut, VoiceModels, VoicePhase, VoiceSession, VoiceTurnMeta, VoiceTurnOptions, } from './turns.ts';

package/dist/types/tokens.d.ts

@@ -0,0 +1,11 @@
+import type { ProviderRequestOptions, Tool } from '@earendil-works/pi-ai';
+import type { RealtimeModel } from './models.ts';
+import type { RealtimeToken } from './realtime.ts';
+export interface RealtimeConstraints {
+    instructions?: string;
+    tools?: Tool[];
+    voice?: string;
+    uses?: number;
+    agentId?: string;
+}
+export declare function mintRealtimeToken(model: RealtimeModel<string>, constraints: RealtimeConstraints, options?: ProviderRequestOptions): Promise<RealtimeToken>;

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

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

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

@@ -2,3 +2,4 @@ import type { RealtimeModel } from '../models.ts';
 import { type AdapterOptions } from '../openai-realtime/session.ts';
 export declare const grokVoiceLatest: RealtimeModel<'xai-realtime'>;
 export declare function registerXaiRealtime(options?: AdapterOptions): () => void;
+export { registerXAI } from './catalog.ts';

package/dist/types/xai/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 listenXAI(model: TranscriptionModel<string>, audio: AsyncIterable<AudioChunk>, options?: ListenOptions): TranscriptionEventStream;

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

package/dist/types/xai/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 transcribeXAI(model: TranscriptionModel<string>, input: AudioInput, options?: ProviderRequestOptions): Promise<Transcript>;

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

@@ -0,0 +1,20 @@
+import type { AudioFormat } from '../audio.ts';
+import type { TranscriptionModel } from '../models.ts';
+export interface XaiFile {
+    data: Uint8Array;
+    mimeType: string;
+    filename: string;
+}
+export declare function mediaType(mimeType: string): string;
+export declare function audioFilename(mimeType: string): string;
+export declare function rawFields(mimeType: string, model: TranscriptionModel<string>): Record<string, string> | undefined;
+export declare function encodeForm(file: XaiFile, fields: Record<string, string>): ArrayBuffer;
+export declare function formType(): string;
+export declare function socketUrl(baseUrl: string, path: string): URL;
+export declare function codecName(encoding: AudioFormat['encoding']): string;
+export declare function encodingName(encoding: AudioFormat['encoding']): string;
+export declare function connectSocket(url: string, headers: Record<string, string>): WebSocket;
+export declare function awaitOpen(socket: WebSocket, label: string, fail: (message: string) => void): Promise<void>;
+export declare function messageText(data: unknown): Promise<string>;
+export declare function sameFormat(left: AudioFormat, right: AudioFormat): boolean;
+export declare function copiedBytes(data: Uint8Array): ArrayBuffer;

package/dist/xai/catalog.js

@@ -0,0 +1,133 @@
+import {
+  registerVoiceModel
+} from "../models.js";
+import { registerSpeechApiProvider, registerTranscriptionApiProvider } from "../registry.js";
+import { listenXAI } from "./listen.js";
+import { speakXAI } from "./speak.js";
+import { transcribeXAI } from "./transcribe.js";
+const tokenCost = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };
+const xai = "https://api.x.ai/v1";
+const languages = [
+  "ar",
+  "bs",
+  "bg",
+  "yue",
+  "ca",
+  "zh",
+  "hr",
+  "cs",
+  "da",
+  "nl",
+  "en",
+  "fil",
+  "fi",
+  "fr",
+  "de",
+  "el",
+  "hi",
+  "hu",
+  "id",
+  "it",
+  "ja",
+  "ko",
+  "mk",
+  "ms",
+  "nb",
+  "fa",
+  "pl",
+  "pt",
+  "ro",
+  "ru",
+  "sk",
+  "es",
+  "sv",
+  "th",
+  "tr",
+  "uk",
+  "ur",
+  "vi"
+];
+const pcm16 = (sampleRate) => ({
+  encoding: "pcm16",
+  sampleRate,
+  channels: 1
+});
+const law = (encoding, sampleRate = 8e3) => ({
+  encoding,
+  sampleRate,
+  channels: 1
+});
+const container = (encoding, sampleRate) => ({
+  encoding,
+  sampleRate,
+  channels: 1
+});
+const voices = ["eve", "ara", "rex", "sal", "leo"].map((id) => ({ id, name: id }));
+function xaiCatalog() {
+  const transcription = {
+    id: "grok-voice-transcribe-2.0",
+    name: "grok-voice-transcribe-2.0",
+    api: "xai-stt",
+    provider: "xai",
+    baseUrl: xai,
+    type: "transcription",
+    streaming: true,
+    turnDetection: "semantic",
+    inputFormats: [
+      pcm16(16e3),
+      pcm16(24e3),
+      law("mulaw"),
+      law("alaw"),
+      container("wav", 16e3),
+      container("mp3", 16e3),
+      container("flac", 16e3)
+    ],
+    languages,
+    diarization: true,
+    cost: tokenCost,
+    audioCost: { perInputMinute: 0.2 / 60 }
+  };
+  const speech = {
+    id: "grok-voice-tts-1.0",
+    name: "grok-voice-tts-1.0",
+    api: "xai-tts",
+    provider: "xai",
+    baseUrl: xai,
+    type: "speech",
+    textStreaming: true,
+    outputFormats: [
+      container("mp3", 24e3),
+      pcm16(24e3),
+      container("wav", 24e3),
+      law("mulaw"),
+      law("alaw")
+    ],
+    voices,
+    cloning: false,
+    cost: tokenCost,
+    audioCost: { perThousandChars: 0.015 }
+  };
+  return [transcription, speech];
+}
+function registerXAI() {
+  const models = xaiCatalog();
+  const unregisterModels = models.map((model) => registerVoiceModel(model));
+  const unregisterStt = registerTranscriptionApiProvider({
+    api: "xai-stt",
+    transcribe: transcribeXAI,
+    listen: listenXAI
+  });
+  const unregisterTts = registerSpeechApiProvider({ api: "xai-tts", speak: speakXAI });
+  let released = false;
+  return () => {
+    if (released) return;
+    released = true;
+    for (const unregister of unregisterModels) unregister();
+    unregisterStt();
+    unregisterTts();
+  };
+}
+export {
+  registerXAI
+};
+//# sourceMappingURL=catalog.js.map

package/dist/xai/catalog.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/xai/catalog.ts"],
+  "sourcesContent": ["import type { AudioFormat } from '../audio.ts';\nimport {\n  registerVoiceModel,\n  type SpeechModel,\n  type TranscriptionModel,\n  type VoiceInfo,\n} from '../models.ts';\nimport { registerSpeechApiProvider, registerTranscriptionApiProvider } from '../registry.ts';\nimport { listenXAI } from './listen.ts';\nimport { speakXAI } from './speak.ts';\nimport { transcribeXAI } from './transcribe.ts';\n\nconst tokenCost = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };\nconst xai = 'https://api.x.ai/v1';\n\nconst languages = [\n  'ar',\n  'bs',\n  'bg',\n  'yue',\n  'ca',\n  'zh',\n  'hr',\n  'cs',\n  'da',\n  'nl',\n  'en',\n  'fil',\n  'fi',\n  'fr',\n  'de',\n  'el',\n  'hi',\n  'hu',\n  'id',\n  'it',\n  'ja',\n  'ko',\n  'mk',\n  'ms',\n  'nb',\n  'fa',\n  'pl',\n  'pt',\n  'ro',\n  'ru',\n  'sk',\n  'es',\n  'sv',\n  'th',\n  'tr',\n  'uk',\n  'ur',\n  'vi',\n];\n\nconst pcm16 = (sampleRate: number): AudioFormat => ({\n  encoding: 'pcm16',\n  sampleRate,\n  channels: 1,\n});\n\nconst law = (encoding: 'mulaw' | 'alaw', sampleRate = 8000): AudioFormat => ({\n  encoding,\n  sampleRate,\n  channels: 1,\n});\n\nconst container = (encoding: 'wav' | 'mp3' | 'flac', sampleRate: number): AudioFormat => ({\n  encoding,\n  sampleRate,\n  channels: 1,\n});\n\nconst voices: VoiceInfo[] = ['eve', 'ara', 'rex', 'sal', 'leo'].map((id) => ({ id, name: id }));\n\nfunction xaiCatalog(): Array<TranscriptionModel<string> | SpeechModel<string>> {\n  const transcription: TranscriptionModel<string> = {\n    id: 'grok-voice-transcribe-2.0',\n    name: 'grok-voice-transcribe-2.0',\n    api: 'xai-stt',\n    provider: 'xai',\n    baseUrl: xai,\n    type: 'transcription',\n    streaming: true,\n    turnDetection: 'semantic',\n    inputFormats: [\n      pcm16(16000),\n      pcm16(24000),\n      law('mulaw'),\n      law('alaw'),\n      container('wav', 16000),\n      container('mp3', 16000),\n      container('flac', 16000),\n    ],\n    languages,\n    diarization: true,\n    cost: tokenCost,\n    audioCost: { perInputMinute: 0.2 / 60 },\n  };\n  const speech: SpeechModel<string> = {\n    id: 'grok-voice-tts-1.0',\n    name: 'grok-voice-tts-1.0',\n    api: 'xai-tts',\n    provider: 'xai',\n    baseUrl: xai,\n    type: 'speech',\n    textStreaming: true,\n    outputFormats: [\n      container('mp3', 24000),\n      pcm16(24000),\n      container('wav', 24000),\n      law('mulaw'),\n      law('alaw'),\n    ],\n    voices,\n    cloning: false,\n    cost: tokenCost,\n    audioCost: { perThousandChars: 0.015 },\n  };\n  return [transcription, speech];\n}\n\nexport function registerXAI(): () => void {\n  const models = xaiCatalog();\n  const unregisterModels = models.map((model) => registerVoiceModel(model));\n  const unregisterStt = registerTranscriptionApiProvider({\n    api: 'xai-stt',\n    transcribe: transcribeXAI,\n    listen: listenXAI,\n  });\n  const unregisterTts = registerSpeechApiProvider({ api: 'xai-tts', speak: speakXAI });\n  let released = false;\n  return () => {\n    if (released) return;\n    released = true;\n    for (const unregister of unregisterModels) unregister();\n    unregisterStt();\n    unregisterTts();\n  };\n}\n"],
+  "mappings": "AACA;AAAA,EACE;AAAA,OAIK;AACP,SAAS,2BAA2B,wCAAwC;AAC5E,SAAS,iBAAiB;AAC1B,SAAS,gBAAgB;AACzB,SAAS,qBAAqB;AAE9B,MAAM,YAAY,EAAE,OAAO,GAAG,QAAQ,GAAG,WAAW,GAAG,YAAY,EAAE;AACrE,MAAM,MAAM;AAEZ,MAAM,YAAY;AAAA,EAChB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAEA,MAAM,QAAQ,CAAC,gBAAqC;AAAA,EAClD,UAAU;AAAA,EACV;AAAA,EACA,UAAU;AACZ;AAEA,MAAM,MAAM,CAAC,UAA4B,aAAa,SAAuB;AAAA,EAC3E;AAAA,EACA;AAAA,EACA,UAAU;AACZ;AAEA,MAAM,YAAY,CAAC,UAAkC,gBAAqC;AAAA,EACxF;AAAA,EACA;AAAA,EACA,UAAU;AACZ;AAEA,MAAM,SAAsB,CAAC,OAAO,OAAO,OAAO,OAAO,KAAK,EAAE,IAAI,CAAC,QAAQ,EAAE,IAAI,MAAM,GAAG,EAAE;AAE9F,SAAS,aAAsE;AAC7E,QAAM,gBAA4C;AAAA,IAChD,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,MAAM,IAAK;AAAA,MACX,MAAM,IAAK;AAAA,MACX,IAAI,OAAO;AAAA,MACX,IAAI,MAAM;AAAA,MACV,UAAU,OAAO,IAAK;AAAA,MACtB,UAAU,OAAO,IAAK;AAAA,MACtB,UAAU,QAAQ,IAAK;AAAA,IACzB;AAAA,IACA;AAAA,IACA,aAAa;AAAA,IACb,MAAM;AAAA,IACN,WAAW,EAAE,gBAAgB,MAAM,GAAG;AAAA,EACxC;AACA,QAAM,SAA8B;AAAA,IAClC,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,KAAK;AAAA,IACL,UAAU;AAAA,IACV,SAAS;AAAA,IACT,MAAM;AAAA,IACN,eAAe;AAAA,IACf,eAAe;AAAA,MACb,UAAU,OAAO,IAAK;AAAA,MACtB,MAAM,IAAK;AAAA,MACX,UAAU,OAAO,IAAK;AAAA,MACtB,IAAI,OAAO;AAAA,MACX,IAAI,MAAM;AAAA,IACZ;AAAA,IACA;AAAA,IACA,SAAS;AAAA,IACT,MAAM;AAAA,IACN,WAAW,EAAE,kBAAkB,MAAM;AAAA,EACvC;AACA,SAAO,CAAC,eAAe,MAAM;AAC/B;AAEO,SAAS,cAA0B;AACxC,QAAM,SAAS,WAAW;AAC1B,QAAM,mBAAmB,OAAO,IAAI,CAAC,UAAU,mBAAmB,KAAK,CAAC;AACxE,QAAM,gBAAgB,iCAAiC;AAAA,IACrD,KAAK;AAAA,IACL,YAAY;AAAA,IACZ,QAAQ;AAAA,EACV,CAAC;AACD,QAAM,gBAAgB,0BAA0B,EAAE,KAAK,WAAW,OAAO,SAAS,CAAC;AACnF,MAAI,WAAW;AACf,SAAO,MAAM;AACX,QAAI,SAAU;AACd,eAAW;AACX,eAAW,cAAc,iBAAkB,YAAW;AACtD,kBAAc;AACd,kBAAc;AAAA,EAChB;AACF;",
+  "names": []
+}

package/dist/xai/index.js

@@ -44,8 +44,10 @@ function registerXaiRealtime(options = {}) {
     unregisterApi();
   };
 }
+import { registerXAI } from "./catalog.js";
 export {
   grokVoiceLatest,
+  registerXAI,
   registerXaiRealtime
 };
 //# sourceMappingURL=index.js.map

package/dist/xai/index.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["../../src/xai/index.ts"],
-  "sourcesContent": ["import type { RealtimeModel } from '../models.ts';\nimport { registerVoiceModel } from '../models.ts';\nimport {\n  type AdapterOptions,\n  connectVendorSession,\n  pcmFormat,\n} from '../openai-realtime/session.ts';\nimport { registerRealtimeApiProvider } from '../registry.ts';\n\nconst voices = [\n  { id: 'eve', name: 'Eve' },\n  { id: 'ara', name: 'Ara' },\n  { id: 'rex', name: 'Rex' },\n  { id: 'sal', name: 'Sal' },\n  { id: 'leo', name: 'Leo' },\n];\n\nexport const grokVoiceLatest: RealtimeModel<'xai-realtime'> = {\n  id: 'grok-voice-latest',\n  name: 'Grok Voice',\n  api: 'xai-realtime',\n  provider: 'xai',\n  baseUrl: 'wss://api.x.ai/v1/realtime',\n  type: 'realtime',\n  transports: ['websocket', 'webrtc'],\n  toolCalls: 'blocking',\n  inputFormat: pcmFormat(24000),\n  outputFormat: pcmFormat(24000),\n  maxSessionMs: 7_200_000,\n  resumable: true,\n  ephemeralAuth: true,\n  voices,\n  provisioning: 'none',\n  delivery: { interrupt: 'native', when_idle: 'approximate', silent: 'native' },\n  truncation: 'client',\n  liveToolUpdate: true,\n  cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },\n  audioCost: { perInputMinute: 0.08, perOutputMinute: 0.08, perTextMessage: 0.004 },\n};\n\nexport function registerXaiRealtime(options: AdapterOptions = {}): () => void {\n  const unregisterModel = registerVoiceModel(grokVoiceLatest);\n  const unregisterApi = registerRealtimeApiProvider({\n    api: 'xai-realtime',\n    connect: (model, context, connectOptions) =>\n      connectVendorSession(model, context, connectOptions, { ...options, dialect: 'xai' }),\n  });\n  return () => {\n    unregisterModel();\n    unregisterApi();\n  };\n}\n"],
-  "mappings": "AACA,SAAS,0BAA0B;AACnC;AAAA,EAEE;AAAA,EACA;AAAA,OACK;AACP,SAAS,mCAAmC;AAE5C,MAAM,SAAS;AAAA,EACb,EAAE,IAAI,OAAO,MAAM,MAAM;AAAA,EACzB,EAAE,IAAI,OAAO,MAAM,MAAM;AAAA,EACzB,EAAE,IAAI,OAAO,MAAM,MAAM;AAAA,EACzB,EAAE,IAAI,OAAO,MAAM,MAAM;AAAA,EACzB,EAAE,IAAI,OAAO,MAAM,MAAM;AAC3B;AAEO,MAAM,kBAAiD;AAAA,EAC5D,IAAI;AAAA,EACJ,MAAM;AAAA,EACN,KAAK;AAAA,EACL,UAAU;AAAA,EACV,SAAS;AAAA,EACT,MAAM;AAAA,EACN,YAAY,CAAC,aAAa,QAAQ;AAAA,EAClC,WAAW;AAAA,EACX,aAAa,UAAU,IAAK;AAAA,EAC5B,cAAc,UAAU,IAAK;AAAA,EAC7B,cAAc;AAAA,EACd,WAAW;AAAA,EACX,eAAe;AAAA,EACf;AAAA,EACA,cAAc;AAAA,EACd,UAAU,EAAE,WAAW,UAAU,WAAW,eAAe,QAAQ,SAAS;AAAA,EAC5E,YAAY;AAAA,EACZ,gBAAgB;AAAA,EAChB,MAAM,EAAE,OAAO,GAAG,QAAQ,GAAG,WAAW,GAAG,YAAY,EAAE;AAAA,EACzD,WAAW,EAAE,gBAAgB,MAAM,iBAAiB,MAAM,gBAAgB,KAAM;AAClF;AAEO,SAAS,oBAAoB,UAA0B,CAAC,GAAe;AAC5E,QAAM,kBAAkB,mBAAmB,eAAe;AAC1D,QAAM,gBAAgB,4BAA4B;AAAA,IAChD,KAAK;AAAA,IACL,SAAS,CAAC,OAAO,SAAS,mBACxB,qBAAqB,OAAO,SAAS,gBAAgB,EAAE,GAAG,SAAS,SAAS,MAAM,CAAC;AAAA,EACvF,CAAC;AACD,SAAO,MAAM;AACX,oBAAgB;AAChB,kBAAc;AAAA,EAChB;AACF;",
+  "sourcesContent": ["import type { RealtimeModel } from '../models.ts';\nimport { registerVoiceModel } from '../models.ts';\nimport {\n  type AdapterOptions,\n  connectVendorSession,\n  pcmFormat,\n} from '../openai-realtime/session.ts';\nimport { registerRealtimeApiProvider } from '../registry.ts';\n\nconst voices = [\n  { id: 'eve', name: 'Eve' },\n  { id: 'ara', name: 'Ara' },\n  { id: 'rex', name: 'Rex' },\n  { id: 'sal', name: 'Sal' },\n  { id: 'leo', name: 'Leo' },\n];\n\nexport const grokVoiceLatest: RealtimeModel<'xai-realtime'> = {\n  id: 'grok-voice-latest',\n  name: 'Grok Voice',\n  api: 'xai-realtime',\n  provider: 'xai',\n  baseUrl: 'wss://api.x.ai/v1/realtime',\n  type: 'realtime',\n  transports: ['websocket', 'webrtc'],\n  toolCalls: 'blocking',\n  inputFormat: pcmFormat(24000),\n  outputFormat: pcmFormat(24000),\n  maxSessionMs: 7_200_000,\n  resumable: true,\n  ephemeralAuth: true,\n  voices,\n  provisioning: 'none',\n  delivery: { interrupt: 'native', when_idle: 'approximate', silent: 'native' },\n  truncation: 'client',\n  liveToolUpdate: true,\n  cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },\n  audioCost: { perInputMinute: 0.08, perOutputMinute: 0.08, perTextMessage: 0.004 },\n};\n\nexport function registerXaiRealtime(options: AdapterOptions = {}): () => void {\n  const unregisterModel = registerVoiceModel(grokVoiceLatest);\n  const unregisterApi = registerRealtimeApiProvider({\n    api: 'xai-realtime',\n    connect: (model, context, connectOptions) =>\n      connectVendorSession(model, context, connectOptions, { ...options, dialect: 'xai' }),\n  });\n  return () => {\n    unregisterModel();\n    unregisterApi();\n  };\n}\n\nexport { registerXAI } from './catalog.ts';\n"],
+  "mappings": "AACA,SAAS,0BAA0B;AACnC;AAAA,EAEE;AAAA,EACA;AAAA,OACK;AACP,SAAS,mCAAmC;AAE5C,MAAM,SAAS;AAAA,EACb,EAAE,IAAI,OAAO,MAAM,MAAM;AAAA,EACzB,EAAE,IAAI,OAAO,MAAM,MAAM;AAAA,EACzB,EAAE,IAAI,OAAO,MAAM,MAAM;AAAA,EACzB,EAAE,IAAI,OAAO,MAAM,MAAM;AAAA,EACzB,EAAE,IAAI,OAAO,MAAM,MAAM;AAC3B;AAEO,MAAM,kBAAiD;AAAA,EAC5D,IAAI;AAAA,EACJ,MAAM;AAAA,EACN,KAAK;AAAA,EACL,UAAU;AAAA,EACV,SAAS;AAAA,EACT,MAAM;AAAA,EACN,YAAY,CAAC,aAAa,QAAQ;AAAA,EAClC,WAAW;AAAA,EACX,aAAa,UAAU,IAAK;AAAA,EAC5B,cAAc,UAAU,IAAK;AAAA,EAC7B,cAAc;AAAA,EACd,WAAW;AAAA,EACX,eAAe;AAAA,EACf;AAAA,EACA,cAAc;AAAA,EACd,UAAU,EAAE,WAAW,UAAU,WAAW,eAAe,QAAQ,SAAS;AAAA,EAC5E,YAAY;AAAA,EACZ,gBAAgB;AAAA,EAChB,MAAM,EAAE,OAAO,GAAG,QAAQ,GAAG,WAAW,GAAG,YAAY,EAAE;AAAA,EACzD,WAAW,EAAE,gBAAgB,MAAM,iBAAiB,MAAM,gBAAgB,KAAM;AAClF;AAEO,SAAS,oBAAoB,UAA0B,CAAC,GAAe;AAC5E,QAAM,kBAAkB,mBAAmB,eAAe;AAC1D,QAAM,gBAAgB,4BAA4B;AAAA,IAChD,KAAK;AAAA,IACL,SAAS,CAAC,OAAO,SAAS,mBACxB,qBAAqB,OAAO,SAAS,gBAAgB,EAAE,GAAG,SAAS,SAAS,MAAM,CAAC;AAAA,EACvF,CAAC;AACD,SAAO,MAAM;AACX,oBAAgB;AAChB,kBAAc;AAAA,EAChB;AACF;AAEA,SAAS,mBAAmB;",
   "names": []
 }