@tanstack/ai-grok 0.14.11 → 0.15.0

package/dist/esm/realtime/adapter.js.map

@@ -1 +1 @@
-{"version":3,"file":"adapter.js","names":[],"sources":["../../../src/realtime/adapter.ts"],"sourcesContent":["import { resolveDebugOption } from '@tanstack/ai/adapter-internals'\nimport type {\n  AnyClientTool,\n  AudioVisualization,\n  RealtimeAdapter,\n  RealtimeConnection,\n  RealtimeEvent,\n  RealtimeEventHandler,\n  RealtimeMessage,\n  RealtimeMode,\n  RealtimeSessionConfig,\n  RealtimeToken,\n} from '@tanstack/ai'\nimport type { InternalLogger } from '@tanstack/ai/adapter-internals'\nimport type { GrokRealtimeOptions } from './types'\n\nconst GROK_REALTIME_URL = 'https://api.x.ai/v1/realtime'\n\n/**\n * Runtime-checked field readers for untyped server events. Replace the\n * drive-by `event.X as string` / `event.X as Record<string, unknown>` casts\n * with readers that return `undefined` when the shape doesn't match, so a\n * malformed frame can't throw a TypeError inside `handleServerEvent`.\n */\nfunction readString(\n  obj: Record<string, unknown>,\n  key: string,\n): string | undefined {\n  const value = obj[key]\n  return typeof value === 'string' ? value : undefined\n}\n\nfunction readObject(\n  obj: Record<string, unknown>,\n  key: string,\n): Record<string, unknown> | undefined {\n  const value = obj[key]\n  return value && typeof value === 'object' && !Array.isArray(value)\n    ? (value as Record<string, unknown>)\n    : undefined\n}\n\nfunction readObjectArray(\n  obj: Record<string, unknown>,\n  key: string,\n): Array<Record<string, unknown>> | undefined {\n  const value = obj[key]\n  if (!Array.isArray(value)) return undefined\n  return value.filter(\n    (item): item is Record<string, unknown> =>\n      item !== null && typeof item === 'object' && !Array.isArray(item),\n  )\n}\n\ntype RealtimeServerError = Error & {\n  code?: string\n  type?: string\n  param?: string\n}\n\n/**\n * Creates a Grok realtime adapter for client-side use.\n *\n * Uses WebRTC for browser connections (default). Mirrors the OpenAI realtime\n * adapter because xAI's Voice Agent API is OpenAI-realtime-compatible — the\n * only differences are the endpoint URL and default model.\n *\n * @example\n * ```typescript\n * import { RealtimeClient } from '@tanstack/ai-client'\n * import { grokRealtime } from '@tanstack/ai-grok'\n *\n * const client = new RealtimeClient({\n *   getToken: () => fetch('/api/realtime-token').then(r => r.json()),\n *   adapter: grokRealtime(),\n * })\n * ```\n */\nexport function grokRealtime(\n  options: GrokRealtimeOptions = {},\n): RealtimeAdapter {\n  const connectionMode = options.connectionMode ?? 'webrtc'\n  const logger = resolveDebugOption(options.debug)\n\n  return {\n    provider: 'grok',\n\n    async connect(\n      token: RealtimeToken,\n      _clientTools?: ReadonlyArray<AnyClientTool>,\n    ): Promise<RealtimeConnection> {\n      const model = token.config.model ?? 'grok-voice-fast-1.0'\n      logger.request(`activity=realtime provider=grok model=${model}`, {\n        provider: 'grok',\n        model,\n      })\n\n      if (connectionMode === 'webrtc') {\n        return createWebRTCConnection(token, logger)\n      }\n      const error = new Error('WebSocket connection mode not yet implemented')\n      logger.errors('grok.realtime fatal', {\n        error,\n        source: 'grok.realtime',\n      })\n      throw error\n    },\n  }\n}\n\n/**\n * Creates a WebRTC connection to xAI's realtime API.\n */\nasync function createWebRTCConnection(\n  token: RealtimeToken,\n  logger: InternalLogger,\n): Promise<RealtimeConnection> {\n  const model = token.config.model ?? 'grok-voice-fast-1.0'\n  const eventHandlers = new Map<RealtimeEvent, Set<RealtimeEventHandler<any>>>()\n\n  const pc = new RTCPeerConnection()\n\n  let audioContext: AudioContext | null = null\n  let inputAnalyser: AnalyserNode | null = null\n  let outputAnalyser: AnalyserNode | null = null\n  let inputSource: MediaStreamAudioSourceNode | null = null\n  let outputSource: MediaStreamAudioSourceNode | null = null\n  let localStream: MediaStream | null = null\n\n  let audioElement: HTMLAudioElement | null = null\n\n  // Captured into a const so closures see a non-nullable reference (teardown\n  // re-points the outer `dataChannel` to null, but in-flight closures still\n  // need to close their own channel).\n  const channel = pc.createDataChannel('oai-events')\n  let dataChannel: RTCDataChannel | null = channel\n\n  let currentMode: RealtimeMode = 'idle'\n  let currentMessageId: string | null = null\n\n  // Flipped by `teardownConnection`. Guards `sendEvent` so post-disconnect\n  // calls (e.g. a React `useEffect` cleanup flushing queued events) are\n  // logged and skipped instead of silently piling up in `pendingEvents`.\n  let isTornDown = false\n\n  // Outbound events queued while the data channel isn't yet open. Declared\n  // here (rather than next to `sendEvent`) so `teardownConnection` — which\n  // lives higher up and can run from the SDP-path catch before `sendEvent`\n  // is defined — can drain it without hitting the TDZ.\n  const pendingEvents: Array<Record<string, unknown>> = []\n\n  // Tracks whether we've sent the first session.update. On the first update\n  // we attach a default input_audio_transcription so the server will emit\n  // user transcripts unless the caller opts out via\n  // `providerOptions.inputAudioTranscription = null | false`.\n  let hasSentInitialSessionUpdate = false\n\n  // Size hints for the fallback buffers returned when an analyser isn't yet\n  // populated. We return a *fresh* `Uint8Array` on each call so a caller\n  // that draws into it (e.g. a canvas visualiser zeroing the buffer) can't\n  // mutate a shared module-level instance for every other consumer.\n  const FALLBACK_FREQUENCY_BIN_COUNT = 1024\n  const FALLBACK_TIME_DOMAIN_SIZE = 2048\n  const FALLBACK_TIME_DOMAIN_FILL = 128\n\n  function emit<TEvent extends RealtimeEvent>(\n    event: TEvent,\n    payload: Parameters<RealtimeEventHandler<TEvent>>[0],\n  ) {\n    const handlers = eventHandlers.get(event)\n    if (handlers) {\n      for (const handler of handlers) {\n        handler(payload)\n      }\n    }\n  }\n\n  let dataChannelOpened = false\n  let rejectDataChannelReady: ((reason: unknown) => void) | null = null\n  let dataChannelReadyTimeout: ReturnType<typeof setTimeout> | null = null\n\n  const dataChannelReady = new Promise<void>((resolve, reject) => {\n    rejectDataChannelReady = (reason) => {\n      if (dataChannelReadyTimeout !== null) {\n        clearTimeout(dataChannelReadyTimeout)\n        dataChannelReadyTimeout = null\n      }\n      // One-shot: null out so later state transitions don't reject twice.\n      rejectDataChannelReady = null\n      reject(reason)\n    }\n\n    dataChannelReadyTimeout = setTimeout(() => {\n      if (!dataChannelOpened) {\n        rejectDataChannelReady?.(\n          new Error(\n            'Data channel did not open within 15000ms — aborting connection',\n          ),\n        )\n      }\n    }, 15000)\n\n    channel.onopen = () => {\n      dataChannelOpened = true\n      if (dataChannelReadyTimeout !== null) {\n        clearTimeout(dataChannelReadyTimeout)\n        dataChannelReadyTimeout = null\n      }\n      // Once resolved, rejecting is a no-op — null out so teardown paths\n      // don't attempt a redundant reject on an already-settled promise.\n      rejectDataChannelReady = null\n      flushPendingEvents()\n      emit('status_change', { status: 'connected' })\n      resolve()\n    }\n  })\n\n  channel.onmessage = (event) => {\n    try {\n      const message = JSON.parse(event.data)\n      const messageRecord: Record<string, unknown> =\n        message !== null && typeof message === 'object' ? message : {}\n      logger.provider(\n        `provider=grok direction=in type=${readString(messageRecord, 'type') ?? '<unknown>'}`,\n        { frame: messageRecord },\n      )\n      handleServerEvent(messageRecord)\n    } catch (parseErr) {\n      logger.errors('grok.realtime fatal', {\n        error: parseErr,\n        source: 'grok.realtime',\n      })\n      emit('error', {\n        error:\n          parseErr instanceof Error ? parseErr : new Error(String(parseErr)),\n      })\n    }\n  }\n\n  channel.onerror = (error) => {\n    // Closing the peer connection cascades into `onerror`/`onclose` on the\n    // data channel. Once teardown has started, re-surfacing those as\n    // `emit('error')` is noise that confuses consumers (they just called\n    // `disconnect()` — they don't want an error toast for it).\n    if (isTornDown) return\n    logger.errors('grok.realtime fatal', {\n      error,\n      source: 'grok.realtime',\n    })\n    // RTCErrorEvent exposes a typed `.error`; fall back to the event type\n    // name, then to a string representation, so the emitted error message\n    // doesn't end up as \"[object Event]\".\n    // `onerror` always fires with an Event (often an RTCErrorEvent), so we\n    // can read it via the untyped helpers without first proving object-ness.\n    // oxlint-disable-next-line eslint-js/no-restricted-syntax -- RTCErrorEvent is a typed DOM class that does not structurally overlap Record<string, unknown>; we duck-type it via readObject/readString\n    const errorRecord = error as unknown as Record<string, unknown>\n    const rtcError = readObject(errorRecord, 'error')\n    const msg =\n      (rtcError && readString(rtcError, 'message')) ?? (error.type || 'unknown')\n    const dcErr = new Error(`Data channel error: ${msg}`)\n    if (!dataChannelOpened) {\n      rejectDataChannelReady?.(dcErr)\n    }\n    emit('error', { error: dcErr })\n  }\n\n  channel.onclose = () => {\n    // Same rationale as `onerror` above: `pc.close()` during teardown\n    // cascades to the data channel's `onclose`. If we've already started\n    // teardown, there's nothing to do here.\n    if (isTornDown) return\n    if (!dataChannelOpened) {\n      rejectDataChannelReady?.(new Error('Data channel closed before opening'))\n    }\n  }\n\n  pc.ontrack = (event) => {\n    if (event.track.kind === 'audio' && event.streams[0]) {\n      setupOutputAudioAnalysis(event.streams[0])\n    }\n  }\n\n  // `status_change` has a single source of truth: `onconnectionstatechange`\n  // (the higher-level aggregate state). `oniceconnectionstatechange` is\n  // responsible only for rejecting `dataChannelReady` on ICE failures so we\n  // surface them without waiting for the 15s timeout.\n  pc.onconnectionstatechange = () => {\n    const state = pc.connectionState\n    logger.provider(`provider=grok pc.connectionState=${state}`, {\n      state,\n    })\n    if (state === 'failed' || state === 'disconnected' || state === 'closed') {\n      // Suppress the `status_change` emission when teardown is in progress:\n      // the user-facing `disconnect()` already emits `status_change: 'idle'`\n      // and then calls `teardownConnection()` → `pc.close()`, which fires\n      // `onconnectionstatechange` with state === 'closed'. Without this\n      // guard listeners would see two `idle` events per disconnect.\n      if (!isTornDown) {\n        emit('status_change', {\n          status: state === 'failed' ? 'error' : 'idle',\n        })\n      }\n      if (!dataChannelOpened) {\n        // Reject on any terminal-ish pre-open state so callers don't hang\n        // for the full 15s timeout. The reject is one-shot — subsequent\n        // state changes become no-ops via the null-out in\n        // `rejectDataChannelReady`.\n        const message =\n          state === 'failed'\n            ? `PeerConnection failed before data channel opened`\n            : `PeerConnection entered state '${state}' before data channel opened`\n        rejectDataChannelReady?.(new Error(message))\n      }\n      // Auto-teardown on `failed`: without this the mic track, pc, and\n      // AudioContext stay allocated after a fatal connection failure, so the\n      // browser's mic indicator stays on and the user sees a broken\n      // \"connected mic\" state. `closed` already means pc was torn down\n      // (usually by teardownConnection itself) so nothing extra to do.\n      // `disconnected` is transient per the WebRTC spec and may recover, so\n      // we leave resources in place. `teardownConnection` is idempotent so\n      // a subsequent consumer `disconnect()` remains safe.\n      if (state === 'failed' && !isTornDown) {\n        void teardownConnection()\n      }\n    }\n  }\n\n  pc.oniceconnectionstatechange = () => {\n    const state = pc.iceConnectionState\n    logger.provider(`provider=grok pc.iceConnectionState=${state}`, {\n      state,\n    })\n    if (\n      !dataChannelOpened &&\n      (state === 'failed' || state === 'closed' || state === 'disconnected')\n    ) {\n      const message =\n        state === 'failed'\n          ? `ICE connection failed before data channel opened`\n          : `ICE connection entered state '${state}' before data channel opened`\n      rejectDataChannelReady?.(new Error(message))\n    }\n  }\n\n  /**\n   * Tear down every resource we may have allocated so the mic/pc/audio\n   * nodes/audio element don't leak on a failed connect. Safe to call from\n   * any point after `new RTCPeerConnection()`; each branch null-guards and\n   * swallows errors because cascading closes (e.g. `pc.close()` closing the\n   * data channel implicitly) are expected.\n   *\n   * Shared between the SDP-path catch, the post-SDP catch, and (implicitly\n   * via idempotency) the `disconnect()` entry point.\n   */\n  async function teardownConnection() {\n    // Flip the teardown flag BEFORE any awaits so handlers that fire during\n    // `await audioContext.close()` (or any other async step below) can guard\n    // on it — otherwise a late `pc.onconnectionstatechange` or `pc.ontrack`\n    // can allocate new resources or re-emit `status_change: idle` after the\n    // user-facing `disconnect()` already emitted one.\n    isTornDown = true\n\n    // Drop any queued events the caller sent before the data channel opened\n    // up front. Without this they'd accumulate across reconnect attempts\n    // (each connect allocates a fresh closure, but a caller holding the old\n    // `connection` reference could otherwise keep appending forever). Done\n    // at the top — before the awaits below — so `sendEvent` calls racing\n    // with teardown don't push into a list we're about to drain.\n    pendingEvents.length = 0\n\n    // Clear the data-channel-open timeout / reject the readiness promise\n    // if it's still pending. `rejectDataChannelReady` is one-shot and nulls\n    // itself on first call, so calling it from `disconnect()` after a\n    // successful open is a no-op.\n    rejectDataChannelReady?.(\n      new Error('Connection torn down before data channel opened'),\n    )\n\n    if (localStream) {\n      for (const track of localStream.getTracks()) {\n        track.stop()\n      }\n      localStream = null\n    }\n\n    // Output audio (populated by `pc.ontrack` → setupOutputAudioAnalysis,\n    // which may have fired during SDP negotiation before we threw).\n    if (audioElement) {\n      try {\n        audioElement.pause()\n      } catch {\n        // ignore — element may already be unloaded\n      }\n      audioElement.srcObject = null\n      audioElement = null\n    }\n    if (outputSource) {\n      try {\n        outputSource.disconnect()\n      } catch {\n        // ignore\n      }\n      outputSource = null\n    }\n    if (outputAnalyser) {\n      try {\n        outputAnalyser.disconnect()\n      } catch {\n        // ignore\n      }\n      outputAnalyser = null\n    }\n\n    // Input audio (populated by setupInputAudioAnalysis after SDP).\n    if (inputSource) {\n      try {\n        inputSource.disconnect()\n      } catch {\n        // ignore\n      }\n      inputSource = null\n    }\n    if (inputAnalyser) {\n      try {\n        inputAnalyser.disconnect()\n      } catch {\n        // ignore\n      }\n      inputAnalyser = null\n    }\n\n    if (dataChannel) {\n      try {\n        dataChannel.close()\n      } catch {\n        // ignore — channel may already be closed by pc.close()\n      }\n      dataChannel = null\n    }\n\n    try {\n      pc.close()\n    } catch {\n      // ignore — pc may already be closed\n    }\n\n    if (audioContext) {\n      try {\n        await audioContext.close()\n      } catch {\n        // ignore — context may already be closed\n      }\n      audioContext = null\n    }\n  }\n\n  // xAI requires an audio track in the SDP offer, same as OpenAI realtime.\n  //\n  // This try/catch also covers `getUserMedia` failure (e.g. the user denies\n  // microphone permission). `pc` + `dataChannel` are already allocated above\n  // and the 15s `dataChannelReady` timeout is already armed, so we MUST\n  // teardown on failure here — otherwise they leak until the tab closes.\n  // `teardownConnection` is idempotent and null-safe (runs fine even if the\n  // mic was never acquired).\n  try {\n    try {\n      localStream = await navigator.mediaDevices.getUserMedia({\n        audio: {\n          echoCancellation: true,\n          noiseSuppression: true,\n          sampleRate: 24000,\n        },\n      })\n    } catch (error) {\n      logger.errors('grok.realtime fatal', {\n        error,\n        source: 'grok.realtime.getUserMedia',\n      })\n      // Re-throw with the descriptive message callers rely on. Teardown runs\n      // in the outer catch below.\n      throw new Error(\n        `Microphone access required for realtime voice: ${error instanceof Error ? error.message : error}`,\n      )\n    }\n\n    for (const track of localStream.getAudioTracks()) {\n      pc.addTrack(track, localStream)\n    }\n\n    const offer = await pc.createOffer()\n    await pc.setLocalDescription(offer)\n\n    const sdpResponse = await fetch(`${GROK_REALTIME_URL}?model=${model}`, {\n      method: 'POST',\n      headers: {\n        Authorization: `Bearer ${token.token}`,\n        'Content-Type': 'application/sdp',\n      },\n      body: offer.sdp,\n    })\n\n    if (!sdpResponse.ok) {\n      const errorText = await sdpResponse.text()\n      const error = new Error(\n        `Failed to establish WebRTC connection: ${sdpResponse.status} - ${errorText}`,\n      )\n      logger.errors('grok.realtime fatal', {\n        error,\n        source: 'grok.realtime.sdp',\n        status: sdpResponse.status,\n      })\n      throw error\n    }\n\n    const answerSdp = await sdpResponse.text()\n    await pc.setRemoteDescription({ type: 'answer', sdp: answerSdp })\n  } catch (err) {\n    await teardownConnection()\n    throw err\n  }\n\n  // Second cleanup scope: after SDP succeeds we still have to set up input\n  // audio analysis and wait for the data channel to open. Both can fail\n  // (AudioContext allocation, 15s timeout, ICE failure, pc.close from the\n  // other end, etc.) and those failures must NOT leave the mic/pc/audio\n  // nodes running.\n  try {\n    setupInputAudioAnalysis(localStream)\n    await dataChannelReady\n  } catch (err) {\n    await teardownConnection()\n    throw err\n  }\n\n  function handleServerEvent(event: Record<string, unknown>) {\n    const type = readString(event, 'type')\n\n    switch (type) {\n      case 'session.created':\n      case 'session.updated':\n        break\n\n      case 'input_audio_buffer.speech_started':\n        currentMode = 'listening'\n        emit('mode_change', { mode: 'listening' })\n        break\n\n      case 'input_audio_buffer.speech_stopped':\n        currentMode = 'thinking'\n        emit('mode_change', { mode: 'thinking' })\n        break\n\n      case 'input_audio_buffer.committed':\n        break\n\n      case 'conversation.item.input_audio_transcription.completed': {\n        const transcript = readString(event, 'transcript')\n        if (transcript === undefined) break\n        emit('transcript', { role: 'user', transcript, isFinal: true })\n        break\n      }\n\n      case 'response.created':\n        // Reset message id so a tool-only response (which never emits\n        // response.output_item.added for a message) can't reuse the previous\n        // turn's id when `response.done` later inspects this flag.\n        currentMessageId = null\n        currentMode = 'thinking'\n        emit('mode_change', { mode: 'thinking' })\n        break\n\n      case 'response.output_item.added': {\n        const item = readObject(event, 'item')\n        if (item && readString(item, 'type') === 'message') {\n          const id = readString(item, 'id')\n          if (id !== undefined) currentMessageId = id\n        }\n        break\n      }\n\n      // xAI realtime per docs uses `response.output_audio_transcript.*`;\n      // accept the legacy OpenAI-realtime `response.audio_transcript.*` as\n      // an alias so this adapter stays compatible across protocol versions.\n      case 'response.output_audio_transcript.delta':\n      case 'response.audio_transcript.delta': {\n        const delta = readString(event, 'delta')\n        if (delta === undefined) break\n        emit('transcript', {\n          role: 'assistant',\n          transcript: delta,\n          isFinal: false,\n        })\n        break\n      }\n\n      case 'response.output_audio_transcript.done':\n      case 'response.audio_transcript.done': {\n        const transcript = readString(event, 'transcript')\n        if (transcript === undefined) break\n        emit('transcript', { role: 'assistant', transcript, isFinal: true })\n        break\n      }\n\n      // xAI realtime per docs uses `response.text.*`; accept the legacy\n      // OpenAI-realtime `response.output_text.*` as an alias.\n      case 'response.text.delta':\n      case 'response.output_text.delta': {\n        const delta = readString(event, 'delta')\n        if (delta === undefined) break\n        emit('transcript', {\n          role: 'assistant',\n          transcript: delta,\n          isFinal: false,\n        })\n        break\n      }\n\n      case 'response.text.done':\n      case 'response.output_text.done': {\n        const text = readString(event, 'text')\n        if (text === undefined) break\n        emit('transcript', {\n          role: 'assistant',\n          transcript: text,\n          isFinal: true,\n        })\n        break\n      }\n\n      // xAI realtime per docs uses `response.output_audio.*`; accept the\n      // legacy OpenAI-realtime `response.audio.*` as an alias.\n      case 'response.output_audio.delta':\n      case 'response.audio.delta':\n        if (currentMode !== 'speaking') {\n          currentMode = 'speaking'\n          emit('mode_change', { mode: 'speaking' })\n        }\n        break\n\n      case 'response.output_audio.done':\n      case 'response.audio.done':\n        break\n\n      case 'response.function_call_arguments.done': {\n        // Only `call_id` is valid for `sendToolResult` correlation. Falling\n        // back to `item_id` would produce a tool-call id the server doesn't\n        // recognise when the result is posted back, silently dropping the\n        // tool execution. If `call_id` is missing we surface an error event\n        // so the UI can react instead of pretending the tool call succeeded.\n        const callId = readString(event, 'call_id')\n        const name = readString(event, 'name') ?? ''\n        const args = readString(event, 'arguments') ?? ''\n        if (!callId) {\n          logger.errors(\n            'grok.realtime tool_call missing call_id — dropping tool_call',\n            {\n              source: 'grok.realtime',\n              event_type: 'response.function_call_arguments.done',\n              item_id: event.item_id,\n            },\n          )\n          emit('error', {\n            error: new Error(\n              'Realtime tool call missing call_id; tool will not execute',\n            ),\n          })\n          break\n        }\n        try {\n          const input = JSON.parse(args)\n          emit('tool_call', { toolCallId: callId, toolName: name, input })\n        } catch {\n          emit('tool_call', { toolCallId: callId, toolName: name, input: args })\n        }\n        break\n      }\n\n      case 'response.done': {\n        const response = readObject(event, 'response') ?? {}\n        const output = readObjectArray(response, 'output')\n\n        // Only transition back to `listening` if the user hasn't already\n        // stopped capture — otherwise we'd override their explicit `idle`\n        // state and re-arm the mic visualisation.\n        if (currentMode !== 'idle') {\n          currentMode = 'listening'\n          emit('mode_change', { mode: 'listening' })\n        }\n\n        if (currentMessageId) {\n          const message: RealtimeMessage = {\n            id: currentMessageId,\n            role: 'assistant',\n            timestamp: Date.now(),\n            parts: [],\n          }\n\n          for (const item of output ?? []) {\n            if (readString(item, 'type') !== 'message') continue\n            const content = readObjectArray(item, 'content')\n            if (!content) continue\n            for (const part of content) {\n              const partType = readString(part, 'type')\n              if (partType === 'audio') {\n                const transcript = readString(part, 'transcript')\n                if (transcript) {\n                  message.parts.push({ type: 'audio', transcript })\n                }\n              } else if (partType === 'text') {\n                const content = readString(part, 'text')\n                if (content) {\n                  message.parts.push({ type: 'text', content })\n                }\n              }\n            }\n          }\n\n          emit('message_complete', { message })\n          currentMessageId = null\n        }\n        break\n      }\n\n      case 'conversation.item.truncated':\n        // Assistant playback was interrupted — flip mode back to `listening`\n        // unless the user already called `stopAudioCapture()` (idle). Without\n        // this the visualisation would stay stuck on `speaking` even though\n        // no audio is playing.\n        if (currentMode !== 'idle') {\n          currentMode = 'listening'\n          emit('mode_change', { mode: 'listening' })\n        }\n        emit('interrupted', {\n          ...(currentMessageId !== null && { messageId: currentMessageId }),\n        })\n        break\n\n      case 'error': {\n        // The realtime server's `error` envelope isn't guaranteed to carry\n        // an `error` object at all (network-layer corruption, protocol\n        // drift, etc.). Validate shape before dereferencing so a malformed\n        // payload can't throw a TypeError inside this handler and stop the\n        // switch from running for the rest of the session.\n        const errorObj = readObject(event, 'error') ?? {}\n        const message =\n          readString(errorObj, 'message') ?? 'Unknown realtime server error'\n        const err: RealtimeServerError = new Error(message)\n        // Preserve `code` / `type` / `param` on the Error as extra props so\n        // consumers can branch on them without re-parsing the raw event.\n        const code = readString(errorObj, 'code')\n        if (code !== undefined) err.code = code\n        const errType = readString(errorObj, 'type')\n        if (errType !== undefined) err.type = errType\n        const param = readString(errorObj, 'param')\n        if (param !== undefined) err.param = param\n        logger.errors('grok.realtime server error', {\n          ...errorObj,\n          source: 'grok.realtime server',\n        })\n        emit('error', { error: err })\n        break\n      }\n\n      case undefined:\n      default:\n        // The xAI realtime protocol is a moving target; log unhandled event\n        // types at provider level so they're visible during debugging without\n        // emitting a user-visible error. `undefined` shares the bucket because\n        // a malformed event without a `type` field is just as unhandleable.\n        logger.provider('grok.realtime unhandled server event', {\n          type: event.type,\n        })\n        break\n    }\n  }\n\n  function setupOutputAudioAnalysis(stream: MediaStream) {\n    // Bail out if teardown has already started. `pc.ontrack` can fire\n    // asynchronously after `teardownConnection()` has flipped `isTornDown`\n    // (e.g. a remote track arriving mid-close); without this guard we'd\n    // allocate a fresh AudioContext / audio element that nothing would ever\n    // clean up.\n    if (isTornDown) return\n\n    // Tear down any prior output audio before allocating new resources.\n    // `pc.ontrack` can fire multiple times over the lifetime of a session\n    // (e.g. after renegotiation), and without this we'd leak audio elements\n    // and analyser nodes.\n    if (audioElement) {\n      try {\n        audioElement.pause()\n      } catch {\n        // ignore — element may already be unloaded\n      }\n      audioElement.srcObject = null\n      audioElement = null\n    }\n    if (outputSource) {\n      try {\n        outputSource.disconnect()\n      } catch {\n        // ignore — may already be disconnected\n      }\n      outputSource = null\n    }\n    if (outputAnalyser) {\n      try {\n        outputAnalyser.disconnect()\n      } catch {\n        // ignore\n      }\n      outputAnalyser = null\n    }\n\n    audioElement = new Audio()\n    audioElement.srcObject = stream\n    audioElement.autoplay = true\n    audioElement.play().catch((e) => {\n      // Autoplay is commonly blocked until the user interacts with the page\n      // (browser gesture requirement). Surfacing this as a fatal `error`\n      // event makes the UI render a red/error state even though the\n      // connection is healthy — the page just needs a click. Log at a\n      // dedicated source tag so it's debuggable, but don't emit `error`.\n      logger.errors('grok.realtime audio autoplay blocked', {\n        error: e,\n        source: 'grok.realtime.audio_permission_required',\n      })\n    })\n\n    if (!audioContext) {\n      audioContext = new AudioContext()\n    }\n\n    if (audioContext.state === 'suspended') {\n      audioContext.resume().catch((err) => {\n        // Same rationale as the autoplay catch: `resume()` failure usually\n        // means the user hasn't interacted yet. Logging only — no error\n        // emit — so the UI doesn't go into a fatal state for a recoverable\n        // condition.\n        logger.errors('grok.realtime audioContext.resume failed', {\n          error: err,\n          source: 'grok.realtime',\n        })\n      })\n    }\n\n    outputAnalyser = audioContext.createAnalyser()\n    outputAnalyser.fftSize = 2048\n    outputAnalyser.smoothingTimeConstant = 0.3\n\n    outputSource = audioContext.createMediaStreamSource(stream)\n    outputSource.connect(outputAnalyser)\n  }\n\n  function setupInputAudioAnalysis(stream: MediaStream) {\n    // Defensive symmetry with `setupOutputAudioAnalysis`. Today this is\n    // only called inline after SDP negotiation, but keeping the guard\n    // means any future caller path (e.g. renegotiation) won't leak a fresh\n    // AudioContext after teardown.\n    if (isTornDown) return\n\n    if (!audioContext) {\n      audioContext = new AudioContext()\n    }\n\n    if (audioContext.state === 'suspended') {\n      audioContext.resume().catch((err) => {\n        // Same rationale as in setupOutputAudioAnalysis: a suspended\n        // AudioContext usually resumes after a user gesture. Log only —\n        // surfacing this as a fatal error makes the UI look broken for a\n        // recoverable condition.\n        logger.errors('grok.realtime audioContext.resume failed', {\n          error: err,\n          source: 'grok.realtime',\n        })\n      })\n    }\n\n    inputAnalyser = audioContext.createAnalyser()\n    inputAnalyser.fftSize = 2048\n    inputAnalyser.smoothingTimeConstant = 0.3\n\n    inputSource = audioContext.createMediaStreamSource(stream)\n    inputSource.connect(inputAnalyser)\n  }\n\n  function sendEvent(event: Record<string, unknown>) {\n    if (isTornDown) {\n      // The caller is holding onto a `connection` object after `disconnect()`\n      // (or a failed connect). Silently queueing would leak memory and the\n      // events would never flush. Log + drop so the misuse is visible in\n      // debug mode without escalating to a throw — throwing from a React\n      // useEffect cleanup path can break teardown ordering in the UI.\n      logger.errors('grok.realtime sendEvent after disconnect', {\n        eventType: readString(event, 'type') ?? '<unknown>',\n        source: 'grok.realtime',\n      })\n      return\n    }\n    if (dataChannel?.readyState === 'open') {\n      logger.provider(\n        `provider=grok direction=out type=${readString(event, 'type') ?? '<unknown>'}`,\n        { frame: event },\n      )\n      // Mirror the try/catch in `flushPendingEvents` — `dataChannel.send`\n      // can synchronously throw if the channel flipped to `closing` between\n      // our readyState check and this call, or if `JSON.stringify` chokes\n      // on a caller-supplied payload. Log + emit error instead of letting\n      // the exception propagate up through public `sendText` / `sendImage`\n      // / `updateSession` call sites.\n      try {\n        dataChannel.send(JSON.stringify(event))\n      } catch (error) {\n        logger.errors('grok.realtime sendEvent failed', {\n          error,\n          eventType: readString(event, 'type') ?? '<unknown>',\n          source: 'grok.realtime',\n        })\n        emit('error', {\n          error: error instanceof Error ? error : new Error(String(error)),\n        })\n      }\n    } else {\n      pendingEvents.push(event)\n    }\n  }\n\n  function flushPendingEvents() {\n    try {\n      for (const event of pendingEvents) {\n        logger.provider(\n          `provider=grok direction=out type=${readString(event, 'type') ?? '<unknown>'}`,\n          { frame: event },\n        )\n        channel.send(JSON.stringify(event))\n      }\n      pendingEvents.length = 0\n    } catch (error) {\n      // A send failure here (e.g. dataChannel went from 'open' back to\n      // 'closing' mid-flush, or JSON.stringify on a caller-provided event\n      // threw) would otherwise be silently swallowed. By the time we're\n      // called, `onopen` has already resolved `dataChannelReady`, so the\n      // consumer-facing signal is `emit('error')` — try rejectDataChannelReady\n      // as a defensive belt-and-braces in case this ever runs pre-resolve.\n      logger.errors('grok.realtime flushPendingEvents failed', {\n        error,\n        source: 'grok.realtime',\n      })\n      const err = error instanceof Error ? error : new Error(String(error))\n      rejectDataChannelReady?.(err)\n      emit('error', { error: err })\n    }\n  }\n\n  const connection: RealtimeConnection = {\n    async disconnect() {\n      // Reuse the same teardown path as the failed-connect branches so\n      // every cleanup site stays in sync (input analyser, output analyser,\n      // output source, audio element, etc.).\n      await teardownConnection()\n      emit('status_change', { status: 'idle' })\n    },\n\n    async startAudioCapture() {\n      if (localStream) {\n        for (const track of localStream.getAudioTracks()) {\n          track.enabled = true\n        }\n      }\n      currentMode = 'listening'\n      emit('mode_change', { mode: 'listening' })\n    },\n\n    stopAudioCapture() {\n      if (localStream) {\n        for (const track of localStream.getAudioTracks()) {\n          track.enabled = false\n        }\n      }\n      currentMode = 'idle'\n      emit('mode_change', { mode: 'idle' })\n    },\n\n    sendText(text: string) {\n      sendEvent({\n        type: 'conversation.item.create',\n        item: {\n          type: 'message',\n          role: 'user',\n          content: [{ type: 'input_text', text }],\n        },\n      })\n      sendEvent({ type: 'response.create' })\n    },\n\n    sendImage(imageData: string, mimeType: string) {\n      // Accept:\n      //  - http(s):// URLs → forward as-is\n      //  - data: URIs (e.g. from FileReader.readAsDataURL) → forward as-is\n      //    so we don't double-wrap into `data:image/png;base64,data:image/png;base64,…`\n      //  - bare base64 → wrap in `data:${mimeType};base64,…`\n      const isAlreadyUrlOrDataUri =\n        imageData.startsWith('http://') ||\n        imageData.startsWith('https://') ||\n        imageData.startsWith('data:')\n      const imageContent = {\n        type: 'input_image',\n        // The OpenAI-realtime content part (which this adapter mirrors) nests\n        // the URL under an `image_url: { url: ... }` object, not a bare\n        // string.\n        image_url: {\n          url: isAlreadyUrlOrDataUri\n            ? imageData\n            : `data:${mimeType};base64,${imageData}`,\n        },\n      }\n\n      sendEvent({\n        type: 'conversation.item.create',\n        item: {\n          type: 'message',\n          role: 'user',\n          content: [imageContent],\n        },\n      })\n      sendEvent({ type: 'response.create' })\n    },\n\n    sendToolResult(callId: string, result: string) {\n      sendEvent({\n        type: 'conversation.item.create',\n        item: {\n          type: 'function_call_output',\n          call_id: callId,\n          output: result,\n        },\n      })\n      sendEvent({ type: 'response.create' })\n    },\n\n    updateSession(config: Partial<RealtimeSessionConfig>) {\n      const sessionUpdate: Record<string, unknown> = {}\n\n      if (config.instructions) {\n        sessionUpdate.instructions = config.instructions\n      }\n\n      if (config.voice) {\n        sessionUpdate.voice = config.voice\n      }\n\n      if (config.vadMode) {\n        if (config.vadMode === 'semantic') {\n          sessionUpdate.turn_detection = {\n            type: 'semantic_vad',\n            eagerness: config.semanticEagerness ?? 'medium',\n          }\n        } else if (config.vadMode === 'server') {\n          sessionUpdate.turn_detection = {\n            type: 'server_vad',\n            threshold: config.vadConfig?.threshold ?? 0.5,\n            prefix_padding_ms: config.vadConfig?.prefixPaddingMs ?? 300,\n            silence_duration_ms: config.vadConfig?.silenceDurationMs ?? 500,\n          }\n        } else {\n          sessionUpdate.turn_detection = null\n        }\n      }\n\n      if (config.tools !== undefined) {\n        sessionUpdate.tools = config.tools.map((t) => ({\n          type: 'function',\n          name: t.name,\n          description: t.description,\n          parameters: t.inputSchema ?? { type: 'object', properties: {} },\n        }))\n        sessionUpdate.tool_choice = 'auto'\n      }\n\n      if (config.outputModalities) {\n        sessionUpdate.modalities = config.outputModalities\n      }\n\n      if (config.temperature !== undefined) {\n        sessionUpdate.temperature = config.temperature\n      }\n\n      if (config.maxOutputTokens !== undefined) {\n        sessionUpdate.max_response_output_tokens = config.maxOutputTokens\n      }\n\n      // Let callers forward an explicit `input_audio_transcription` value\n      // through `providerOptions` — including `null` / `false` to disable\n      // the feature. Only apply our `grok-stt` default on the first\n      // session.update and only if the caller hasn't set it themselves.\n      const providerOptions: Record<string, unknown> =\n        config.providerOptions ?? {}\n      const callerTranscription =\n        'inputAudioTranscription' in providerOptions\n          ? providerOptions.inputAudioTranscription\n          : 'input_audio_transcription' in providerOptions\n            ? providerOptions.input_audio_transcription\n            : undefined\n      if (callerTranscription !== undefined) {\n        sessionUpdate.input_audio_transcription =\n          callerTranscription === false ? null : callerTranscription\n      } else if (!hasSentInitialSessionUpdate) {\n        sessionUpdate.input_audio_transcription = { model: 'grok-stt' }\n      }\n\n      if (Object.keys(sessionUpdate).length > 0) {\n        sendEvent({\n          type: 'session.update',\n          session: sessionUpdate,\n        })\n        hasSentInitialSessionUpdate = true\n      }\n    },\n\n    interrupt() {\n      sendEvent({ type: 'response.cancel' })\n      currentMode = 'listening'\n      emit('mode_change', { mode: 'listening' })\n      emit('interrupted', {\n        ...(currentMessageId !== null && { messageId: currentMessageId }),\n      })\n    },\n\n    on<TEvent extends RealtimeEvent>(\n      event: TEvent,\n      handler: RealtimeEventHandler<TEvent>,\n    ): () => void {\n      let handlers = eventHandlers.get(event)\n      if (!handlers) {\n        handlers = new Set()\n        eventHandlers.set(event, handlers)\n      }\n      handlers.add(handler)\n\n      return () => {\n        eventHandlers.get(event)?.delete(handler)\n      }\n    },\n\n    getAudioVisualization(): AudioVisualization {\n      function calculateLevel(analyser: AnalyserNode): number {\n        const data = new Uint8Array(analyser.fftSize)\n        analyser.getByteTimeDomainData(data)\n\n        let maxDeviation = 0\n        for (const sample of data) {\n          const deviation = Math.abs(sample - 128)\n          if (deviation > maxDeviation) {\n            maxDeviation = deviation\n          }\n        }\n\n        const normalized = maxDeviation / 128\n        return Math.min(1, normalized * 1.5)\n      }\n\n      return {\n        get inputLevel() {\n          if (!inputAnalyser) return 0\n          return calculateLevel(inputAnalyser)\n        },\n\n        get outputLevel() {\n          if (!outputAnalyser) return 0\n          return calculateLevel(outputAnalyser)\n        },\n\n        getInputFrequencyData() {\n          if (!inputAnalyser)\n            return new Uint8Array(FALLBACK_FREQUENCY_BIN_COUNT)\n          const data = new Uint8Array(inputAnalyser.frequencyBinCount)\n          inputAnalyser.getByteFrequencyData(data)\n          return data\n        },\n\n        getOutputFrequencyData() {\n          if (!outputAnalyser)\n            return new Uint8Array(FALLBACK_FREQUENCY_BIN_COUNT)\n          const data = new Uint8Array(outputAnalyser.frequencyBinCount)\n          outputAnalyser.getByteFrequencyData(data)\n          return data\n        },\n\n        getInputTimeDomainData() {\n          if (!inputAnalyser)\n            return new Uint8Array(FALLBACK_TIME_DOMAIN_SIZE).fill(\n              FALLBACK_TIME_DOMAIN_FILL,\n            )\n          const data = new Uint8Array(inputAnalyser.fftSize)\n          inputAnalyser.getByteTimeDomainData(data)\n          return data\n        },\n\n        getOutputTimeDomainData() {\n          if (!outputAnalyser)\n            return new Uint8Array(FALLBACK_TIME_DOMAIN_SIZE).fill(\n              FALLBACK_TIME_DOMAIN_FILL,\n            )\n          const data = new Uint8Array(outputAnalyser.fftSize)\n          outputAnalyser.getByteTimeDomainData(data)\n          return data\n        },\n\n        get inputSampleRate() {\n          return 24000\n        },\n\n        get outputSampleRate() {\n          return 24000\n        },\n      }\n    },\n  }\n\n  // `dataChannelReady` was already awaited inside the post-SDP try/catch\n  // above so we can short-circuit on failures with full teardown.\n  return connection\n}\n"],"mappings":";;AAgBA,IAAM,oBAAoB;;;;;;;AAQ1B,SAAS,WACP,KACA,KACoB;CACpB,MAAM,QAAQ,IAAI;CAClB,OAAO,OAAO,UAAU,WAAW,QAAQ,KAAA;AAC7C;AAEA,SAAS,WACP,KACA,KACqC;CACrC,MAAM,QAAQ,IAAI;CAClB,OAAO,SAAS,OAAO,UAAU,YAAY,CAAC,MAAM,QAAQ,KAAK,IAC5D,QACD,KAAA;AACN;AAEA,SAAS,gBACP,KACA,KAC4C;CAC5C,MAAM,QAAQ,IAAI;CAClB,IAAI,CAAC,MAAM,QAAQ,KAAK,GAAG,OAAO,KAAA;CAClC,OAAO,MAAM,QACV,SACC,SAAS,QAAQ,OAAO,SAAS,YAAY,CAAC,MAAM,QAAQ,IAAI,CACpE;AACF;;;;;;;;;;;;;;;;;;;AA0BA,SAAgB,aACd,UAA+B,CAAC,GACf;CACjB,MAAM,iBAAiB,QAAQ,kBAAkB;CACjD,MAAM,SAAS,mBAAmB,QAAQ,KAAK;CAE/C,OAAO;EACL,UAAU;EAEV,MAAM,QACJ,OACA,cAC6B;GAC7B,MAAM,QAAQ,MAAM,OAAO,SAAS;GACpC,OAAO,QAAQ,yCAAyC,SAAS;IAC/D,UAAU;IACV;GACF,CAAC;GAED,IAAI,mBAAmB,UACrB,OAAO,uBAAuB,OAAO,MAAM;GAE7C,MAAM,wBAAQ,IAAI,MAAM,+CAA+C;GACvE,OAAO,OAAO,uBAAuB;IACnC;IACA,QAAQ;GACV,CAAC;GACD,MAAM;EACR;CACF;AACF;;;;AAKA,eAAe,uBACb,OACA,QAC6B;CAC7B,MAAM,QAAQ,MAAM,OAAO,SAAS;CACpC,MAAM,gCAAgB,IAAI,IAAmD;CAE7E,MAAM,KAAK,IAAI,kBAAkB;CAEjC,IAAI,eAAoC;CACxC,IAAI,gBAAqC;CACzC,IAAI,iBAAsC;CAC1C,IAAI,cAAiD;CACrD,IAAI,eAAkD;CACtD,IAAI,cAAkC;CAEtC,IAAI,eAAwC;CAK5C,MAAM,UAAU,GAAG,kBAAkB,YAAY;CACjD,IAAI,cAAqC;CAEzC,IAAI,cAA4B;CAChC,IAAI,mBAAkC;CAKtC,IAAI,aAAa;CAMjB,MAAM,gBAAgD,CAAC;CAMvD,IAAI,8BAA8B;CAMlC,MAAM,+BAA+B;CACrC,MAAM,4BAA4B;CAClC,MAAM,4BAA4B;CAElC,SAAS,KACP,OACA,SACA;EACA,MAAM,WAAW,cAAc,IAAI,KAAK;EACxC,IAAI,UACF,KAAK,MAAM,WAAW,UACpB,QAAQ,OAAO;CAGrB;CAEA,IAAI,oBAAoB;CACxB,IAAI,yBAA6D;CACjE,IAAI,0BAAgE;CAEpE,MAAM,mBAAmB,IAAI,SAAe,SAAS,WAAW;EAC9D,0BAA0B,WAAW;GACnC,IAAI,4BAA4B,MAAM;IACpC,aAAa,uBAAuB;IACpC,0BAA0B;GAC5B;GAEA,yBAAyB;GACzB,OAAO,MAAM;EACf;EAEA,0BAA0B,iBAAiB;GACzC,IAAI,CAAC,mBACH,yCACE,IAAI,MACF,gEACF,CACF;EAEJ,GAAG,IAAK;EAER,QAAQ,eAAe;GACrB,oBAAoB;GACpB,IAAI,4BAA4B,MAAM;IACpC,aAAa,uBAAuB;IACpC,0BAA0B;GAC5B;GAGA,yBAAyB;GACzB,mBAAmB;GACnB,KAAK,iBAAiB,EAAE,QAAQ,YAAY,CAAC;GAC7C,QAAQ;EACV;CACF,CAAC;CAED,QAAQ,aAAa,UAAU;EAC7B,IAAI;GACF,MAAM,UAAU,KAAK,MAAM,MAAM,IAAI;GACrC,MAAM,gBACJ,YAAY,QAAQ,OAAO,YAAY,WAAW,UAAU,CAAC;GAC/D,OAAO,SACL,mCAAmC,WAAW,eAAe,MAAM,KAAK,eACxE,EAAE,OAAO,cAAc,CACzB;GACA,kBAAkB,aAAa;EACjC,SAAS,UAAU;GACjB,OAAO,OAAO,uBAAuB;IACnC,OAAO;IACP,QAAQ;GACV,CAAC;GACD,KAAK,SAAS,EACZ,OACE,oBAAoB,QAAQ,WAAW,IAAI,MAAM,OAAO,QAAQ,CAAC,EACrE,CAAC;EACH;CACF;CAEA,QAAQ,WAAW,UAAU;EAK3B,IAAI,YAAY;EAChB,OAAO,OAAO,uBAAuB;GACnC;GACA,QAAQ;EACV,CAAC;EAQD,MAAM,WAAW,WAAW,OAAa,OAAO;EAChD,MAAM,OACH,YAAY,WAAW,UAAU,SAAS,OAAO,MAAM,QAAQ;EAClE,MAAM,wBAAQ,IAAI,MAAM,uBAAuB,KAAK;EACpD,IAAI,CAAC,mBACH,yBAAyB,KAAK;EAEhC,KAAK,SAAS,EAAE,OAAO,MAAM,CAAC;CAChC;CAEA,QAAQ,gBAAgB;EAItB,IAAI,YAAY;EAChB,IAAI,CAAC,mBACH,yCAAyB,IAAI,MAAM,oCAAoC,CAAC;CAE5E;CAEA,GAAG,WAAW,UAAU;EACtB,IAAI,MAAM,MAAM,SAAS,WAAW,MAAM,QAAQ,IAChD,yBAAyB,MAAM,QAAQ,EAAE;CAE7C;CAMA,GAAG,gCAAgC;EACjC,MAAM,QAAQ,GAAG;EACjB,OAAO,SAAS,oCAAoC,SAAS,EAC3D,MACF,CAAC;EACD,IAAI,UAAU,YAAY,UAAU,kBAAkB,UAAU,UAAU;GAMxE,IAAI,CAAC,YACH,KAAK,iBAAiB,EACpB,QAAQ,UAAU,WAAW,UAAU,OACzC,CAAC;GAEH,IAAI,CAAC,mBAAmB;IAKtB,MAAM,UACJ,UAAU,WACN,qDACA,iCAAiC,MAAM;IAC7C,yBAAyB,IAAI,MAAM,OAAO,CAAC;GAC7C;GASA,IAAI,UAAU,YAAY,CAAC,YACzB,mBAAwB;EAE5B;CACF;CAEA,GAAG,mCAAmC;EACpC,MAAM,QAAQ,GAAG;EACjB,OAAO,SAAS,uCAAuC,SAAS,EAC9D,MACF,CAAC;EACD,IACE,CAAC,sBACA,UAAU,YAAY,UAAU,YAAY,UAAU,iBACvD;GACA,MAAM,UACJ,UAAU,WACN,qDACA,iCAAiC,MAAM;GAC7C,yBAAyB,IAAI,MAAM,OAAO,CAAC;EAC7C;CACF;;;;;;;;;;;CAYA,eAAe,qBAAqB;EAMlC,aAAa;EAQb,cAAc,SAAS;EAMvB,yCACE,IAAI,MAAM,iDAAiD,CAC7D;EAEA,IAAI,aAAa;GACf,KAAK,MAAM,SAAS,YAAY,UAAU,GACxC,MAAM,KAAK;GAEb,cAAc;EAChB;EAIA,IAAI,cAAc;GAChB,IAAI;IACF,aAAa,MAAM;GACrB,QAAQ,CAER;GACA,aAAa,YAAY;GACzB,eAAe;EACjB;EACA,IAAI,cAAc;GAChB,IAAI;IACF,aAAa,WAAW;GAC1B,QAAQ,CAER;GACA,eAAe;EACjB;EACA,IAAI,gBAAgB;GAClB,IAAI;IACF,eAAe,WAAW;GAC5B,QAAQ,CAER;GACA,iBAAiB;EACnB;EAGA,IAAI,aAAa;GACf,IAAI;IACF,YAAY,WAAW;GACzB,QAAQ,CAER;GACA,cAAc;EAChB;EACA,IAAI,eAAe;GACjB,IAAI;IACF,cAAc,WAAW;GAC3B,QAAQ,CAER;GACA,gBAAgB;EAClB;EAEA,IAAI,aAAa;GACf,IAAI;IACF,YAAY,MAAM;GACpB,QAAQ,CAER;GACA,cAAc;EAChB;EAEA,IAAI;GACF,GAAG,MAAM;EACX,QAAQ,CAER;EAEA,IAAI,cAAc;GAChB,IAAI;IACF,MAAM,aAAa,MAAM;GAC3B,QAAQ,CAER;GACA,eAAe;EACjB;CACF;CAUA,IAAI;EACF,IAAI;GACF,cAAc,MAAM,UAAU,aAAa,aAAa,EACtD,OAAO;IACL,kBAAkB;IAClB,kBAAkB;IAClB,YAAY;GACd,EACF,CAAC;EACH,SAAS,OAAO;GACd,OAAO,OAAO,uBAAuB;IACnC;IACA,QAAQ;GACV,CAAC;GAGD,MAAM,IAAI,MACR,kDAAkD,iBAAiB,QAAQ,MAAM,UAAU,OAC7F;EACF;EAEA,KAAK,MAAM,SAAS,YAAY,eAAe,GAC7C,GAAG,SAAS,OAAO,WAAW;EAGhC,MAAM,QAAQ,MAAM,GAAG,YAAY;EACnC,MAAM,GAAG,oBAAoB,KAAK;EAElC,MAAM,cAAc,MAAM,MAAM,GAAG,kBAAkB,SAAS,SAAS;GACrE,QAAQ;GACR,SAAS;IACP,eAAe,UAAU,MAAM;IAC/B,gBAAgB;GAClB;GACA,MAAM,MAAM;EACd,CAAC;EAED,IAAI,CAAC,YAAY,IAAI;GACnB,MAAM,YAAY,MAAM,YAAY,KAAK;GACzC,MAAM,wBAAQ,IAAI,MAChB,0CAA0C,YAAY,OAAO,KAAK,WACpE;GACA,OAAO,OAAO,uBAAuB;IACnC;IACA,QAAQ;IACR,QAAQ,YAAY;GACtB,CAAC;GACD,MAAM;EACR;EAEA,MAAM,YAAY,MAAM,YAAY,KAAK;EACzC,MAAM,GAAG,qBAAqB;GAAE,MAAM;GAAU,KAAK;EAAU,CAAC;CAClE,SAAS,KAAK;EACZ,MAAM,mBAAmB;EACzB,MAAM;CACR;CAOA,IAAI;EACF,wBAAwB,WAAW;EACnC,MAAM;CACR,SAAS,KAAK;EACZ,MAAM,mBAAmB;EACzB,MAAM;CACR;CAEA,SAAS,kBAAkB,OAAgC;EAGzD,QAFa,WAAW,OAAO,MAEvB,GAAR;GACE,KAAK;GACL,KAAK,mBACH;GAEF,KAAK;IACH,cAAc;IACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;IACzC;GAEF,KAAK;IACH,cAAc;IACd,KAAK,eAAe,EAAE,MAAM,WAAW,CAAC;IACxC;GAEF,KAAK,gCACH;GAEF,KAAK,yDAAyD;IAC5D,MAAM,aAAa,WAAW,OAAO,YAAY;IACjD,IAAI,eAAe,KAAA,GAAW;IAC9B,KAAK,cAAc;KAAE,MAAM;KAAQ;KAAY,SAAS;IAAK,CAAC;IAC9D;GACF;GAEA,KAAK;IAIH,mBAAmB;IACnB,cAAc;IACd,KAAK,eAAe,EAAE,MAAM,WAAW,CAAC;IACxC;GAEF,KAAK,8BAA8B;IACjC,MAAM,OAAO,WAAW,OAAO,MAAM;IACrC,IAAI,QAAQ,WAAW,MAAM,MAAM,MAAM,WAAW;KAClD,MAAM,KAAK,WAAW,MAAM,IAAI;KAChC,IAAI,OAAO,KAAA,GAAW,mBAAmB;IAC3C;IACA;GACF;GAKA,KAAK;GACL,KAAK,mCAAmC;IACtC,MAAM,QAAQ,WAAW,OAAO,OAAO;IACvC,IAAI,UAAU,KAAA,GAAW;IACzB,KAAK,cAAc;KACjB,MAAM;KACN,YAAY;KACZ,SAAS;IACX,CAAC;IACD;GACF;GAEA,KAAK;GACL,KAAK,kCAAkC;IACrC,MAAM,aAAa,WAAW,OAAO,YAAY;IACjD,IAAI,eAAe,KAAA,GAAW;IAC9B,KAAK,cAAc;KAAE,MAAM;KAAa;KAAY,SAAS;IAAK,CAAC;IACnE;GACF;GAIA,KAAK;GACL,KAAK,8BAA8B;IACjC,MAAM,QAAQ,WAAW,OAAO,OAAO;IACvC,IAAI,UAAU,KAAA,GAAW;IACzB,KAAK,cAAc;KACjB,MAAM;KACN,YAAY;KACZ,SAAS;IACX,CAAC;IACD;GACF;GAEA,KAAK;GACL,KAAK,6BAA6B;IAChC,MAAM,OAAO,WAAW,OAAO,MAAM;IACrC,IAAI,SAAS,KAAA,GAAW;IACxB,KAAK,cAAc;KACjB,MAAM;KACN,YAAY;KACZ,SAAS;IACX,CAAC;IACD;GACF;GAIA,KAAK;GACL,KAAK;IACH,IAAI,gBAAgB,YAAY;KAC9B,cAAc;KACd,KAAK,eAAe,EAAE,MAAM,WAAW,CAAC;IAC1C;IACA;GAEF,KAAK;GACL,KAAK,uBACH;GAEF,KAAK,yCAAyC;IAM5C,MAAM,SAAS,WAAW,OAAO,SAAS;IAC1C,MAAM,OAAO,WAAW,OAAO,MAAM,KAAK;IAC1C,MAAM,OAAO,WAAW,OAAO,WAAW,KAAK;IAC/C,IAAI,CAAC,QAAQ;KACX,OAAO,OACL,gEACA;MACE,QAAQ;MACR,YAAY;MACZ,SAAS,MAAM;KACjB,CACF;KACA,KAAK,SAAS,EACZ,uBAAO,IAAI,MACT,2DACF,EACF,CAAC;KACD;IACF;IACA,IAAI;KAEF,KAAK,aAAa;MAAE,YAAY;MAAQ,UAAU;MAAM,OAD1C,KAAK,MAAM,IAC+B;KAAM,CAAC;IACjE,QAAQ;KACN,KAAK,aAAa;MAAE,YAAY;MAAQ,UAAU;MAAM,OAAO;KAAK,CAAC;IACvE;IACA;GACF;GAEA,KAAK,iBAAiB;IAEpB,MAAM,SAAS,gBADE,WAAW,OAAO,UAAU,KAAK,CAAC,GACV,QAAQ;IAKjD,IAAI,gBAAgB,QAAQ;KAC1B,cAAc;KACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;IAC3C;IAEA,IAAI,kBAAkB;KACpB,MAAM,UAA2B;MAC/B,IAAI;MACJ,MAAM;MACN,WAAW,KAAK,IAAI;MACpB,OAAO,CAAC;KACV;KAEA,KAAK,MAAM,QAAQ,UAAU,CAAC,GAAG;MAC/B,IAAI,WAAW,MAAM,MAAM,MAAM,WAAW;MAC5C,MAAM,UAAU,gBAAgB,MAAM,SAAS;MAC/C,IAAI,CAAC,SAAS;MACd,KAAK,MAAM,QAAQ,SAAS;OAC1B,MAAM,WAAW,WAAW,MAAM,MAAM;OACxC,IAAI,aAAa,SAAS;QACxB,MAAM,aAAa,WAAW,MAAM,YAAY;QAChD,IAAI,YACF,QAAQ,MAAM,KAAK;SAAE,MAAM;SAAS;QAAW,CAAC;OAEpD,OAAO,IAAI,aAAa,QAAQ;QAC9B,MAAM,UAAU,WAAW,MAAM,MAAM;QACvC,IAAI,SACF,QAAQ,MAAM,KAAK;SAAE,MAAM;SAAQ;QAAQ,CAAC;OAEhD;MACF;KACF;KAEA,KAAK,oBAAoB,EAAE,QAAQ,CAAC;KACpC,mBAAmB;IACrB;IACA;GACF;GAEA,KAAK;IAKH,IAAI,gBAAgB,QAAQ;KAC1B,cAAc;KACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;IAC3C;IACA,KAAK,eAAe,EAClB,GAAI,qBAAqB,QAAQ,EAAE,WAAW,iBAAiB,EACjE,CAAC;IACD;GAEF,KAAK,SAAS;IAMZ,MAAM,WAAW,WAAW,OAAO,OAAO,KAAK,CAAC;IAChD,MAAM,UACJ,WAAW,UAAU,SAAS,KAAK;IACrC,MAAM,MAA2B,IAAI,MAAM,OAAO;IAGlD,MAAM,OAAO,WAAW,UAAU,MAAM;IACxC,IAAI,SAAS,KAAA,GAAW,IAAI,OAAO;IACnC,MAAM,UAAU,WAAW,UAAU,MAAM;IAC3C,IAAI,YAAY,KAAA,GAAW,IAAI,OAAO;IACtC,MAAM,QAAQ,WAAW,UAAU,OAAO;IAC1C,IAAI,UAAU,KAAA,GAAW,IAAI,QAAQ;IACrC,OAAO,OAAO,8BAA8B;KAC1C,GAAG;KACH,QAAQ;IACV,CAAC;IACD,KAAK,SAAS,EAAE,OAAO,IAAI,CAAC;IAC5B;GACF;GAEA,KAAK,KAAA;GACL;IAKE,OAAO,SAAS,wCAAwC,EACtD,MAAM,MAAM,KACd,CAAC;IACD;EACJ;CACF;CAEA,SAAS,yBAAyB,QAAqB;EAMrD,IAAI,YAAY;EAMhB,IAAI,cAAc;GAChB,IAAI;IACF,aAAa,MAAM;GACrB,QAAQ,CAER;GACA,aAAa,YAAY;GACzB,eAAe;EACjB;EACA,IAAI,cAAc;GAChB,IAAI;IACF,aAAa,WAAW;GAC1B,QAAQ,CAER;GACA,eAAe;EACjB;EACA,IAAI,gBAAgB;GAClB,IAAI;IACF,eAAe,WAAW;GAC5B,QAAQ,CAER;GACA,iBAAiB;EACnB;EAEA,eAAe,IAAI,MAAM;EACzB,aAAa,YAAY;EACzB,aAAa,WAAW;EACxB,aAAa,KAAK,CAAC,CAAC,OAAO,MAAM;GAM/B,OAAO,OAAO,wCAAwC;IACpD,OAAO;IACP,QAAQ;GACV,CAAC;EACH,CAAC;EAED,IAAI,CAAC,cACH,eAAe,IAAI,aAAa;EAGlC,IAAI,aAAa,UAAU,aACzB,aAAa,OAAO,CAAC,CAAC,OAAO,QAAQ;GAKnC,OAAO,OAAO,4CAA4C;IACxD,OAAO;IACP,QAAQ;GACV,CAAC;EACH,CAAC;EAGH,iBAAiB,aAAa,eAAe;EAC7C,eAAe,UAAU;EACzB,eAAe,wBAAwB;EAEvC,eAAe,aAAa,wBAAwB,MAAM;EAC1D,aAAa,QAAQ,cAAc;CACrC;CAEA,SAAS,wBAAwB,QAAqB;EAKpD,IAAI,YAAY;EAEhB,IAAI,CAAC,cACH,eAAe,IAAI,aAAa;EAGlC,IAAI,aAAa,UAAU,aACzB,aAAa,OAAO,CAAC,CAAC,OAAO,QAAQ;GAKnC,OAAO,OAAO,4CAA4C;IACxD,OAAO;IACP,QAAQ;GACV,CAAC;EACH,CAAC;EAGH,gBAAgB,aAAa,eAAe;EAC5C,cAAc,UAAU;EACxB,cAAc,wBAAwB;EAEtC,cAAc,aAAa,wBAAwB,MAAM;EACzD,YAAY,QAAQ,aAAa;CACnC;CAEA,SAAS,UAAU,OAAgC;EACjD,IAAI,YAAY;GAMd,OAAO,OAAO,4CAA4C;IACxD,WAAW,WAAW,OAAO,MAAM,KAAK;IACxC,QAAQ;GACV,CAAC;GACD;EACF;EACA,IAAI,aAAa,eAAe,QAAQ;GACtC,OAAO,SACL,oCAAoC,WAAW,OAAO,MAAM,KAAK,eACjE,EAAE,OAAO,MAAM,CACjB;GAOA,IAAI;IACF,YAAY,KAAK,KAAK,UAAU,KAAK,CAAC;GACxC,SAAS,OAAO;IACd,OAAO,OAAO,kCAAkC;KAC9C;KACA,WAAW,WAAW,OAAO,MAAM,KAAK;KACxC,QAAQ;IACV,CAAC;IACD,KAAK,SAAS,EACZ,OAAO,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,KAAK,CAAC,EACjE,CAAC;GACH;EACF,OACE,cAAc,KAAK,KAAK;CAE5B;CAEA,SAAS,qBAAqB;EAC5B,IAAI;GACF,KAAK,MAAM,SAAS,eAAe;IACjC,OAAO,SACL,oCAAoC,WAAW,OAAO,MAAM,KAAK,eACjE,EAAE,OAAO,MAAM,CACjB;IACA,QAAQ,KAAK,KAAK,UAAU,KAAK,CAAC;GACpC;GACA,cAAc,SAAS;EACzB,SAAS,OAAO;GAOd,OAAO,OAAO,2CAA2C;IACvD;IACA,QAAQ;GACV,CAAC;GACD,MAAM,MAAM,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,KAAK,CAAC;GACpE,yBAAyB,GAAG;GAC5B,KAAK,SAAS,EAAE,OAAO,IAAI,CAAC;EAC9B;CACF;CA6QA,OAAO;EA1QL,MAAM,aAAa;GAIjB,MAAM,mBAAmB;GACzB,KAAK,iBAAiB,EAAE,QAAQ,OAAO,CAAC;EAC1C;EAEA,MAAM,oBAAoB;GACxB,IAAI,aACF,KAAK,MAAM,SAAS,YAAY,eAAe,GAC7C,MAAM,UAAU;GAGpB,cAAc;GACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;EAC3C;EAEA,mBAAmB;GACjB,IAAI,aACF,KAAK,MAAM,SAAS,YAAY,eAAe,GAC7C,MAAM,UAAU;GAGpB,cAAc;GACd,KAAK,eAAe,EAAE,MAAM,OAAO,CAAC;EACtC;EAEA,SAAS,MAAc;GACrB,UAAU;IACR,MAAM;IACN,MAAM;KACJ,MAAM;KACN,MAAM;KACN,SAAS,CAAC;MAAE,MAAM;MAAc;KAAK,CAAC;IACxC;GACF,CAAC;GACD,UAAU,EAAE,MAAM,kBAAkB,CAAC;EACvC;EAEA,UAAU,WAAmB,UAAkB;GAsB7C,UAAU;IACR,MAAM;IACN,MAAM;KACJ,MAAM;KACN,MAAM;KACN,SAAS,CAAC;MAhBZ,MAAM;MAIN,WAAW,EACT,KATF,UAAU,WAAW,SAAS,KAC9B,UAAU,WAAW,UAAU,KAC/B,UAAU,WAAW,OAAO,IAQtB,YACA,QAAQ,SAAS,UAAU,YACjC;KAQY,CAAY;IACxB;GACF,CAAC;GACD,UAAU,EAAE,MAAM,kBAAkB,CAAC;EACvC;EAEA,eAAe,QAAgB,QAAgB;GAC7C,UAAU;IACR,MAAM;IACN,MAAM;KACJ,MAAM;KACN,SAAS;KACT,QAAQ;IACV;GACF,CAAC;GACD,UAAU,EAAE,MAAM,kBAAkB,CAAC;EACvC;EAEA,cAAc,QAAwC;GACpD,MAAM,gBAAyC,CAAC;GAEhD,IAAI,OAAO,cACT,cAAc,eAAe,OAAO;GAGtC,IAAI,OAAO,OACT,cAAc,QAAQ,OAAO;GAG/B,IAAI,OAAO,SACT,IAAI,OAAO,YAAY,YACrB,cAAc,iBAAiB;IAC7B,MAAM;IACN,WAAW,OAAO,qBAAqB;GACzC;QACK,IAAI,OAAO,YAAY,UAC5B,cAAc,iBAAiB;IAC7B,MAAM;IACN,WAAW,OAAO,WAAW,aAAa;IAC1C,mBAAmB,OAAO,WAAW,mBAAmB;IACxD,qBAAqB,OAAO,WAAW,qBAAqB;GAC9D;QAEA,cAAc,iBAAiB;GAInC,IAAI,OAAO,UAAU,KAAA,GAAW;IAC9B,cAAc,QAAQ,OAAO,MAAM,KAAK,OAAO;KAC7C,MAAM;KACN,MAAM,EAAE;KACR,aAAa,EAAE;KACf,YAAY,EAAE,eAAe;MAAE,MAAM;MAAU,YAAY,CAAC;KAAE;IAChE,EAAE;IACF,cAAc,cAAc;GAC9B;GAEA,IAAI,OAAO,kBACT,cAAc,aAAa,OAAO;GAGpC,IAAI,OAAO,gBAAgB,KAAA,GACzB,cAAc,cAAc,OAAO;GAGrC,IAAI,OAAO,oBAAoB,KAAA,GAC7B,cAAc,6BAA6B,OAAO;GAOpD,MAAM,kBACJ,OAAO,mBAAmB,CAAC;GAC7B,MAAM,sBACJ,6BAA6B,kBACzB,gBAAgB,0BAChB,+BAA+B,kBAC7B,gBAAgB,4BAChB,KAAA;GACR,IAAI,wBAAwB,KAAA,GAC1B,cAAc,4BACZ,wBAAwB,QAAQ,OAAO;QACpC,IAAI,CAAC,6BACV,cAAc,4BAA4B,EAAE,OAAO,WAAW;GAGhE,IAAI,OAAO,KAAK,aAAa,CAAC,CAAC,SAAS,GAAG;IACzC,UAAU;KACR,MAAM;KACN,SAAS;IACX,CAAC;IACD,8BAA8B;GAChC;EACF;EAEA,YAAY;GACV,UAAU,EAAE,MAAM,kBAAkB,CAAC;GACrC,cAAc;GACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;GACzC,KAAK,eAAe,EAClB,GAAI,qBAAqB,QAAQ,EAAE,WAAW,iBAAiB,EACjE,CAAC;EACH;EAEA,GACE,OACA,SACY;GACZ,IAAI,WAAW,cAAc,IAAI,KAAK;GACtC,IAAI,CAAC,UAAU;IACb,2BAAW,IAAI,IAAI;IACnB,cAAc,IAAI,OAAO,QAAQ;GACnC;GACA,SAAS,IAAI,OAAO;GAEpB,aAAa;IACX,cAAc,IAAI,KAAK,CAAC,EAAE,OAAO,OAAO;GAC1C;EACF;EAEA,wBAA4C;GAC1C,SAAS,eAAe,UAAgC;IACtD,MAAM,OAAO,IAAI,WAAW,SAAS,OAAO;IAC5C,SAAS,sBAAsB,IAAI;IAEnC,IAAI,eAAe;IACnB,KAAK,MAAM,UAAU,MAAM;KACzB,MAAM,YAAY,KAAK,IAAI,SAAS,GAAG;KACvC,IAAI,YAAY,cACd,eAAe;IAEnB;IAEA,MAAM,aAAa,eAAe;IAClC,OAAO,KAAK,IAAI,GAAG,aAAa,GAAG;GACrC;GAEA,OAAO;IACL,IAAI,aAAa;KACf,IAAI,CAAC,eAAe,OAAO;KAC3B,OAAO,eAAe,aAAa;IACrC;IAEA,IAAI,cAAc;KAChB,IAAI,CAAC,gBAAgB,OAAO;KAC5B,OAAO,eAAe,cAAc;IACtC;IAEA,wBAAwB;KACtB,IAAI,CAAC,eACH,OAAO,IAAI,WAAW,4BAA4B;KACpD,MAAM,OAAO,IAAI,WAAW,cAAc,iBAAiB;KAC3D,cAAc,qBAAqB,IAAI;KACvC,OAAO;IACT;IAEA,yBAAyB;KACvB,IAAI,CAAC,gBACH,OAAO,IAAI,WAAW,4BAA4B;KACpD,MAAM,OAAO,IAAI,WAAW,eAAe,iBAAiB;KAC5D,eAAe,qBAAqB,IAAI;KACxC,OAAO;IACT;IAEA,yBAAyB;KACvB,IAAI,CAAC,eACH,OAAO,IAAI,WAAW,yBAAyB,CAAC,CAAC,KAC/C,yBACF;KACF,MAAM,OAAO,IAAI,WAAW,cAAc,OAAO;KACjD,cAAc,sBAAsB,IAAI;KACxC,OAAO;IACT;IAEA,0BAA0B;KACxB,IAAI,CAAC,gBACH,OAAO,IAAI,WAAW,yBAAyB,CAAC,CAAC,KAC/C,yBACF;KACF,MAAM,OAAO,IAAI,WAAW,eAAe,OAAO;KAClD,eAAe,sBAAsB,IAAI;KACzC,OAAO;IACT;IAEA,IAAI,kBAAkB;KACpB,OAAO;IACT;IAEA,IAAI,mBAAmB;KACrB,OAAO;IACT;GACF;EACF;CAKK;AACT"}
+{"version":3,"file":"adapter.js","names":[],"sources":["../../../src/realtime/adapter.ts"],"sourcesContent":["import { resolveDebugOption } from '@tanstack/ai/adapter-internals'\nimport { GROK_DEFAULT_REALTIME_MODEL } from '../model-meta'\nimport type {\n  AnyClientTool,\n  AudioVisualization,\n  RealtimeAdapter,\n  RealtimeConnection,\n  RealtimeEvent,\n  RealtimeEventHandler,\n  RealtimeMessage,\n  RealtimeMode,\n  RealtimeSessionConfig,\n  RealtimeToken,\n} from '@tanstack/ai'\nimport type { InternalLogger } from '@tanstack/ai/adapter-internals'\nimport type { GrokRealtimeOptions } from './types'\n\nconst GROK_REALTIME_URL = 'https://api.x.ai/v1/realtime'\n\n/**\n * Runtime-checked field readers for untyped server events. Replace the\n * drive-by `event.X as string` / `event.X as Record<string, unknown>` casts\n * with readers that return `undefined` when the shape doesn't match, so a\n * malformed frame can't throw a TypeError inside `handleServerEvent`.\n */\nfunction readString(\n  obj: Record<string, unknown>,\n  key: string,\n): string | undefined {\n  const value = obj[key]\n  return typeof value === 'string' ? value : undefined\n}\n\nfunction readObject(\n  obj: Record<string, unknown>,\n  key: string,\n): Record<string, unknown> | undefined {\n  const value = obj[key]\n  return value && typeof value === 'object' && !Array.isArray(value)\n    ? (value as Record<string, unknown>)\n    : undefined\n}\n\nfunction readObjectArray(\n  obj: Record<string, unknown>,\n  key: string,\n): Array<Record<string, unknown>> | undefined {\n  const value = obj[key]\n  if (!Array.isArray(value)) return undefined\n  return value.filter(\n    (item): item is Record<string, unknown> =>\n      item !== null && typeof item === 'object' && !Array.isArray(item),\n  )\n}\n\ntype RealtimeServerError = Error & {\n  code?: string\n  type?: string\n  param?: string\n}\n\n/**\n * Creates a Grok realtime adapter for client-side use.\n *\n * Uses WebRTC for browser connections (default). Mirrors the OpenAI realtime\n * adapter because xAI's Voice Agent API is OpenAI-realtime-compatible — the\n * only differences are the endpoint URL and default model.\n *\n * @example\n * ```typescript\n * import { RealtimeClient } from '@tanstack/ai-client'\n * import { grokRealtime } from '@tanstack/ai-grok'\n *\n * const client = new RealtimeClient({\n *   getToken: () => fetch('/api/realtime-token').then(r => r.json()),\n *   adapter: grokRealtime(),\n * })\n * ```\n */\nexport function grokRealtime(\n  options: GrokRealtimeOptions = {},\n): RealtimeAdapter {\n  const connectionMode = options.connectionMode ?? 'webrtc'\n  const logger = resolveDebugOption(options.debug)\n\n  return {\n    provider: 'grok',\n\n    async connect(\n      token: RealtimeToken,\n      _clientTools?: ReadonlyArray<AnyClientTool>,\n    ): Promise<RealtimeConnection> {\n      const model = token.config.model ?? GROK_DEFAULT_REALTIME_MODEL\n      logger.request(`activity=realtime provider=grok model=${model}`, {\n        provider: 'grok',\n        model,\n      })\n\n      if (connectionMode === 'webrtc') {\n        return createWebRTCConnection(token, logger)\n      }\n      const error = new Error('WebSocket connection mode not yet implemented')\n      logger.errors('grok.realtime fatal', {\n        error,\n        source: 'grok.realtime',\n      })\n      throw error\n    },\n  }\n}\n\n/**\n * Creates a WebRTC connection to xAI's realtime API.\n */\nasync function createWebRTCConnection(\n  token: RealtimeToken,\n  logger: InternalLogger,\n): Promise<RealtimeConnection> {\n  const model = token.config.model ?? GROK_DEFAULT_REALTIME_MODEL\n  const eventHandlers = new Map<RealtimeEvent, Set<RealtimeEventHandler<any>>>()\n\n  const pc = new RTCPeerConnection()\n\n  let audioContext: AudioContext | null = null\n  let inputAnalyser: AnalyserNode | null = null\n  let outputAnalyser: AnalyserNode | null = null\n  let inputSource: MediaStreamAudioSourceNode | null = null\n  let outputSource: MediaStreamAudioSourceNode | null = null\n  let localStream: MediaStream | null = null\n\n  let audioElement: HTMLAudioElement | null = null\n\n  // Captured into a const so closures see a non-nullable reference (teardown\n  // re-points the outer `dataChannel` to null, but in-flight closures still\n  // need to close their own channel).\n  const channel = pc.createDataChannel('oai-events')\n  let dataChannel: RTCDataChannel | null = channel\n\n  let currentMode: RealtimeMode = 'idle'\n  let currentMessageId: string | null = null\n\n  // Flipped by `teardownConnection`. Guards `sendEvent` so post-disconnect\n  // calls (e.g. a React `useEffect` cleanup flushing queued events) are\n  // logged and skipped instead of silently piling up in `pendingEvents`.\n  let isTornDown = false\n\n  // Outbound events queued while the data channel isn't yet open. Declared\n  // here (rather than next to `sendEvent`) so `teardownConnection` — which\n  // lives higher up and can run from the SDP-path catch before `sendEvent`\n  // is defined — can drain it without hitting the TDZ.\n  const pendingEvents: Array<Record<string, unknown>> = []\n\n  // Tracks whether we've sent the first session.update. On the first update\n  // we attach a default input_audio_transcription so the server will emit\n  // user transcripts unless the caller opts out via\n  // `providerOptions.inputAudioTranscription = null | false`.\n  let hasSentInitialSessionUpdate = false\n\n  // Size hints for the fallback buffers returned when an analyser isn't yet\n  // populated. We return a *fresh* `Uint8Array` on each call so a caller\n  // that draws into it (e.g. a canvas visualiser zeroing the buffer) can't\n  // mutate a shared module-level instance for every other consumer.\n  const FALLBACK_FREQUENCY_BIN_COUNT = 1024\n  const FALLBACK_TIME_DOMAIN_SIZE = 2048\n  const FALLBACK_TIME_DOMAIN_FILL = 128\n\n  function emit<TEvent extends RealtimeEvent>(\n    event: TEvent,\n    payload: Parameters<RealtimeEventHandler<TEvent>>[0],\n  ) {\n    const handlers = eventHandlers.get(event)\n    if (handlers) {\n      for (const handler of handlers) {\n        handler(payload)\n      }\n    }\n  }\n\n  let dataChannelOpened = false\n  let rejectDataChannelReady: ((reason: unknown) => void) | null = null\n  let dataChannelReadyTimeout: ReturnType<typeof setTimeout> | null = null\n\n  const dataChannelReady = new Promise<void>((resolve, reject) => {\n    rejectDataChannelReady = (reason) => {\n      if (dataChannelReadyTimeout !== null) {\n        clearTimeout(dataChannelReadyTimeout)\n        dataChannelReadyTimeout = null\n      }\n      // One-shot: null out so later state transitions don't reject twice.\n      rejectDataChannelReady = null\n      reject(reason)\n    }\n\n    dataChannelReadyTimeout = setTimeout(() => {\n      if (!dataChannelOpened) {\n        rejectDataChannelReady?.(\n          new Error(\n            'Data channel did not open within 15000ms — aborting connection',\n          ),\n        )\n      }\n    }, 15000)\n\n    channel.onopen = () => {\n      dataChannelOpened = true\n      if (dataChannelReadyTimeout !== null) {\n        clearTimeout(dataChannelReadyTimeout)\n        dataChannelReadyTimeout = null\n      }\n      // Once resolved, rejecting is a no-op — null out so teardown paths\n      // don't attempt a redundant reject on an already-settled promise.\n      rejectDataChannelReady = null\n      flushPendingEvents()\n      emit('status_change', { status: 'connected' })\n      resolve()\n    }\n  })\n\n  channel.onmessage = (event) => {\n    try {\n      const message = JSON.parse(event.data)\n      const messageRecord: Record<string, unknown> =\n        message !== null && typeof message === 'object' ? message : {}\n      logger.provider(\n        `provider=grok direction=in type=${readString(messageRecord, 'type') ?? '<unknown>'}`,\n        { frame: messageRecord },\n      )\n      handleServerEvent(messageRecord)\n    } catch (parseErr) {\n      logger.errors('grok.realtime fatal', {\n        error: parseErr,\n        source: 'grok.realtime',\n      })\n      emit('error', {\n        error:\n          parseErr instanceof Error ? parseErr : new Error(String(parseErr)),\n      })\n    }\n  }\n\n  channel.onerror = (error) => {\n    // Closing the peer connection cascades into `onerror`/`onclose` on the\n    // data channel. Once teardown has started, re-surfacing those as\n    // `emit('error')` is noise that confuses consumers (they just called\n    // `disconnect()` — they don't want an error toast for it).\n    if (isTornDown) return\n    logger.errors('grok.realtime fatal', {\n      error,\n      source: 'grok.realtime',\n    })\n    // RTCErrorEvent exposes a typed `.error`; fall back to the event type\n    // name, then to a string representation, so the emitted error message\n    // doesn't end up as \"[object Event]\".\n    // `onerror` always fires with an Event (often an RTCErrorEvent), so we\n    // can read it via the untyped helpers without first proving object-ness.\n    // oxlint-disable-next-line eslint-js/no-restricted-syntax -- RTCErrorEvent is a typed DOM class that does not structurally overlap Record<string, unknown>; we duck-type it via readObject/readString\n    const errorRecord = error as unknown as Record<string, unknown>\n    const rtcError = readObject(errorRecord, 'error')\n    const msg =\n      (rtcError && readString(rtcError, 'message')) ?? (error.type || 'unknown')\n    const dcErr = new Error(`Data channel error: ${msg}`)\n    if (!dataChannelOpened) {\n      rejectDataChannelReady?.(dcErr)\n    }\n    emit('error', { error: dcErr })\n  }\n\n  channel.onclose = () => {\n    // Same rationale as `onerror` above: `pc.close()` during teardown\n    // cascades to the data channel's `onclose`. If we've already started\n    // teardown, there's nothing to do here.\n    if (isTornDown) return\n    if (!dataChannelOpened) {\n      rejectDataChannelReady?.(new Error('Data channel closed before opening'))\n    }\n  }\n\n  pc.ontrack = (event) => {\n    if (event.track.kind === 'audio' && event.streams[0]) {\n      setupOutputAudioAnalysis(event.streams[0])\n    }\n  }\n\n  // `status_change` has a single source of truth: `onconnectionstatechange`\n  // (the higher-level aggregate state). `oniceconnectionstatechange` is\n  // responsible only for rejecting `dataChannelReady` on ICE failures so we\n  // surface them without waiting for the 15s timeout.\n  pc.onconnectionstatechange = () => {\n    const state = pc.connectionState\n    logger.provider(`provider=grok pc.connectionState=${state}`, {\n      state,\n    })\n    if (state === 'failed' || state === 'disconnected' || state === 'closed') {\n      // Suppress the `status_change` emission when teardown is in progress:\n      // the user-facing `disconnect()` already emits `status_change: 'idle'`\n      // and then calls `teardownConnection()` → `pc.close()`, which fires\n      // `onconnectionstatechange` with state === 'closed'. Without this\n      // guard listeners would see two `idle` events per disconnect.\n      if (!isTornDown) {\n        emit('status_change', {\n          status: state === 'failed' ? 'error' : 'idle',\n        })\n      }\n      if (!dataChannelOpened) {\n        // Reject on any terminal-ish pre-open state so callers don't hang\n        // for the full 15s timeout. The reject is one-shot — subsequent\n        // state changes become no-ops via the null-out in\n        // `rejectDataChannelReady`.\n        const message =\n          state === 'failed'\n            ? `PeerConnection failed before data channel opened`\n            : `PeerConnection entered state '${state}' before data channel opened`\n        rejectDataChannelReady?.(new Error(message))\n      }\n      // Auto-teardown on `failed`: without this the mic track, pc, and\n      // AudioContext stay allocated after a fatal connection failure, so the\n      // browser's mic indicator stays on and the user sees a broken\n      // \"connected mic\" state. `closed` already means pc was torn down\n      // (usually by teardownConnection itself) so nothing extra to do.\n      // `disconnected` is transient per the WebRTC spec and may recover, so\n      // we leave resources in place. `teardownConnection` is idempotent so\n      // a subsequent consumer `disconnect()` remains safe.\n      if (state === 'failed' && !isTornDown) {\n        void teardownConnection()\n      }\n    }\n  }\n\n  pc.oniceconnectionstatechange = () => {\n    const state = pc.iceConnectionState\n    logger.provider(`provider=grok pc.iceConnectionState=${state}`, {\n      state,\n    })\n    if (\n      !dataChannelOpened &&\n      (state === 'failed' || state === 'closed' || state === 'disconnected')\n    ) {\n      const message =\n        state === 'failed'\n          ? `ICE connection failed before data channel opened`\n          : `ICE connection entered state '${state}' before data channel opened`\n      rejectDataChannelReady?.(new Error(message))\n    }\n  }\n\n  /**\n   * Tear down every resource we may have allocated so the mic/pc/audio\n   * nodes/audio element don't leak on a failed connect. Safe to call from\n   * any point after `new RTCPeerConnection()`; each branch null-guards and\n   * swallows errors because cascading closes (e.g. `pc.close()` closing the\n   * data channel implicitly) are expected.\n   *\n   * Shared between the SDP-path catch, the post-SDP catch, and (implicitly\n   * via idempotency) the `disconnect()` entry point.\n   */\n  async function teardownConnection() {\n    // Flip the teardown flag BEFORE any awaits so handlers that fire during\n    // `await audioContext.close()` (or any other async step below) can guard\n    // on it — otherwise a late `pc.onconnectionstatechange` or `pc.ontrack`\n    // can allocate new resources or re-emit `status_change: idle` after the\n    // user-facing `disconnect()` already emitted one.\n    isTornDown = true\n\n    // Drop any queued events the caller sent before the data channel opened\n    // up front. Without this they'd accumulate across reconnect attempts\n    // (each connect allocates a fresh closure, but a caller holding the old\n    // `connection` reference could otherwise keep appending forever). Done\n    // at the top — before the awaits below — so `sendEvent` calls racing\n    // with teardown don't push into a list we're about to drain.\n    pendingEvents.length = 0\n\n    // Clear the data-channel-open timeout / reject the readiness promise\n    // if it's still pending. `rejectDataChannelReady` is one-shot and nulls\n    // itself on first call, so calling it from `disconnect()` after a\n    // successful open is a no-op.\n    rejectDataChannelReady?.(\n      new Error('Connection torn down before data channel opened'),\n    )\n\n    if (localStream) {\n      for (const track of localStream.getTracks()) {\n        track.stop()\n      }\n      localStream = null\n    }\n\n    // Output audio (populated by `pc.ontrack` → setupOutputAudioAnalysis,\n    // which may have fired during SDP negotiation before we threw).\n    if (audioElement) {\n      try {\n        audioElement.pause()\n      } catch {\n        // ignore — element may already be unloaded\n      }\n      audioElement.srcObject = null\n      audioElement = null\n    }\n    if (outputSource) {\n      try {\n        outputSource.disconnect()\n      } catch {\n        // ignore\n      }\n      outputSource = null\n    }\n    if (outputAnalyser) {\n      try {\n        outputAnalyser.disconnect()\n      } catch {\n        // ignore\n      }\n      outputAnalyser = null\n    }\n\n    // Input audio (populated by setupInputAudioAnalysis after SDP).\n    if (inputSource) {\n      try {\n        inputSource.disconnect()\n      } catch {\n        // ignore\n      }\n      inputSource = null\n    }\n    if (inputAnalyser) {\n      try {\n        inputAnalyser.disconnect()\n      } catch {\n        // ignore\n      }\n      inputAnalyser = null\n    }\n\n    if (dataChannel) {\n      try {\n        dataChannel.close()\n      } catch {\n        // ignore — channel may already be closed by pc.close()\n      }\n      dataChannel = null\n    }\n\n    try {\n      pc.close()\n    } catch {\n      // ignore — pc may already be closed\n    }\n\n    if (audioContext) {\n      try {\n        await audioContext.close()\n      } catch {\n        // ignore — context may already be closed\n      }\n      audioContext = null\n    }\n  }\n\n  // xAI requires an audio track in the SDP offer, same as OpenAI realtime.\n  //\n  // This try/catch also covers `getUserMedia` failure (e.g. the user denies\n  // microphone permission). `pc` + `dataChannel` are already allocated above\n  // and the 15s `dataChannelReady` timeout is already armed, so we MUST\n  // teardown on failure here — otherwise they leak until the tab closes.\n  // `teardownConnection` is idempotent and null-safe (runs fine even if the\n  // mic was never acquired).\n  try {\n    try {\n      localStream = await navigator.mediaDevices.getUserMedia({\n        audio: {\n          echoCancellation: true,\n          noiseSuppression: true,\n          sampleRate: 24000,\n        },\n      })\n    } catch (error) {\n      logger.errors('grok.realtime fatal', {\n        error,\n        source: 'grok.realtime.getUserMedia',\n      })\n      // Re-throw with the descriptive message callers rely on. Teardown runs\n      // in the outer catch below.\n      throw new Error(\n        `Microphone access required for realtime voice: ${error instanceof Error ? error.message : error}`,\n      )\n    }\n\n    for (const track of localStream.getAudioTracks()) {\n      pc.addTrack(track, localStream)\n    }\n\n    const offer = await pc.createOffer()\n    await pc.setLocalDescription(offer)\n\n    const sdpResponse = await fetch(`${GROK_REALTIME_URL}?model=${model}`, {\n      method: 'POST',\n      headers: {\n        Authorization: `Bearer ${token.token}`,\n        'Content-Type': 'application/sdp',\n      },\n      body: offer.sdp,\n    })\n\n    if (!sdpResponse.ok) {\n      const errorText = await sdpResponse.text()\n      const error = new Error(\n        `Failed to establish WebRTC connection: ${sdpResponse.status} - ${errorText}`,\n      )\n      logger.errors('grok.realtime fatal', {\n        error,\n        source: 'grok.realtime.sdp',\n        status: sdpResponse.status,\n      })\n      throw error\n    }\n\n    const answerSdp = await sdpResponse.text()\n    await pc.setRemoteDescription({ type: 'answer', sdp: answerSdp })\n  } catch (err) {\n    await teardownConnection()\n    throw err\n  }\n\n  // Second cleanup scope: after SDP succeeds we still have to set up input\n  // audio analysis and wait for the data channel to open. Both can fail\n  // (AudioContext allocation, 15s timeout, ICE failure, pc.close from the\n  // other end, etc.) and those failures must NOT leave the mic/pc/audio\n  // nodes running.\n  try {\n    setupInputAudioAnalysis(localStream)\n    await dataChannelReady\n  } catch (err) {\n    await teardownConnection()\n    throw err\n  }\n\n  function handleServerEvent(event: Record<string, unknown>) {\n    const type = readString(event, 'type')\n\n    switch (type) {\n      case 'session.created':\n      case 'session.updated':\n        break\n\n      case 'input_audio_buffer.speech_started':\n        currentMode = 'listening'\n        emit('mode_change', { mode: 'listening' })\n        break\n\n      case 'input_audio_buffer.speech_stopped':\n        currentMode = 'thinking'\n        emit('mode_change', { mode: 'thinking' })\n        break\n\n      case 'input_audio_buffer.committed':\n        break\n\n      case 'conversation.item.input_audio_transcription.completed': {\n        const transcript = readString(event, 'transcript')\n        if (transcript === undefined) break\n        emit('transcript', { role: 'user', transcript, isFinal: true })\n        break\n      }\n\n      case 'response.created':\n        // Reset message id so a tool-only response (which never emits\n        // response.output_item.added for a message) can't reuse the previous\n        // turn's id when `response.done` later inspects this flag.\n        currentMessageId = null\n        currentMode = 'thinking'\n        emit('mode_change', { mode: 'thinking' })\n        break\n\n      case 'response.output_item.added': {\n        const item = readObject(event, 'item')\n        if (item && readString(item, 'type') === 'message') {\n          const id = readString(item, 'id')\n          if (id !== undefined) currentMessageId = id\n        }\n        break\n      }\n\n      // xAI realtime per docs uses `response.output_audio_transcript.*`;\n      // accept the legacy OpenAI-realtime `response.audio_transcript.*` as\n      // an alias so this adapter stays compatible across protocol versions.\n      case 'response.output_audio_transcript.delta':\n      case 'response.audio_transcript.delta': {\n        const delta = readString(event, 'delta')\n        if (delta === undefined) break\n        emit('transcript', {\n          role: 'assistant',\n          transcript: delta,\n          isFinal: false,\n        })\n        break\n      }\n\n      case 'response.output_audio_transcript.done':\n      case 'response.audio_transcript.done': {\n        const transcript = readString(event, 'transcript')\n        if (transcript === undefined) break\n        emit('transcript', { role: 'assistant', transcript, isFinal: true })\n        break\n      }\n\n      // xAI realtime per docs uses `response.text.*`; accept the legacy\n      // OpenAI-realtime `response.output_text.*` as an alias.\n      case 'response.text.delta':\n      case 'response.output_text.delta': {\n        const delta = readString(event, 'delta')\n        if (delta === undefined) break\n        emit('transcript', {\n          role: 'assistant',\n          transcript: delta,\n          isFinal: false,\n        })\n        break\n      }\n\n      case 'response.text.done':\n      case 'response.output_text.done': {\n        const text = readString(event, 'text')\n        if (text === undefined) break\n        emit('transcript', {\n          role: 'assistant',\n          transcript: text,\n          isFinal: true,\n        })\n        break\n      }\n\n      // xAI realtime per docs uses `response.output_audio.*`; accept the\n      // legacy OpenAI-realtime `response.audio.*` as an alias.\n      case 'response.output_audio.delta':\n      case 'response.audio.delta':\n        if (currentMode !== 'speaking') {\n          currentMode = 'speaking'\n          emit('mode_change', { mode: 'speaking' })\n        }\n        break\n\n      case 'response.output_audio.done':\n      case 'response.audio.done':\n        break\n\n      case 'response.function_call_arguments.done': {\n        // Only `call_id` is valid for `sendToolResult` correlation. Falling\n        // back to `item_id` would produce a tool-call id the server doesn't\n        // recognise when the result is posted back, silently dropping the\n        // tool execution. If `call_id` is missing we surface an error event\n        // so the UI can react instead of pretending the tool call succeeded.\n        const callId = readString(event, 'call_id')\n        const name = readString(event, 'name') ?? ''\n        const args = readString(event, 'arguments') ?? ''\n        if (!callId) {\n          logger.errors(\n            'grok.realtime tool_call missing call_id — dropping tool_call',\n            {\n              source: 'grok.realtime',\n              event_type: 'response.function_call_arguments.done',\n              item_id: event.item_id,\n            },\n          )\n          emit('error', {\n            error: new Error(\n              'Realtime tool call missing call_id; tool will not execute',\n            ),\n          })\n          break\n        }\n        try {\n          const input = JSON.parse(args)\n          emit('tool_call', { toolCallId: callId, toolName: name, input })\n        } catch {\n          emit('tool_call', { toolCallId: callId, toolName: name, input: args })\n        }\n        break\n      }\n\n      case 'response.done': {\n        const response = readObject(event, 'response') ?? {}\n        const output = readObjectArray(response, 'output')\n\n        // Only transition back to `listening` if the user hasn't already\n        // stopped capture — otherwise we'd override their explicit `idle`\n        // state and re-arm the mic visualisation.\n        if (currentMode !== 'idle') {\n          currentMode = 'listening'\n          emit('mode_change', { mode: 'listening' })\n        }\n\n        if (currentMessageId) {\n          const message: RealtimeMessage = {\n            id: currentMessageId,\n            role: 'assistant',\n            timestamp: Date.now(),\n            parts: [],\n          }\n\n          for (const item of output ?? []) {\n            if (readString(item, 'type') !== 'message') continue\n            const content = readObjectArray(item, 'content')\n            if (!content) continue\n            for (const part of content) {\n              const partType = readString(part, 'type')\n              if (partType === 'audio') {\n                const transcript = readString(part, 'transcript')\n                if (transcript) {\n                  message.parts.push({ type: 'audio', transcript })\n                }\n              } else if (partType === 'text') {\n                const content = readString(part, 'text')\n                if (content) {\n                  message.parts.push({ type: 'text', content })\n                }\n              }\n            }\n          }\n\n          emit('message_complete', { message })\n          currentMessageId = null\n        }\n        break\n      }\n\n      case 'conversation.item.truncated':\n        // Assistant playback was interrupted — flip mode back to `listening`\n        // unless the user already called `stopAudioCapture()` (idle). Without\n        // this the visualisation would stay stuck on `speaking` even though\n        // no audio is playing.\n        if (currentMode !== 'idle') {\n          currentMode = 'listening'\n          emit('mode_change', { mode: 'listening' })\n        }\n        emit('interrupted', {\n          ...(currentMessageId !== null && { messageId: currentMessageId }),\n        })\n        break\n\n      case 'error': {\n        // The realtime server's `error` envelope isn't guaranteed to carry\n        // an `error` object at all (network-layer corruption, protocol\n        // drift, etc.). Validate shape before dereferencing so a malformed\n        // payload can't throw a TypeError inside this handler and stop the\n        // switch from running for the rest of the session.\n        const errorObj = readObject(event, 'error') ?? {}\n        const message =\n          readString(errorObj, 'message') ?? 'Unknown realtime server error'\n        const err: RealtimeServerError = new Error(message)\n        // Preserve `code` / `type` / `param` on the Error as extra props so\n        // consumers can branch on them without re-parsing the raw event.\n        const code = readString(errorObj, 'code')\n        if (code !== undefined) err.code = code\n        const errType = readString(errorObj, 'type')\n        if (errType !== undefined) err.type = errType\n        const param = readString(errorObj, 'param')\n        if (param !== undefined) err.param = param\n        logger.errors('grok.realtime server error', {\n          ...errorObj,\n          source: 'grok.realtime server',\n        })\n        emit('error', { error: err })\n        break\n      }\n\n      case undefined:\n      default:\n        // The xAI realtime protocol is a moving target; log unhandled event\n        // types at provider level so they're visible during debugging without\n        // emitting a user-visible error. `undefined` shares the bucket because\n        // a malformed event without a `type` field is just as unhandleable.\n        logger.provider('grok.realtime unhandled server event', {\n          type: event.type,\n        })\n        break\n    }\n  }\n\n  function setupOutputAudioAnalysis(stream: MediaStream) {\n    // Bail out if teardown has already started. `pc.ontrack` can fire\n    // asynchronously after `teardownConnection()` has flipped `isTornDown`\n    // (e.g. a remote track arriving mid-close); without this guard we'd\n    // allocate a fresh AudioContext / audio element that nothing would ever\n    // clean up.\n    if (isTornDown) return\n\n    // Tear down any prior output audio before allocating new resources.\n    // `pc.ontrack` can fire multiple times over the lifetime of a session\n    // (e.g. after renegotiation), and without this we'd leak audio elements\n    // and analyser nodes.\n    if (audioElement) {\n      try {\n        audioElement.pause()\n      } catch {\n        // ignore — element may already be unloaded\n      }\n      audioElement.srcObject = null\n      audioElement = null\n    }\n    if (outputSource) {\n      try {\n        outputSource.disconnect()\n      } catch {\n        // ignore — may already be disconnected\n      }\n      outputSource = null\n    }\n    if (outputAnalyser) {\n      try {\n        outputAnalyser.disconnect()\n      } catch {\n        // ignore\n      }\n      outputAnalyser = null\n    }\n\n    audioElement = new Audio()\n    audioElement.srcObject = stream\n    audioElement.autoplay = true\n    audioElement.play().catch((e) => {\n      // Autoplay is commonly blocked until the user interacts with the page\n      // (browser gesture requirement). Surfacing this as a fatal `error`\n      // event makes the UI render a red/error state even though the\n      // connection is healthy — the page just needs a click. Log at a\n      // dedicated source tag so it's debuggable, but don't emit `error`.\n      logger.errors('grok.realtime audio autoplay blocked', {\n        error: e,\n        source: 'grok.realtime.audio_permission_required',\n      })\n    })\n\n    if (!audioContext) {\n      audioContext = new AudioContext()\n    }\n\n    if (audioContext.state === 'suspended') {\n      audioContext.resume().catch((err) => {\n        // Same rationale as the autoplay catch: `resume()` failure usually\n        // means the user hasn't interacted yet. Logging only — no error\n        // emit — so the UI doesn't go into a fatal state for a recoverable\n        // condition.\n        logger.errors('grok.realtime audioContext.resume failed', {\n          error: err,\n          source: 'grok.realtime',\n        })\n      })\n    }\n\n    outputAnalyser = audioContext.createAnalyser()\n    outputAnalyser.fftSize = 2048\n    outputAnalyser.smoothingTimeConstant = 0.3\n\n    outputSource = audioContext.createMediaStreamSource(stream)\n    outputSource.connect(outputAnalyser)\n  }\n\n  function setupInputAudioAnalysis(stream: MediaStream) {\n    // Defensive symmetry with `setupOutputAudioAnalysis`. Today this is\n    // only called inline after SDP negotiation, but keeping the guard\n    // means any future caller path (e.g. renegotiation) won't leak a fresh\n    // AudioContext after teardown.\n    if (isTornDown) return\n\n    if (!audioContext) {\n      audioContext = new AudioContext()\n    }\n\n    if (audioContext.state === 'suspended') {\n      audioContext.resume().catch((err) => {\n        // Same rationale as in setupOutputAudioAnalysis: a suspended\n        // AudioContext usually resumes after a user gesture. Log only —\n        // surfacing this as a fatal error makes the UI look broken for a\n        // recoverable condition.\n        logger.errors('grok.realtime audioContext.resume failed', {\n          error: err,\n          source: 'grok.realtime',\n        })\n      })\n    }\n\n    inputAnalyser = audioContext.createAnalyser()\n    inputAnalyser.fftSize = 2048\n    inputAnalyser.smoothingTimeConstant = 0.3\n\n    inputSource = audioContext.createMediaStreamSource(stream)\n    inputSource.connect(inputAnalyser)\n  }\n\n  function sendEvent(event: Record<string, unknown>) {\n    if (isTornDown) {\n      // The caller is holding onto a `connection` object after `disconnect()`\n      // (or a failed connect). Silently queueing would leak memory and the\n      // events would never flush. Log + drop so the misuse is visible in\n      // debug mode without escalating to a throw — throwing from a React\n      // useEffect cleanup path can break teardown ordering in the UI.\n      logger.errors('grok.realtime sendEvent after disconnect', {\n        eventType: readString(event, 'type') ?? '<unknown>',\n        source: 'grok.realtime',\n      })\n      return\n    }\n    if (dataChannel?.readyState === 'open') {\n      logger.provider(\n        `provider=grok direction=out type=${readString(event, 'type') ?? '<unknown>'}`,\n        { frame: event },\n      )\n      // Mirror the try/catch in `flushPendingEvents` — `dataChannel.send`\n      // can synchronously throw if the channel flipped to `closing` between\n      // our readyState check and this call, or if `JSON.stringify` chokes\n      // on a caller-supplied payload. Log + emit error instead of letting\n      // the exception propagate up through public `sendText` / `sendImage`\n      // / `updateSession` call sites.\n      try {\n        dataChannel.send(JSON.stringify(event))\n      } catch (error) {\n        logger.errors('grok.realtime sendEvent failed', {\n          error,\n          eventType: readString(event, 'type') ?? '<unknown>',\n          source: 'grok.realtime',\n        })\n        emit('error', {\n          error: error instanceof Error ? error : new Error(String(error)),\n        })\n      }\n    } else {\n      pendingEvents.push(event)\n    }\n  }\n\n  function flushPendingEvents() {\n    try {\n      for (const event of pendingEvents) {\n        logger.provider(\n          `provider=grok direction=out type=${readString(event, 'type') ?? '<unknown>'}`,\n          { frame: event },\n        )\n        channel.send(JSON.stringify(event))\n      }\n      pendingEvents.length = 0\n    } catch (error) {\n      // A send failure here (e.g. dataChannel went from 'open' back to\n      // 'closing' mid-flush, or JSON.stringify on a caller-provided event\n      // threw) would otherwise be silently swallowed. By the time we're\n      // called, `onopen` has already resolved `dataChannelReady`, so the\n      // consumer-facing signal is `emit('error')` — try rejectDataChannelReady\n      // as a defensive belt-and-braces in case this ever runs pre-resolve.\n      logger.errors('grok.realtime flushPendingEvents failed', {\n        error,\n        source: 'grok.realtime',\n      })\n      const err = error instanceof Error ? error : new Error(String(error))\n      rejectDataChannelReady?.(err)\n      emit('error', { error: err })\n    }\n  }\n\n  const connection: RealtimeConnection = {\n    async disconnect() {\n      // Reuse the same teardown path as the failed-connect branches so\n      // every cleanup site stays in sync (input analyser, output analyser,\n      // output source, audio element, etc.).\n      await teardownConnection()\n      emit('status_change', { status: 'idle' })\n    },\n\n    async startAudioCapture() {\n      if (localStream) {\n        for (const track of localStream.getAudioTracks()) {\n          track.enabled = true\n        }\n      }\n      currentMode = 'listening'\n      emit('mode_change', { mode: 'listening' })\n    },\n\n    stopAudioCapture() {\n      if (localStream) {\n        for (const track of localStream.getAudioTracks()) {\n          track.enabled = false\n        }\n      }\n      currentMode = 'idle'\n      emit('mode_change', { mode: 'idle' })\n    },\n\n    sendText(text: string) {\n      sendEvent({\n        type: 'conversation.item.create',\n        item: {\n          type: 'message',\n          role: 'user',\n          content: [{ type: 'input_text', text }],\n        },\n      })\n      sendEvent({ type: 'response.create' })\n    },\n\n    sendImage(imageData: string, mimeType: string) {\n      // Accept:\n      //  - http(s):// URLs → forward as-is\n      //  - data: URIs (e.g. from FileReader.readAsDataURL) → forward as-is\n      //    so we don't double-wrap into `data:image/png;base64,data:image/png;base64,…`\n      //  - bare base64 → wrap in `data:${mimeType};base64,…`\n      const isAlreadyUrlOrDataUri =\n        imageData.startsWith('http://') ||\n        imageData.startsWith('https://') ||\n        imageData.startsWith('data:')\n      const imageContent = {\n        type: 'input_image',\n        // The OpenAI-realtime content part (which this adapter mirrors) nests\n        // the URL under an `image_url: { url: ... }` object, not a bare\n        // string.\n        image_url: {\n          url: isAlreadyUrlOrDataUri\n            ? imageData\n            : `data:${mimeType};base64,${imageData}`,\n        },\n      }\n\n      sendEvent({\n        type: 'conversation.item.create',\n        item: {\n          type: 'message',\n          role: 'user',\n          content: [imageContent],\n        },\n      })\n      sendEvent({ type: 'response.create' })\n    },\n\n    sendToolResult(callId: string, result: string) {\n      sendEvent({\n        type: 'conversation.item.create',\n        item: {\n          type: 'function_call_output',\n          call_id: callId,\n          output: result,\n        },\n      })\n      sendEvent({ type: 'response.create' })\n    },\n\n    updateSession(config: Partial<RealtimeSessionConfig>) {\n      const sessionUpdate: Record<string, unknown> = {}\n\n      if (config.instructions) {\n        sessionUpdate.instructions = config.instructions\n      }\n\n      if (config.voice) {\n        sessionUpdate.voice = config.voice\n      }\n\n      if (config.vadMode) {\n        if (config.vadMode === 'semantic') {\n          sessionUpdate.turn_detection = {\n            type: 'semantic_vad',\n            eagerness: config.semanticEagerness ?? 'medium',\n          }\n        } else if (config.vadMode === 'server') {\n          sessionUpdate.turn_detection = {\n            type: 'server_vad',\n            threshold: config.vadConfig?.threshold ?? 0.5,\n            prefix_padding_ms: config.vadConfig?.prefixPaddingMs ?? 300,\n            silence_duration_ms: config.vadConfig?.silenceDurationMs ?? 500,\n          }\n        } else {\n          sessionUpdate.turn_detection = null\n        }\n      }\n\n      if (config.tools !== undefined) {\n        sessionUpdate.tools = config.tools.map((t) => ({\n          type: 'function',\n          name: t.name,\n          description: t.description,\n          parameters: t.inputSchema ?? { type: 'object', properties: {} },\n        }))\n        sessionUpdate.tool_choice = 'auto'\n      }\n\n      if (config.outputModalities) {\n        sessionUpdate.modalities = config.outputModalities\n      }\n\n      if (config.temperature !== undefined) {\n        sessionUpdate.temperature = config.temperature\n      }\n\n      if (config.maxOutputTokens !== undefined) {\n        sessionUpdate.max_response_output_tokens = config.maxOutputTokens\n      }\n\n      // Let callers forward an explicit `input_audio_transcription` value\n      // through `providerOptions` — including `null` / `false` to disable\n      // the feature. Only apply our `grok-stt` default on the first\n      // session.update and only if the caller hasn't set it themselves.\n      const providerOptions: Record<string, unknown> =\n        config.providerOptions ?? {}\n      const callerTranscription =\n        'inputAudioTranscription' in providerOptions\n          ? providerOptions.inputAudioTranscription\n          : 'input_audio_transcription' in providerOptions\n            ? providerOptions.input_audio_transcription\n            : undefined\n      if (callerTranscription !== undefined) {\n        sessionUpdate.input_audio_transcription =\n          callerTranscription === false ? null : callerTranscription\n      } else if (!hasSentInitialSessionUpdate) {\n        sessionUpdate.input_audio_transcription = { model: 'grok-stt' }\n      }\n\n      if (Object.keys(sessionUpdate).length > 0) {\n        sendEvent({\n          type: 'session.update',\n          session: sessionUpdate,\n        })\n        hasSentInitialSessionUpdate = true\n      }\n    },\n\n    interrupt() {\n      sendEvent({ type: 'response.cancel' })\n      currentMode = 'listening'\n      emit('mode_change', { mode: 'listening' })\n      emit('interrupted', {\n        ...(currentMessageId !== null && { messageId: currentMessageId }),\n      })\n    },\n\n    on<TEvent extends RealtimeEvent>(\n      event: TEvent,\n      handler: RealtimeEventHandler<TEvent>,\n    ): () => void {\n      let handlers = eventHandlers.get(event)\n      if (!handlers) {\n        handlers = new Set()\n        eventHandlers.set(event, handlers)\n      }\n      handlers.add(handler)\n\n      return () => {\n        eventHandlers.get(event)?.delete(handler)\n      }\n    },\n\n    getAudioVisualization(): AudioVisualization {\n      function calculateLevel(analyser: AnalyserNode): number {\n        const data = new Uint8Array(analyser.fftSize)\n        analyser.getByteTimeDomainData(data)\n\n        let maxDeviation = 0\n        for (const sample of data) {\n          const deviation = Math.abs(sample - 128)\n          if (deviation > maxDeviation) {\n            maxDeviation = deviation\n          }\n        }\n\n        const normalized = maxDeviation / 128\n        return Math.min(1, normalized * 1.5)\n      }\n\n      return {\n        get inputLevel() {\n          if (!inputAnalyser) return 0\n          return calculateLevel(inputAnalyser)\n        },\n\n        get outputLevel() {\n          if (!outputAnalyser) return 0\n          return calculateLevel(outputAnalyser)\n        },\n\n        getInputFrequencyData() {\n          if (!inputAnalyser)\n            return new Uint8Array(FALLBACK_FREQUENCY_BIN_COUNT)\n          const data = new Uint8Array(inputAnalyser.frequencyBinCount)\n          inputAnalyser.getByteFrequencyData(data)\n          return data\n        },\n\n        getOutputFrequencyData() {\n          if (!outputAnalyser)\n            return new Uint8Array(FALLBACK_FREQUENCY_BIN_COUNT)\n          const data = new Uint8Array(outputAnalyser.frequencyBinCount)\n          outputAnalyser.getByteFrequencyData(data)\n          return data\n        },\n\n        getInputTimeDomainData() {\n          if (!inputAnalyser)\n            return new Uint8Array(FALLBACK_TIME_DOMAIN_SIZE).fill(\n              FALLBACK_TIME_DOMAIN_FILL,\n            )\n          const data = new Uint8Array(inputAnalyser.fftSize)\n          inputAnalyser.getByteTimeDomainData(data)\n          return data\n        },\n\n        getOutputTimeDomainData() {\n          if (!outputAnalyser)\n            return new Uint8Array(FALLBACK_TIME_DOMAIN_SIZE).fill(\n              FALLBACK_TIME_DOMAIN_FILL,\n            )\n          const data = new Uint8Array(outputAnalyser.fftSize)\n          outputAnalyser.getByteTimeDomainData(data)\n          return data\n        },\n\n        get inputSampleRate() {\n          return 24000\n        },\n\n        get outputSampleRate() {\n          return 24000\n        },\n      }\n    },\n  }\n\n  // `dataChannelReady` was already awaited inside the post-SDP try/catch\n  // above so we can short-circuit on failures with full teardown.\n  return connection\n}\n"],"mappings":";;;AAiBA,IAAM,oBAAoB;;;;;;;AAQ1B,SAAS,WACP,KACA,KACoB;CACpB,MAAM,QAAQ,IAAI;CAClB,OAAO,OAAO,UAAU,WAAW,QAAQ,KAAA;AAC7C;AAEA,SAAS,WACP,KACA,KACqC;CACrC,MAAM,QAAQ,IAAI;CAClB,OAAO,SAAS,OAAO,UAAU,YAAY,CAAC,MAAM,QAAQ,KAAK,IAC5D,QACD,KAAA;AACN;AAEA,SAAS,gBACP,KACA,KAC4C;CAC5C,MAAM,QAAQ,IAAI;CAClB,IAAI,CAAC,MAAM,QAAQ,KAAK,GAAG,OAAO,KAAA;CAClC,OAAO,MAAM,QACV,SACC,SAAS,QAAQ,OAAO,SAAS,YAAY,CAAC,MAAM,QAAQ,IAAI,CACpE;AACF;;;;;;;;;;;;;;;;;;;AA0BA,SAAgB,aACd,UAA+B,CAAC,GACf;CACjB,MAAM,iBAAiB,QAAQ,kBAAkB;CACjD,MAAM,SAAS,mBAAmB,QAAQ,KAAK;CAE/C,OAAO;EACL,UAAU;EAEV,MAAM,QACJ,OACA,cAC6B;GAC7B,MAAM,QAAQ,MAAM,OAAO,SAAA;GAC3B,OAAO,QAAQ,yCAAyC,SAAS;IAC/D,UAAU;IACV;GACF,CAAC;GAED,IAAI,mBAAmB,UACrB,OAAO,uBAAuB,OAAO,MAAM;GAE7C,MAAM,wBAAQ,IAAI,MAAM,+CAA+C;GACvE,OAAO,OAAO,uBAAuB;IACnC;IACA,QAAQ;GACV,CAAC;GACD,MAAM;EACR;CACF;AACF;;;;AAKA,eAAe,uBACb,OACA,QAC6B;CAC7B,MAAM,QAAQ,MAAM,OAAO,SAAA;CAC3B,MAAM,gCAAgB,IAAI,IAAmD;CAE7E,MAAM,KAAK,IAAI,kBAAkB;CAEjC,IAAI,eAAoC;CACxC,IAAI,gBAAqC;CACzC,IAAI,iBAAsC;CAC1C,IAAI,cAAiD;CACrD,IAAI,eAAkD;CACtD,IAAI,cAAkC;CAEtC,IAAI,eAAwC;CAK5C,MAAM,UAAU,GAAG,kBAAkB,YAAY;CACjD,IAAI,cAAqC;CAEzC,IAAI,cAA4B;CAChC,IAAI,mBAAkC;CAKtC,IAAI,aAAa;CAMjB,MAAM,gBAAgD,CAAC;CAMvD,IAAI,8BAA8B;CAMlC,MAAM,+BAA+B;CACrC,MAAM,4BAA4B;CAClC,MAAM,4BAA4B;CAElC,SAAS,KACP,OACA,SACA;EACA,MAAM,WAAW,cAAc,IAAI,KAAK;EACxC,IAAI,UACF,KAAK,MAAM,WAAW,UACpB,QAAQ,OAAO;CAGrB;CAEA,IAAI,oBAAoB;CACxB,IAAI,yBAA6D;CACjE,IAAI,0BAAgE;CAEpE,MAAM,mBAAmB,IAAI,SAAe,SAAS,WAAW;EAC9D,0BAA0B,WAAW;GACnC,IAAI,4BAA4B,MAAM;IACpC,aAAa,uBAAuB;IACpC,0BAA0B;GAC5B;GAEA,yBAAyB;GACzB,OAAO,MAAM;EACf;EAEA,0BAA0B,iBAAiB;GACzC,IAAI,CAAC,mBACH,yCACE,IAAI,MACF,gEACF,CACF;EAEJ,GAAG,IAAK;EAER,QAAQ,eAAe;GACrB,oBAAoB;GACpB,IAAI,4BAA4B,MAAM;IACpC,aAAa,uBAAuB;IACpC,0BAA0B;GAC5B;GAGA,yBAAyB;GACzB,mBAAmB;GACnB,KAAK,iBAAiB,EAAE,QAAQ,YAAY,CAAC;GAC7C,QAAQ;EACV;CACF,CAAC;CAED,QAAQ,aAAa,UAAU;EAC7B,IAAI;GACF,MAAM,UAAU,KAAK,MAAM,MAAM,IAAI;GACrC,MAAM,gBACJ,YAAY,QAAQ,OAAO,YAAY,WAAW,UAAU,CAAC;GAC/D,OAAO,SACL,mCAAmC,WAAW,eAAe,MAAM,KAAK,eACxE,EAAE,OAAO,cAAc,CACzB;GACA,kBAAkB,aAAa;EACjC,SAAS,UAAU;GACjB,OAAO,OAAO,uBAAuB;IACnC,OAAO;IACP,QAAQ;GACV,CAAC;GACD,KAAK,SAAS,EACZ,OACE,oBAAoB,QAAQ,WAAW,IAAI,MAAM,OAAO,QAAQ,CAAC,EACrE,CAAC;EACH;CACF;CAEA,QAAQ,WAAW,UAAU;EAK3B,IAAI,YAAY;EAChB,OAAO,OAAO,uBAAuB;GACnC;GACA,QAAQ;EACV,CAAC;EAQD,MAAM,WAAW,WAAW,OAAa,OAAO;EAChD,MAAM,OACH,YAAY,WAAW,UAAU,SAAS,OAAO,MAAM,QAAQ;EAClE,MAAM,wBAAQ,IAAI,MAAM,uBAAuB,KAAK;EACpD,IAAI,CAAC,mBACH,yBAAyB,KAAK;EAEhC,KAAK,SAAS,EAAE,OAAO,MAAM,CAAC;CAChC;CAEA,QAAQ,gBAAgB;EAItB,IAAI,YAAY;EAChB,IAAI,CAAC,mBACH,yCAAyB,IAAI,MAAM,oCAAoC,CAAC;CAE5E;CAEA,GAAG,WAAW,UAAU;EACtB,IAAI,MAAM,MAAM,SAAS,WAAW,MAAM,QAAQ,IAChD,yBAAyB,MAAM,QAAQ,EAAE;CAE7C;CAMA,GAAG,gCAAgC;EACjC,MAAM,QAAQ,GAAG;EACjB,OAAO,SAAS,oCAAoC,SAAS,EAC3D,MACF,CAAC;EACD,IAAI,UAAU,YAAY,UAAU,kBAAkB,UAAU,UAAU;GAMxE,IAAI,CAAC,YACH,KAAK,iBAAiB,EACpB,QAAQ,UAAU,WAAW,UAAU,OACzC,CAAC;GAEH,IAAI,CAAC,mBAAmB;IAKtB,MAAM,UACJ,UAAU,WACN,qDACA,iCAAiC,MAAM;IAC7C,yBAAyB,IAAI,MAAM,OAAO,CAAC;GAC7C;GASA,IAAI,UAAU,YAAY,CAAC,YACzB,mBAAwB;EAE5B;CACF;CAEA,GAAG,mCAAmC;EACpC,MAAM,QAAQ,GAAG;EACjB,OAAO,SAAS,uCAAuC,SAAS,EAC9D,MACF,CAAC;EACD,IACE,CAAC,sBACA,UAAU,YAAY,UAAU,YAAY,UAAU,iBACvD;GACA,MAAM,UACJ,UAAU,WACN,qDACA,iCAAiC,MAAM;GAC7C,yBAAyB,IAAI,MAAM,OAAO,CAAC;EAC7C;CACF;;;;;;;;;;;CAYA,eAAe,qBAAqB;EAMlC,aAAa;EAQb,cAAc,SAAS;EAMvB,yCACE,IAAI,MAAM,iDAAiD,CAC7D;EAEA,IAAI,aAAa;GACf,KAAK,MAAM,SAAS,YAAY,UAAU,GACxC,MAAM,KAAK;GAEb,cAAc;EAChB;EAIA,IAAI,cAAc;GAChB,IAAI;IACF,aAAa,MAAM;GACrB,QAAQ,CAER;GACA,aAAa,YAAY;GACzB,eAAe;EACjB;EACA,IAAI,cAAc;GAChB,IAAI;IACF,aAAa,WAAW;GAC1B,QAAQ,CAER;GACA,eAAe;EACjB;EACA,IAAI,gBAAgB;GAClB,IAAI;IACF,eAAe,WAAW;GAC5B,QAAQ,CAER;GACA,iBAAiB;EACnB;EAGA,IAAI,aAAa;GACf,IAAI;IACF,YAAY,WAAW;GACzB,QAAQ,CAER;GACA,cAAc;EAChB;EACA,IAAI,eAAe;GACjB,IAAI;IACF,cAAc,WAAW;GAC3B,QAAQ,CAER;GACA,gBAAgB;EAClB;EAEA,IAAI,aAAa;GACf,IAAI;IACF,YAAY,MAAM;GACpB,QAAQ,CAER;GACA,cAAc;EAChB;EAEA,IAAI;GACF,GAAG,MAAM;EACX,QAAQ,CAER;EAEA,IAAI,cAAc;GAChB,IAAI;IACF,MAAM,aAAa,MAAM;GAC3B,QAAQ,CAER;GACA,eAAe;EACjB;CACF;CAUA,IAAI;EACF,IAAI;GACF,cAAc,MAAM,UAAU,aAAa,aAAa,EACtD,OAAO;IACL,kBAAkB;IAClB,kBAAkB;IAClB,YAAY;GACd,EACF,CAAC;EACH,SAAS,OAAO;GACd,OAAO,OAAO,uBAAuB;IACnC;IACA,QAAQ;GACV,CAAC;GAGD,MAAM,IAAI,MACR,kDAAkD,iBAAiB,QAAQ,MAAM,UAAU,OAC7F;EACF;EAEA,KAAK,MAAM,SAAS,YAAY,eAAe,GAC7C,GAAG,SAAS,OAAO,WAAW;EAGhC,MAAM,QAAQ,MAAM,GAAG,YAAY;EACnC,MAAM,GAAG,oBAAoB,KAAK;EAElC,MAAM,cAAc,MAAM,MAAM,GAAG,kBAAkB,SAAS,SAAS;GACrE,QAAQ;GACR,SAAS;IACP,eAAe,UAAU,MAAM;IAC/B,gBAAgB;GAClB;GACA,MAAM,MAAM;EACd,CAAC;EAED,IAAI,CAAC,YAAY,IAAI;GACnB,MAAM,YAAY,MAAM,YAAY,KAAK;GACzC,MAAM,wBAAQ,IAAI,MAChB,0CAA0C,YAAY,OAAO,KAAK,WACpE;GACA,OAAO,OAAO,uBAAuB;IACnC;IACA,QAAQ;IACR,QAAQ,YAAY;GACtB,CAAC;GACD,MAAM;EACR;EAEA,MAAM,YAAY,MAAM,YAAY,KAAK;EACzC,MAAM,GAAG,qBAAqB;GAAE,MAAM;GAAU,KAAK;EAAU,CAAC;CAClE,SAAS,KAAK;EACZ,MAAM,mBAAmB;EACzB,MAAM;CACR;CAOA,IAAI;EACF,wBAAwB,WAAW;EACnC,MAAM;CACR,SAAS,KAAK;EACZ,MAAM,mBAAmB;EACzB,MAAM;CACR;CAEA,SAAS,kBAAkB,OAAgC;EAGzD,QAFa,WAAW,OAAO,MAEvB,GAAR;GACE,KAAK;GACL,KAAK,mBACH;GAEF,KAAK;IACH,cAAc;IACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;IACzC;GAEF,KAAK;IACH,cAAc;IACd,KAAK,eAAe,EAAE,MAAM,WAAW,CAAC;IACxC;GAEF,KAAK,gCACH;GAEF,KAAK,yDAAyD;IAC5D,MAAM,aAAa,WAAW,OAAO,YAAY;IACjD,IAAI,eAAe,KAAA,GAAW;IAC9B,KAAK,cAAc;KAAE,MAAM;KAAQ;KAAY,SAAS;IAAK,CAAC;IAC9D;GACF;GAEA,KAAK;IAIH,mBAAmB;IACnB,cAAc;IACd,KAAK,eAAe,EAAE,MAAM,WAAW,CAAC;IACxC;GAEF,KAAK,8BAA8B;IACjC,MAAM,OAAO,WAAW,OAAO,MAAM;IACrC,IAAI,QAAQ,WAAW,MAAM,MAAM,MAAM,WAAW;KAClD,MAAM,KAAK,WAAW,MAAM,IAAI;KAChC,IAAI,OAAO,KAAA,GAAW,mBAAmB;IAC3C;IACA;GACF;GAKA,KAAK;GACL,KAAK,mCAAmC;IACtC,MAAM,QAAQ,WAAW,OAAO,OAAO;IACvC,IAAI,UAAU,KAAA,GAAW;IACzB,KAAK,cAAc;KACjB,MAAM;KACN,YAAY;KACZ,SAAS;IACX,CAAC;IACD;GACF;GAEA,KAAK;GACL,KAAK,kCAAkC;IACrC,MAAM,aAAa,WAAW,OAAO,YAAY;IACjD,IAAI,eAAe,KAAA,GAAW;IAC9B,KAAK,cAAc;KAAE,MAAM;KAAa;KAAY,SAAS;IAAK,CAAC;IACnE;GACF;GAIA,KAAK;GACL,KAAK,8BAA8B;IACjC,MAAM,QAAQ,WAAW,OAAO,OAAO;IACvC,IAAI,UAAU,KAAA,GAAW;IACzB,KAAK,cAAc;KACjB,MAAM;KACN,YAAY;KACZ,SAAS;IACX,CAAC;IACD;GACF;GAEA,KAAK;GACL,KAAK,6BAA6B;IAChC,MAAM,OAAO,WAAW,OAAO,MAAM;IACrC,IAAI,SAAS,KAAA,GAAW;IACxB,KAAK,cAAc;KACjB,MAAM;KACN,YAAY;KACZ,SAAS;IACX,CAAC;IACD;GACF;GAIA,KAAK;GACL,KAAK;IACH,IAAI,gBAAgB,YAAY;KAC9B,cAAc;KACd,KAAK,eAAe,EAAE,MAAM,WAAW,CAAC;IAC1C;IACA;GAEF,KAAK;GACL,KAAK,uBACH;GAEF,KAAK,yCAAyC;IAM5C,MAAM,SAAS,WAAW,OAAO,SAAS;IAC1C,MAAM,OAAO,WAAW,OAAO,MAAM,KAAK;IAC1C,MAAM,OAAO,WAAW,OAAO,WAAW,KAAK;IAC/C,IAAI,CAAC,QAAQ;KACX,OAAO,OACL,gEACA;MACE,QAAQ;MACR,YAAY;MACZ,SAAS,MAAM;KACjB,CACF;KACA,KAAK,SAAS,EACZ,uBAAO,IAAI,MACT,2DACF,EACF,CAAC;KACD;IACF;IACA,IAAI;KAEF,KAAK,aAAa;MAAE,YAAY;MAAQ,UAAU;MAAM,OAD1C,KAAK,MAAM,IAC+B;KAAM,CAAC;IACjE,QAAQ;KACN,KAAK,aAAa;MAAE,YAAY;MAAQ,UAAU;MAAM,OAAO;KAAK,CAAC;IACvE;IACA;GACF;GAEA,KAAK,iBAAiB;IAEpB,MAAM,SAAS,gBADE,WAAW,OAAO,UAAU,KAAK,CAAC,GACV,QAAQ;IAKjD,IAAI,gBAAgB,QAAQ;KAC1B,cAAc;KACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;IAC3C;IAEA,IAAI,kBAAkB;KACpB,MAAM,UAA2B;MAC/B,IAAI;MACJ,MAAM;MACN,WAAW,KAAK,IAAI;MACpB,OAAO,CAAC;KACV;KAEA,KAAK,MAAM,QAAQ,UAAU,CAAC,GAAG;MAC/B,IAAI,WAAW,MAAM,MAAM,MAAM,WAAW;MAC5C,MAAM,UAAU,gBAAgB,MAAM,SAAS;MAC/C,IAAI,CAAC,SAAS;MACd,KAAK,MAAM,QAAQ,SAAS;OAC1B,MAAM,WAAW,WAAW,MAAM,MAAM;OACxC,IAAI,aAAa,SAAS;QACxB,MAAM,aAAa,WAAW,MAAM,YAAY;QAChD,IAAI,YACF,QAAQ,MAAM,KAAK;SAAE,MAAM;SAAS;QAAW,CAAC;OAEpD,OAAO,IAAI,aAAa,QAAQ;QAC9B,MAAM,UAAU,WAAW,MAAM,MAAM;QACvC,IAAI,SACF,QAAQ,MAAM,KAAK;SAAE,MAAM;SAAQ;QAAQ,CAAC;OAEhD;MACF;KACF;KAEA,KAAK,oBAAoB,EAAE,QAAQ,CAAC;KACpC,mBAAmB;IACrB;IACA;GACF;GAEA,KAAK;IAKH,IAAI,gBAAgB,QAAQ;KAC1B,cAAc;KACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;IAC3C;IACA,KAAK,eAAe,EAClB,GAAI,qBAAqB,QAAQ,EAAE,WAAW,iBAAiB,EACjE,CAAC;IACD;GAEF,KAAK,SAAS;IAMZ,MAAM,WAAW,WAAW,OAAO,OAAO,KAAK,CAAC;IAChD,MAAM,UACJ,WAAW,UAAU,SAAS,KAAK;IACrC,MAAM,MAA2B,IAAI,MAAM,OAAO;IAGlD,MAAM,OAAO,WAAW,UAAU,MAAM;IACxC,IAAI,SAAS,KAAA,GAAW,IAAI,OAAO;IACnC,MAAM,UAAU,WAAW,UAAU,MAAM;IAC3C,IAAI,YAAY,KAAA,GAAW,IAAI,OAAO;IACtC,MAAM,QAAQ,WAAW,UAAU,OAAO;IAC1C,IAAI,UAAU,KAAA,GAAW,IAAI,QAAQ;IACrC,OAAO,OAAO,8BAA8B;KAC1C,GAAG;KACH,QAAQ;IACV,CAAC;IACD,KAAK,SAAS,EAAE,OAAO,IAAI,CAAC;IAC5B;GACF;GAEA,KAAK,KAAA;GACL,SAKE,OAAO,SAAS,wCAAwC,EACtD,MAAM,MAAM,KACd,CAAC;EAEL;CACF;CAEA,SAAS,yBAAyB,QAAqB;EAMrD,IAAI,YAAY;EAMhB,IAAI,cAAc;GAChB,IAAI;IACF,aAAa,MAAM;GACrB,QAAQ,CAER;GACA,aAAa,YAAY;GACzB,eAAe;EACjB;EACA,IAAI,cAAc;GAChB,IAAI;IACF,aAAa,WAAW;GAC1B,QAAQ,CAER;GACA,eAAe;EACjB;EACA,IAAI,gBAAgB;GAClB,IAAI;IACF,eAAe,WAAW;GAC5B,QAAQ,CAER;GACA,iBAAiB;EACnB;EAEA,eAAe,IAAI,MAAM;EACzB,aAAa,YAAY;EACzB,aAAa,WAAW;EACxB,aAAa,KAAK,CAAC,CAAC,OAAO,MAAM;GAM/B,OAAO,OAAO,wCAAwC;IACpD,OAAO;IACP,QAAQ;GACV,CAAC;EACH,CAAC;EAED,IAAI,CAAC,cACH,eAAe,IAAI,aAAa;EAGlC,IAAI,aAAa,UAAU,aACzB,aAAa,OAAO,CAAC,CAAC,OAAO,QAAQ;GAKnC,OAAO,OAAO,4CAA4C;IACxD,OAAO;IACP,QAAQ;GACV,CAAC;EACH,CAAC;EAGH,iBAAiB,aAAa,eAAe;EAC7C,eAAe,UAAU;EACzB,eAAe,wBAAwB;EAEvC,eAAe,aAAa,wBAAwB,MAAM;EAC1D,aAAa,QAAQ,cAAc;CACrC;CAEA,SAAS,wBAAwB,QAAqB;EAKpD,IAAI,YAAY;EAEhB,IAAI,CAAC,cACH,eAAe,IAAI,aAAa;EAGlC,IAAI,aAAa,UAAU,aACzB,aAAa,OAAO,CAAC,CAAC,OAAO,QAAQ;GAKnC,OAAO,OAAO,4CAA4C;IACxD,OAAO;IACP,QAAQ;GACV,CAAC;EACH,CAAC;EAGH,gBAAgB,aAAa,eAAe;EAC5C,cAAc,UAAU;EACxB,cAAc,wBAAwB;EAEtC,cAAc,aAAa,wBAAwB,MAAM;EACzD,YAAY,QAAQ,aAAa;CACnC;CAEA,SAAS,UAAU,OAAgC;EACjD,IAAI,YAAY;GAMd,OAAO,OAAO,4CAA4C;IACxD,WAAW,WAAW,OAAO,MAAM,KAAK;IACxC,QAAQ;GACV,CAAC;GACD;EACF;EACA,IAAI,aAAa,eAAe,QAAQ;GACtC,OAAO,SACL,oCAAoC,WAAW,OAAO,MAAM,KAAK,eACjE,EAAE,OAAO,MAAM,CACjB;GAOA,IAAI;IACF,YAAY,KAAK,KAAK,UAAU,KAAK,CAAC;GACxC,SAAS,OAAO;IACd,OAAO,OAAO,kCAAkC;KAC9C;KACA,WAAW,WAAW,OAAO,MAAM,KAAK;KACxC,QAAQ;IACV,CAAC;IACD,KAAK,SAAS,EACZ,OAAO,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,KAAK,CAAC,EACjE,CAAC;GACH;EACF,OACE,cAAc,KAAK,KAAK;CAE5B;CAEA,SAAS,qBAAqB;EAC5B,IAAI;GACF,KAAK,MAAM,SAAS,eAAe;IACjC,OAAO,SACL,oCAAoC,WAAW,OAAO,MAAM,KAAK,eACjE,EAAE,OAAO,MAAM,CACjB;IACA,QAAQ,KAAK,KAAK,UAAU,KAAK,CAAC;GACpC;GACA,cAAc,SAAS;EACzB,SAAS,OAAO;GAOd,OAAO,OAAO,2CAA2C;IACvD;IACA,QAAQ;GACV,CAAC;GACD,MAAM,MAAM,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,KAAK,CAAC;GACpE,yBAAyB,GAAG;GAC5B,KAAK,SAAS,EAAE,OAAO,IAAI,CAAC;EAC9B;CACF;CA6QA,OAAO;EA1QL,MAAM,aAAa;GAIjB,MAAM,mBAAmB;GACzB,KAAK,iBAAiB,EAAE,QAAQ,OAAO,CAAC;EAC1C;EAEA,MAAM,oBAAoB;GACxB,IAAI,aACF,KAAK,MAAM,SAAS,YAAY,eAAe,GAC7C,MAAM,UAAU;GAGpB,cAAc;GACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;EAC3C;EAEA,mBAAmB;GACjB,IAAI,aACF,KAAK,MAAM,SAAS,YAAY,eAAe,GAC7C,MAAM,UAAU;GAGpB,cAAc;GACd,KAAK,eAAe,EAAE,MAAM,OAAO,CAAC;EACtC;EAEA,SAAS,MAAc;GACrB,UAAU;IACR,MAAM;IACN,MAAM;KACJ,MAAM;KACN,MAAM;KACN,SAAS,CAAC;MAAE,MAAM;MAAc;KAAK,CAAC;IACxC;GACF,CAAC;GACD,UAAU,EAAE,MAAM,kBAAkB,CAAC;EACvC;EAEA,UAAU,WAAmB,UAAkB;GAsB7C,UAAU;IACR,MAAM;IACN,MAAM;KACJ,MAAM;KACN,MAAM;KACN,SAAS,CAAC;MAhBZ,MAAM;MAIN,WAAW,EACT,KATF,UAAU,WAAW,SAAS,KAC9B,UAAU,WAAW,UAAU,KAC/B,UAAU,WAAW,OAAO,IAQtB,YACA,QAAQ,SAAS,UAAU,YACjC;KAQY,CAAY;IACxB;GACF,CAAC;GACD,UAAU,EAAE,MAAM,kBAAkB,CAAC;EACvC;EAEA,eAAe,QAAgB,QAAgB;GAC7C,UAAU;IACR,MAAM;IACN,MAAM;KACJ,MAAM;KACN,SAAS;KACT,QAAQ;IACV;GACF,CAAC;GACD,UAAU,EAAE,MAAM,kBAAkB,CAAC;EACvC;EAEA,cAAc,QAAwC;GACpD,MAAM,gBAAyC,CAAC;GAEhD,IAAI,OAAO,cACT,cAAc,eAAe,OAAO;GAGtC,IAAI,OAAO,OACT,cAAc,QAAQ,OAAO;GAG/B,IAAI,OAAO,SAAS;IAClB,IAAI,OAAO,YAAY,YACrB,cAAc,iBAAiB;KAC7B,MAAM;KACN,WAAW,OAAO,qBAAqB;IACzC;SACK,IAAI,OAAO,YAAY,UAC5B,cAAc,iBAAiB;KAC7B,MAAM;KACN,WAAW,OAAO,WAAW,aAAa;KAC1C,mBAAmB,OAAO,WAAW,mBAAmB;KACxD,qBAAqB,OAAO,WAAW,qBAAqB;IAC9D;SAEA,cAAc,iBAAiB;GAEnC;GAEA,IAAI,OAAO,UAAU,KAAA,GAAW;IAC9B,cAAc,QAAQ,OAAO,MAAM,KAAK,OAAO;KAC7C,MAAM;KACN,MAAM,EAAE;KACR,aAAa,EAAE;KACf,YAAY,EAAE,eAAe;MAAE,MAAM;MAAU,YAAY,CAAC;KAAE;IAChE,EAAE;IACF,cAAc,cAAc;GAC9B;GAEA,IAAI,OAAO,kBACT,cAAc,aAAa,OAAO;GAGpC,IAAI,OAAO,gBAAgB,KAAA,GACzB,cAAc,cAAc,OAAO;GAGrC,IAAI,OAAO,oBAAoB,KAAA,GAC7B,cAAc,6BAA6B,OAAO;GAOpD,MAAM,kBACJ,OAAO,mBAAmB,CAAC;GAC7B,MAAM,sBACJ,6BAA6B,kBACzB,gBAAgB,0BAChB,+BAA+B,kBAC7B,gBAAgB,4BAChB,KAAA;GACR,IAAI,wBAAwB,KAAA,GAC1B,cAAc,4BACZ,wBAAwB,QAAQ,OAAO;QACpC,IAAI,CAAC,6BACV,cAAc,4BAA4B,EAAE,OAAO,WAAW;GAGhE,IAAI,OAAO,KAAK,aAAa,CAAC,CAAC,SAAS,GAAG;IACzC,UAAU;KACR,MAAM;KACN,SAAS;IACX,CAAC;IACD,8BAA8B;GAChC;EACF;EAEA,YAAY;GACV,UAAU,EAAE,MAAM,kBAAkB,CAAC;GACrC,cAAc;GACd,KAAK,eAAe,EAAE,MAAM,YAAY,CAAC;GACzC,KAAK,eAAe,EAClB,GAAI,qBAAqB,QAAQ,EAAE,WAAW,iBAAiB,EACjE,CAAC;EACH;EAEA,GACE,OACA,SACY;GACZ,IAAI,WAAW,cAAc,IAAI,KAAK;GACtC,IAAI,CAAC,UAAU;IACb,2BAAW,IAAI,IAAI;IACnB,cAAc,IAAI,OAAO,QAAQ;GACnC;GACA,SAAS,IAAI,OAAO;GAEpB,aAAa;IACX,cAAc,IAAI,KAAK,CAAC,EAAE,OAAO,OAAO;GAC1C;EACF;EAEA,wBAA4C;GAC1C,SAAS,eAAe,UAAgC;IACtD,MAAM,OAAO,IAAI,WAAW,SAAS,OAAO;IAC5C,SAAS,sBAAsB,IAAI;IAEnC,IAAI,eAAe;IACnB,KAAK,MAAM,UAAU,MAAM;KACzB,MAAM,YAAY,KAAK,IAAI,SAAS,GAAG;KACvC,IAAI,YAAY,cACd,eAAe;IAEnB;IAEA,MAAM,aAAa,eAAe;IAClC,OAAO,KAAK,IAAI,GAAG,aAAa,GAAG;GACrC;GAEA,OAAO;IACL,IAAI,aAAa;KACf,IAAI,CAAC,eAAe,OAAO;KAC3B,OAAO,eAAe,aAAa;IACrC;IAEA,IAAI,cAAc;KAChB,IAAI,CAAC,gBAAgB,OAAO;KAC5B,OAAO,eAAe,cAAc;IACtC;IAEA,wBAAwB;KACtB,IAAI,CAAC,eACH,OAAO,IAAI,WAAW,4BAA4B;KACpD,MAAM,OAAO,IAAI,WAAW,cAAc,iBAAiB;KAC3D,cAAc,qBAAqB,IAAI;KACvC,OAAO;IACT;IAEA,yBAAyB;KACvB,IAAI,CAAC,gBACH,OAAO,IAAI,WAAW,4BAA4B;KACpD,MAAM,OAAO,IAAI,WAAW,eAAe,iBAAiB;KAC5D,eAAe,qBAAqB,IAAI;KACxC,OAAO;IACT;IAEA,yBAAyB;KACvB,IAAI,CAAC,eACH,OAAO,IAAI,WAAW,yBAAyB,CAAC,CAAC,KAC/C,yBACF;KACF,MAAM,OAAO,IAAI,WAAW,cAAc,OAAO;KACjD,cAAc,sBAAsB,IAAI;KACxC,OAAO;IACT;IAEA,0BAA0B;KACxB,IAAI,CAAC,gBACH,OAAO,IAAI,WAAW,yBAAyB,CAAC,CAAC,KAC/C,yBACF;KACF,MAAM,OAAO,IAAI,WAAW,eAAe,OAAO;KAClD,eAAe,sBAAsB,IAAI;KACzC,OAAO;IACT;IAEA,IAAI,kBAAkB;KACpB,OAAO;IACT;IAEA,IAAI,mBAAmB;KACrB,OAAO;IACT;GACF;EACF;CAKK;AACT"}

package/dist/esm/realtime/token.d.ts

@@ -15,7 +15,7 @@ import { GrokRealtimeTokenOptions } from './types.js';
  * import { grokRealtimeToken } from '@tanstack/ai-grok'
  *
  * const token = await realtimeToken({
- *   adapter: grokRealtimeToken({ model: 'grok-voice-fast-1.0' }),
+ *   adapter: grokRealtimeToken({ model: 'grok-voice-think-fast-2.0' }),
  * })
  * ```
  */

package/dist/esm/realtime/token.js

@@ -1,5 +1,6 @@
 import { getGrokApiKeyFromEnv } from "../utils/client.js";
 import "../utils/index.js";
+import "../model-meta.js";
 import { resolveDebugOption } from "@tanstack/ai/adapter-internals";
 //#region src/realtime/token.ts
 var GROK_REALTIME_CLIENT_SECRETS_URL = "https://api.x.ai/v1/realtime/client_secrets";
@@ -19,7 +20,7 @@ var DEFAULT_TOKEN_FETCH_TIMEOUT_MS = 15e3;
 * import { grokRealtimeToken } from '@tanstack/ai-grok'
 *
 * const token = await realtimeToken({
-*   adapter: grokRealtimeToken({ model: 'grok-voice-fast-1.0' }),
+*   adapter: grokRealtimeToken({ model: 'grok-voice-think-fast-2.0' }),
 * })
 * ```
 */
@@ -29,7 +30,7 @@ function grokRealtimeToken(options = {}) {
 	return {
 		provider: "grok",
 		async generateToken() {
-			const model = options.model ?? "grok-voice-fast-1.0";
+			const model = options.model ?? "grok-voice-think-fast-2.0";
 			logger.request(`activity=realtimeToken provider=grok model=${model}`, {
 				provider: "grok",
 				model

package/dist/esm/realtime/token.js.map

@@ -1 +1 @@
-{"version":3,"file":"token.js","names":[],"sources":["../../../src/realtime/token.ts"],"sourcesContent":["import { resolveDebugOption } from '@tanstack/ai/adapter-internals'\nimport { getGrokApiKeyFromEnv } from '../utils'\nimport type { RealtimeToken, RealtimeTokenAdapter } from '@tanstack/ai'\nimport type { GrokRealtimeModel } from '../model-meta'\nimport type {\n  GrokRealtimeSessionResponse,\n  GrokRealtimeTokenOptions,\n} from './types'\n\nconst GROK_REALTIME_CLIENT_SECRETS_URL =\n  'https://api.x.ai/v1/realtime/client_secrets'\n\nconst DEFAULT_TOKEN_FETCH_TIMEOUT_MS = 15_000\n\n/**\n * Creates a Grok realtime token adapter.\n *\n * Generates ephemeral client secrets for browser-side WebRTC connections to\n * the xAI Voice Agent API.\n *\n * @param options - Configuration options for the realtime session.\n * @returns A RealtimeTokenAdapter for use with `realtimeToken()`.\n *\n * @example\n * ```typescript\n * import { realtimeToken } from '@tanstack/ai'\n * import { grokRealtimeToken } from '@tanstack/ai-grok'\n *\n * const token = await realtimeToken({\n *   adapter: grokRealtimeToken({ model: 'grok-voice-fast-1.0' }),\n * })\n * ```\n */\nexport function grokRealtimeToken(\n  options: GrokRealtimeTokenOptions = {},\n): RealtimeTokenAdapter {\n  const apiKey = getGrokApiKeyFromEnv()\n  const logger = resolveDebugOption(options.debug)\n\n  return {\n    provider: 'grok',\n\n    async generateToken(): Promise<RealtimeToken> {\n      const model: GrokRealtimeModel = options.model ?? 'grok-voice-fast-1.0'\n\n      logger.request(`activity=realtimeToken provider=grok model=${model}`, {\n        provider: 'grok',\n        model,\n      })\n\n      // xAI docs (docs.x.ai/developers/rest-api-reference/inference/voice)\n      // specify the body as `{ session: { model } }`. `expires_after` is\n      // available to shorten the default 600s TTL but we don't expose it\n      // yet — the caller can still call `generateToken()` more often if\n      // they want a shorter-lived session.\n      const requestBody: Record<string, unknown> = {\n        session: { model },\n      }\n\n      // Abort the fetch if xAI never responds. Without this the whole\n      // realtime connect flow hangs forever on a dead endpoint.\n      const controller = new AbortController()\n      const timeout = setTimeout(\n        () =>\n          controller.abort(new Error('Grok realtime token request timed out')),\n        DEFAULT_TOKEN_FETCH_TIMEOUT_MS,\n      )\n\n      try {\n        const response = await fetch(GROK_REALTIME_CLIENT_SECRETS_URL, {\n          method: 'POST',\n          headers: {\n            Authorization: `Bearer ${apiKey}`,\n            'Content-Type': 'application/json',\n          },\n          body: JSON.stringify(requestBody),\n          signal: controller.signal,\n        })\n\n        if (!response.ok) {\n          const errorText = await response.text()\n          throw new Error(\n            `Grok realtime session creation failed: ${response.status} ${errorText}`,\n          )\n        }\n\n        const sessionData = (await response.json()) as\n          | Partial<GrokRealtimeSessionResponse>\n          | undefined\n\n        // Validate shape before dereferencing — xAI could return an error\n        // envelope with 200 status, or a partial response under protocol drift.\n        const clientSecret = sessionData?.client_secret\n        if (\n          !clientSecret ||\n          typeof clientSecret.value !== 'string' ||\n          typeof clientSecret.expires_at !== 'number' ||\n          !Number.isFinite(clientSecret.expires_at)\n        ) {\n          throw new Error(\n            'Grok realtime session response missing or malformed `client_secret`',\n          )\n        }\n        const sessionModel = sessionData.model ?? model\n\n        // xAI docs describe `expires_at` as a unix timestamp in seconds, but\n        // in practice different deployments have returned milliseconds. Treat\n        // any value that already looks like ms (>1e12 ≈ Sep 2001 in ms) as ms.\n        const raw = clientSecret.expires_at\n        const expiresAt = raw > 1e12 ? raw : raw * 1000\n\n        return {\n          provider: 'grok',\n          token: clientSecret.value,\n          expiresAt,\n          config: {\n            model: sessionModel,\n          },\n        }\n      } catch (error) {\n        logger.errors('grok.realtimeToken fatal', {\n          error,\n          source: 'grok.realtimeToken',\n        })\n        throw error\n      } finally {\n        clearTimeout(timeout)\n      }\n    },\n  }\n}\n"],"mappings":";;;;AASA,IAAM,mCACJ;AAEF,IAAM,iCAAiC;;;;;;;;;;;;;;;;;;;;AAqBvC,SAAgB,kBACd,UAAoC,CAAC,GACf;CACtB,MAAM,SAAS,qBAAqB;CACpC,MAAM,SAAS,mBAAmB,QAAQ,KAAK;CAE/C,OAAO;EACL,UAAU;EAEV,MAAM,gBAAwC;GAC5C,MAAM,QAA2B,QAAQ,SAAS;GAElD,OAAO,QAAQ,8CAA8C,SAAS;IACpE,UAAU;IACV;GACF,CAAC;GAOD,MAAM,cAAuC,EAC3C,SAAS,EAAE,MAAM,EACnB;GAIA,MAAM,aAAa,IAAI,gBAAgB;GACvC,MAAM,UAAU,iBAEZ,WAAW,sBAAM,IAAI,MAAM,uCAAuC,CAAC,GACrE,8BACF;GAEA,IAAI;IACF,MAAM,WAAW,MAAM,MAAM,kCAAkC;KAC7D,QAAQ;KACR,SAAS;MACP,eAAe,UAAU;MACzB,gBAAgB;KAClB;KACA,MAAM,KAAK,UAAU,WAAW;KAChC,QAAQ,WAAW;IACrB,CAAC;IAED,IAAI,CAAC,SAAS,IAAI;KAChB,MAAM,YAAY,MAAM,SAAS,KAAK;KACtC,MAAM,IAAI,MACR,0CAA0C,SAAS,OAAO,GAAG,WAC/D;IACF;IAEA,MAAM,cAAe,MAAM,SAAS,KAAK;IAMzC,MAAM,eAAe,aAAa;IAClC,IACE,CAAC,gBACD,OAAO,aAAa,UAAU,YAC9B,OAAO,aAAa,eAAe,YACnC,CAAC,OAAO,SAAS,aAAa,UAAU,GAExC,MAAM,IAAI,MACR,qEACF;IAEF,MAAM,eAAe,YAAY,SAAS;IAK1C,MAAM,MAAM,aAAa;IACzB,MAAM,YAAY,MAAM,eAAO,MAAM,MAAM;IAE3C,OAAO;KACL,UAAU;KACV,OAAO,aAAa;KACpB;KACA,QAAQ,EACN,OAAO,aACT;IACF;GACF,SAAS,OAAO;IACd,OAAO,OAAO,4BAA4B;KACxC;KACA,QAAQ;IACV,CAAC;IACD,MAAM;GACR,UAAU;IACR,aAAa,OAAO;GACtB;EACF;CACF;AACF"}
+{"version":3,"file":"token.js","names":[],"sources":["../../../src/realtime/token.ts"],"sourcesContent":["import { resolveDebugOption } from '@tanstack/ai/adapter-internals'\nimport { GROK_DEFAULT_REALTIME_MODEL } from '../model-meta'\nimport { getGrokApiKeyFromEnv } from '../utils'\nimport type { RealtimeToken, RealtimeTokenAdapter } from '@tanstack/ai'\nimport type { GrokRealtimeModel } from '../model-meta'\nimport type {\n  GrokRealtimeSessionResponse,\n  GrokRealtimeTokenOptions,\n} from './types'\n\nconst GROK_REALTIME_CLIENT_SECRETS_URL =\n  'https://api.x.ai/v1/realtime/client_secrets'\n\nconst DEFAULT_TOKEN_FETCH_TIMEOUT_MS = 15_000\n\n/**\n * Creates a Grok realtime token adapter.\n *\n * Generates ephemeral client secrets for browser-side WebRTC connections to\n * the xAI Voice Agent API.\n *\n * @param options - Configuration options for the realtime session.\n * @returns A RealtimeTokenAdapter for use with `realtimeToken()`.\n *\n * @example\n * ```typescript\n * import { realtimeToken } from '@tanstack/ai'\n * import { grokRealtimeToken } from '@tanstack/ai-grok'\n *\n * const token = await realtimeToken({\n *   adapter: grokRealtimeToken({ model: 'grok-voice-think-fast-2.0' }),\n * })\n * ```\n */\nexport function grokRealtimeToken(\n  options: GrokRealtimeTokenOptions = {},\n): RealtimeTokenAdapter {\n  const apiKey = getGrokApiKeyFromEnv()\n  const logger = resolveDebugOption(options.debug)\n\n  return {\n    provider: 'grok',\n\n    async generateToken(): Promise<RealtimeToken> {\n      const model: GrokRealtimeModel =\n        options.model ?? GROK_DEFAULT_REALTIME_MODEL\n\n      logger.request(`activity=realtimeToken provider=grok model=${model}`, {\n        provider: 'grok',\n        model,\n      })\n\n      // xAI docs (docs.x.ai/developers/rest-api-reference/inference/voice)\n      // specify the body as `{ session: { model } }`. `expires_after` is\n      // available to shorten the default 600s TTL but we don't expose it\n      // yet — the caller can still call `generateToken()` more often if\n      // they want a shorter-lived session.\n      const requestBody: Record<string, unknown> = {\n        session: { model },\n      }\n\n      // Abort the fetch if xAI never responds. Without this the whole\n      // realtime connect flow hangs forever on a dead endpoint.\n      const controller = new AbortController()\n      const timeout = setTimeout(\n        () =>\n          controller.abort(new Error('Grok realtime token request timed out')),\n        DEFAULT_TOKEN_FETCH_TIMEOUT_MS,\n      )\n\n      try {\n        const response = await fetch(GROK_REALTIME_CLIENT_SECRETS_URL, {\n          method: 'POST',\n          headers: {\n            Authorization: `Bearer ${apiKey}`,\n            'Content-Type': 'application/json',\n          },\n          body: JSON.stringify(requestBody),\n          signal: controller.signal,\n        })\n\n        if (!response.ok) {\n          const errorText = await response.text()\n          throw new Error(\n            `Grok realtime session creation failed: ${response.status} ${errorText}`,\n          )\n        }\n\n        const sessionData = (await response.json()) as\n          | Partial<GrokRealtimeSessionResponse>\n          | undefined\n\n        // Validate shape before dereferencing — xAI could return an error\n        // envelope with 200 status, or a partial response under protocol drift.\n        const clientSecret = sessionData?.client_secret\n        if (\n          !clientSecret ||\n          typeof clientSecret.value !== 'string' ||\n          typeof clientSecret.expires_at !== 'number' ||\n          !Number.isFinite(clientSecret.expires_at)\n        ) {\n          throw new Error(\n            'Grok realtime session response missing or malformed `client_secret`',\n          )\n        }\n        const sessionModel = sessionData.model ?? model\n\n        // xAI docs describe `expires_at` as a unix timestamp in seconds, but\n        // in practice different deployments have returned milliseconds. Treat\n        // any value that already looks like ms (>1e12 ≈ Sep 2001 in ms) as ms.\n        const raw = clientSecret.expires_at\n        const expiresAt = raw > 1e12 ? raw : raw * 1000\n\n        return {\n          provider: 'grok',\n          token: clientSecret.value,\n          expiresAt,\n          config: {\n            model: sessionModel,\n          },\n        }\n      } catch (error) {\n        logger.errors('grok.realtimeToken fatal', {\n          error,\n          source: 'grok.realtimeToken',\n        })\n        throw error\n      } finally {\n        clearTimeout(timeout)\n      }\n    },\n  }\n}\n"],"mappings":";;;;;AAUA,IAAM,mCACJ;AAEF,IAAM,iCAAiC;;;;;;;;;;;;;;;;;;;;AAqBvC,SAAgB,kBACd,UAAoC,CAAC,GACf;CACtB,MAAM,SAAS,qBAAqB;CACpC,MAAM,SAAS,mBAAmB,QAAQ,KAAK;CAE/C,OAAO;EACL,UAAU;EAEV,MAAM,gBAAwC;GAC5C,MAAM,QACJ,QAAQ,SAAA;GAEV,OAAO,QAAQ,8CAA8C,SAAS;IACpE,UAAU;IACV;GACF,CAAC;GAOD,MAAM,cAAuC,EAC3C,SAAS,EAAE,MAAM,EACnB;GAIA,MAAM,aAAa,IAAI,gBAAgB;GACvC,MAAM,UAAU,iBAEZ,WAAW,sBAAM,IAAI,MAAM,uCAAuC,CAAC,GACrE,8BACF;GAEA,IAAI;IACF,MAAM,WAAW,MAAM,MAAM,kCAAkC;KAC7D,QAAQ;KACR,SAAS;MACP,eAAe,UAAU;MACzB,gBAAgB;KAClB;KACA,MAAM,KAAK,UAAU,WAAW;KAChC,QAAQ,WAAW;IACrB,CAAC;IAED,IAAI,CAAC,SAAS,IAAI;KAChB,MAAM,YAAY,MAAM,SAAS,KAAK;KACtC,MAAM,IAAI,MACR,0CAA0C,SAAS,OAAO,GAAG,WAC/D;IACF;IAEA,MAAM,cAAe,MAAM,SAAS,KAAK;IAMzC,MAAM,eAAe,aAAa;IAClC,IACE,CAAC,gBACD,OAAO,aAAa,UAAU,YAC9B,OAAO,aAAa,eAAe,YACnC,CAAC,OAAO,SAAS,aAAa,UAAU,GAExC,MAAM,IAAI,MACR,qEACF;IAEF,MAAM,eAAe,YAAY,SAAS;IAK1C,MAAM,MAAM,aAAa;IACzB,MAAM,YAAY,MAAM,eAAO,MAAM,MAAM;IAE3C,OAAO;KACL,UAAU;KACV,OAAO,aAAa;KACpB;KACA,QAAQ,EACN,OAAO,aACT;IACF;GACF,SAAS,OAAO;IACd,OAAO,OAAO,4BAA4B;KACxC;KACA,QAAQ;IACV,CAAC;IACD,MAAM;GACR,UAAU;IACR,aAAa,OAAO;GACtB;EACF;CACF;AACF"}

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

@@ -27,7 +27,7 @@ export type GrokTurnDetection = GrokSemanticVADConfig | GrokServerVADConfig | nu
  * Options for the Grok realtime token adapter.
  */
 export interface GrokRealtimeTokenOptions {
-    /** Model to use (default: 'grok-voice-fast-1.0'). */
+    /** Model to use (default: 'grok-voice-think-fast-2.0'). */
     model?: GrokRealtimeModel;
     /**
      * Enable debug logging for token creation.

package/dist/esm/tools/index.js.map

@@ -1 +1 @@
-{"version":3,"file":"index.js","names":[],"sources":["../../../src/tools/index.ts"],"sourcesContent":["import { brandProviderTool } from '@tanstack/ai'\nimport { convertFunctionToolToResponsesFormat } from '@tanstack/openai-base'\nimport type { ProviderTool, Tool } from '@tanstack/ai'\nimport type { ResponsesFunctionTool } from '@tanstack/openai-base'\nimport type { GrokProviderToolKind } from '../model-meta'\n\nexport type FunctionTool = ResponsesFunctionTool\n\nexport { convertFunctionToolToResponsesFormat as convertFunctionToolToAdapterFormat }\n\nexport type GrokProviderTool<TKind extends GrokProviderToolKind> = ProviderTool<\n  'grok',\n  TKind\n>\n\ntype GrokToolKindMarker<TKind extends GrokProviderToolKind> = `grok.${TKind}`\n\nexport interface GrokWebSearchToolConfig {\n  type: 'web_search'\n  filters?: {\n    allowed_domains?: Array<string>\n    excluded_domains?: Array<string>\n  }\n  enable_image_understanding?: boolean\n  enable_image_search?: boolean\n}\n\nexport interface GrokXSearchToolConfig {\n  type: 'x_search'\n  allowed_x_handles?: Array<string>\n  excluded_x_handles?: Array<string>\n  from_date?: string\n  to_date?: string\n  enable_image_understanding?: boolean\n  enable_video_understanding?: boolean\n}\n\nexport interface GrokFileSearchToolConfig {\n  type: 'file_search'\n  vector_store_ids: Array<string>\n  max_num_results?: number\n}\n\nexport interface GrokMCPToolConfig {\n  type: 'mcp'\n  server_label: string\n  server_url: string\n  allowed_tools?: Array<string>\n  server_description?: string\n  authorization?: string\n  headers?: Record<string, string>\n}\n\nexport type GrokServerTool =\n  | GrokWebSearchToolConfig\n  | GrokXSearchToolConfig\n  | GrokFileSearchToolConfig\n  | GrokMCPToolConfig\n\ntype GrokProviderToolMetadata<TKind extends GrokProviderToolKind> = Extract<\n  GrokServerTool,\n  { type: TKind }\n> & {\n  __kind: GrokToolKindMarker<TKind>\n}\n\nexport type GrokResponsesTool = GrokServerTool | ResponsesFunctionTool\n\nfunction providerTool<TKind extends GrokProviderToolKind>(\n  kind: TKind,\n  description: string,\n  metadata: Extract<GrokServerTool, { type: TKind }>,\n): GrokProviderTool<TKind> {\n  return brandProviderTool<GrokProviderTool<TKind>>({\n    name: kind,\n    description,\n    metadata: {\n      __kind: `grok.${kind}`,\n      ...metadata,\n    },\n  })\n}\n\nexport function grokWebSearchTool(\n  config: Omit<GrokWebSearchToolConfig, 'type'> = {},\n): GrokProviderTool<'web_search'> {\n  if (\n    config.filters?.allowed_domains !== undefined &&\n    config.filters.excluded_domains !== undefined\n  ) {\n    throw new Error(\n      'allowed_domains and excluded_domains cannot both be provided.',\n    )\n  }\n  if (\n    config.filters?.allowed_domains !== undefined &&\n    config.filters.allowed_domains.length > 5\n  ) {\n    throw new Error('allowed_domains supports at most 5 domains.')\n  }\n  if (\n    config.filters?.excluded_domains !== undefined &&\n    config.filters.excluded_domains.length > 5\n  ) {\n    throw new Error('excluded_domains supports at most 5 domains.')\n  }\n  return providerTool('web_search', 'Search the web', {\n    type: 'web_search',\n    ...config,\n  })\n}\n\nexport function grokXSearchTool(\n  config: Omit<GrokXSearchToolConfig, 'type'> = {},\n): GrokProviderTool<'x_search'> {\n  if (\n    config.allowed_x_handles !== undefined &&\n    config.excluded_x_handles !== undefined\n  ) {\n    throw new Error(\n      'allowed_x_handles and excluded_x_handles cannot both be provided.',\n    )\n  }\n  if (\n    config.allowed_x_handles !== undefined &&\n    config.allowed_x_handles.length > 20\n  ) {\n    throw new Error('allowed_x_handles supports at most 20 handles.')\n  }\n  if (\n    config.excluded_x_handles !== undefined &&\n    config.excluded_x_handles.length > 20\n  ) {\n    throw new Error('excluded_x_handles supports at most 20 handles.')\n  }\n  return providerTool('x_search', 'Search X posts', {\n    type: 'x_search',\n    ...config,\n  })\n}\n\nexport function grokFileSearchTool(\n  config: Omit<GrokFileSearchToolConfig, 'type'>,\n): GrokProviderTool<'file_search'> {\n  if (config.vector_store_ids.length === 0) {\n    throw new Error('vector_store_ids must contain at least one collection id.')\n  }\n  if (config.max_num_results !== undefined) {\n    if (config.max_num_results < 1 || config.max_num_results > 50) {\n      throw new Error('max_num_results must be between 1 and 50.')\n    }\n  }\n  return providerTool('file_search', 'Search xAI file collections', {\n    type: 'file_search',\n    ...config,\n  })\n}\n\nexport function grokMCPTool(\n  config: Omit<GrokMCPToolConfig, 'type'>,\n): GrokProviderTool<'mcp'> {\n  if (!config.server_url) {\n    throw new Error('server_url must be provided.')\n  }\n  return providerTool('mcp', config.server_description || 'Remote MCP server', {\n    type: 'mcp',\n    ...config,\n  })\n}\n\nfunction getGrokProviderToolKind(tool: Tool): GrokProviderToolKind | undefined {\n  const kind = (tool.metadata as { __kind?: unknown } | undefined)?.__kind\n  switch (kind) {\n    case 'grok.web_search':\n      return 'web_search'\n    case 'grok.x_search':\n      return 'x_search'\n    case 'grok.file_search':\n      return 'file_search'\n    case 'grok.mcp':\n      return 'mcp'\n    default:\n      return undefined\n  }\n}\n\nfunction convertGrokProviderToolToAdapterFormat(\n  tool: Tool,\n  kind: GrokProviderToolKind,\n): GrokServerTool {\n  const metadata = tool.metadata as GrokProviderToolMetadata<typeof kind>\n  if (metadata.type !== kind) {\n    throw new Error(\n      `convertGrokProviderToolToAdapterFormat: tool \"${tool.name}\" has mismatched Grok tool metadata.`,\n    )\n  }\n  const { __kind: _kind, ...toolConfig } = metadata\n  void _kind\n  return toolConfig\n}\n\nexport function convertToolsToProviderFormat(\n  tools: Array<Tool>,\n): Array<GrokResponsesTool> {\n  return tools.map((tool) => {\n    const grokProviderToolKind = getGrokProviderToolKind(tool)\n    if (grokProviderToolKind) {\n      return convertGrokProviderToolToAdapterFormat(tool, grokProviderToolKind)\n    }\n    return convertFunctionToolToResponsesFormat(tool)\n  })\n}\n"],"mappings":";;;AAoEA,SAAS,aACP,MACA,aACA,UACyB;CACzB,OAAO,kBAA2C;EAChD,MAAM;EACN;EACA,UAAU;GACR,QAAQ,QAAQ;GAChB,GAAG;EACL;CACF,CAAC;AACH;AAEA,SAAgB,kBACd,SAAgD,CAAC,GACjB;CAChC,IACE,OAAO,SAAS,oBAAoB,KAAA,KACpC,OAAO,QAAQ,qBAAqB,KAAA,GAEpC,MAAM,IAAI,MACR,+DACF;CAEF,IACE,OAAO,SAAS,oBAAoB,KAAA,KACpC,OAAO,QAAQ,gBAAgB,SAAS,GAExC,MAAM,IAAI,MAAM,6CAA6C;CAE/D,IACE,OAAO,SAAS,qBAAqB,KAAA,KACrC,OAAO,QAAQ,iBAAiB,SAAS,GAEzC,MAAM,IAAI,MAAM,8CAA8C;CAEhE,OAAO,aAAa,cAAc,kBAAkB;EAClD,MAAM;EACN,GAAG;CACL,CAAC;AACH;AAEA,SAAgB,gBACd,SAA8C,CAAC,GACjB;CAC9B,IACE,OAAO,sBAAsB,KAAA,KAC7B,OAAO,uBAAuB,KAAA,GAE9B,MAAM,IAAI,MACR,mEACF;CAEF,IACE,OAAO,sBAAsB,KAAA,KAC7B,OAAO,kBAAkB,SAAS,IAElC,MAAM,IAAI,MAAM,gDAAgD;CAElE,IACE,OAAO,uBAAuB,KAAA,KAC9B,OAAO,mBAAmB,SAAS,IAEnC,MAAM,IAAI,MAAM,iDAAiD;CAEnE,OAAO,aAAa,YAAY,kBAAkB;EAChD,MAAM;EACN,GAAG;CACL,CAAC;AACH;AAEA,SAAgB,mBACd,QACiC;CACjC,IAAI,OAAO,iBAAiB,WAAW,GACrC,MAAM,IAAI,MAAM,2DAA2D;CAE7E,IAAI,OAAO,oBAAoB,KAAA;MACzB,OAAO,kBAAkB,KAAK,OAAO,kBAAkB,IACzD,MAAM,IAAI,MAAM,2CAA2C;CAAA;CAG/D,OAAO,aAAa,eAAe,+BAA+B;EAChE,MAAM;EACN,GAAG;CACL,CAAC;AACH;AAEA,SAAgB,YACd,QACyB;CACzB,IAAI,CAAC,OAAO,YACV,MAAM,IAAI,MAAM,8BAA8B;CAEhD,OAAO,aAAa,OAAO,OAAO,sBAAsB,qBAAqB;EAC3E,MAAM;EACN,GAAG;CACL,CAAC;AACH;AAEA,SAAS,wBAAwB,MAA8C;CAE7E,QADc,KAAK,UAA+C,QAClE;EACE,KAAK,mBACH,OAAO;EACT,KAAK,iBACH,OAAO;EACT,KAAK,oBACH,OAAO;EACT,KAAK,YACH,OAAO;EACT,SACE;CACJ;AACF;AAEA,SAAS,uCACP,MACA,MACgB;CAChB,MAAM,WAAW,KAAK;CACtB,IAAI,SAAS,SAAS,MACpB,MAAM,IAAI,MACR,iDAAiD,KAAK,KAAK,qCAC7D;CAEF,MAAM,EAAE,QAAQ,OAAO,GAAG,eAAe;CAEzC,OAAO;AACT;AAEA,SAAgB,6BACd,OAC0B;CAC1B,OAAO,MAAM,KAAK,SAAS;EACzB,MAAM,uBAAuB,wBAAwB,IAAI;EACzD,IAAI,sBACF,OAAO,uCAAuC,MAAM,oBAAoB;EAE1E,OAAO,qCAAqC,IAAI;CAClD,CAAC;AACH"}
+{"version":3,"file":"index.js","names":[],"sources":["../../../src/tools/index.ts"],"sourcesContent":["import { brandProviderTool } from '@tanstack/ai'\nimport { convertFunctionToolToResponsesFormat } from '@tanstack/openai-base'\nimport type { ProviderTool, Tool } from '@tanstack/ai'\nimport type { ResponsesFunctionTool } from '@tanstack/openai-base'\nimport type { GrokProviderToolKind } from '../model-meta'\n\nexport type FunctionTool = ResponsesFunctionTool\n\nexport { convertFunctionToolToResponsesFormat as convertFunctionToolToAdapterFormat }\n\nexport type GrokProviderTool<TKind extends GrokProviderToolKind> = ProviderTool<\n  'grok',\n  TKind\n>\n\ntype GrokToolKindMarker<TKind extends GrokProviderToolKind> = `grok.${TKind}`\n\nexport interface GrokWebSearchToolConfig {\n  type: 'web_search'\n  filters?: {\n    allowed_domains?: Array<string>\n    excluded_domains?: Array<string>\n  }\n  enable_image_understanding?: boolean\n  enable_image_search?: boolean\n}\n\nexport interface GrokXSearchToolConfig {\n  type: 'x_search'\n  allowed_x_handles?: Array<string>\n  excluded_x_handles?: Array<string>\n  from_date?: string\n  to_date?: string\n  enable_image_understanding?: boolean\n  enable_video_understanding?: boolean\n}\n\nexport interface GrokFileSearchToolConfig {\n  type: 'file_search'\n  vector_store_ids: Array<string>\n  max_num_results?: number\n}\n\nexport interface GrokMCPToolConfig {\n  type: 'mcp'\n  server_label: string\n  server_url: string\n  allowed_tools?: Array<string>\n  server_description?: string\n  authorization?: string\n  headers?: Record<string, string>\n}\n\nexport type GrokServerTool =\n  | GrokWebSearchToolConfig\n  | GrokXSearchToolConfig\n  | GrokFileSearchToolConfig\n  | GrokMCPToolConfig\n\ntype GrokProviderToolMetadata<TKind extends GrokProviderToolKind> = Extract<\n  GrokServerTool,\n  { type: TKind }\n> & {\n  __kind: GrokToolKindMarker<TKind>\n}\n\nexport type GrokResponsesTool = GrokServerTool | ResponsesFunctionTool\n\nfunction providerTool<TKind extends GrokProviderToolKind>(\n  kind: TKind,\n  description: string,\n  metadata: Extract<GrokServerTool, { type: TKind }>,\n): GrokProviderTool<TKind> {\n  return brandProviderTool<GrokProviderTool<TKind>>({\n    name: kind,\n    description,\n    metadata: {\n      __kind: `grok.${kind}`,\n      ...metadata,\n    },\n  })\n}\n\nexport function grokWebSearchTool(\n  config: Omit<GrokWebSearchToolConfig, 'type'> = {},\n): GrokProviderTool<'web_search'> {\n  if (\n    config.filters?.allowed_domains !== undefined &&\n    config.filters.excluded_domains !== undefined\n  ) {\n    throw new Error(\n      'allowed_domains and excluded_domains cannot both be provided.',\n    )\n  }\n  if (\n    config.filters?.allowed_domains !== undefined &&\n    config.filters.allowed_domains.length > 5\n  ) {\n    throw new Error('allowed_domains supports at most 5 domains.')\n  }\n  if (\n    config.filters?.excluded_domains !== undefined &&\n    config.filters.excluded_domains.length > 5\n  ) {\n    throw new Error('excluded_domains supports at most 5 domains.')\n  }\n  return providerTool('web_search', 'Search the web', {\n    type: 'web_search',\n    ...config,\n  })\n}\n\nexport function grokXSearchTool(\n  config: Omit<GrokXSearchToolConfig, 'type'> = {},\n): GrokProviderTool<'x_search'> {\n  if (\n    config.allowed_x_handles !== undefined &&\n    config.excluded_x_handles !== undefined\n  ) {\n    throw new Error(\n      'allowed_x_handles and excluded_x_handles cannot both be provided.',\n    )\n  }\n  if (\n    config.allowed_x_handles !== undefined &&\n    config.allowed_x_handles.length > 20\n  ) {\n    throw new Error('allowed_x_handles supports at most 20 handles.')\n  }\n  if (\n    config.excluded_x_handles !== undefined &&\n    config.excluded_x_handles.length > 20\n  ) {\n    throw new Error('excluded_x_handles supports at most 20 handles.')\n  }\n  return providerTool('x_search', 'Search X posts', {\n    type: 'x_search',\n    ...config,\n  })\n}\n\nexport function grokFileSearchTool(\n  config: Omit<GrokFileSearchToolConfig, 'type'>,\n): GrokProviderTool<'file_search'> {\n  if (config.vector_store_ids.length === 0) {\n    throw new Error('vector_store_ids must contain at least one collection id.')\n  }\n  if (config.max_num_results !== undefined) {\n    if (config.max_num_results < 1 || config.max_num_results > 50) {\n      throw new Error('max_num_results must be between 1 and 50.')\n    }\n  }\n  return providerTool('file_search', 'Search xAI file collections', {\n    type: 'file_search',\n    ...config,\n  })\n}\n\nexport function grokMCPTool(\n  config: Omit<GrokMCPToolConfig, 'type'>,\n): GrokProviderTool<'mcp'> {\n  if (!config.server_url) {\n    throw new Error('server_url must be provided.')\n  }\n  return providerTool('mcp', config.server_description || 'Remote MCP server', {\n    type: 'mcp',\n    ...config,\n  })\n}\n\nfunction getGrokProviderToolKind(tool: Tool): GrokProviderToolKind | undefined {\n  const kind = (tool.metadata as { __kind?: unknown } | undefined)?.__kind\n  switch (kind) {\n    case 'grok.web_search':\n      return 'web_search'\n    case 'grok.x_search':\n      return 'x_search'\n    case 'grok.file_search':\n      return 'file_search'\n    case 'grok.mcp':\n      return 'mcp'\n    default:\n      return undefined\n  }\n}\n\nfunction convertGrokProviderToolToAdapterFormat(\n  tool: Tool,\n  kind: GrokProviderToolKind,\n): GrokServerTool {\n  const metadata = tool.metadata as GrokProviderToolMetadata<typeof kind>\n  if (metadata.type !== kind) {\n    throw new Error(\n      `convertGrokProviderToolToAdapterFormat: tool \"${tool.name}\" has mismatched Grok tool metadata.`,\n    )\n  }\n  const { __kind: _kind, ...toolConfig } = metadata\n  void _kind\n  return toolConfig\n}\n\nexport function convertToolsToProviderFormat(\n  tools: Array<Tool>,\n): Array<GrokResponsesTool> {\n  return tools.map((tool) => {\n    const grokProviderToolKind = getGrokProviderToolKind(tool)\n    if (grokProviderToolKind) {\n      return convertGrokProviderToolToAdapterFormat(tool, grokProviderToolKind)\n    }\n    return convertFunctionToolToResponsesFormat(tool)\n  })\n}\n"],"mappings":";;;AAoEA,SAAS,aACP,MACA,aACA,UACyB;CACzB,OAAO,kBAA2C;EAChD,MAAM;EACN;EACA,UAAU;GACR,QAAQ,QAAQ;GAChB,GAAG;EACL;CACF,CAAC;AACH;AAEA,SAAgB,kBACd,SAAgD,CAAC,GACjB;CAChC,IACE,OAAO,SAAS,oBAAoB,KAAA,KACpC,OAAO,QAAQ,qBAAqB,KAAA,GAEpC,MAAM,IAAI,MACR,+DACF;CAEF,IACE,OAAO,SAAS,oBAAoB,KAAA,KACpC,OAAO,QAAQ,gBAAgB,SAAS,GAExC,MAAM,IAAI,MAAM,6CAA6C;CAE/D,IACE,OAAO,SAAS,qBAAqB,KAAA,KACrC,OAAO,QAAQ,iBAAiB,SAAS,GAEzC,MAAM,IAAI,MAAM,8CAA8C;CAEhE,OAAO,aAAa,cAAc,kBAAkB;EAClD,MAAM;EACN,GAAG;CACL,CAAC;AACH;AAEA,SAAgB,gBACd,SAA8C,CAAC,GACjB;CAC9B,IACE,OAAO,sBAAsB,KAAA,KAC7B,OAAO,uBAAuB,KAAA,GAE9B,MAAM,IAAI,MACR,mEACF;CAEF,IACE,OAAO,sBAAsB,KAAA,KAC7B,OAAO,kBAAkB,SAAS,IAElC,MAAM,IAAI,MAAM,gDAAgD;CAElE,IACE,OAAO,uBAAuB,KAAA,KAC9B,OAAO,mBAAmB,SAAS,IAEnC,MAAM,IAAI,MAAM,iDAAiD;CAEnE,OAAO,aAAa,YAAY,kBAAkB;EAChD,MAAM;EACN,GAAG;CACL,CAAC;AACH;AAEA,SAAgB,mBACd,QACiC;CACjC,IAAI,OAAO,iBAAiB,WAAW,GACrC,MAAM,IAAI,MAAM,2DAA2D;CAE7E,IAAI,OAAO,oBAAoB,KAAA,GACzB;MAAA,OAAO,kBAAkB,KAAK,OAAO,kBAAkB,IACzD,MAAM,IAAI,MAAM,2CAA2C;CAAA;CAG/D,OAAO,aAAa,eAAe,+BAA+B;EAChE,MAAM;EACN,GAAG;CACL,CAAC;AACH;AAEA,SAAgB,YACd,QACyB;CACzB,IAAI,CAAC,OAAO,YACV,MAAM,IAAI,MAAM,8BAA8B;CAEhD,OAAO,aAAa,OAAO,OAAO,sBAAsB,qBAAqB;EAC3E,MAAM;EACN,GAAG;CACL,CAAC;AACH;AAEA,SAAS,wBAAwB,MAA8C;CAE7E,QADc,KAAK,UAA+C,QAClE;EACE,KAAK,mBACH,OAAO;EACT,KAAK,iBACH,OAAO;EACT,KAAK,oBACH,OAAO;EACT,KAAK,YACH,OAAO;EACT,SACE;CACJ;AACF;AAEA,SAAS,uCACP,MACA,MACgB;CAChB,MAAM,WAAW,KAAK;CACtB,IAAI,SAAS,SAAS,MACpB,MAAM,IAAI,MACR,iDAAiD,KAAK,KAAK,qCAC7D;CAEF,MAAM,EAAE,QAAQ,OAAO,GAAG,eAAe;CAEzC,OAAO;AACT;AAEA,SAAgB,6BACd,OAC0B;CAC1B,OAAO,MAAM,KAAK,SAAS;EACzB,MAAM,uBAAuB,wBAAwB,IAAI;EACzD,IAAI,sBACF,OAAO,uCAAuC,MAAM,oBAAoB;EAE1E,OAAO,qCAAqC,IAAI;CAClD,CAAC;AACH"}

package/dist/esm/video/video-provider-options.d.ts

@@ -16,6 +16,13 @@ export type GrokVideoAspectRatio = '1:1' | '16:9' | '9:16' | '4:3' | '3:4' | '3:
  * @experimental Video generation is an experimental feature and may change.
  */
 export type GrokVideoResolution = '480p' | '720p' | '1080p';
+/**
+ * Resolutions accepted by grok-imagine-video (v1.0). Native 1080p is a
+ * grok-imagine-video-1.5 generation feature.
+ *
+ * @experimental Video generation is an experimental feature and may change.
+ */
+export type GrokVideoResolutionV1 = '480p' | '720p';
 /**
  * Size strings for grok-imagine video models. The Imagine API is
  * aspect-ratio based rather than pixel-size based; like the grok-imagine
@@ -26,6 +33,12 @@ export type GrokVideoResolution = '480p' | '720p' | '1080p';
  * @experimental Video generation is an experimental feature and may change.
  */
 export type GrokVideoSize = GrokVideoAspectRatio | `${GrokVideoAspectRatio}_${GrokVideoResolution}`;
+/**
+ * Size strings for grok-imagine-video (v1.0) — 1080p is not in the type.
+ *
+ * @experimental Video generation is an experimental feature and may change.
+ */
+export type GrokVideoSizeV1 = GrokVideoAspectRatio | `${GrokVideoAspectRatio}_${GrokVideoResolutionV1}`;
 /**
  * Video duration limits enforced by the Imagine API (seconds).
  */
@@ -41,6 +54,13 @@ export declare function parseGrokVideoSize(size: string): {
     aspectRatio: string;
     resolution?: string;
 } | undefined;
+/**
+ * Models that accept native 1080p on text-to-video and image-to-video.
+ * Reference-to-video stays capped at 720p even on these models.
+ *
+ * @experimental Video generation is an experimental feature and may change.
+ */
+export declare function isGrokVideoNative1080pModel(model: string): boolean;
 /**
  * Validate the `size` template for a given grok video model.
  *
@@ -76,33 +96,128 @@ export declare const GROK_VIDEO_DURATIONS: {
  */
 export declare function getGrokVideoDurationOptions<TModel extends GrokVideoModel>(model: TModel): DurationOptions<GrokVideoModelDurationByName[TModel]>;
 /**
- * Provider-specific options for grok video generation. These map directly
- * onto the Imagine API request body and take precedence over the generic
- * `size` / `duration` options when both are provided.
+ * Request mode for a source-video job. `'edit'` posts to `/v1/videos/edits`
+ * (modify the source clip in place); `'extend'` posts to
+ * `/v1/videos/extensions` (continue the source clip — `duration` is the
+ * length of the **added tail**, not the total). Both require exactly one
+ * video prompt part carrying the source clip, and both are
+ * `grok-imagine-video` (v1.0) only.
+ *
+ * Output geometry (aspect ratio / resolution) is inherited from the source
+ * clip in both modes, capped at 720p, and edit outputs also inherit the
+ * source length — the adapter rejects `size`, `aspect_ratio`, `resolution`,
+ * and (in edit mode) `duration` rather than sending fields the API ignores.
  *
  * @experimental Video generation is an experimental feature and may change.
  */
-export interface GrokVideoProviderOptions {
+export type GrokVideoMode = 'edit' | 'extend';
+/**
+ * Provider options shared by both grok-imagine video models. These map
+ * directly onto the Imagine API request body and take precedence over the
+ * generic `size` / `duration` options when both are provided.
+ *
+ * @experimental Video generation is an experimental feature and may change.
+ */
+export interface GrokVideoBaseProviderOptions {
     /**
-     * Output aspect ratio.
+     * Output aspect ratio. Generation only — edit / extend outputs inherit
+     * the source clip's geometry and the adapter rejects this in those modes.
      */
     aspect_ratio?: GrokVideoAspectRatio;
     /**
-     * Output resolution tier.
+     * Output resolution tier. Generation only — edit / extend outputs inherit
+     * the source clip's geometry and the adapter rejects this in those modes.
+     * `1080p` is grok-imagine-video-1.5 generation only; reference-to-video
+     * is capped at 720p.
      */
     resolution?: GrokVideoResolution;
     /**
-     * Video duration in integer seconds (1–15).
+     * Video duration in integer seconds (1–15). In `'extend'` mode this is
+     * the length of the added tail only, not the total output length. Not
+     * valid in `'edit'` mode (the output inherits the source clip's length).
      */
     duration?: number;
 }
 /**
- * Type-only map from model name to its specific provider options.
+ * Provider options for grok-imagine-video (v1.0), which is the only model
+ * that accepts a source-video edit / extend job.
+ *
+ * @experimental Video generation is an experimental feature and may change.
+ */
+export interface GrokVideoSourceProviderOptions extends GrokVideoBaseProviderOptions {
+    /**
+     * Selects the request mode for a source-video prompt part: `'edit'`
+     * (`/v1/videos/edits`) or `'extend'` (`/v1/videos/extensions`). Required
+     * when the prompt carries a video part; not valid without one. Omit for
+     * plain generation (`/v1/videos/generations`). grok-imagine-video only.
+     */
+    mode?: GrokVideoMode;
+}
+/**
+ * Provider options for grok-imagine-video-1.5, which adds the
+ * reference-to-video inputs on top of the shared options.
+ *
+ * @experimental Video generation is an experimental feature and may change.
+ */
+export interface GrokVideoProviderOptions extends GrokVideoBaseProviderOptions {
+    /**
+     * Reference images for reference-to-video generation (output capped at
+     * 720p). Usually populated from image prompt parts with
+     * `metadata.role: 'reference'` (or `'character'`); set explicitly to
+     * replace the part-derived list. Reference images are addressed from the
+     * prompt text as `<IMAGE_0>`, `<IMAGE_1>`, … in request order, and do not
+     * lock the first frame.
+     */
+    reference_images?: Array<{
+        url: string;
+    }>;
+    /**
+     * Preset TTS voices to reference for generated speech (max 3). Voice ids
+     * come from the xAI TTS voice roster (e.g. 'eve', 'rex') or a custom
+     * voice id, and are addressed from the prompt text as `<AUDIO_0>`,
+     * `<AUDIO_1>`, `<AUDIO_2>`.
+     */
+    reference_audios?: Array<{
+        voice_id: string;
+    }>;
+}
+/**
+ * Widest option surface. Used when `modelOptions` arrives as deserialized
+ * JSON and the adapter must validate fields the per-model map already
+ * hides at compile time.
+ *
+ * @experimental Video generation is an experimental feature and may change.
+ */
+export type GrokVideoRuntimeOptions = GrokVideoSourceProviderOptions & GrokVideoProviderOptions;
+/**
+ * Maximum reference voices accepted by the Imagine video endpoint.
+ */
+export declare const GROK_VIDEO_MAX_REFERENCE_AUDIOS = 3;
+/**
+ * Maximum reference images accepted by the Imagine video endpoint.
+ */
+export declare const GROK_VIDEO_MAX_REFERENCE_IMAGES = 7;
+/**
+ * True when the model accepts reference-to-video inputs.
+ *
+ * @experimental Video generation is an experimental feature and may change.
+ */
+export declare function isGrokVideoReferenceModel(model: string): boolean;
+/**
+ * True when the model accepts edit / extend source-video jobs.
+ *
+ * @experimental Video generation is an experimental feature and may change.
+ */
+export declare function isGrokVideoSourceModel(model: string): boolean;
+/**
+ * Type-only map from model name to its specific provider options. Only
+ * grok-imagine-video-1.5 exposes the reference-to-video fields. Only
+ * grok-imagine-video (v1.0) exposes `mode` for edit / extend.
  *
  * @experimental Video generation is an experimental feature and may change.
  */
 export type GrokVideoModelProviderOptionsByName = {
-    'grok-imagine-video': GrokVideoProviderOptions;
+    'grok-imagine-video': GrokVideoSourceProviderOptions;
     'grok-imagine-video-1.5': GrokVideoProviderOptions;
 };
 /**
@@ -111,25 +226,20 @@ export type GrokVideoModelProviderOptionsByName = {
  * @experimental Video generation is an experimental feature and may change.
  */
 export type GrokVideoModelSizeByName = {
-    'grok-imagine-video': GrokVideoSize;
+    'grok-imagine-video': GrokVideoSizeV1;
     'grok-imagine-video-1.5': GrokVideoSize;
 };
 /**
  * Type-only map from model name to the non-text prompt modalities it accepts.
- * Both models accept an `image` prompt part as the starting frame:
- * `grok-imagine-video` (v1.0) does text-to-video and image-to-video, while
- * `grok-imagine-video-1.5` is image-to-video only (the image is required).
+ * Both models support text-to-video and accept an optional `image` prompt
+ * part as the starting frame; image parts with `metadata.role: 'reference'`
+ * or `'character'` become `reference_images` (grok-imagine-video-1.5 only).
+ * A `video` prompt part carries the source clip for edit / extension mode
+ * on grok-imagine-video only (`modelOptions.mode: 'edit' | 'extend'`).
  *
  * @experimental Video generation is an experimental feature and may change.
  */
 export type GrokVideoModelInputModalitiesByName = {
-    'grok-imagine-video': readonly ['image'];
+    'grok-imagine-video': readonly ['image', 'video'];
     'grok-imagine-video-1.5': readonly ['image'];
 };
-/**
- * True when the model only supports image-to-video (a starting frame is
- * required).
- *
- * @experimental Video generation is an experimental feature and may change.
- */
-export declare function isImageToVideoOnlyModel(model: string): boolean;