@tanstack/ai-elevenlabs 0.2.28 → 0.2.30

package/dist/esm/adapters/transcription.js

@@ -0,0 +1,143 @@
+import { BaseTranscriptionAdapter } from "@tanstack/ai/adapters";
+import { createElevenLabsClient, generateId, dataUrlToBlob } from "../utils/client.js";
+class ElevenLabsTranscriptionAdapter extends BaseTranscriptionAdapter {
+  name = "elevenlabs";
+  client;
+  constructor(model, config) {
+    super(model, config ?? {});
+    this.client = createElevenLabsClient(config);
+  }
+  async transcribe(options) {
+    const { logger } = options;
+    logger.request(
+      `activity=generateTranscription provider=elevenlabs model=${this.model}`,
+      { provider: "elevenlabs", model: this.model }
+    );
+    try {
+      const modelOpts = options.modelOptions ?? {};
+      const audioInput = normalizeAudioInput(options.audio);
+      const response = await this.client.speechToText.convert({
+        modelId: this.model,
+        ...audioInput.kind === "file" ? { file: audioInput.value } : { cloudStorageUrl: audioInput.value },
+        ...options.language ? { languageCode: options.language } : {},
+        ...modelOpts.tagAudioEvents != null ? { tagAudioEvents: modelOpts.tagAudioEvents } : {},
+        ...modelOpts.numSpeakers != null ? { numSpeakers: modelOpts.numSpeakers } : {},
+        ...modelOpts.timestampsGranularity ? { timestampsGranularity: modelOpts.timestampsGranularity } : {},
+        ...modelOpts.diarize != null ? { diarize: modelOpts.diarize } : {},
+        ...modelOpts.diarizationThreshold != null ? { diarizationThreshold: modelOpts.diarizationThreshold } : {},
+        ...modelOpts.detectSpeakerRoles != null ? { detectSpeakerRoles: modelOpts.detectSpeakerRoles } : {},
+        ...modelOpts.keyterms ? { keyterms: modelOpts.keyterms } : {},
+        ...modelOpts.entityDetection ? { entityDetection: modelOpts.entityDetection } : {},
+        ...modelOpts.entityRedaction ? { entityRedaction: modelOpts.entityRedaction } : {},
+        ...modelOpts.entityRedactionMode ? { entityRedactionMode: modelOpts.entityRedactionMode } : {},
+        ...modelOpts.noVerbatim != null ? { noVerbatim: modelOpts.noVerbatim } : {},
+        ...modelOpts.temperature != null ? { temperature: modelOpts.temperature } : {},
+        ...modelOpts.seed != null ? { seed: modelOpts.seed } : {},
+        ...modelOpts.enableLogging != null ? { enableLogging: modelOpts.enableLogging } : {},
+        ...modelOpts.useMultiChannel != null ? { useMultiChannel: modelOpts.useMultiChannel } : {},
+        ...modelOpts.fileFormat ? { fileFormat: modelOpts.fileFormat } : {}
+      });
+      return this.transformResponse(response);
+    } catch (error) {
+      logger.errors("elevenlabs.generateTranscription fatal", {
+        error,
+        source: "elevenlabs.generateTranscription"
+      });
+      throw error;
+    }
+  }
+  transformResponse(response) {
+    const data = response;
+    if (data.transcripts) {
+      const joinedText = data.transcripts.map((t) => t.text ?? "").filter(Boolean).join("\n");
+      const joinedWords = data.transcripts.flatMap((t) => t.words ?? []);
+      const duration = data.transcripts.reduce(
+        (max, t) => Math.max(max, t.audioDurationSecs ?? 0),
+        0
+      );
+      const firstLang = data.transcripts.find(
+        (t) => t.languageCode
+      )?.languageCode;
+      return {
+        id: generateId(this.name),
+        model: this.model,
+        text: joinedText,
+        ...firstLang ? { language: firstLang } : {},
+        ...duration ? { duration } : {},
+        ...buildWordsAndSegments(joinedWords)
+      };
+    }
+    return {
+      id: generateId(this.name),
+      model: this.model,
+      text: data.text ?? "",
+      ...data.languageCode ? { language: data.languageCode } : {},
+      ...data.audioDurationSecs ? { duration: data.audioDurationSecs } : {},
+      ...buildWordsAndSegments(data.words ?? [])
+    };
+  }
+  generateId() {
+    return generateId(this.name);
+  }
+}
+function normalizeAudioInput(audio) {
+  if (audio instanceof ArrayBuffer) {
+    return { kind: "file", value: new Blob([audio]) };
+  }
+  if (typeof audio === "string") {
+    const blob = dataUrlToBlob(audio);
+    if (blob) return { kind: "file", value: blob };
+    return { kind: "url", value: audio };
+  }
+  return { kind: "file", value: audio };
+}
+function buildWordsAndSegments(words) {
+  const timedWords = words.filter(
+    (w) => typeof w.start === "number" && typeof w.end === "number" && w.type !== "spacing"
+  );
+  if (timedWords.length === 0) return {};
+  const outWords = timedWords.map((w) => ({
+    word: w.text,
+    start: w.start,
+    end: w.end
+  }));
+  const segments = [];
+  let current = null;
+  for (const w of timedWords) {
+    if (!current) {
+      current = {
+        start: w.start,
+        end: w.end,
+        text: w.text,
+        ...w.speakerId ? { speaker: w.speakerId } : {}
+      };
+      continue;
+    }
+    if (w.speakerId && current.speaker !== w.speakerId) {
+      segments.push({ id: segments.length, ...current });
+      current = {
+        start: w.start,
+        end: w.end,
+        text: w.text,
+        speaker: w.speakerId
+      };
+      continue;
+    }
+    current.end = w.end;
+    current.text = current.text ? `${current.text} ${w.text}` : w.text;
+  }
+  if (current) segments.push({ id: segments.length, ...current });
+  return { words: outWords, segments };
+}
+function elevenlabsTranscription(model, config) {
+  return new ElevenLabsTranscriptionAdapter(model, config);
+}
+function createElevenLabsTranscription(model, apiKey, config) {
+  return new ElevenLabsTranscriptionAdapter(model, { apiKey, ...config });
+}
+export {
+  ElevenLabsTranscriptionAdapter,
+  createElevenLabsTranscription,
+  elevenlabsTranscription
+};
+//# sourceMappingURL=transcription.js.map

package/dist/esm/adapters/transcription.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcription.js","sources":["../../../src/adapters/transcription.ts"],"sourcesContent":["import { BaseTranscriptionAdapter } from '@tanstack/ai/adapters'\nimport {\n  createElevenLabsClient,\n  dataUrlToBlob,\n  generateId,\n} from '../utils/client'\nimport type { ElevenLabsClient } from '@elevenlabs/elevenlabs-js'\nimport type {\n  TranscriptionOptions,\n  TranscriptionResult,\n  TranscriptionSegment,\n  TranscriptionWord,\n} from '@tanstack/ai'\nimport type { ElevenLabsClientConfig } from '../utils/client'\nimport type { ElevenLabsTranscriptionModel } from '../model-meta'\n\n/**\n * Provider-specific options for ElevenLabs Scribe transcription. Fields map\n * 1:1 onto the SDK's `BodySpeechToTextV1SpeechToTextPost` — mirroring the\n * names so documentation stays useful.\n * @see https://elevenlabs.io/docs/api-reference/speech-to-text/convert\n */\nexport interface ElevenLabsTranscriptionProviderOptions {\n  /** Annotate non-speech events like (laughter), (footsteps), …. */\n  tagAudioEvents?: boolean\n  /** Maximum number of speakers in the audio (1..32). */\n  numSpeakers?: number\n  /** Timestamp granularity for words. */\n  timestampsGranularity?: 'word' | 'character' | 'none'\n  /** Enable speaker diarization. */\n  diarize?: boolean\n  /** Diarization threshold (requires `diarize=true` and no `numSpeakers`). */\n  diarizationThreshold?: number\n  /** Detect speaker roles (agent/customer). Requires diarize=true. */\n  detectSpeakerRoles?: boolean\n  /** Bias the model towards these keyterms (max 1000). */\n  keyterms?: Array<string>\n  /**\n   * Entity detection: `'all'`, a category (`'pii'`, `'phi'`, `'pci'`,\n   * `'other'`, `'offensive_language'`), or a specific entity type.\n   */\n  entityDetection?: string\n  /** Redact entities from the transcript text. Must be a subset of `entityDetection`. */\n  entityRedaction?: string\n  /** How redacted entities are formatted. */\n  entityRedactionMode?: string\n  /** Whether to skip filler words / non-speech sounds (scribe_v2 only). */\n  noVerbatim?: boolean\n  /** Sampling temperature (0..2). */\n  temperature?: number\n  /** Deterministic sampling seed (0..2147483647). */\n  seed?: number\n  /** Use `false` for zero-retention mode (enterprise only). */\n  enableLogging?: boolean\n  /** Multi-channel audio with one speaker per channel. Max 5 channels. */\n  useMultiChannel?: boolean\n  /**\n   * Hint for audio format. Use `'pcm_s16le_16'` to skip encoding for 16-bit\n   * PCM @ 16kHz mono little-endian inputs (lower latency).\n   */\n  fileFormat?: 'pcm_s16le_16' | 'other'\n}\n\n/**\n * ElevenLabs speech-to-text adapter built on the official SDK's Scribe family.\n *\n * @example\n * ```ts\n * const adapter = elevenlabsTranscription('scribe_v1')\n * const result = await generateTranscription({\n *   adapter,\n *   audio: fileInput,\n *   language: 'en',\n * })\n * ```\n */\nexport class ElevenLabsTranscriptionAdapter<\n  TModel extends ElevenLabsTranscriptionModel,\n> extends BaseTranscriptionAdapter<\n  TModel,\n  ElevenLabsTranscriptionProviderOptions\n> {\n  readonly name = 'elevenlabs' as const\n\n  private readonly client: ElevenLabsClient\n\n  constructor(model: TModel, config?: ElevenLabsClientConfig) {\n    super(model, config ?? {})\n    this.client = createElevenLabsClient(config)\n  }\n\n  async transcribe(\n    options: TranscriptionOptions<ElevenLabsTranscriptionProviderOptions>,\n  ): Promise<TranscriptionResult> {\n    const { logger } = options\n    logger.request(\n      `activity=generateTranscription provider=elevenlabs model=${this.model}`,\n      { provider: 'elevenlabs', model: this.model },\n    )\n    try {\n      const modelOpts = options.modelOptions ?? {}\n      const audioInput = normalizeAudioInput(options.audio)\n\n      const response = await this.client.speechToText.convert({\n        modelId: this.model,\n        ...(audioInput.kind === 'file'\n          ? { file: audioInput.value }\n          : { cloudStorageUrl: audioInput.value }),\n        ...(options.language ? { languageCode: options.language } : {}),\n        ...(modelOpts.tagAudioEvents != null\n          ? { tagAudioEvents: modelOpts.tagAudioEvents }\n          : {}),\n        ...(modelOpts.numSpeakers != null\n          ? { numSpeakers: modelOpts.numSpeakers }\n          : {}),\n        ...(modelOpts.timestampsGranularity\n          ? { timestampsGranularity: modelOpts.timestampsGranularity }\n          : {}),\n        ...(modelOpts.diarize != null ? { diarize: modelOpts.diarize } : {}),\n        ...(modelOpts.diarizationThreshold != null\n          ? { diarizationThreshold: modelOpts.diarizationThreshold }\n          : {}),\n        ...(modelOpts.detectSpeakerRoles != null\n          ? { detectSpeakerRoles: modelOpts.detectSpeakerRoles }\n          : {}),\n        ...(modelOpts.keyterms ? { keyterms: modelOpts.keyterms } : {}),\n        ...(modelOpts.entityDetection\n          ? { entityDetection: modelOpts.entityDetection }\n          : {}),\n        ...(modelOpts.entityRedaction\n          ? { entityRedaction: modelOpts.entityRedaction }\n          : {}),\n        ...(modelOpts.entityRedactionMode\n          ? { entityRedactionMode: modelOpts.entityRedactionMode }\n          : {}),\n        ...(modelOpts.noVerbatim != null\n          ? { noVerbatim: modelOpts.noVerbatim }\n          : {}),\n        ...(modelOpts.temperature != null\n          ? { temperature: modelOpts.temperature }\n          : {}),\n        ...(modelOpts.seed != null ? { seed: modelOpts.seed } : {}),\n        ...(modelOpts.enableLogging != null\n          ? { enableLogging: modelOpts.enableLogging }\n          : {}),\n        ...(modelOpts.useMultiChannel != null\n          ? { useMultiChannel: modelOpts.useMultiChannel }\n          : {}),\n        ...(modelOpts.fileFormat ? { fileFormat: modelOpts.fileFormat } : {}),\n      } as Parameters<ElevenLabsClient['speechToText']['convert']>[0])\n\n      return this.transformResponse(response)\n    } catch (error) {\n      logger.errors('elevenlabs.generateTranscription fatal', {\n        error,\n        source: 'elevenlabs.generateTranscription',\n      })\n      throw error\n    }\n  }\n\n  private transformResponse(\n    response: Awaited<ReturnType<ElevenLabsClient['speechToText']['convert']>>,\n  ): TranscriptionResult {\n    // The SDK types this as a union of single- and multi-channel responses.\n    // We treat multi-channel as \"join the channel transcripts\" — consumers\n    // who care about per-channel detail can re-parse from `modelOptions`.\n    // eslint-disable-next-line no-restricted-syntax -- bridges SpeechToTextConvertResponse union (incl. webhook variant with no text/words/transcripts) to a flattened duck-typed shape we discriminate at runtime\n    const data = response as unknown as {\n      text?: string\n      languageCode?: string\n      languageProbability?: number\n      words?: Array<{\n        text: string\n        start?: number\n        end?: number\n        type: string\n        speakerId?: string\n      }>\n      audioDurationSecs?: number\n      transcripts?: Array<{\n        text?: string\n        languageCode?: string\n        words?: Array<{\n          text: string\n          start?: number\n          end?: number\n          type: string\n          speakerId?: string\n        }>\n        audioDurationSecs?: number\n      }>\n    }\n\n    if (data.transcripts) {\n      const joinedText = data.transcripts\n        .map((t) => t.text ?? '')\n        .filter(Boolean)\n        .join('\\n')\n      const joinedWords = data.transcripts.flatMap((t) => t.words ?? [])\n      const duration = data.transcripts.reduce(\n        (max, t) => Math.max(max, t.audioDurationSecs ?? 0),\n        0,\n      )\n      const firstLang = data.transcripts.find(\n        (t) => t.languageCode,\n      )?.languageCode\n      return {\n        id: generateId(this.name),\n        model: this.model,\n        text: joinedText,\n        ...(firstLang ? { language: firstLang } : {}),\n        ...(duration ? { duration } : {}),\n        ...buildWordsAndSegments(joinedWords),\n      }\n    }\n\n    return {\n      id: generateId(this.name),\n      model: this.model,\n      text: data.text ?? '',\n      ...(data.languageCode ? { language: data.languageCode } : {}),\n      ...(data.audioDurationSecs ? { duration: data.audioDurationSecs } : {}),\n      ...buildWordsAndSegments(data.words ?? []),\n    }\n  }\n\n  protected override generateId(): string {\n    return generateId(this.name)\n  }\n}\n\ntype NormalizedAudio =\n  | { kind: 'file'; value: Blob }\n  | { kind: 'url'; value: string }\n\nfunction normalizeAudioInput(\n  audio: TranscriptionOptions['audio'],\n): NormalizedAudio {\n  if (audio instanceof ArrayBuffer) {\n    return { kind: 'file', value: new Blob([audio]) }\n  }\n  if (typeof audio === 'string') {\n    const blob = dataUrlToBlob(audio)\n    if (blob) return { kind: 'file', value: blob }\n    return { kind: 'url', value: audio }\n  }\n  // Blob or File both fit the SDK's `Uploadable` contract.\n  return { kind: 'file', value: audio }\n}\n\nfunction buildWordsAndSegments(\n  words: Array<{\n    text: string\n    start?: number\n    end?: number\n    type: string\n    speakerId?: string\n  }>,\n): {\n  words?: Array<TranscriptionWord>\n  segments?: Array<TranscriptionSegment>\n} {\n  const timedWords = words.filter(\n    (w): w is typeof w & { start: number; end: number } =>\n      typeof w.start === 'number' &&\n      typeof w.end === 'number' &&\n      w.type !== 'spacing',\n  )\n  if (timedWords.length === 0) return {}\n\n  const outWords: Array<TranscriptionWord> = timedWords.map((w) => ({\n    word: w.text,\n    start: w.start,\n    end: w.end,\n  }))\n\n  // Group contiguous words that share a speaker into segments. If no speaker\n  // is ever set, we still emit one segment per sentence-ish grouping.\n  const segments: Array<TranscriptionSegment> = []\n  let current: {\n    start: number\n    end: number\n    text: string\n    speaker?: string\n  } | null = null\n\n  for (const w of timedWords) {\n    if (!current) {\n      current = {\n        start: w.start,\n        end: w.end,\n        text: w.text,\n        ...(w.speakerId ? { speaker: w.speakerId } : {}),\n      }\n      continue\n    }\n    if (w.speakerId && current.speaker !== w.speakerId) {\n      segments.push({ id: segments.length, ...current })\n      current = {\n        start: w.start,\n        end: w.end,\n        text: w.text,\n        speaker: w.speakerId,\n      }\n      continue\n    }\n    current.end = w.end\n    current.text = current.text ? `${current.text} ${w.text}` : w.text\n  }\n  if (current) segments.push({ id: segments.length, ...current })\n\n  return { words: outWords, segments }\n}\n\n/**\n * Create an ElevenLabs transcription adapter using `ELEVENLABS_API_KEY` from env.\n */\nexport function elevenlabsTranscription<\n  TModel extends ElevenLabsTranscriptionModel,\n>(\n  model: TModel,\n  config?: ElevenLabsClientConfig,\n): ElevenLabsTranscriptionAdapter<TModel> {\n  return new ElevenLabsTranscriptionAdapter(model, config)\n}\n\n/**\n * Create an ElevenLabs transcription adapter with an explicit API key.\n */\nexport function createElevenLabsTranscription<\n  TModel extends ElevenLabsTranscriptionModel,\n>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<ElevenLabsClientConfig, 'apiKey'>,\n): ElevenLabsTranscriptionAdapter<TModel> {\n  return new ElevenLabsTranscriptionAdapter(model, { apiKey, ...config })\n}\n"],"names":[],"mappings":";;AA4EO,MAAM,uCAEH,yBAGR;AAAA,EACS,OAAO;AAAA,EAEC;AAAA,EAEjB,YAAY,OAAe,QAAiC;AAC1D,UAAM,OAAO,UAAU,EAAE;AACzB,SAAK,SAAS,uBAAuB,MAAM;AAAA,EAC7C;AAAA,EAEA,MAAM,WACJ,SAC8B;AAC9B,UAAM,EAAE,WAAW;AACnB,WAAO;AAAA,MACL,4DAA4D,KAAK,KAAK;AAAA,MACtE,EAAE,UAAU,cAAc,OAAO,KAAK,MAAA;AAAA,IAAM;AAE9C,QAAI;AACF,YAAM,YAAY,QAAQ,gBAAgB,CAAA;AAC1C,YAAM,aAAa,oBAAoB,QAAQ,KAAK;AAEpD,YAAM,WAAW,MAAM,KAAK,OAAO,aAAa,QAAQ;AAAA,QACtD,SAAS,KAAK;AAAA,QACd,GAAI,WAAW,SAAS,SACpB,EAAE,MAAM,WAAW,UACnB,EAAE,iBAAiB,WAAW,MAAA;AAAA,QAClC,GAAI,QAAQ,WAAW,EAAE,cAAc,QAAQ,SAAA,IAAa,CAAA;AAAA,QAC5D,GAAI,UAAU,kBAAkB,OAC5B,EAAE,gBAAgB,UAAU,eAAA,IAC5B,CAAA;AAAA,QACJ,GAAI,UAAU,eAAe,OACzB,EAAE,aAAa,UAAU,YAAA,IACzB,CAAA;AAAA,QACJ,GAAI,UAAU,wBACV,EAAE,uBAAuB,UAAU,sBAAA,IACnC,CAAA;AAAA,QACJ,GAAI,UAAU,WAAW,OAAO,EAAE,SAAS,UAAU,QAAA,IAAY,CAAA;AAAA,QACjE,GAAI,UAAU,wBAAwB,OAClC,EAAE,sBAAsB,UAAU,qBAAA,IAClC,CAAA;AAAA,QACJ,GAAI,UAAU,sBAAsB,OAChC,EAAE,oBAAoB,UAAU,mBAAA,IAChC,CAAA;AAAA,QACJ,GAAI,UAAU,WAAW,EAAE,UAAU,UAAU,SAAA,IAAa,CAAA;AAAA,QAC5D,GAAI,UAAU,kBACV,EAAE,iBAAiB,UAAU,gBAAA,IAC7B,CAAA;AAAA,QACJ,GAAI,UAAU,kBACV,EAAE,iBAAiB,UAAU,gBAAA,IAC7B,CAAA;AAAA,QACJ,GAAI,UAAU,sBACV,EAAE,qBAAqB,UAAU,oBAAA,IACjC,CAAA;AAAA,QACJ,GAAI,UAAU,cAAc,OACxB,EAAE,YAAY,UAAU,WAAA,IACxB,CAAA;AAAA,QACJ,GAAI,UAAU,eAAe,OACzB,EAAE,aAAa,UAAU,YAAA,IACzB,CAAA;AAAA,QACJ,GAAI,UAAU,QAAQ,OAAO,EAAE,MAAM,UAAU,KAAA,IAAS,CAAA;AAAA,QACxD,GAAI,UAAU,iBAAiB,OAC3B,EAAE,eAAe,UAAU,cAAA,IAC3B,CAAA;AAAA,QACJ,GAAI,UAAU,mBAAmB,OAC7B,EAAE,iBAAiB,UAAU,gBAAA,IAC7B,CAAA;AAAA,QACJ,GAAI,UAAU,aAAa,EAAE,YAAY,UAAU,WAAA,IAAe,CAAA;AAAA,MAAC,CACN;AAE/D,aAAO,KAAK,kBAAkB,QAAQ;AAAA,IACxC,SAAS,OAAO;AACd,aAAO,OAAO,0CAA0C;AAAA,QACtD;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEQ,kBACN,UACqB;AAKrB,UAAM,OAAO;AA0Bb,QAAI,KAAK,aAAa;AACpB,YAAM,aAAa,KAAK,YACrB,IAAI,CAAC,MAAM,EAAE,QAAQ,EAAE,EACvB,OAAO,OAAO,EACd,KAAK,IAAI;AACZ,YAAM,cAAc,KAAK,YAAY,QAAQ,CAAC,MAAM,EAAE,SAAS,EAAE;AACjE,YAAM,WAAW,KAAK,YAAY;AAAA,QAChC,CAAC,KAAK,MAAM,KAAK,IAAI,KAAK,EAAE,qBAAqB,CAAC;AAAA,QAClD;AAAA,MAAA;AAEF,YAAM,YAAY,KAAK,YAAY;AAAA,QACjC,CAAC,MAAM,EAAE;AAAA,MAAA,GACR;AACH,aAAO;AAAA,QACL,IAAI,WAAW,KAAK,IAAI;AAAA,QACxB,OAAO,KAAK;AAAA,QACZ,MAAM;AAAA,QACN,GAAI,YAAY,EAAE,UAAU,UAAA,IAAc,CAAA;AAAA,QAC1C,GAAI,WAAW,EAAE,SAAA,IAAa,CAAA;AAAA,QAC9B,GAAG,sBAAsB,WAAW;AAAA,MAAA;AAAA,IAExC;AAEA,WAAO;AAAA,MACL,IAAI,WAAW,KAAK,IAAI;AAAA,MACxB,OAAO,KAAK;AAAA,MACZ,MAAM,KAAK,QAAQ;AAAA,MACnB,GAAI,KAAK,eAAe,EAAE,UAAU,KAAK,aAAA,IAAiB,CAAA;AAAA,MAC1D,GAAI,KAAK,oBAAoB,EAAE,UAAU,KAAK,kBAAA,IAAsB,CAAA;AAAA,MACpE,GAAG,sBAAsB,KAAK,SAAS,CAAA,CAAE;AAAA,IAAA;AAAA,EAE7C;AAAA,EAEmB,aAAqB;AACtC,WAAO,WAAW,KAAK,IAAI;AAAA,EAC7B;AACF;AAMA,SAAS,oBACP,OACiB;AACjB,MAAI,iBAAiB,aAAa;AAChC,WAAO,EAAE,MAAM,QAAQ,OAAO,IAAI,KAAK,CAAC,KAAK,CAAC,EAAA;AAAA,EAChD;AACA,MAAI,OAAO,UAAU,UAAU;AAC7B,UAAM,OAAO,cAAc,KAAK;AAChC,QAAI,KAAM,QAAO,EAAE,MAAM,QAAQ,OAAO,KAAA;AACxC,WAAO,EAAE,MAAM,OAAO,OAAO,MAAA;AAAA,EAC/B;AAEA,SAAO,EAAE,MAAM,QAAQ,OAAO,MAAA;AAChC;AAEA,SAAS,sBACP,OAUA;AACA,QAAM,aAAa,MAAM;AAAA,IACvB,CAAC,MACC,OAAO,EAAE,UAAU,YACnB,OAAO,EAAE,QAAQ,YACjB,EAAE,SAAS;AAAA,EAAA;AAEf,MAAI,WAAW,WAAW,EAAG,QAAO,CAAA;AAEpC,QAAM,WAAqC,WAAW,IAAI,CAAC,OAAO;AAAA,IAChE,MAAM,EAAE;AAAA,IACR,OAAO,EAAE;AAAA,IACT,KAAK,EAAE;AAAA,EAAA,EACP;AAIF,QAAM,WAAwC,CAAA;AAC9C,MAAI,UAKO;AAEX,aAAW,KAAK,YAAY;AAC1B,QAAI,CAAC,SAAS;AACZ,gBAAU;AAAA,QACR,OAAO,EAAE;AAAA,QACT,KAAK,EAAE;AAAA,QACP,MAAM,EAAE;AAAA,QACR,GAAI,EAAE,YAAY,EAAE,SAAS,EAAE,UAAA,IAAc,CAAA;AAAA,MAAC;AAEhD;AAAA,IACF;AACA,QAAI,EAAE,aAAa,QAAQ,YAAY,EAAE,WAAW;AAClD,eAAS,KAAK,EAAE,IAAI,SAAS,QAAQ,GAAG,SAAS;AACjD,gBAAU;AAAA,QACR,OAAO,EAAE;AAAA,QACT,KAAK,EAAE;AAAA,QACP,MAAM,EAAE;AAAA,QACR,SAAS,EAAE;AAAA,MAAA;AAEb;AAAA,IACF;AACA,YAAQ,MAAM,EAAE;AAChB,YAAQ,OAAO,QAAQ,OAAO,GAAG,QAAQ,IAAI,IAAI,EAAE,IAAI,KAAK,EAAE;AAAA,EAChE;AACA,MAAI,kBAAkB,KAAK,EAAE,IAAI,SAAS,QAAQ,GAAG,SAAS;AAE9D,SAAO,EAAE,OAAO,UAAU,SAAA;AAC5B;AAKO,SAAS,wBAGd,OACA,QACwC;AACxC,SAAO,IAAI,+BAA+B,OAAO,MAAM;AACzD;AAKO,SAAS,8BAGd,OACA,QACA,QACwC;AACxC,SAAO,IAAI,+BAA+B,OAAO,EAAE,QAAQ,GAAG,QAAQ;AACxE;"}

package/dist/esm/index.d.ts

@@ -0,0 +1,7 @@
+export { elevenlabsRealtimeToken, elevenlabsRealtime } from './realtime/index.js';
+export type { ElevenLabsRealtimeTokenOptions, ElevenLabsRealtimeOptions, ElevenLabsConversationMode, ElevenLabsVADConfig, ElevenLabsClientTool, } from './realtime/index.js';
+export { ElevenLabsSpeechAdapter, createElevenLabsSpeech, elevenlabsSpeech, type ElevenLabsSpeechProviderOptions, type ElevenLabsVoiceSettings, } from './adapters/speech.js';
+export { ElevenLabsAudioAdapter, createElevenLabsAudio, elevenlabsAudio, type ElevenLabsAudioProviderOptions, type ElevenLabsMusicProviderOptions, type ElevenLabsSoundEffectsProviderOptions, type ElevenLabsMusicCompositionPlan, } from './adapters/audio.js';
+export { ElevenLabsTranscriptionAdapter, createElevenLabsTranscription, elevenlabsTranscription, type ElevenLabsTranscriptionProviderOptions, } from './adapters/transcription.js';
+export { ELEVENLABS_TTS_MODELS, ELEVENLABS_AUDIO_MODELS, ELEVENLABS_TRANSCRIPTION_MODELS, isElevenLabsMusicModel, isElevenLabsSoundEffectsModel, type ElevenLabsTTSModel, type ElevenLabsAudioModel, type ElevenLabsMusicModel, type ElevenLabsSoundEffectsModel, type ElevenLabsTranscriptionModel, type ElevenLabsOutputFormat, } from './model-meta.js';
+export { getElevenLabsApiKeyFromEnv, type ElevenLabsClientConfig, } from './utils/index.js';

package/dist/esm/index.js

@@ -0,0 +1,27 @@
+import { elevenlabsRealtimeToken } from "./realtime/token.js";
+import { elevenlabsRealtime } from "./realtime/adapter.js";
+import { ElevenLabsSpeechAdapter, createElevenLabsSpeech, elevenlabsSpeech } from "./adapters/speech.js";
+import { ElevenLabsAudioAdapter, createElevenLabsAudio, elevenlabsAudio } from "./adapters/audio.js";
+import { ElevenLabsTranscriptionAdapter, createElevenLabsTranscription, elevenlabsTranscription } from "./adapters/transcription.js";
+import { ELEVENLABS_AUDIO_MODELS, ELEVENLABS_TRANSCRIPTION_MODELS, ELEVENLABS_TTS_MODELS, isElevenLabsMusicModel, isElevenLabsSoundEffectsModel } from "./model-meta.js";
+import { getElevenLabsApiKeyFromEnv } from "./utils/client.js";
+export {
+  ELEVENLABS_AUDIO_MODELS,
+  ELEVENLABS_TRANSCRIPTION_MODELS,
+  ELEVENLABS_TTS_MODELS,
+  ElevenLabsAudioAdapter,
+  ElevenLabsSpeechAdapter,
+  ElevenLabsTranscriptionAdapter,
+  createElevenLabsAudio,
+  createElevenLabsSpeech,
+  createElevenLabsTranscription,
+  elevenlabsAudio,
+  elevenlabsRealtime,
+  elevenlabsRealtimeToken,
+  elevenlabsSpeech,
+  elevenlabsTranscription,
+  getElevenLabsApiKeyFromEnv,
+  isElevenLabsMusicModel,
+  isElevenLabsSoundEffectsModel
+};
+//# sourceMappingURL=index.js.map

package/dist/esm/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sources":[],"sourcesContent":[],"names":[],"mappings":";;;;;;;"}

package/dist/esm/model-meta.d.ts

@@ -0,0 +1,42 @@
+import { ElevenLabs } from '@elevenlabs/elevenlabs-js';
+/**
+ * ElevenLabs model identifiers. The lists below are the source of truth —
+ * callers are blocked from passing unknown model IDs. Keep them in sync with
+ * the ElevenLabs SDK via the automated update pipeline.
+ */
+/**
+ * Text-to-speech models.
+ * @see https://elevenlabs.io/docs/models
+ */
+export declare const ELEVENLABS_TTS_MODELS: readonly ["eleven_v3", "eleven_multilingual_v2", "eleven_flash_v2_5", "eleven_flash_v2", "eleven_turbo_v2_5", "eleven_turbo_v2", "eleven_monolingual_v1"];
+export type ElevenLabsTTSModel = (typeof ELEVENLABS_TTS_MODELS)[number];
+/**
+ * Audio generation models — music (`music_v1`) + sound effects
+ * (`eleven_text_to_sound_v*`) share one `generateAudio` adapter.
+ * The adapter dispatches by model id so callers pick behavior via the model.
+ *
+ * @see https://elevenlabs.io/docs/overview/capabilities/music
+ * @see https://elevenlabs.io/docs/overview/capabilities/sound-effects
+ */
+export declare const ELEVENLABS_AUDIO_MODELS: readonly ["music_v1", "eleven_text_to_sound_v2", "eleven_text_to_sound_v1"];
+export type ElevenLabsAudioModel = (typeof ELEVENLABS_AUDIO_MODELS)[number];
+/** Music models within the audio family. */
+export type ElevenLabsMusicModel = 'music_v1';
+/** SFX models within the audio family. */
+export type ElevenLabsSoundEffectsModel = 'eleven_text_to_sound_v2' | 'eleven_text_to_sound_v1';
+export declare function isElevenLabsMusicModel(model: string): model is ElevenLabsMusicModel;
+export declare function isElevenLabsSoundEffectsModel(model: string): model is ElevenLabsSoundEffectsModel;
+/**
+ * Speech-to-text (transcription) models — Scribe family.
+ * @see https://elevenlabs.io/docs/overview/capabilities/speech-to-text
+ */
+export declare const ELEVENLABS_TRANSCRIPTION_MODELS: readonly ["scribe_v2", "scribe_v1"];
+export type ElevenLabsTranscriptionModel = (typeof ELEVENLABS_TRANSCRIPTION_MODELS)[number];
+/**
+ * Supported `output_format` strings, encoded as `codec_samplerate[_bitrate]`.
+ * Aliased to the SDK's `AllowedOutputFormats` so the list stays in sync
+ * automatically whenever the `@elevenlabs/elevenlabs-js` dependency is bumped.
+ *
+ * @see https://elevenlabs.io/docs/api-reference/text-to-speech/convert
+ */
+export type ElevenLabsOutputFormat = ElevenLabs.AllowedOutputFormats;

package/dist/esm/model-meta.js

@@ -0,0 +1,32 @@
+const ELEVENLABS_TTS_MODELS = [
+  "eleven_v3",
+  "eleven_multilingual_v2",
+  "eleven_flash_v2_5",
+  "eleven_flash_v2",
+  "eleven_turbo_v2_5",
+  "eleven_turbo_v2",
+  "eleven_monolingual_v1"
+];
+const ELEVENLABS_AUDIO_MODELS = [
+  "music_v1",
+  "eleven_text_to_sound_v2",
+  "eleven_text_to_sound_v1"
+];
+function isElevenLabsMusicModel(model) {
+  return model === "music_v1";
+}
+function isElevenLabsSoundEffectsModel(model) {
+  return model.startsWith("eleven_text_to_sound_");
+}
+const ELEVENLABS_TRANSCRIPTION_MODELS = [
+  "scribe_v2",
+  "scribe_v1"
+];
+export {
+  ELEVENLABS_AUDIO_MODELS,
+  ELEVENLABS_TRANSCRIPTION_MODELS,
+  ELEVENLABS_TTS_MODELS,
+  isElevenLabsMusicModel,
+  isElevenLabsSoundEffectsModel
+};
+//# sourceMappingURL=model-meta.js.map

package/dist/esm/model-meta.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"model-meta.js","sources":["../../src/model-meta.ts"],"sourcesContent":["import type { ElevenLabs } from '@elevenlabs/elevenlabs-js'\n\n/**\n * ElevenLabs model identifiers. The lists below are the source of truth —\n * callers are blocked from passing unknown model IDs. Keep them in sync with\n * the ElevenLabs SDK via the automated update pipeline.\n */\n\n/**\n * Text-to-speech models.\n * @see https://elevenlabs.io/docs/models\n */\nexport const ELEVENLABS_TTS_MODELS = [\n  'eleven_v3',\n  'eleven_multilingual_v2',\n  'eleven_flash_v2_5',\n  'eleven_flash_v2',\n  'eleven_turbo_v2_5',\n  'eleven_turbo_v2',\n  'eleven_monolingual_v1',\n] as const\n\nexport type ElevenLabsTTSModel = (typeof ELEVENLABS_TTS_MODELS)[number]\n\n/**\n * Audio generation models — music (`music_v1`) + sound effects\n * (`eleven_text_to_sound_v*`) share one `generateAudio` adapter.\n * The adapter dispatches by model id so callers pick behavior via the model.\n *\n * @see https://elevenlabs.io/docs/overview/capabilities/music\n * @see https://elevenlabs.io/docs/overview/capabilities/sound-effects\n */\nexport const ELEVENLABS_AUDIO_MODELS = [\n  'music_v1',\n  'eleven_text_to_sound_v2',\n  'eleven_text_to_sound_v1',\n] as const\n\nexport type ElevenLabsAudioModel = (typeof ELEVENLABS_AUDIO_MODELS)[number]\n\n/** Music models within the audio family. */\nexport type ElevenLabsMusicModel = 'music_v1'\n/** SFX models within the audio family. */\nexport type ElevenLabsSoundEffectsModel =\n  | 'eleven_text_to_sound_v2'\n  | 'eleven_text_to_sound_v1'\n\nexport function isElevenLabsMusicModel(\n  model: string,\n): model is ElevenLabsMusicModel {\n  return model === 'music_v1'\n}\n\nexport function isElevenLabsSoundEffectsModel(\n  model: string,\n): model is ElevenLabsSoundEffectsModel {\n  return model.startsWith('eleven_text_to_sound_')\n}\n\n/**\n * Speech-to-text (transcription) models — Scribe family.\n * @see https://elevenlabs.io/docs/overview/capabilities/speech-to-text\n */\nexport const ELEVENLABS_TRANSCRIPTION_MODELS = [\n  'scribe_v2',\n  'scribe_v1',\n] as const\n\nexport type ElevenLabsTranscriptionModel =\n  (typeof ELEVENLABS_TRANSCRIPTION_MODELS)[number]\n\n/**\n * Supported `output_format` strings, encoded as `codec_samplerate[_bitrate]`.\n * Aliased to the SDK's `AllowedOutputFormats` so the list stays in sync\n * automatically whenever the `@elevenlabs/elevenlabs-js` dependency is bumped.\n *\n * @see https://elevenlabs.io/docs/api-reference/text-to-speech/convert\n */\nexport type ElevenLabsOutputFormat = ElevenLabs.AllowedOutputFormats\n"],"names":[],"mappings":"AAYO,MAAM,wBAAwB;AAAA,EACnC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAYO,MAAM,0BAA0B;AAAA,EACrC;AAAA,EACA;AAAA,EACA;AACF;AAWO,SAAS,uBACd,OAC+B;AAC/B,SAAO,UAAU;AACnB;AAEO,SAAS,8BACd,OACsC;AACtC,SAAO,MAAM,WAAW,uBAAuB;AACjD;AAMO,MAAM,kCAAkC;AAAA,EAC7C;AAAA,EACA;AACF;"}

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

@@ -0,0 +1,22 @@
+import { RealtimeAdapter } from '@tanstack/ai';
+import { ElevenLabsRealtimeOptions } from './types.js';
+/**
+ * Creates an ElevenLabs realtime adapter for client-side use.
+ *
+ * Wraps the @elevenlabs/client SDK for voice conversations.
+ *
+ * @param options - Optional configuration
+ * @returns A RealtimeAdapter for use with RealtimeClient
+ *
+ * @example
+ * ```typescript
+ * import { RealtimeClient } from '@tanstack/ai-client'
+ * import { elevenlabsRealtime } from '@tanstack/ai-elevenlabs'
+ *
+ * const client = new RealtimeClient({
+ *   getToken: () => fetch('/api/realtime-token').then(r => r.json()),
+ *   adapter: elevenlabsRealtime(),
+ * })
+ * ```
+ */
+export declare function elevenlabsRealtime(options?: ElevenLabsRealtimeOptions): RealtimeAdapter;

package/dist/esm/realtime/adapter.js

@@ -0,0 +1,218 @@
+import { Conversation } from "@elevenlabs/client";
+import { resolveDebugOption } from "@tanstack/ai/adapter-internals";
+function elevenlabsRealtime(options = {}) {
+  const logger = resolveDebugOption(options.debug);
+  return {
+    provider: "elevenlabs",
+    async connect(token, clientToolDefs) {
+      logger.request(`activity=realtime provider=elevenlabs`, {
+        provider: "elevenlabs"
+      });
+      return createElevenLabsConnection(token, options, logger, clientToolDefs);
+    }
+  };
+}
+async function createElevenLabsConnection(token, _options, logger, clientToolDefs) {
+  const eventHandlers = /* @__PURE__ */ new Map();
+  let conversation = null;
+  let messageIdCounter = 0;
+  const emptyFrequencyData = new Uint8Array(128);
+  const emptyTimeDomainData = new Uint8Array(128).fill(128);
+  function emit(event, payload) {
+    const handlers = eventHandlers.get(event);
+    if (handlers) {
+      for (const handler of handlers) {
+        handler(payload);
+      }
+    }
+  }
+  function generateMessageId() {
+    return `el-msg-${Date.now()}-${++messageIdCounter}`;
+  }
+  const elevenLabsClientTools = {};
+  if (clientToolDefs) {
+    for (const tool of clientToolDefs) {
+      elevenLabsClientTools[tool.name] = async (params) => {
+        if (tool.execute) {
+          const result = await tool.execute(params);
+          return typeof result === "string" ? result : JSON.stringify(result);
+        }
+        return JSON.stringify({
+          error: `No execute function for tool ${tool.name}`
+        });
+      };
+    }
+  }
+  const sessionOptions = {
+    signedUrl: token.token,
+    onConnect: () => {
+      logger.provider(`provider=elevenlabs direction=in type=connect`, {
+        frame: { type: "connect" }
+      });
+      emit("status_change", { status: "connected" });
+      emit("mode_change", { mode: "listening" });
+    },
+    onDisconnect: () => {
+      logger.provider(`provider=elevenlabs direction=in type=disconnect`, {
+        frame: { type: "disconnect" }
+      });
+      emit("status_change", { status: "idle" });
+      emit("mode_change", { mode: "idle" });
+    },
+    onModeChange: ({ mode }) => {
+      logger.provider(`provider=elevenlabs direction=in type=mode_change`, {
+        frame: { type: "mode_change", mode }
+      });
+      const mappedMode = mode === "speaking" ? "speaking" : "listening";
+      emit("mode_change", { mode: mappedMode });
+    },
+    onMessage: ({ message, source }) => {
+      logger.provider(`provider=elevenlabs direction=in type=message`, {
+        frame: { type: "message", source, message }
+      });
+      const role = source === "user" ? "user" : "assistant";
+      if (role === "user") {
+        emit("transcript", {
+          role,
+          transcript: message,
+          isFinal: true
+        });
+      } else {
+        const realtimeMessage = {
+          id: generateMessageId(),
+          role,
+          timestamp: Date.now(),
+          parts: [{ type: "audio", transcript: message }]
+        };
+        emit("message_complete", { message: realtimeMessage });
+      }
+    },
+    onError: (error) => {
+      logger.errors("elevenlabs.realtime fatal", {
+        error,
+        source: "elevenlabs.realtime"
+      });
+      emit("error", {
+        error: new Error(
+          typeof error === "string" ? error : error.message || "Unknown error"
+        )
+      });
+    }
+  };
+  if (Object.keys(elevenLabsClientTools).length > 0) {
+    sessionOptions.clientTools = elevenLabsClientTools;
+  }
+  try {
+    conversation = await Conversation.startSession(
+      sessionOptions
+    );
+  } catch (error) {
+    logger.errors("elevenlabs.realtime fatal", {
+      error,
+      source: "elevenlabs.realtime"
+    });
+    throw error;
+  }
+  const connection = {
+    async disconnect() {
+      if (conversation) {
+        await conversation.endSession();
+        conversation = null;
+      }
+      emit("status_change", { status: "idle" });
+    },
+    async startAudioCapture() {
+      emit("mode_change", { mode: "listening" });
+    },
+    stopAudioCapture() {
+      emit("mode_change", { mode: "idle" });
+    },
+    sendText(text) {
+      if (!conversation) return;
+      logger.provider(`provider=elevenlabs direction=out type=user_message`, {
+        frame: { type: "user_message", text }
+      });
+      conversation.sendUserMessage(text);
+    },
+    sendImage(_imageData, _mimeType) {
+      logger.errors("elevenlabs.realtime sendImage not supported", {
+        source: "elevenlabs.realtime"
+      });
+    },
+    sendToolResult(_callId, _result) {
+    },
+    updateSession(_config) {
+      logger.errors("elevenlabs.realtime updateSession not supported", {
+        source: "elevenlabs.realtime"
+      });
+    },
+    interrupt() {
+      emit("mode_change", { mode: "listening" });
+      emit("interrupted", {});
+    },
+    on(event, handler) {
+      let handlers = eventHandlers.get(event);
+      if (!handlers) {
+        handlers = /* @__PURE__ */ new Set();
+        eventHandlers.set(event, handlers);
+      }
+      handlers.add(handler);
+      return () => {
+        eventHandlers.get(event)?.delete(handler);
+      };
+    },
+    getAudioVisualization() {
+      return {
+        get inputLevel() {
+          if (!conversation) return 0;
+          try {
+            return conversation.getInputVolume();
+          } catch {
+            return 0;
+          }
+        },
+        get outputLevel() {
+          if (!conversation) return 0;
+          try {
+            return conversation.getOutputVolume();
+          } catch {
+            return 0;
+          }
+        },
+        getInputFrequencyData() {
+          if (!conversation) return emptyFrequencyData;
+          try {
+            return conversation.getInputByteFrequencyData();
+          } catch {
+            return emptyFrequencyData;
+          }
+        },
+        getOutputFrequencyData() {
+          if (!conversation) return emptyFrequencyData;
+          try {
+            return conversation.getOutputByteFrequencyData();
+          } catch {
+            return emptyFrequencyData;
+          }
+        },
+        getInputTimeDomainData() {
+          return emptyTimeDomainData;
+        },
+        getOutputTimeDomainData() {
+          return emptyTimeDomainData;
+        },
+        get inputSampleRate() {
+          return 16e3;
+        },
+        get outputSampleRate() {
+          return 16e3;
+        }
+      };
+    }
+  };
+  return connection;
+}
+export {
+  elevenlabsRealtime
+};
+//# sourceMappingURL=adapter.js.map

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

@@ -0,0 +1 @@
+{"version":3,"file":"adapter.js","sources":["../../../src/realtime/adapter.ts"],"sourcesContent":["import { Conversation } from '@elevenlabs/client'\nimport { 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 { ElevenLabsRealtimeOptions } from './types'\n\n/**\n * Creates an ElevenLabs realtime adapter for client-side use.\n *\n * Wraps the @elevenlabs/client SDK for voice conversations.\n *\n * @param options - Optional configuration\n * @returns A RealtimeAdapter for use with RealtimeClient\n *\n * @example\n * ```typescript\n * import { RealtimeClient } from '@tanstack/ai-client'\n * import { elevenlabsRealtime } from '@tanstack/ai-elevenlabs'\n *\n * const client = new RealtimeClient({\n *   getToken: () => fetch('/api/realtime-token').then(r => r.json()),\n *   adapter: elevenlabsRealtime(),\n * })\n * ```\n */\nexport function elevenlabsRealtime(\n  options: ElevenLabsRealtimeOptions = {},\n): RealtimeAdapter {\n  const logger = resolveDebugOption(options.debug)\n\n  return {\n    provider: 'elevenlabs',\n\n    async connect(\n      token: RealtimeToken,\n      clientToolDefs?: ReadonlyArray<AnyClientTool>,\n    ): Promise<RealtimeConnection> {\n      logger.request(`activity=realtime provider=elevenlabs`, {\n        provider: 'elevenlabs',\n      })\n      return createElevenLabsConnection(token, options, logger, clientToolDefs)\n    },\n  }\n}\n\n/**\n * Creates a connection to ElevenLabs conversational AI\n */\nasync function createElevenLabsConnection(\n  token: RealtimeToken,\n  _options: ElevenLabsRealtimeOptions,\n  logger: InternalLogger,\n  clientToolDefs?: ReadonlyArray<AnyClientTool>,\n): Promise<RealtimeConnection> {\n  const eventHandlers = new Map<RealtimeEvent, Set<RealtimeEventHandler<any>>>()\n  let conversation: Awaited<\n    ReturnType<typeof Conversation.startSession>\n  > | null = null\n  let messageIdCounter = 0\n\n  // Empty arrays for when visualization isn't available\n  const emptyFrequencyData = new Uint8Array(128)\n  const emptyTimeDomainData = new Uint8Array(128).fill(128)\n\n  // Helper to emit events\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  function generateMessageId(): string {\n    return `el-msg-${Date.now()}-${++messageIdCounter}`\n  }\n\n  // Convert TanStack tool definitions to ElevenLabs clientTools format.\n  // @elevenlabs/client expects plain async functions, not objects.\n  const elevenLabsClientTools: Record<\n    string,\n    (params: unknown) => Promise<string>\n  > = {}\n\n  if (clientToolDefs) {\n    for (const tool of clientToolDefs) {\n      elevenLabsClientTools[tool.name] = async (params: unknown) => {\n        if (tool.execute) {\n          const result = await tool.execute(params)\n          return typeof result === 'string' ? result : JSON.stringify(result)\n        }\n        return JSON.stringify({\n          error: `No execute function for tool ${tool.name}`,\n        })\n      }\n    }\n  }\n\n  // Build session options\n  const sessionOptions: Record<string, unknown> = {\n    signedUrl: token.token,\n\n    onConnect: () => {\n      logger.provider(`provider=elevenlabs direction=in type=connect`, {\n        frame: { type: 'connect' },\n      })\n      emit('status_change', { status: 'connected' })\n      emit('mode_change', { mode: 'listening' })\n    },\n\n    onDisconnect: () => {\n      logger.provider(`provider=elevenlabs direction=in type=disconnect`, {\n        frame: { type: 'disconnect' },\n      })\n      emit('status_change', { status: 'idle' })\n      emit('mode_change', { mode: 'idle' })\n    },\n\n    onModeChange: ({ mode }: { mode: string }) => {\n      logger.provider(`provider=elevenlabs direction=in type=mode_change`, {\n        frame: { type: 'mode_change', mode },\n      })\n      const mappedMode: RealtimeMode =\n        mode === 'speaking' ? 'speaking' : 'listening'\n      emit('mode_change', { mode: mappedMode })\n    },\n\n    onMessage: ({ message, source }: { message: string; source: string }) => {\n      logger.provider(`provider=elevenlabs direction=in type=message`, {\n        frame: { type: 'message', source, message },\n      })\n      const role = source === 'user' ? 'user' : 'assistant'\n\n      if (role === 'user') {\n        // User transcripts: only emit transcript — the RealtimeClient\n        // creates the user message when it receives a final transcript\n        emit('transcript', {\n          role,\n          transcript: message,\n          isFinal: true,\n        })\n      } else {\n        // Assistant messages: only emit message_complete — the canonical\n        // event for finalized assistant messages\n        const realtimeMessage: RealtimeMessage = {\n          id: generateMessageId(),\n          role,\n          timestamp: Date.now(),\n          parts: [{ type: 'audio', transcript: message }],\n        }\n        emit('message_complete', { message: realtimeMessage })\n      }\n    },\n\n    onError: (error: string | Error) => {\n      logger.errors('elevenlabs.realtime fatal', {\n        error,\n        source: 'elevenlabs.realtime',\n      })\n      emit('error', {\n        error: new Error(\n          typeof error === 'string' ? error : error.message || 'Unknown error',\n        ),\n      })\n    },\n  }\n\n  // Only add clientTools if we have any\n  if (Object.keys(elevenLabsClientTools).length > 0) {\n    sessionOptions.clientTools = elevenLabsClientTools\n  }\n\n  // Start the conversation session\n  try {\n    conversation = await Conversation.startSession(\n      sessionOptions as Parameters<typeof Conversation.startSession>[0],\n    )\n  } catch (error) {\n    logger.errors('elevenlabs.realtime fatal', {\n      error,\n      source: 'elevenlabs.realtime',\n    })\n    throw error\n  }\n\n  // Connection implementation\n  const connection: RealtimeConnection = {\n    async disconnect() {\n      if (conversation) {\n        await conversation.endSession()\n        conversation = null\n      }\n      emit('status_change', { status: 'idle' })\n    },\n\n    async startAudioCapture() {\n      // ElevenLabs SDK handles audio capture automatically\n      // This is called when the session starts\n      emit('mode_change', { mode: 'listening' })\n    },\n\n    stopAudioCapture() {\n      // ElevenLabs SDK handles this\n      emit('mode_change', { mode: 'idle' })\n    },\n\n    sendText(text: string) {\n      if (!conversation) return\n      logger.provider(`provider=elevenlabs direction=out type=user_message`, {\n        frame: { type: 'user_message', text },\n      })\n      conversation.sendUserMessage(text)\n    },\n\n    sendImage(_imageData: string, _mimeType: string) {\n      // ElevenLabs does not support direct image input in the conversation API\n      logger.errors('elevenlabs.realtime sendImage not supported', {\n        source: 'elevenlabs.realtime',\n      })\n    },\n\n    sendToolResult(_callId: string, _result: string) {\n      // ElevenLabs client tools are handled via the clientTools handlers\n      // registered at session start — results are returned automatically\n    },\n\n    updateSession(_config: Partial<RealtimeSessionConfig>) {\n      // ElevenLabs session config is set at creation time\n      logger.errors('elevenlabs.realtime updateSession not supported', {\n        source: 'elevenlabs.realtime',\n      })\n    },\n\n    interrupt() {\n      // ElevenLabs handles interruption automatically via barge-in\n      // No explicit API to call\n      emit('mode_change', { mode: 'listening' })\n      emit('interrupted', {})\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      return {\n        get inputLevel() {\n          if (!conversation) return 0\n          try {\n            return conversation.getInputVolume()\n          } catch {\n            return 0\n          }\n        },\n\n        get outputLevel() {\n          if (!conversation) return 0\n          try {\n            return conversation.getOutputVolume()\n          } catch {\n            return 0\n          }\n        },\n\n        getInputFrequencyData() {\n          if (!conversation) return emptyFrequencyData\n          try {\n            return conversation.getInputByteFrequencyData()\n          } catch {\n            return emptyFrequencyData\n          }\n        },\n\n        getOutputFrequencyData() {\n          if (!conversation) return emptyFrequencyData\n          try {\n            return conversation.getOutputByteFrequencyData()\n          } catch {\n            return emptyFrequencyData\n          }\n        },\n\n        getInputTimeDomainData() {\n          // ElevenLabs SDK doesn't expose time domain data\n          return emptyTimeDomainData\n        },\n\n        getOutputTimeDomainData() {\n          // ElevenLabs SDK doesn't expose time domain data\n          return emptyTimeDomainData\n        },\n\n        get inputSampleRate() {\n          return 16000\n        },\n\n        get outputSampleRate() {\n          return 16000\n        },\n      }\n    },\n  }\n\n  return connection\n}\n"],"names":[],"mappings":";;AAoCO,SAAS,mBACd,UAAqC,IACpB;AACjB,QAAM,SAAS,mBAAmB,QAAQ,KAAK;AAE/C,SAAO;AAAA,IACL,UAAU;AAAA,IAEV,MAAM,QACJ,OACA,gBAC6B;AAC7B,aAAO,QAAQ,yCAAyC;AAAA,QACtD,UAAU;AAAA,MAAA,CACX;AACD,aAAO,2BAA2B,OAAO,SAAS,QAAQ,cAAc;AAAA,IAC1E;AAAA,EAAA;AAEJ;AAKA,eAAe,2BACb,OACA,UACA,QACA,gBAC6B;AAC7B,QAAM,oCAAoB,IAAA;AAC1B,MAAI,eAEO;AACX,MAAI,mBAAmB;AAGvB,QAAM,qBAAqB,IAAI,WAAW,GAAG;AAC7C,QAAM,sBAAsB,IAAI,WAAW,GAAG,EAAE,KAAK,GAAG;AAGxD,WAAS,KACP,OACA,SACA;AACA,UAAM,WAAW,cAAc,IAAI,KAAK;AACxC,QAAI,UAAU;AACZ,iBAAW,WAAW,UAAU;AAC9B,gBAAQ,OAAO;AAAA,MACjB;AAAA,IACF;AAAA,EACF;AAEA,WAAS,oBAA4B;AACnC,WAAO,UAAU,KAAK,IAAA,CAAK,IAAI,EAAE,gBAAgB;AAAA,EACnD;AAIA,QAAM,wBAGF,CAAA;AAEJ,MAAI,gBAAgB;AAClB,eAAW,QAAQ,gBAAgB;AACjC,4BAAsB,KAAK,IAAI,IAAI,OAAO,WAAoB;AAC5D,YAAI,KAAK,SAAS;AAChB,gBAAM,SAAS,MAAM,KAAK,QAAQ,MAAM;AACxC,iBAAO,OAAO,WAAW,WAAW,SAAS,KAAK,UAAU,MAAM;AAAA,QACpE;AACA,eAAO,KAAK,UAAU;AAAA,UACpB,OAAO,gCAAgC,KAAK,IAAI;AAAA,QAAA,CACjD;AAAA,MACH;AAAA,IACF;AAAA,EACF;AAGA,QAAM,iBAA0C;AAAA,IAC9C,WAAW,MAAM;AAAA,IAEjB,WAAW,MAAM;AACf,aAAO,SAAS,iDAAiD;AAAA,QAC/D,OAAO,EAAE,MAAM,UAAA;AAAA,MAAU,CAC1B;AACD,WAAK,iBAAiB,EAAE,QAAQ,YAAA,CAAa;AAC7C,WAAK,eAAe,EAAE,MAAM,YAAA,CAAa;AAAA,IAC3C;AAAA,IAEA,cAAc,MAAM;AAClB,aAAO,SAAS,oDAAoD;AAAA,QAClE,OAAO,EAAE,MAAM,aAAA;AAAA,MAAa,CAC7B;AACD,WAAK,iBAAiB,EAAE,QAAQ,OAAA,CAAQ;AACxC,WAAK,eAAe,EAAE,MAAM,OAAA,CAAQ;AAAA,IACtC;AAAA,IAEA,cAAc,CAAC,EAAE,WAA6B;AAC5C,aAAO,SAAS,qDAAqD;AAAA,QACnE,OAAO,EAAE,MAAM,eAAe,KAAA;AAAA,MAAK,CACpC;AACD,YAAM,aACJ,SAAS,aAAa,aAAa;AACrC,WAAK,eAAe,EAAE,MAAM,WAAA,CAAY;AAAA,IAC1C;AAAA,IAEA,WAAW,CAAC,EAAE,SAAS,aAAkD;AACvE,aAAO,SAAS,iDAAiD;AAAA,QAC/D,OAAO,EAAE,MAAM,WAAW,QAAQ,QAAA;AAAA,MAAQ,CAC3C;AACD,YAAM,OAAO,WAAW,SAAS,SAAS;AAE1C,UAAI,SAAS,QAAQ;AAGnB,aAAK,cAAc;AAAA,UACjB;AAAA,UACA,YAAY;AAAA,UACZ,SAAS;AAAA,QAAA,CACV;AAAA,MACH,OAAO;AAGL,cAAM,kBAAmC;AAAA,UACvC,IAAI,kBAAA;AAAA,UACJ;AAAA,UACA,WAAW,KAAK,IAAA;AAAA,UAChB,OAAO,CAAC,EAAE,MAAM,SAAS,YAAY,SAAS;AAAA,QAAA;AAEhD,aAAK,oBAAoB,EAAE,SAAS,gBAAA,CAAiB;AAAA,MACvD;AAAA,IACF;AAAA,IAEA,SAAS,CAAC,UAA0B;AAClC,aAAO,OAAO,6BAA6B;AAAA,QACzC;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,WAAK,SAAS;AAAA,QACZ,OAAO,IAAI;AAAA,UACT,OAAO,UAAU,WAAW,QAAQ,MAAM,WAAW;AAAA,QAAA;AAAA,MACvD,CACD;AAAA,IACH;AAAA,EAAA;AAIF,MAAI,OAAO,KAAK,qBAAqB,EAAE,SAAS,GAAG;AACjD,mBAAe,cAAc;AAAA,EAC/B;AAGA,MAAI;AACF,mBAAe,MAAM,aAAa;AAAA,MAChC;AAAA,IAAA;AAAA,EAEJ,SAAS,OAAO;AACd,WAAO,OAAO,6BAA6B;AAAA,MACzC;AAAA,MACA,QAAQ;AAAA,IAAA,CACT;AACD,UAAM;AAAA,EACR;AAGA,QAAM,aAAiC;AAAA,IACrC,MAAM,aAAa;AACjB,UAAI,cAAc;AAChB,cAAM,aAAa,WAAA;AACnB,uBAAe;AAAA,MACjB;AACA,WAAK,iBAAiB,EAAE,QAAQ,OAAA,CAAQ;AAAA,IAC1C;AAAA,IAEA,MAAM,oBAAoB;AAGxB,WAAK,eAAe,EAAE,MAAM,YAAA,CAAa;AAAA,IAC3C;AAAA,IAEA,mBAAmB;AAEjB,WAAK,eAAe,EAAE,MAAM,OAAA,CAAQ;AAAA,IACtC;AAAA,IAEA,SAAS,MAAc;AACrB,UAAI,CAAC,aAAc;AACnB,aAAO,SAAS,uDAAuD;AAAA,QACrE,OAAO,EAAE,MAAM,gBAAgB,KAAA;AAAA,MAAK,CACrC;AACD,mBAAa,gBAAgB,IAAI;AAAA,IACnC;AAAA,IAEA,UAAU,YAAoB,WAAmB;AAE/C,aAAO,OAAO,+CAA+C;AAAA,QAC3D,QAAQ;AAAA,MAAA,CACT;AAAA,IACH;AAAA,IAEA,eAAe,SAAiB,SAAiB;AAAA,IAGjD;AAAA,IAEA,cAAc,SAAyC;AAErD,aAAO,OAAO,mDAAmD;AAAA,QAC/D,QAAQ;AAAA,MAAA,CACT;AAAA,IACH;AAAA,IAEA,YAAY;AAGV,WAAK,eAAe,EAAE,MAAM,YAAA,CAAa;AACzC,WAAK,eAAe,EAAE;AAAA,IACxB;AAAA,IAEA,GACE,OACA,SACY;AACZ,UAAI,WAAW,cAAc,IAAI,KAAK;AACtC,UAAI,CAAC,UAAU;AACb,uCAAe,IAAA;AACf,sBAAc,IAAI,OAAO,QAAQ;AAAA,MACnC;AACA,eAAS,IAAI,OAAO;AAEpB,aAAO,MAAM;AACX,sBAAc,IAAI,KAAK,GAAG,OAAO,OAAO;AAAA,MAC1C;AAAA,IACF;AAAA,IAEA,wBAA4C;AAC1C,aAAO;AAAA,QACL,IAAI,aAAa;AACf,cAAI,CAAC,aAAc,QAAO;AAC1B,cAAI;AACF,mBAAO,aAAa,eAAA;AAAA,UACtB,QAAQ;AACN,mBAAO;AAAA,UACT;AAAA,QACF;AAAA,QAEA,IAAI,cAAc;AAChB,cAAI,CAAC,aAAc,QAAO;AAC1B,cAAI;AACF,mBAAO,aAAa,gBAAA;AAAA,UACtB,QAAQ;AACN,mBAAO;AAAA,UACT;AAAA,QACF;AAAA,QAEA,wBAAwB;AACtB,cAAI,CAAC,aAAc,QAAO;AAC1B,cAAI;AACF,mBAAO,aAAa,0BAAA;AAAA,UACtB,QAAQ;AACN,mBAAO;AAAA,UACT;AAAA,QACF;AAAA,QAEA,yBAAyB;AACvB,cAAI,CAAC,aAAc,QAAO;AAC1B,cAAI;AACF,mBAAO,aAAa,2BAAA;AAAA,UACtB,QAAQ;AACN,mBAAO;AAAA,UACT;AAAA,QACF;AAAA,QAEA,yBAAyB;AAEvB,iBAAO;AAAA,QACT;AAAA,QAEA,0BAA0B;AAExB,iBAAO;AAAA,QACT;AAAA,QAEA,IAAI,kBAAkB;AACpB,iBAAO;AAAA,QACT;AAAA,QAEA,IAAI,mBAAmB;AACrB,iBAAO;AAAA,QACT;AAAA,MAAA;AAAA,IAEJ;AAAA,EAAA;AAGF,SAAO;AACT;"}

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

@@ -0,0 +1,3 @@
+export { elevenlabsRealtimeToken } from './token.js';
+export { elevenlabsRealtime } from './adapter.js';
+export type { ElevenLabsRealtimeTokenOptions, ElevenLabsRealtimeOptions, ElevenLabsConversationMode, ElevenLabsVADConfig, ElevenLabsClientTool, } from './types.js';

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

@@ -0,0 +1,28 @@
+import { RealtimeTokenAdapter } from '@tanstack/ai';
+import { ElevenLabsRealtimeTokenOptions } from './types.js';
+/**
+ * Creates an ElevenLabs realtime token adapter.
+ *
+ * Uses the official `@elevenlabs/elevenlabs-js` SDK to request a signed URL
+ * for client-side conversation connections. The signed URL is valid for
+ * 30 minutes.
+ *
+ * @param options - Configuration. `agentId` falls back to
+ *   `ELEVENLABS_AGENT_ID` in the environment when omitted.
+ * @returns A RealtimeTokenAdapter for use with realtimeToken()
+ *
+ * @example
+ * ```typescript
+ * import { realtimeToken } from '@tanstack/ai'
+ * import { elevenlabsRealtimeToken } from '@tanstack/ai-elevenlabs'
+ *
+ * // Reads ELEVENLABS_AGENT_ID from env:
+ * const token = await realtimeToken({ adapter: elevenlabsRealtimeToken() })
+ *
+ * // Or pass explicitly:
+ * const token = await realtimeToken({
+ *   adapter: elevenlabsRealtimeToken({ agentId: 'your-agent-id' }),
+ * })
+ * ```
+ */
+export declare function elevenlabsRealtimeToken(options?: ElevenLabsRealtimeTokenOptions): RealtimeTokenAdapter;

package/dist/esm/realtime/token.js

@@ -0,0 +1,39 @@
+import { createElevenLabsClient, getElevenLabsAgentIdFromEnv } from "../utils/client.js";
+function elevenlabsRealtimeToken(options = {}) {
+  const client = createElevenLabsClient();
+  return {
+    provider: "elevenlabs",
+    async generateToken() {
+      const { overrides } = options;
+      const agentId = options.agentId ?? getElevenLabsAgentIdFromEnv();
+      const response = await client.conversationalAi.conversations.getSignedUrl(
+        { agentId }
+      );
+      const expiresAt = Date.now() + 30 * 60 * 1e3;
+      return {
+        provider: "elevenlabs",
+        token: response.signedUrl,
+        expiresAt,
+        config: {
+          ...overrides?.voiceId !== void 0 && { voice: overrides.voiceId },
+          ...overrides?.systemPrompt !== void 0 && {
+            instructions: overrides.systemPrompt
+          },
+          providerOptions: {
+            agentId,
+            ...overrides?.firstMessage !== void 0 && {
+              firstMessage: overrides.firstMessage
+            },
+            ...overrides?.language !== void 0 && {
+              language: overrides.language
+            }
+          }
+        }
+      };
+    }
+  };
+}
+export {
+  elevenlabsRealtimeToken
+};
+//# sourceMappingURL=token.js.map