@tanstack/ai-elevenlabs 0.2.7 → 0.2.9

package/dist/esm/adapters/audio.d.ts

@@ -74,7 +74,7 @@ export type ElevenLabsAudioProviderOptions = (ElevenLabsMusicProviderOptions & E
  */
 export declare class ElevenLabsAudioAdapter<TModel extends ElevenLabsAudioModel> extends BaseAudioAdapter<TModel, ElevenLabsAudioProviderOptions> {
     readonly name: "elevenlabs";
-    private client;
+    private readonly client;
     constructor(model: TModel, config?: ElevenLabsClientConfig);
     generateAudio(options: AudioGenerationOptions<ElevenLabsAudioProviderOptions>): Promise<AudioGenerationResult>;
     private runMusic;

package/dist/esm/adapters/audio.js

@@ -2,9 +2,10 @@ import { BaseAudioAdapter } from "@tanstack/ai/adapters";
 import { createElevenLabsClient, readStreamToArrayBuffer, arrayBufferToBase64, parseOutputFormat, generateId } from "../utils/client.js";
 import { isElevenLabsMusicModel, isElevenLabsSoundEffectsModel } from "../model-meta.js";
 class ElevenLabsAudioAdapter extends BaseAudioAdapter {
+  name = "elevenlabs";
+  client;
   constructor(model, config) {
     super(model, config ?? {});
-    this.name = "elevenlabs";
     this.client = createElevenLabsClient(config);
   }
   async generateAudio(options) {

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

@@ -1 +1 @@
-{"version":3,"file":"audio.js","sources":["../../../src/adapters/audio.ts"],"sourcesContent":["import { BaseAudioAdapter } from '@tanstack/ai/adapters'\nimport {\n  arrayBufferToBase64,\n  createElevenLabsClient,\n  generateId,\n  parseOutputFormat,\n  readStreamToArrayBuffer,\n} from '../utils/client'\nimport {\n  isElevenLabsMusicModel,\n  isElevenLabsSoundEffectsModel,\n} from '../model-meta'\nimport type { ElevenLabsClient } from '@elevenlabs/elevenlabs-js'\nimport type {\n  AudioGenerationOptions,\n  AudioGenerationResult,\n} from '@tanstack/ai'\nimport type { ElevenLabsClientConfig } from '../utils/client'\nimport type {\n  ElevenLabsAudioModel,\n  ElevenLabsMusicModel,\n  ElevenLabsOutputFormat,\n  ElevenLabsSoundEffectsModel,\n} from '../model-meta'\n\n/**\n * Structured composition plan for ElevenLabs music generation. Mutually\n * exclusive with a free-form `prompt` on the `generateAudio()` call — when\n * supplied, `prompt` is ignored by ElevenLabs.\n *\n * We mirror the SDK's camelCase naming. Lengths are in milliseconds.\n * @see https://elevenlabs.io/docs/api-reference/music/compose\n */\nexport interface ElevenLabsMusicCompositionPlan {\n  /** Positive global style descriptors (mood, instruments, tempo, …). */\n  positiveGlobalStyles?: Array<string>\n  /** Negative global style descriptors — styles to avoid. */\n  negativeGlobalStyles?: Array<string>\n  /** Section definitions (verse/chorus/bridge/…) with local style hints. */\n  sections?: Array<{\n    sectionName: string\n    positiveLocalStyles?: Array<string>\n    negativeLocalStyles?: Array<string>\n    durationMs?: number\n    lines?: Array<string>\n  }>\n}\n\n/**\n * Provider options common to all ElevenLabs audio endpoints.\n */\ninterface CommonAudioOptions {\n  /** Output audio format. Defaults to `mp3_44100_128`. */\n  outputFormat?: ElevenLabsOutputFormat\n}\n\n/**\n * Provider options for music generation (`music_v1`).\n */\nexport interface ElevenLabsMusicProviderOptions extends CommonAudioOptions {\n  /** Structured composition plan. Mutually exclusive with `prompt`/`duration`. */\n  compositionPlan?: ElevenLabsMusicCompositionPlan\n  /** Deterministic sampling seed (incompatible with `prompt`). */\n  seed?: number\n  /** Force the output to be purely instrumental (prompt-mode only). */\n  forceInstrumental?: boolean\n  /** Strictly respect section durations in `compositionPlan`. */\n  respectSectionsDurations?: boolean\n}\n\n/**\n * Provider options for sound-effect generation (`eleven_text_to_sound_v*`).\n */\nexport interface ElevenLabsSoundEffectsProviderOptions extends CommonAudioOptions {\n  /** Prompt influence, 0..1. Default 0.3. Higher = more prompt adherence. */\n  promptInfluence?: number\n  /** Generate a loopable SFX (v2 only). */\n  loop?: boolean\n}\n\n/**\n * Union of per-model provider options. We keep both branches on one type so\n * the adapter stays tree-shakeable; callers narrow by model at the factory.\n */\nexport type ElevenLabsAudioProviderOptions =\n  | (ElevenLabsMusicProviderOptions & ElevenLabsSoundEffectsProviderOptions)\n  | ElevenLabsMusicProviderOptions\n  | ElevenLabsSoundEffectsProviderOptions\n\n/**\n * ElevenLabs audio generation adapter. Dispatches to music or SFX endpoints\n * based on the model id. Music → `client.music.compose`, SFX →\n * `client.textToSoundEffects.convert`.\n *\n * @example\n * ```ts\n * const music = elevenlabsAudio('music_v1')\n * await generateAudio({ adapter: music, prompt: 'lo-fi beat', duration: 15 })\n *\n * const sfx = elevenlabsAudio('eleven_text_to_sound_v2')\n * await generateAudio({ adapter: sfx, prompt: 'glass shattering', duration: 3 })\n * ```\n */\nexport class ElevenLabsAudioAdapter<\n  TModel extends ElevenLabsAudioModel,\n> extends BaseAudioAdapter<TModel, ElevenLabsAudioProviderOptions> {\n  readonly name = 'elevenlabs' as const\n\n  private client: ElevenLabsClient\n\n  constructor(model: TModel, config?: ElevenLabsClientConfig) {\n    super(model, config ?? {})\n    this.client = createElevenLabsClient(config)\n  }\n\n  async generateAudio(\n    options: AudioGenerationOptions<ElevenLabsAudioProviderOptions>,\n  ): Promise<AudioGenerationResult> {\n    const { logger } = options\n    logger.request(\n      `activity=generateAudio provider=elevenlabs model=${this.model}`,\n      { provider: 'elevenlabs', model: this.model },\n    )\n    try {\n      if (isElevenLabsMusicModel(this.model)) {\n        return await this.runMusic(options)\n      }\n      if (isElevenLabsSoundEffectsModel(this.model)) {\n        return await this.runSoundEffects(options)\n      }\n      throw new Error(\n        `Unsupported ElevenLabs audio model \"${this.model}\". Expected one of: music_v1, eleven_text_to_sound_v2, eleven_text_to_sound_v1.`,\n      )\n    } catch (error) {\n      logger.errors('elevenlabs.generateAudio fatal', {\n        error,\n        source: 'elevenlabs.generateAudio',\n      })\n      throw error\n    }\n  }\n\n  private async runMusic(\n    options: AudioGenerationOptions<ElevenLabsAudioProviderOptions>,\n  ): Promise<AudioGenerationResult> {\n    // Gated by isElevenLabsMusicModel() in generateAudio().\n    const modelId = this.model as ElevenLabsMusicModel\n    const music = (options.modelOptions ?? {}) as ElevenLabsMusicProviderOptions\n    const outputFormat = music.outputFormat\n\n    const stream = await this.client.music.compose({\n      modelId,\n      ...(options.prompt && !music.compositionPlan\n        ? { prompt: options.prompt }\n        : {}),\n      ...(music.compositionPlan\n        ? { compositionPlan: toMusicPrompt(music.compositionPlan) }\n        : {}),\n      ...(options.duration != null && !music.compositionPlan\n        ? { musicLengthMs: Math.round(options.duration * 1000) }\n        : {}),\n      ...(outputFormat ? { outputFormat } : {}),\n      ...(music.seed != null ? { seed: music.seed } : {}),\n      ...(music.forceInstrumental != null\n        ? { forceInstrumental: music.forceInstrumental }\n        : {}),\n      ...(music.respectSectionsDurations != null\n        ? { respectSectionsDurations: music.respectSectionsDurations }\n        : {}),\n    })\n\n    return this.finalize(stream, outputFormat, options.duration)\n  }\n\n  private async runSoundEffects(\n    options: AudioGenerationOptions<ElevenLabsAudioProviderOptions>,\n  ): Promise<AudioGenerationResult> {\n    // Gated by isElevenLabsSoundEffectsModel() in generateAudio().\n    const modelId = this.model as ElevenLabsSoundEffectsModel\n    const sfx = (options.modelOptions ??\n      {}) as ElevenLabsSoundEffectsProviderOptions\n    const outputFormat = sfx.outputFormat\n\n    const stream = await this.client.textToSoundEffects.convert({\n      text: options.prompt,\n      modelId,\n      ...(options.duration != null\n        ? { durationSeconds: options.duration }\n        : {}),\n      ...(outputFormat ? { outputFormat } : {}),\n      ...(sfx.promptInfluence != null\n        ? { promptInfluence: sfx.promptInfluence }\n        : {}),\n      ...(sfx.loop != null ? { loop: sfx.loop } : {}),\n    })\n\n    return this.finalize(stream, outputFormat, options.duration)\n  }\n\n  private async finalize(\n    stream: ReadableStream<Uint8Array>,\n    outputFormat: ElevenLabsOutputFormat | undefined,\n    duration: number | undefined,\n  ): Promise<AudioGenerationResult> {\n    const buffer = await readStreamToArrayBuffer(stream)\n    const base64 = arrayBufferToBase64(buffer)\n    const { contentType } = parseOutputFormat(outputFormat)\n    return {\n      id: generateId(this.name),\n      model: this.model,\n      audio: {\n        b64Json: base64,\n        contentType,\n        ...(duration != null ? { duration } : {}),\n      },\n    }\n  }\n\n  protected override generateId(): string {\n    return generateId(this.name)\n  }\n}\n\nfunction toMusicPrompt(plan: ElevenLabsMusicCompositionPlan) {\n  return {\n    positiveGlobalStyles: plan.positiveGlobalStyles ?? [],\n    negativeGlobalStyles: plan.negativeGlobalStyles ?? [],\n    sections: (plan.sections ?? []).map((section) => ({\n      sectionName: section.sectionName,\n      positiveLocalStyles: section.positiveLocalStyles ?? [],\n      negativeLocalStyles: section.negativeLocalStyles ?? [],\n      durationMs: section.durationMs ?? 10000,\n      lines: section.lines ?? [],\n    })),\n  }\n}\n\n/**\n * Create an ElevenLabs audio adapter using `ELEVENLABS_API_KEY` from env.\n */\nexport function elevenlabsAudio<TModel extends ElevenLabsAudioModel>(\n  model: TModel,\n  config?: ElevenLabsClientConfig,\n): ElevenLabsAudioAdapter<TModel> {\n  return new ElevenLabsAudioAdapter(model, config)\n}\n\n/**\n * Create an ElevenLabs audio adapter with an explicit API key.\n */\nexport function createElevenLabsAudio<TModel extends ElevenLabsAudioModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<ElevenLabsClientConfig, 'apiKey'>,\n): ElevenLabsAudioAdapter<TModel> {\n  return new ElevenLabsAudioAdapter(model, { apiKey, ...config })\n}\n"],"names":[],"mappings":";;;AAuGO,MAAM,+BAEH,iBAAyD;AAAA,EAKjE,YAAY,OAAe,QAAiC;AAC1D,UAAM,OAAO,UAAU,EAAE;AAL3B,SAAS,OAAO;AAMd,SAAK,SAAS,uBAAuB,MAAM;AAAA,EAC7C;AAAA,EAEA,MAAM,cACJ,SACgC;AAChC,UAAM,EAAE,WAAW;AACnB,WAAO;AAAA,MACL,oDAAoD,KAAK,KAAK;AAAA,MAC9D,EAAE,UAAU,cAAc,OAAO,KAAK,MAAA;AAAA,IAAM;AAE9C,QAAI;AACF,UAAI,uBAAuB,KAAK,KAAK,GAAG;AACtC,eAAO,MAAM,KAAK,SAAS,OAAO;AAAA,MACpC;AACA,UAAI,8BAA8B,KAAK,KAAK,GAAG;AAC7C,eAAO,MAAM,KAAK,gBAAgB,OAAO;AAAA,MAC3C;AACA,YAAM,IAAI;AAAA,QACR,uCAAuC,KAAK,KAAK;AAAA,MAAA;AAAA,IAErD,SAAS,OAAO;AACd,aAAO,OAAO,kCAAkC;AAAA,QAC9C;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEA,MAAc,SACZ,SACgC;AAEhC,UAAM,UAAU,KAAK;AACrB,UAAM,QAAS,QAAQ,gBAAgB,CAAA;AACvC,UAAM,eAAe,MAAM;AAE3B,UAAM,SAAS,MAAM,KAAK,OAAO,MAAM,QAAQ;AAAA,MAC7C;AAAA,MACA,GAAI,QAAQ,UAAU,CAAC,MAAM,kBACzB,EAAE,QAAQ,QAAQ,OAAA,IAClB,CAAA;AAAA,MACJ,GAAI,MAAM,kBACN,EAAE,iBAAiB,cAAc,MAAM,eAAe,EAAA,IACtD,CAAA;AAAA,MACJ,GAAI,QAAQ,YAAY,QAAQ,CAAC,MAAM,kBACnC,EAAE,eAAe,KAAK,MAAM,QAAQ,WAAW,GAAI,EAAA,IACnD,CAAA;AAAA,MACJ,GAAI,eAAe,EAAE,aAAA,IAAiB,CAAA;AAAA,MACtC,GAAI,MAAM,QAAQ,OAAO,EAAE,MAAM,MAAM,KAAA,IAAS,CAAA;AAAA,MAChD,GAAI,MAAM,qBAAqB,OAC3B,EAAE,mBAAmB,MAAM,kBAAA,IAC3B,CAAA;AAAA,MACJ,GAAI,MAAM,4BAA4B,OAClC,EAAE,0BAA0B,MAAM,6BAClC,CAAA;AAAA,IAAC,CACN;AAED,WAAO,KAAK,SAAS,QAAQ,cAAc,QAAQ,QAAQ;AAAA,EAC7D;AAAA,EAEA,MAAc,gBACZ,SACgC;AAEhC,UAAM,UAAU,KAAK;AACrB,UAAM,MAAO,QAAQ,gBACnB,CAAA;AACF,UAAM,eAAe,IAAI;AAEzB,UAAM,SAAS,MAAM,KAAK,OAAO,mBAAmB,QAAQ;AAAA,MAC1D,MAAM,QAAQ;AAAA,MACd;AAAA,MACA,GAAI,QAAQ,YAAY,OACpB,EAAE,iBAAiB,QAAQ,SAAA,IAC3B,CAAA;AAAA,MACJ,GAAI,eAAe,EAAE,aAAA,IAAiB,CAAA;AAAA,MACtC,GAAI,IAAI,mBAAmB,OACvB,EAAE,iBAAiB,IAAI,gBAAA,IACvB,CAAA;AAAA,MACJ,GAAI,IAAI,QAAQ,OAAO,EAAE,MAAM,IAAI,SAAS,CAAA;AAAA,IAAC,CAC9C;AAED,WAAO,KAAK,SAAS,QAAQ,cAAc,QAAQ,QAAQ;AAAA,EAC7D;AAAA,EAEA,MAAc,SACZ,QACA,cACA,UACgC;AAChC,UAAM,SAAS,MAAM,wBAAwB,MAAM;AACnD,UAAM,SAAS,oBAAoB,MAAM;AACzC,UAAM,EAAE,YAAA,IAAgB,kBAAkB,YAAY;AACtD,WAAO;AAAA,MACL,IAAI,WAAW,KAAK,IAAI;AAAA,MACxB,OAAO,KAAK;AAAA,MACZ,OAAO;AAAA,QACL,SAAS;AAAA,QACT;AAAA,QACA,GAAI,YAAY,OAAO,EAAE,aAAa,CAAA;AAAA,MAAC;AAAA,IACzC;AAAA,EAEJ;AAAA,EAEmB,aAAqB;AACtC,WAAO,WAAW,KAAK,IAAI;AAAA,EAC7B;AACF;AAEA,SAAS,cAAc,MAAsC;AAC3D,SAAO;AAAA,IACL,sBAAsB,KAAK,wBAAwB,CAAA;AAAA,IACnD,sBAAsB,KAAK,wBAAwB,CAAA;AAAA,IACnD,WAAW,KAAK,YAAY,CAAA,GAAI,IAAI,CAAC,aAAa;AAAA,MAChD,aAAa,QAAQ;AAAA,MACrB,qBAAqB,QAAQ,uBAAuB,CAAA;AAAA,MACpD,qBAAqB,QAAQ,uBAAuB,CAAA;AAAA,MACpD,YAAY,QAAQ,cAAc;AAAA,MAClC,OAAO,QAAQ,SAAS,CAAA;AAAA,IAAC,EACzB;AAAA,EAAA;AAEN;AAKO,SAAS,gBACd,OACA,QACgC;AAChC,SAAO,IAAI,uBAAuB,OAAO,MAAM;AACjD;AAKO,SAAS,sBACd,OACA,QACA,QACgC;AAChC,SAAO,IAAI,uBAAuB,OAAO,EAAE,QAAQ,GAAG,QAAQ;AAChE;"}
+{"version":3,"file":"audio.js","sources":["../../../src/adapters/audio.ts"],"sourcesContent":["import { BaseAudioAdapter } from '@tanstack/ai/adapters'\nimport {\n  arrayBufferToBase64,\n  createElevenLabsClient,\n  generateId,\n  parseOutputFormat,\n  readStreamToArrayBuffer,\n} from '../utils/client'\nimport {\n  isElevenLabsMusicModel,\n  isElevenLabsSoundEffectsModel,\n} from '../model-meta'\nimport type { ElevenLabsClient } from '@elevenlabs/elevenlabs-js'\nimport type {\n  AudioGenerationOptions,\n  AudioGenerationResult,\n} from '@tanstack/ai'\nimport type { ElevenLabsClientConfig } from '../utils/client'\nimport type {\n  ElevenLabsAudioModel,\n  ElevenLabsMusicModel,\n  ElevenLabsOutputFormat,\n  ElevenLabsSoundEffectsModel,\n} from '../model-meta'\n\n/**\n * Structured composition plan for ElevenLabs music generation. Mutually\n * exclusive with a free-form `prompt` on the `generateAudio()` call — when\n * supplied, `prompt` is ignored by ElevenLabs.\n *\n * We mirror the SDK's camelCase naming. Lengths are in milliseconds.\n * @see https://elevenlabs.io/docs/api-reference/music/compose\n */\nexport interface ElevenLabsMusicCompositionPlan {\n  /** Positive global style descriptors (mood, instruments, tempo, …). */\n  positiveGlobalStyles?: Array<string>\n  /** Negative global style descriptors — styles to avoid. */\n  negativeGlobalStyles?: Array<string>\n  /** Section definitions (verse/chorus/bridge/…) with local style hints. */\n  sections?: Array<{\n    sectionName: string\n    positiveLocalStyles?: Array<string>\n    negativeLocalStyles?: Array<string>\n    durationMs?: number\n    lines?: Array<string>\n  }>\n}\n\n/**\n * Provider options common to all ElevenLabs audio endpoints.\n */\ninterface CommonAudioOptions {\n  /** Output audio format. Defaults to `mp3_44100_128`. */\n  outputFormat?: ElevenLabsOutputFormat\n}\n\n/**\n * Provider options for music generation (`music_v1`).\n */\nexport interface ElevenLabsMusicProviderOptions extends CommonAudioOptions {\n  /** Structured composition plan. Mutually exclusive with `prompt`/`duration`. */\n  compositionPlan?: ElevenLabsMusicCompositionPlan\n  /** Deterministic sampling seed (incompatible with `prompt`). */\n  seed?: number\n  /** Force the output to be purely instrumental (prompt-mode only). */\n  forceInstrumental?: boolean\n  /** Strictly respect section durations in `compositionPlan`. */\n  respectSectionsDurations?: boolean\n}\n\n/**\n * Provider options for sound-effect generation (`eleven_text_to_sound_v*`).\n */\nexport interface ElevenLabsSoundEffectsProviderOptions extends CommonAudioOptions {\n  /** Prompt influence, 0..1. Default 0.3. Higher = more prompt adherence. */\n  promptInfluence?: number\n  /** Generate a loopable SFX (v2 only). */\n  loop?: boolean\n}\n\n/**\n * Union of per-model provider options. We keep both branches on one type so\n * the adapter stays tree-shakeable; callers narrow by model at the factory.\n */\nexport type ElevenLabsAudioProviderOptions =\n  | (ElevenLabsMusicProviderOptions & ElevenLabsSoundEffectsProviderOptions)\n  | ElevenLabsMusicProviderOptions\n  | ElevenLabsSoundEffectsProviderOptions\n\n/**\n * ElevenLabs audio generation adapter. Dispatches to music or SFX endpoints\n * based on the model id. Music → `client.music.compose`, SFX →\n * `client.textToSoundEffects.convert`.\n *\n * @example\n * ```ts\n * const music = elevenlabsAudio('music_v1')\n * await generateAudio({ adapter: music, prompt: 'lo-fi beat', duration: 15 })\n *\n * const sfx = elevenlabsAudio('eleven_text_to_sound_v2')\n * await generateAudio({ adapter: sfx, prompt: 'glass shattering', duration: 3 })\n * ```\n */\nexport class ElevenLabsAudioAdapter<\n  TModel extends ElevenLabsAudioModel,\n> extends BaseAudioAdapter<TModel, ElevenLabsAudioProviderOptions> {\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 generateAudio(\n    options: AudioGenerationOptions<ElevenLabsAudioProviderOptions>,\n  ): Promise<AudioGenerationResult> {\n    const { logger } = options\n    logger.request(\n      `activity=generateAudio provider=elevenlabs model=${this.model}`,\n      { provider: 'elevenlabs', model: this.model },\n    )\n    try {\n      if (isElevenLabsMusicModel(this.model)) {\n        return await this.runMusic(options)\n      }\n      if (isElevenLabsSoundEffectsModel(this.model)) {\n        return await this.runSoundEffects(options)\n      }\n      throw new Error(\n        `Unsupported ElevenLabs audio model \"${this.model}\". Expected one of: music_v1, eleven_text_to_sound_v2, eleven_text_to_sound_v1.`,\n      )\n    } catch (error) {\n      logger.errors('elevenlabs.generateAudio fatal', {\n        error,\n        source: 'elevenlabs.generateAudio',\n      })\n      throw error\n    }\n  }\n\n  private async runMusic(\n    options: AudioGenerationOptions<ElevenLabsAudioProviderOptions>,\n  ): Promise<AudioGenerationResult> {\n    // Gated by isElevenLabsMusicModel() in generateAudio().\n    const modelId = this.model as ElevenLabsMusicModel\n    const music = (options.modelOptions ?? {}) as ElevenLabsMusicProviderOptions\n    const outputFormat = music.outputFormat\n\n    const stream = await this.client.music.compose({\n      modelId,\n      ...(options.prompt && !music.compositionPlan\n        ? { prompt: options.prompt }\n        : {}),\n      ...(music.compositionPlan\n        ? { compositionPlan: toMusicPrompt(music.compositionPlan) }\n        : {}),\n      ...(options.duration != null && !music.compositionPlan\n        ? { musicLengthMs: Math.round(options.duration * 1000) }\n        : {}),\n      ...(outputFormat ? { outputFormat } : {}),\n      ...(music.seed != null ? { seed: music.seed } : {}),\n      ...(music.forceInstrumental != null\n        ? { forceInstrumental: music.forceInstrumental }\n        : {}),\n      ...(music.respectSectionsDurations != null\n        ? { respectSectionsDurations: music.respectSectionsDurations }\n        : {}),\n    })\n\n    return this.finalize(stream, outputFormat, options.duration)\n  }\n\n  private async runSoundEffects(\n    options: AudioGenerationOptions<ElevenLabsAudioProviderOptions>,\n  ): Promise<AudioGenerationResult> {\n    // Gated by isElevenLabsSoundEffectsModel() in generateAudio().\n    const modelId = this.model as ElevenLabsSoundEffectsModel\n    const sfx = (options.modelOptions ??\n      {}) as ElevenLabsSoundEffectsProviderOptions\n    const outputFormat = sfx.outputFormat\n\n    const stream = await this.client.textToSoundEffects.convert({\n      text: options.prompt,\n      modelId,\n      ...(options.duration != null\n        ? { durationSeconds: options.duration }\n        : {}),\n      ...(outputFormat ? { outputFormat } : {}),\n      ...(sfx.promptInfluence != null\n        ? { promptInfluence: sfx.promptInfluence }\n        : {}),\n      ...(sfx.loop != null ? { loop: sfx.loop } : {}),\n    })\n\n    return this.finalize(stream, outputFormat, options.duration)\n  }\n\n  private async finalize(\n    stream: ReadableStream<Uint8Array>,\n    outputFormat: ElevenLabsOutputFormat | undefined,\n    duration: number | undefined,\n  ): Promise<AudioGenerationResult> {\n    const buffer = await readStreamToArrayBuffer(stream)\n    const base64 = arrayBufferToBase64(buffer)\n    const { contentType } = parseOutputFormat(outputFormat)\n    return {\n      id: generateId(this.name),\n      model: this.model,\n      audio: {\n        b64Json: base64,\n        contentType,\n        ...(duration != null ? { duration } : {}),\n      },\n    }\n  }\n\n  protected override generateId(): string {\n    return generateId(this.name)\n  }\n}\n\nfunction toMusicPrompt(plan: ElevenLabsMusicCompositionPlan) {\n  return {\n    positiveGlobalStyles: plan.positiveGlobalStyles ?? [],\n    negativeGlobalStyles: plan.negativeGlobalStyles ?? [],\n    sections: (plan.sections ?? []).map((section) => ({\n      sectionName: section.sectionName,\n      positiveLocalStyles: section.positiveLocalStyles ?? [],\n      negativeLocalStyles: section.negativeLocalStyles ?? [],\n      durationMs: section.durationMs ?? 10000,\n      lines: section.lines ?? [],\n    })),\n  }\n}\n\n/**\n * Create an ElevenLabs audio adapter using `ELEVENLABS_API_KEY` from env.\n */\nexport function elevenlabsAudio<TModel extends ElevenLabsAudioModel>(\n  model: TModel,\n  config?: ElevenLabsClientConfig,\n): ElevenLabsAudioAdapter<TModel> {\n  return new ElevenLabsAudioAdapter(model, config)\n}\n\n/**\n * Create an ElevenLabs audio adapter with an explicit API key.\n */\nexport function createElevenLabsAudio<TModel extends ElevenLabsAudioModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<ElevenLabsClientConfig, 'apiKey'>,\n): ElevenLabsAudioAdapter<TModel> {\n  return new ElevenLabsAudioAdapter(model, { apiKey, ...config })\n}\n"],"names":[],"mappings":";;;AAuGO,MAAM,+BAEH,iBAAyD;AAAA,EACxD,OAAO;AAAA,EAEC;AAAA,EAEjB,YAAY,OAAe,QAAiC;AAC1D,UAAM,OAAO,UAAU,EAAE;AACzB,SAAK,SAAS,uBAAuB,MAAM;AAAA,EAC7C;AAAA,EAEA,MAAM,cACJ,SACgC;AAChC,UAAM,EAAE,WAAW;AACnB,WAAO;AAAA,MACL,oDAAoD,KAAK,KAAK;AAAA,MAC9D,EAAE,UAAU,cAAc,OAAO,KAAK,MAAA;AAAA,IAAM;AAE9C,QAAI;AACF,UAAI,uBAAuB,KAAK,KAAK,GAAG;AACtC,eAAO,MAAM,KAAK,SAAS,OAAO;AAAA,MACpC;AACA,UAAI,8BAA8B,KAAK,KAAK,GAAG;AAC7C,eAAO,MAAM,KAAK,gBAAgB,OAAO;AAAA,MAC3C;AACA,YAAM,IAAI;AAAA,QACR,uCAAuC,KAAK,KAAK;AAAA,MAAA;AAAA,IAErD,SAAS,OAAO;AACd,aAAO,OAAO,kCAAkC;AAAA,QAC9C;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEA,MAAc,SACZ,SACgC;AAEhC,UAAM,UAAU,KAAK;AACrB,UAAM,QAAS,QAAQ,gBAAgB,CAAA;AACvC,UAAM,eAAe,MAAM;AAE3B,UAAM,SAAS,MAAM,KAAK,OAAO,MAAM,QAAQ;AAAA,MAC7C;AAAA,MACA,GAAI,QAAQ,UAAU,CAAC,MAAM,kBACzB,EAAE,QAAQ,QAAQ,OAAA,IAClB,CAAA;AAAA,MACJ,GAAI,MAAM,kBACN,EAAE,iBAAiB,cAAc,MAAM,eAAe,EAAA,IACtD,CAAA;AAAA,MACJ,GAAI,QAAQ,YAAY,QAAQ,CAAC,MAAM,kBACnC,EAAE,eAAe,KAAK,MAAM,QAAQ,WAAW,GAAI,EAAA,IACnD,CAAA;AAAA,MACJ,GAAI,eAAe,EAAE,aAAA,IAAiB,CAAA;AAAA,MACtC,GAAI,MAAM,QAAQ,OAAO,EAAE,MAAM,MAAM,KAAA,IAAS,CAAA;AAAA,MAChD,GAAI,MAAM,qBAAqB,OAC3B,EAAE,mBAAmB,MAAM,kBAAA,IAC3B,CAAA;AAAA,MACJ,GAAI,MAAM,4BAA4B,OAClC,EAAE,0BAA0B,MAAM,6BAClC,CAAA;AAAA,IAAC,CACN;AAED,WAAO,KAAK,SAAS,QAAQ,cAAc,QAAQ,QAAQ;AAAA,EAC7D;AAAA,EAEA,MAAc,gBACZ,SACgC;AAEhC,UAAM,UAAU,KAAK;AACrB,UAAM,MAAO,QAAQ,gBACnB,CAAA;AACF,UAAM,eAAe,IAAI;AAEzB,UAAM,SAAS,MAAM,KAAK,OAAO,mBAAmB,QAAQ;AAAA,MAC1D,MAAM,QAAQ;AAAA,MACd;AAAA,MACA,GAAI,QAAQ,YAAY,OACpB,EAAE,iBAAiB,QAAQ,SAAA,IAC3B,CAAA;AAAA,MACJ,GAAI,eAAe,EAAE,aAAA,IAAiB,CAAA;AAAA,MACtC,GAAI,IAAI,mBAAmB,OACvB,EAAE,iBAAiB,IAAI,gBAAA,IACvB,CAAA;AAAA,MACJ,GAAI,IAAI,QAAQ,OAAO,EAAE,MAAM,IAAI,SAAS,CAAA;AAAA,IAAC,CAC9C;AAED,WAAO,KAAK,SAAS,QAAQ,cAAc,QAAQ,QAAQ;AAAA,EAC7D;AAAA,EAEA,MAAc,SACZ,QACA,cACA,UACgC;AAChC,UAAM,SAAS,MAAM,wBAAwB,MAAM;AACnD,UAAM,SAAS,oBAAoB,MAAM;AACzC,UAAM,EAAE,YAAA,IAAgB,kBAAkB,YAAY;AACtD,WAAO;AAAA,MACL,IAAI,WAAW,KAAK,IAAI;AAAA,MACxB,OAAO,KAAK;AAAA,MACZ,OAAO;AAAA,QACL,SAAS;AAAA,QACT;AAAA,QACA,GAAI,YAAY,OAAO,EAAE,aAAa,CAAA;AAAA,MAAC;AAAA,IACzC;AAAA,EAEJ;AAAA,EAEmB,aAAqB;AACtC,WAAO,WAAW,KAAK,IAAI;AAAA,EAC7B;AACF;AAEA,SAAS,cAAc,MAAsC;AAC3D,SAAO;AAAA,IACL,sBAAsB,KAAK,wBAAwB,CAAA;AAAA,IACnD,sBAAsB,KAAK,wBAAwB,CAAA;AAAA,IACnD,WAAW,KAAK,YAAY,CAAA,GAAI,IAAI,CAAC,aAAa;AAAA,MAChD,aAAa,QAAQ;AAAA,MACrB,qBAAqB,QAAQ,uBAAuB,CAAA;AAAA,MACpD,qBAAqB,QAAQ,uBAAuB,CAAA;AAAA,MACpD,YAAY,QAAQ,cAAc;AAAA,MAClC,OAAO,QAAQ,SAAS,CAAA;AAAA,IAAC,EACzB;AAAA,EAAA;AAEN;AAKO,SAAS,gBACd,OACA,QACgC;AAChC,SAAO,IAAI,uBAAuB,OAAO,MAAM;AACjD;AAKO,SAAS,sBACd,OACA,QACA,QACgC;AAChC,SAAO,IAAI,uBAAuB,OAAO,EAAE,QAAQ,GAAG,QAAQ;AAChE;"}

package/dist/esm/adapters/speech.d.ts

@@ -68,7 +68,7 @@ export interface ElevenLabsSpeechProviderOptions {
  */
 export declare class ElevenLabsSpeechAdapter<TModel extends ElevenLabsTTSModel> extends BaseTTSAdapter<TModel, ElevenLabsSpeechProviderOptions> {
     readonly name: "elevenlabs";
-    private client;
+    private readonly client;
     constructor(model: TModel, config?: ElevenLabsClientConfig);
     generateSpeech(options: TTSOptions<ElevenLabsSpeechProviderOptions>): Promise<TTSResult>;
     protected generateId(): string;

package/dist/esm/adapters/speech.js

@@ -1,9 +1,10 @@
 import { BaseTTSAdapter } from "@tanstack/ai/adapters";
 import { createElevenLabsClient, readStreamToArrayBuffer, arrayBufferToBase64, parseOutputFormat, generateId } from "../utils/client.js";
 class ElevenLabsSpeechAdapter extends BaseTTSAdapter {
+  name = "elevenlabs";
+  client;
   constructor(model, config) {
     super(model, config ?? {});
-    this.name = "elevenlabs";
     this.client = createElevenLabsClient(config);
   }
   async generateSpeech(options) {
@@ -91,6 +92,9 @@ function inferOutputFormatFromResponseFormat(format) {
       return "opus_48000_128";
     case void 0:
       return void 0;
+    case "aac":
+    case "flac":
+    case "wav":
     default:
       return "mp3_44100_128";
   }

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

@@ -1 +1 @@
-{"version":3,"file":"speech.js","sources":["../../../src/adapters/speech.ts"],"sourcesContent":["import { BaseTTSAdapter } from '@tanstack/ai/adapters'\nimport {\n  arrayBufferToBase64,\n  createElevenLabsClient,\n  generateId,\n  parseOutputFormat,\n  readStreamToArrayBuffer,\n} from '../utils/client'\nimport type { ElevenLabsClient } from '@elevenlabs/elevenlabs-js'\nimport type { TTSOptions, TTSResult } from '@tanstack/ai'\nimport type { ElevenLabsClientConfig } from '../utils/client'\nimport type { ElevenLabsOutputFormat, ElevenLabsTTSModel } from '../model-meta'\n\n/**\n * ElevenLabs voice settings overrides. All fields are optional — omitted\n * values fall back to the voice's stored defaults.\n * @see https://elevenlabs.io/docs/api-reference/text-to-speech/convert\n */\nexport interface ElevenLabsVoiceSettings {\n  /** Voice stability, 0..1. Default 0.5. */\n  stability?: number\n  /** Similarity boost, 0..1. Default 0.75. */\n  similarityBoost?: number\n  /** Style exaggeration, 0..1. Default 0. */\n  style?: number\n  /** Playback speed. Default 1.0. */\n  speed?: number\n  /** Clarity/presence boost. Default true. */\n  useSpeakerBoost?: boolean\n}\n\n/**\n * Provider-specific TTS options. `voice` on `generateSpeech()` takes priority\n * over `voiceId` here, but we expose the same field for callers that prefer\n * to keep voice configuration inside the adapter config.\n */\nexport interface ElevenLabsSpeechProviderOptions {\n  /** ElevenLabs voice ID to synthesize. Required if `generateSpeech().voice` is not set. */\n  voiceId?: string\n  /** Output audio format encoded as `codec_samplerate[_bitrate]`. Defaults to `mp3_44100_128`. */\n  outputFormat?: ElevenLabsOutputFormat\n  /** Voice-settings overrides for this request only. */\n  voiceSettings?: ElevenLabsVoiceSettings\n  /** ISO-639-1 language code to enforce (e.g. `'en'`, `'ja'`). */\n  languageCode?: string\n  /** Deterministic sampling seed, 0..4294967295. */\n  seed?: number\n  /** Previous text for stitching adjacent clips. */\n  previousText?: string\n  /** Next text for stitching adjacent clips. */\n  nextText?: string\n  /** Previous request IDs for stitching (max 3). */\n  previousRequestIds?: Array<string>\n  /** Next request IDs for stitching (max 3). */\n  nextRequestIds?: Array<string>\n  /** Text normalization toggle. Default `'auto'`. */\n  applyTextNormalization?: 'auto' | 'on' | 'off'\n  /** Language-specific text normalization (currently Japanese only, adds latency). */\n  applyLanguageTextNormalization?: boolean\n  /** Latency optimization level, 0..4. */\n  optimizeStreamingLatency?: number\n  /** Enable logging. Set false for zero-retention mode (enterprise only). */\n  enableLogging?: boolean\n}\n\n/**\n * ElevenLabs text-to-speech adapter built on the official\n * `@elevenlabs/elevenlabs-js` SDK.\n *\n * @example\n * ```ts\n * const adapter = elevenlabsSpeech('eleven_multilingual_v2')\n * const result = await generateSpeech({\n *   adapter,\n *   text: 'Hello, world!',\n *   voice: '21m00Tcm4TlvDq8ikWAM',\n * })\n * ```\n */\nexport class ElevenLabsSpeechAdapter<\n  TModel extends ElevenLabsTTSModel,\n> extends BaseTTSAdapter<TModel, ElevenLabsSpeechProviderOptions> {\n  readonly name = 'elevenlabs' as const\n\n  private client: ElevenLabsClient\n\n  constructor(model: TModel, config?: ElevenLabsClientConfig) {\n    super(model, config ?? {})\n    this.client = createElevenLabsClient(config)\n  }\n\n  async generateSpeech(\n    options: TTSOptions<ElevenLabsSpeechProviderOptions>,\n  ): Promise<TTSResult> {\n    const { logger } = options\n    logger.request(\n      `activity=generateSpeech provider=elevenlabs model=${this.model}`,\n      { provider: 'elevenlabs', model: this.model },\n    )\n    try {\n      const voiceId = options.voice ?? options.modelOptions?.voiceId\n      if (!voiceId) {\n        throw new Error(\n          'ElevenLabs TTS requires a voice. Pass `voice` on generateSpeech() or `voiceId` in modelOptions.',\n        )\n      }\n      const {\n        outputFormat,\n        voiceSettings,\n        languageCode,\n        seed,\n        previousText,\n        nextText,\n        previousRequestIds,\n        nextRequestIds,\n        applyTextNormalization,\n        applyLanguageTextNormalization,\n        optimizeStreamingLatency,\n        enableLogging,\n      } = options.modelOptions ?? {}\n      const effectiveOutputFormat =\n        outputFormat ?? inferOutputFormatFromResponseFormat(options.format)\n\n      const stream = await this.client.textToSpeech.convert(voiceId, {\n        text: options.text,\n        modelId: this.model,\n        ...(effectiveOutputFormat\n          ? { outputFormat: effectiveOutputFormat }\n          : {}),\n        ...(voiceSettings\n          ? { voiceSettings: mapVoiceSettings(voiceSettings, options.speed) }\n          : options.speed != null\n            ? { voiceSettings: { speed: options.speed } }\n            : {}),\n        ...(languageCode ? { languageCode } : {}),\n        ...(seed != null ? { seed } : {}),\n        ...(previousText ? { previousText } : {}),\n        ...(nextText ? { nextText } : {}),\n        ...(previousRequestIds ? { previousRequestIds } : {}),\n        ...(nextRequestIds ? { nextRequestIds } : {}),\n        ...(applyTextNormalization ? { applyTextNormalization } : {}),\n        ...(applyLanguageTextNormalization != null\n          ? { applyLanguageTextNormalization }\n          : {}),\n        ...(optimizeStreamingLatency != null\n          ? { optimizeStreamingLatency }\n          : {}),\n        ...(enableLogging != null ? { enableLogging } : {}),\n      })\n\n      const buffer = await readStreamToArrayBuffer(stream)\n      const base64 = arrayBufferToBase64(buffer)\n      const { format, contentType } = parseOutputFormat(effectiveOutputFormat)\n\n      return {\n        id: generateId(this.name),\n        model: this.model,\n        audio: base64,\n        format,\n        contentType,\n      }\n    } catch (error) {\n      logger.errors('elevenlabs.generateSpeech fatal', {\n        error,\n        source: 'elevenlabs.generateSpeech',\n      })\n      throw error\n    }\n  }\n\n  protected override generateId(): string {\n    return generateId(this.name)\n  }\n}\n\nfunction mapVoiceSettings(\n  settings: ElevenLabsVoiceSettings,\n  speedOverride: number | undefined,\n): Record<string, unknown> {\n  return {\n    ...(settings.stability != null ? { stability: settings.stability } : {}),\n    ...(settings.similarityBoost != null\n      ? { similarityBoost: settings.similarityBoost }\n      : {}),\n    ...(settings.style != null ? { style: settings.style } : {}),\n    ...(speedOverride != null\n      ? { speed: speedOverride }\n      : settings.speed != null\n        ? { speed: settings.speed }\n        : {}),\n    ...(settings.useSpeakerBoost != null\n      ? { useSpeakerBoost: settings.useSpeakerBoost }\n      : {}),\n  }\n}\n\n/**\n * Map the standard TTSOptions `format` (mp3/opus/aac/flac/wav/pcm) to a\n * reasonable ElevenLabs `outputFormat` so callers don't need to know the\n * full codec/samplerate string for the common case.\n */\nfunction inferOutputFormatFromResponseFormat(\n  format: TTSOptions['format'] | undefined,\n): ElevenLabsOutputFormat | undefined {\n  switch (format) {\n    case 'mp3':\n      return 'mp3_44100_128'\n    case 'pcm':\n      return 'pcm_44100'\n    case 'opus':\n      return 'opus_48000_128'\n    case undefined:\n      return undefined\n    default:\n      // `aac` / `flac` / `wav` are not native ElevenLabs formats —\n      // fall back to mp3 rather than blowing up mid-request.\n      return 'mp3_44100_128'\n  }\n}\n\n/**\n * Create an ElevenLabs speech adapter using `ELEVENLABS_API_KEY` from env.\n */\nexport function elevenlabsSpeech<TModel extends ElevenLabsTTSModel>(\n  model: TModel,\n  config?: ElevenLabsClientConfig,\n): ElevenLabsSpeechAdapter<TModel> {\n  return new ElevenLabsSpeechAdapter(model, config)\n}\n\n/**\n * Create an ElevenLabs speech adapter with an explicit API key.\n */\nexport function createElevenLabsSpeech<TModel extends ElevenLabsTTSModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<ElevenLabsClientConfig, 'apiKey'>,\n): ElevenLabsSpeechAdapter<TModel> {\n  return new ElevenLabsSpeechAdapter(model, { apiKey, ...config })\n}\n"],"names":[],"mappings":";;AA+EO,MAAM,gCAEH,eAAwD;AAAA,EAKhE,YAAY,OAAe,QAAiC;AAC1D,UAAM,OAAO,UAAU,EAAE;AAL3B,SAAS,OAAO;AAMd,SAAK,SAAS,uBAAuB,MAAM;AAAA,EAC7C;AAAA,EAEA,MAAM,eACJ,SACoB;AACpB,UAAM,EAAE,WAAW;AACnB,WAAO;AAAA,MACL,qDAAqD,KAAK,KAAK;AAAA,MAC/D,EAAE,UAAU,cAAc,OAAO,KAAK,MAAA;AAAA,IAAM;AAE9C,QAAI;AACF,YAAM,UAAU,QAAQ,SAAS,QAAQ,cAAc;AACvD,UAAI,CAAC,SAAS;AACZ,cAAM,IAAI;AAAA,UACR;AAAA,QAAA;AAAA,MAEJ;AACA,YAAM;AAAA,QACJ;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MAAA,IACE,QAAQ,gBAAgB,CAAA;AAC5B,YAAM,wBACJ,gBAAgB,oCAAoC,QAAQ,MAAM;AAEpE,YAAM,SAAS,MAAM,KAAK,OAAO,aAAa,QAAQ,SAAS;AAAA,QAC7D,MAAM,QAAQ;AAAA,QACd,SAAS,KAAK;AAAA,QACd,GAAI,wBACA,EAAE,cAAc,sBAAA,IAChB,CAAA;AAAA,QACJ,GAAI,gBACA,EAAE,eAAe,iBAAiB,eAAe,QAAQ,KAAK,MAC9D,QAAQ,SAAS,OACf,EAAE,eAAe,EAAE,OAAO,QAAQ,MAAA,EAAM,IACxC,CAAA;AAAA,QACN,GAAI,eAAe,EAAE,aAAA,IAAiB,CAAA;AAAA,QACtC,GAAI,QAAQ,OAAO,EAAE,KAAA,IAAS,CAAA;AAAA,QAC9B,GAAI,eAAe,EAAE,aAAA,IAAiB,CAAA;AAAA,QACtC,GAAI,WAAW,EAAE,SAAA,IAAa,CAAA;AAAA,QAC9B,GAAI,qBAAqB,EAAE,mBAAA,IAAuB,CAAA;AAAA,QAClD,GAAI,iBAAiB,EAAE,eAAA,IAAmB,CAAA;AAAA,QAC1C,GAAI,yBAAyB,EAAE,uBAAA,IAA2B,CAAA;AAAA,QAC1D,GAAI,kCAAkC,OAClC,EAAE,+BAAA,IACF,CAAA;AAAA,QACJ,GAAI,4BAA4B,OAC5B,EAAE,yBAAA,IACF,CAAA;AAAA,QACJ,GAAI,iBAAiB,OAAO,EAAE,kBAAkB,CAAA;AAAA,MAAC,CAClD;AAED,YAAM,SAAS,MAAM,wBAAwB,MAAM;AACnD,YAAM,SAAS,oBAAoB,MAAM;AACzC,YAAM,EAAE,QAAQ,gBAAgB,kBAAkB,qBAAqB;AAEvE,aAAO;AAAA,QACL,IAAI,WAAW,KAAK,IAAI;AAAA,QACxB,OAAO,KAAK;AAAA,QACZ,OAAO;AAAA,QACP;AAAA,QACA;AAAA,MAAA;AAAA,IAEJ,SAAS,OAAO;AACd,aAAO,OAAO,mCAAmC;AAAA,QAC/C;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEmB,aAAqB;AACtC,WAAO,WAAW,KAAK,IAAI;AAAA,EAC7B;AACF;AAEA,SAAS,iBACP,UACA,eACyB;AACzB,SAAO;AAAA,IACL,GAAI,SAAS,aAAa,OAAO,EAAE,WAAW,SAAS,UAAA,IAAc,CAAA;AAAA,IACrE,GAAI,SAAS,mBAAmB,OAC5B,EAAE,iBAAiB,SAAS,gBAAA,IAC5B,CAAA;AAAA,IACJ,GAAI,SAAS,SAAS,OAAO,EAAE,OAAO,SAAS,MAAA,IAAU,CAAA;AAAA,IACzD,GAAI,iBAAiB,OACjB,EAAE,OAAO,cAAA,IACT,SAAS,SAAS,OAChB,EAAE,OAAO,SAAS,MAAA,IAClB,CAAA;AAAA,IACN,GAAI,SAAS,mBAAmB,OAC5B,EAAE,iBAAiB,SAAS,oBAC5B,CAAA;AAAA,EAAC;AAET;AAOA,SAAS,oCACP,QACoC;AACpC,UAAQ,QAAA;AAAA,IACN,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT;AAGE,aAAO;AAAA,EAAA;AAEb;AAKO,SAAS,iBACd,OACA,QACiC;AACjC,SAAO,IAAI,wBAAwB,OAAO,MAAM;AAClD;AAKO,SAAS,uBACd,OACA,QACA,QACiC;AACjC,SAAO,IAAI,wBAAwB,OAAO,EAAE,QAAQ,GAAG,QAAQ;AACjE;"}
+{"version":3,"file":"speech.js","sources":["../../../src/adapters/speech.ts"],"sourcesContent":["import { BaseTTSAdapter } from '@tanstack/ai/adapters'\nimport {\n  arrayBufferToBase64,\n  createElevenLabsClient,\n  generateId,\n  parseOutputFormat,\n  readStreamToArrayBuffer,\n} from '../utils/client'\nimport type { ElevenLabsClient } from '@elevenlabs/elevenlabs-js'\nimport type { TTSOptions, TTSResult } from '@tanstack/ai'\nimport type { ElevenLabsClientConfig } from '../utils/client'\nimport type { ElevenLabsOutputFormat, ElevenLabsTTSModel } from '../model-meta'\n\n/**\n * ElevenLabs voice settings overrides. All fields are optional — omitted\n * values fall back to the voice's stored defaults.\n * @see https://elevenlabs.io/docs/api-reference/text-to-speech/convert\n */\nexport interface ElevenLabsVoiceSettings {\n  /** Voice stability, 0..1. Default 0.5. */\n  stability?: number\n  /** Similarity boost, 0..1. Default 0.75. */\n  similarityBoost?: number\n  /** Style exaggeration, 0..1. Default 0. */\n  style?: number\n  /** Playback speed. Default 1.0. */\n  speed?: number\n  /** Clarity/presence boost. Default true. */\n  useSpeakerBoost?: boolean\n}\n\n/**\n * Provider-specific TTS options. `voice` on `generateSpeech()` takes priority\n * over `voiceId` here, but we expose the same field for callers that prefer\n * to keep voice configuration inside the adapter config.\n */\nexport interface ElevenLabsSpeechProviderOptions {\n  /** ElevenLabs voice ID to synthesize. Required if `generateSpeech().voice` is not set. */\n  voiceId?: string\n  /** Output audio format encoded as `codec_samplerate[_bitrate]`. Defaults to `mp3_44100_128`. */\n  outputFormat?: ElevenLabsOutputFormat\n  /** Voice-settings overrides for this request only. */\n  voiceSettings?: ElevenLabsVoiceSettings\n  /** ISO-639-1 language code to enforce (e.g. `'en'`, `'ja'`). */\n  languageCode?: string\n  /** Deterministic sampling seed, 0..4294967295. */\n  seed?: number\n  /** Previous text for stitching adjacent clips. */\n  previousText?: string\n  /** Next text for stitching adjacent clips. */\n  nextText?: string\n  /** Previous request IDs for stitching (max 3). */\n  previousRequestIds?: Array<string>\n  /** Next request IDs for stitching (max 3). */\n  nextRequestIds?: Array<string>\n  /** Text normalization toggle. Default `'auto'`. */\n  applyTextNormalization?: 'auto' | 'on' | 'off'\n  /** Language-specific text normalization (currently Japanese only, adds latency). */\n  applyLanguageTextNormalization?: boolean\n  /** Latency optimization level, 0..4. */\n  optimizeStreamingLatency?: number\n  /** Enable logging. Set false for zero-retention mode (enterprise only). */\n  enableLogging?: boolean\n}\n\n/**\n * ElevenLabs text-to-speech adapter built on the official\n * `@elevenlabs/elevenlabs-js` SDK.\n *\n * @example\n * ```ts\n * const adapter = elevenlabsSpeech('eleven_multilingual_v2')\n * const result = await generateSpeech({\n *   adapter,\n *   text: 'Hello, world!',\n *   voice: '21m00Tcm4TlvDq8ikWAM',\n * })\n * ```\n */\nexport class ElevenLabsSpeechAdapter<\n  TModel extends ElevenLabsTTSModel,\n> extends BaseTTSAdapter<TModel, ElevenLabsSpeechProviderOptions> {\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 generateSpeech(\n    options: TTSOptions<ElevenLabsSpeechProviderOptions>,\n  ): Promise<TTSResult> {\n    const { logger } = options\n    logger.request(\n      `activity=generateSpeech provider=elevenlabs model=${this.model}`,\n      { provider: 'elevenlabs', model: this.model },\n    )\n    try {\n      const voiceId = options.voice ?? options.modelOptions?.voiceId\n      if (!voiceId) {\n        throw new Error(\n          'ElevenLabs TTS requires a voice. Pass `voice` on generateSpeech() or `voiceId` in modelOptions.',\n        )\n      }\n      const {\n        outputFormat,\n        voiceSettings,\n        languageCode,\n        seed,\n        previousText,\n        nextText,\n        previousRequestIds,\n        nextRequestIds,\n        applyTextNormalization,\n        applyLanguageTextNormalization,\n        optimizeStreamingLatency,\n        enableLogging,\n      } = options.modelOptions ?? {}\n      const effectiveOutputFormat =\n        outputFormat ?? inferOutputFormatFromResponseFormat(options.format)\n\n      const stream = await this.client.textToSpeech.convert(voiceId, {\n        text: options.text,\n        modelId: this.model,\n        ...(effectiveOutputFormat\n          ? { outputFormat: effectiveOutputFormat }\n          : {}),\n        ...(voiceSettings\n          ? { voiceSettings: mapVoiceSettings(voiceSettings, options.speed) }\n          : options.speed != null\n            ? { voiceSettings: { speed: options.speed } }\n            : {}),\n        ...(languageCode ? { languageCode } : {}),\n        ...(seed != null ? { seed } : {}),\n        ...(previousText ? { previousText } : {}),\n        ...(nextText ? { nextText } : {}),\n        ...(previousRequestIds ? { previousRequestIds } : {}),\n        ...(nextRequestIds ? { nextRequestIds } : {}),\n        ...(applyTextNormalization ? { applyTextNormalization } : {}),\n        ...(applyLanguageTextNormalization != null\n          ? { applyLanguageTextNormalization }\n          : {}),\n        ...(optimizeStreamingLatency != null\n          ? { optimizeStreamingLatency }\n          : {}),\n        ...(enableLogging != null ? { enableLogging } : {}),\n      })\n\n      const buffer = await readStreamToArrayBuffer(stream)\n      const base64 = arrayBufferToBase64(buffer)\n      const { format, contentType } = parseOutputFormat(effectiveOutputFormat)\n\n      return {\n        id: generateId(this.name),\n        model: this.model,\n        audio: base64,\n        format,\n        contentType,\n      }\n    } catch (error) {\n      logger.errors('elevenlabs.generateSpeech fatal', {\n        error,\n        source: 'elevenlabs.generateSpeech',\n      })\n      throw error\n    }\n  }\n\n  protected override generateId(): string {\n    return generateId(this.name)\n  }\n}\n\nfunction mapVoiceSettings(\n  settings: ElevenLabsVoiceSettings,\n  speedOverride: number | undefined,\n): Record<string, unknown> {\n  return {\n    ...(settings.stability != null ? { stability: settings.stability } : {}),\n    ...(settings.similarityBoost != null\n      ? { similarityBoost: settings.similarityBoost }\n      : {}),\n    ...(settings.style != null ? { style: settings.style } : {}),\n    ...(speedOverride != null\n      ? { speed: speedOverride }\n      : settings.speed != null\n        ? { speed: settings.speed }\n        : {}),\n    ...(settings.useSpeakerBoost != null\n      ? { useSpeakerBoost: settings.useSpeakerBoost }\n      : {}),\n  }\n}\n\n/**\n * Map the standard TTSOptions `format` (mp3/opus/aac/flac/wav/pcm) to a\n * reasonable ElevenLabs `outputFormat` so callers don't need to know the\n * full codec/samplerate string for the common case.\n */\nfunction inferOutputFormatFromResponseFormat(\n  format: TTSOptions['format'] | undefined,\n): ElevenLabsOutputFormat | undefined {\n  switch (format) {\n    case 'mp3':\n      return 'mp3_44100_128'\n    case 'pcm':\n      return 'pcm_44100'\n    case 'opus':\n      return 'opus_48000_128'\n    case undefined:\n      return undefined\n    case 'aac':\n    case 'flac':\n    case 'wav':\n    default:\n      // `aac` / `flac` / `wav` are not native ElevenLabs formats —\n      // fall back to mp3 rather than blowing up mid-request.\n      return 'mp3_44100_128'\n  }\n}\n\n/**\n * Create an ElevenLabs speech adapter using `ELEVENLABS_API_KEY` from env.\n */\nexport function elevenlabsSpeech<TModel extends ElevenLabsTTSModel>(\n  model: TModel,\n  config?: ElevenLabsClientConfig,\n): ElevenLabsSpeechAdapter<TModel> {\n  return new ElevenLabsSpeechAdapter(model, config)\n}\n\n/**\n * Create an ElevenLabs speech adapter with an explicit API key.\n */\nexport function createElevenLabsSpeech<TModel extends ElevenLabsTTSModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<ElevenLabsClientConfig, 'apiKey'>,\n): ElevenLabsSpeechAdapter<TModel> {\n  return new ElevenLabsSpeechAdapter(model, { apiKey, ...config })\n}\n"],"names":[],"mappings":";;AA+EO,MAAM,gCAEH,eAAwD;AAAA,EACvD,OAAO;AAAA,EAEC;AAAA,EAEjB,YAAY,OAAe,QAAiC;AAC1D,UAAM,OAAO,UAAU,EAAE;AACzB,SAAK,SAAS,uBAAuB,MAAM;AAAA,EAC7C;AAAA,EAEA,MAAM,eACJ,SACoB;AACpB,UAAM,EAAE,WAAW;AACnB,WAAO;AAAA,MACL,qDAAqD,KAAK,KAAK;AAAA,MAC/D,EAAE,UAAU,cAAc,OAAO,KAAK,MAAA;AAAA,IAAM;AAE9C,QAAI;AACF,YAAM,UAAU,QAAQ,SAAS,QAAQ,cAAc;AACvD,UAAI,CAAC,SAAS;AACZ,cAAM,IAAI;AAAA,UACR;AAAA,QAAA;AAAA,MAEJ;AACA,YAAM;AAAA,QACJ;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MAAA,IACE,QAAQ,gBAAgB,CAAA;AAC5B,YAAM,wBACJ,gBAAgB,oCAAoC,QAAQ,MAAM;AAEpE,YAAM,SAAS,MAAM,KAAK,OAAO,aAAa,QAAQ,SAAS;AAAA,QAC7D,MAAM,QAAQ;AAAA,QACd,SAAS,KAAK;AAAA,QACd,GAAI,wBACA,EAAE,cAAc,sBAAA,IAChB,CAAA;AAAA,QACJ,GAAI,gBACA,EAAE,eAAe,iBAAiB,eAAe,QAAQ,KAAK,MAC9D,QAAQ,SAAS,OACf,EAAE,eAAe,EAAE,OAAO,QAAQ,MAAA,EAAM,IACxC,CAAA;AAAA,QACN,GAAI,eAAe,EAAE,aAAA,IAAiB,CAAA;AAAA,QACtC,GAAI,QAAQ,OAAO,EAAE,KAAA,IAAS,CAAA;AAAA,QAC9B,GAAI,eAAe,EAAE,aAAA,IAAiB,CAAA;AAAA,QACtC,GAAI,WAAW,EAAE,SAAA,IAAa,CAAA;AAAA,QAC9B,GAAI,qBAAqB,EAAE,mBAAA,IAAuB,CAAA;AAAA,QAClD,GAAI,iBAAiB,EAAE,eAAA,IAAmB,CAAA;AAAA,QAC1C,GAAI,yBAAyB,EAAE,uBAAA,IAA2B,CAAA;AAAA,QAC1D,GAAI,kCAAkC,OAClC,EAAE,+BAAA,IACF,CAAA;AAAA,QACJ,GAAI,4BAA4B,OAC5B,EAAE,yBAAA,IACF,CAAA;AAAA,QACJ,GAAI,iBAAiB,OAAO,EAAE,kBAAkB,CAAA;AAAA,MAAC,CAClD;AAED,YAAM,SAAS,MAAM,wBAAwB,MAAM;AACnD,YAAM,SAAS,oBAAoB,MAAM;AACzC,YAAM,EAAE,QAAQ,gBAAgB,kBAAkB,qBAAqB;AAEvE,aAAO;AAAA,QACL,IAAI,WAAW,KAAK,IAAI;AAAA,QACxB,OAAO,KAAK;AAAA,QACZ,OAAO;AAAA,QACP;AAAA,QACA;AAAA,MAAA;AAAA,IAEJ,SAAS,OAAO;AACd,aAAO,OAAO,mCAAmC;AAAA,QAC/C;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEmB,aAAqB;AACtC,WAAO,WAAW,KAAK,IAAI;AAAA,EAC7B;AACF;AAEA,SAAS,iBACP,UACA,eACyB;AACzB,SAAO;AAAA,IACL,GAAI,SAAS,aAAa,OAAO,EAAE,WAAW,SAAS,UAAA,IAAc,CAAA;AAAA,IACrE,GAAI,SAAS,mBAAmB,OAC5B,EAAE,iBAAiB,SAAS,gBAAA,IAC5B,CAAA;AAAA,IACJ,GAAI,SAAS,SAAS,OAAO,EAAE,OAAO,SAAS,MAAA,IAAU,CAAA;AAAA,IACzD,GAAI,iBAAiB,OACjB,EAAE,OAAO,cAAA,IACT,SAAS,SAAS,OAChB,EAAE,OAAO,SAAS,MAAA,IAClB,CAAA;AAAA,IACN,GAAI,SAAS,mBAAmB,OAC5B,EAAE,iBAAiB,SAAS,oBAC5B,CAAA;AAAA,EAAC;AAET;AAOA,SAAS,oCACP,QACoC;AACpC,UAAQ,QAAA;AAAA,IACN,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK;AAAA,IACL;AAGE,aAAO;AAAA,EAAA;AAEb;AAKO,SAAS,iBACd,OACA,QACiC;AACjC,SAAO,IAAI,wBAAwB,OAAO,MAAM;AAClD;AAKO,SAAS,uBACd,OACA,QACA,QACiC;AACjC,SAAO,IAAI,wBAAwB,OAAO,EAAE,QAAQ,GAAG,QAAQ;AACjE;"}

package/dist/esm/adapters/transcription.d.ts

@@ -63,7 +63,7 @@ export interface ElevenLabsTranscriptionProviderOptions {
  */
 export declare class ElevenLabsTranscriptionAdapter<TModel extends ElevenLabsTranscriptionModel> extends BaseTranscriptionAdapter<TModel, ElevenLabsTranscriptionProviderOptions> {
     readonly name: "elevenlabs";
-    private client;
+    private readonly client;
     constructor(model: TModel, config?: ElevenLabsClientConfig);
     transcribe(options: TranscriptionOptions<ElevenLabsTranscriptionProviderOptions>): Promise<TranscriptionResult>;
     private transformResponse;

package/dist/esm/adapters/transcription.js

@@ -1,9 +1,10 @@
 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.name = "elevenlabs";
     this.client = createElevenLabsClient(config);
   }
   async transcribe(options) {

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

@@ -1 +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 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    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) =>\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,EAKA,YAAY,OAAe,QAAiC;AAC1D,UAAM,OAAO,UAAU,EAAE;AAL3B,SAAS,OAAO;AAMd,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;AAIrB,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;"}
+{"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/realtime/adapter.js

@@ -151,10 +151,12 @@ async function createElevenLabsConnection(token, _options, logger, clientToolDef
       emit("interrupted", {});
     },
     on(event, handler) {
-      if (!eventHandlers.has(event)) {
-        eventHandlers.set(event, /* @__PURE__ */ new Set());
+      let handlers = eventHandlers.get(event);
+      if (!handlers) {
+        handlers = /* @__PURE__ */ new Set();
+        eventHandlers.set(event, handlers);
       }
-      eventHandlers.get(event).add(handler);
+      handlers.add(handler);
       return () => {
         eventHandlers.get(event)?.delete(handler);
       };

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

@@ -1 +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  RealtimeEvent,\n  RealtimeEventHandler,\n  RealtimeMessage,\n  RealtimeMode,\n  RealtimeSessionConfig,\n  RealtimeStatus,\n  RealtimeToken,\n} from '@tanstack/ai'\nimport type { InternalLogger } from '@tanstack/ai/adapter-internals'\nimport type { RealtimeAdapter, RealtimeConnection } from '@tanstack/ai-client'\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' as RealtimeStatus })\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' as RealtimeStatus })\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' as RealtimeStatus })\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      if (!eventHandlers.has(event)) {\n        eventHandlers.set(event, new Set())\n      }\n      eventHandlers.get(event)!.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,CAA+B;AAC/D,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,CAA0B;AAC1D,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,CAA0B;AAAA,IAC5D;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,CAAC,cAAc,IAAI,KAAK,GAAG;AAC7B,sBAAc,IAAI,OAAO,oBAAI,IAAA,CAAK;AAAA,MACpC;AACA,oBAAc,IAAI,KAAK,EAAG,IAAI,OAAO;AAErC,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;"}
+{"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  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 { RealtimeAdapter, RealtimeConnection } from '@tanstack/ai-client'\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":";;AAmCO,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/token.js

@@ -15,12 +15,18 @@ function elevenlabsRealtimeToken(options = {}) {
         token: response.signedUrl,
         expiresAt,
         config: {
-          voice: overrides?.voiceId,
-          instructions: overrides?.systemPrompt,
+          ...overrides?.voiceId !== void 0 && { voice: overrides.voiceId },
+          ...overrides?.systemPrompt !== void 0 && {
+            instructions: overrides.systemPrompt
+          },
           providerOptions: {
             agentId,
-            firstMessage: overrides?.firstMessage,
-            language: overrides?.language
+            ...overrides?.firstMessage !== void 0 && {
+              firstMessage: overrides.firstMessage
+            },
+            ...overrides?.language !== void 0 && {
+              language: overrides.language
+            }
           }
         }
       };

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

@@ -1 +1 @@
-{"version":3,"file":"token.js","sources":["../../../src/realtime/token.ts"],"sourcesContent":["import {\n  createElevenLabsClient,\n  getElevenLabsAgentIdFromEnv,\n} from '../utils/client'\nimport type { RealtimeToken, RealtimeTokenAdapter } from '@tanstack/ai'\nimport type { ElevenLabsRealtimeTokenOptions } from './types'\n\n/**\n * Creates an ElevenLabs realtime token adapter.\n *\n * Uses the official `@elevenlabs/elevenlabs-js` SDK to request a signed URL\n * for client-side conversation connections. The signed URL is valid for\n * 30 minutes.\n *\n * @param options - Configuration. `agentId` falls back to\n *   `ELEVENLABS_AGENT_ID` in the environment when omitted.\n * @returns A RealtimeTokenAdapter for use with realtimeToken()\n *\n * @example\n * ```typescript\n * import { realtimeToken } from '@tanstack/ai'\n * import { elevenlabsRealtimeToken } from '@tanstack/ai-elevenlabs'\n *\n * // Reads ELEVENLABS_AGENT_ID from env:\n * const token = await realtimeToken({ adapter: elevenlabsRealtimeToken() })\n *\n * // Or pass explicitly:\n * const token = await realtimeToken({\n *   adapter: elevenlabsRealtimeToken({ agentId: 'your-agent-id' }),\n * })\n * ```\n */\nexport function elevenlabsRealtimeToken(\n  options: ElevenLabsRealtimeTokenOptions = {},\n): RealtimeTokenAdapter {\n  const client = createElevenLabsClient()\n\n  return {\n    provider: 'elevenlabs',\n\n    async generateToken(): Promise<RealtimeToken> {\n      const { overrides } = options\n      const agentId = options.agentId ?? getElevenLabsAgentIdFromEnv()\n\n      const response = await client.conversationalAi.conversations.getSignedUrl(\n        { agentId },\n      )\n\n      // Signed URLs are valid for 30 minutes\n      const expiresAt = Date.now() + 30 * 60 * 1000\n\n      return {\n        provider: 'elevenlabs',\n        token: response.signedUrl,\n        expiresAt,\n        config: {\n          voice: overrides?.voiceId,\n          instructions: overrides?.systemPrompt,\n          providerOptions: {\n            agentId,\n            firstMessage: overrides?.firstMessage,\n            language: overrides?.language,\n          },\n        },\n      }\n    },\n  }\n}\n"],"names":[],"mappings":";AAgCO,SAAS,wBACd,UAA0C,IACpB;AACtB,QAAM,SAAS,uBAAA;AAEf,SAAO;AAAA,IACL,UAAU;AAAA,IAEV,MAAM,gBAAwC;AAC5C,YAAM,EAAE,cAAc;AACtB,YAAM,UAAU,QAAQ,WAAW,4BAAA;AAEnC,YAAM,WAAW,MAAM,OAAO,iBAAiB,cAAc;AAAA,QAC3D,EAAE,QAAA;AAAA,MAAQ;AAIZ,YAAM,YAAY,KAAK,IAAA,IAAQ,KAAK,KAAK;AAEzC,aAAO;AAAA,QACL,UAAU;AAAA,QACV,OAAO,SAAS;AAAA,QAChB;AAAA,QACA,QAAQ;AAAA,UACN,OAAO,WAAW;AAAA,UAClB,cAAc,WAAW;AAAA,UACzB,iBAAiB;AAAA,YACf;AAAA,YACA,cAAc,WAAW;AAAA,YACzB,UAAU,WAAW;AAAA,UAAA;AAAA,QACvB;AAAA,MACF;AAAA,IAEJ;AAAA,EAAA;AAEJ;"}
+{"version":3,"file":"token.js","sources":["../../../src/realtime/token.ts"],"sourcesContent":["import {\n  createElevenLabsClient,\n  getElevenLabsAgentIdFromEnv,\n} from '../utils/client'\nimport type { RealtimeToken, RealtimeTokenAdapter } from '@tanstack/ai'\nimport type { ElevenLabsRealtimeTokenOptions } from './types'\n\n/**\n * Creates an ElevenLabs realtime token adapter.\n *\n * Uses the official `@elevenlabs/elevenlabs-js` SDK to request a signed URL\n * for client-side conversation connections. The signed URL is valid for\n * 30 minutes.\n *\n * @param options - Configuration. `agentId` falls back to\n *   `ELEVENLABS_AGENT_ID` in the environment when omitted.\n * @returns A RealtimeTokenAdapter for use with realtimeToken()\n *\n * @example\n * ```typescript\n * import { realtimeToken } from '@tanstack/ai'\n * import { elevenlabsRealtimeToken } from '@tanstack/ai-elevenlabs'\n *\n * // Reads ELEVENLABS_AGENT_ID from env:\n * const token = await realtimeToken({ adapter: elevenlabsRealtimeToken() })\n *\n * // Or pass explicitly:\n * const token = await realtimeToken({\n *   adapter: elevenlabsRealtimeToken({ agentId: 'your-agent-id' }),\n * })\n * ```\n */\nexport function elevenlabsRealtimeToken(\n  options: ElevenLabsRealtimeTokenOptions = {},\n): RealtimeTokenAdapter {\n  const client = createElevenLabsClient()\n\n  return {\n    provider: 'elevenlabs',\n\n    async generateToken(): Promise<RealtimeToken> {\n      const { overrides } = options\n      const agentId = options.agentId ?? getElevenLabsAgentIdFromEnv()\n\n      const response = await client.conversationalAi.conversations.getSignedUrl(\n        { agentId },\n      )\n\n      // Signed URLs are valid for 30 minutes\n      const expiresAt = Date.now() + 30 * 60 * 1000\n\n      // Conditionally spread optional fields — the vendor target type\n      // (`RealtimeSessionConfig` in `@tanstack/ai`) uses strict\n      // `field?: T` semantics under `exactOptionalPropertyTypes`, so we\n      // omit keys entirely when the override is absent rather than\n      // setting them to `undefined`.\n      return {\n        provider: 'elevenlabs',\n        token: response.signedUrl,\n        expiresAt,\n        config: {\n          ...(overrides?.voiceId !== undefined && { voice: overrides.voiceId }),\n          ...(overrides?.systemPrompt !== undefined && {\n            instructions: overrides.systemPrompt,\n          }),\n          providerOptions: {\n            agentId,\n            ...(overrides?.firstMessage !== undefined && {\n              firstMessage: overrides.firstMessage,\n            }),\n            ...(overrides?.language !== undefined && {\n              language: overrides.language,\n            }),\n          },\n        },\n      }\n    },\n  }\n}\n"],"names":[],"mappings":";AAgCO,SAAS,wBACd,UAA0C,IACpB;AACtB,QAAM,SAAS,uBAAA;AAEf,SAAO;AAAA,IACL,UAAU;AAAA,IAEV,MAAM,gBAAwC;AAC5C,YAAM,EAAE,cAAc;AACtB,YAAM,UAAU,QAAQ,WAAW,4BAAA;AAEnC,YAAM,WAAW,MAAM,OAAO,iBAAiB,cAAc;AAAA,QAC3D,EAAE,QAAA;AAAA,MAAQ;AAIZ,YAAM,YAAY,KAAK,IAAA,IAAQ,KAAK,KAAK;AAOzC,aAAO;AAAA,QACL,UAAU;AAAA,QACV,OAAO,SAAS;AAAA,QAChB;AAAA,QACA,QAAQ;AAAA,UACN,GAAI,WAAW,YAAY,UAAa,EAAE,OAAO,UAAU,QAAA;AAAA,UAC3D,GAAI,WAAW,iBAAiB,UAAa;AAAA,YAC3C,cAAc,UAAU;AAAA,UAAA;AAAA,UAE1B,iBAAiB;AAAA,YACf;AAAA,YACA,GAAI,WAAW,iBAAiB,UAAa;AAAA,cAC3C,cAAc,UAAU;AAAA,YAAA;AAAA,YAE1B,GAAI,WAAW,aAAa,UAAa;AAAA,cACvC,UAAU,UAAU;AAAA,YAAA;AAAA,UACtB;AAAA,QACF;AAAA,MACF;AAAA,IAEJ;AAAA,EAAA;AAEJ;"}

package/dist/esm/utils/client.js

@@ -39,8 +39,8 @@ function arrayBufferToBase64(buffer) {
   }
   const view = new Uint8Array(buffer);
   let binary = "";
-  for (let i = 0; i < view.byteLength; i += 1) {
-    binary += String.fromCharCode(view[i]);
+  for (const byte of view) {
+    binary += String.fromCharCode(byte);
   }
   return btoa(binary);
 }
@@ -100,6 +100,7 @@ function parseOutputFormat(fmt) {
       return { format: "ulaw", contentType: "audio/basic" };
     case "alaw":
       return { format: "alaw", contentType: "audio/x-alaw-basic" };
+    case void 0:
     default:
       return { format: "mp3", contentType: "audio/mpeg" };
   }

package/dist/esm/utils/client.js.map

@@ -1 +1 @@
-{"version":3,"file":"client.js","sources":["../../../src/utils/client.ts"],"sourcesContent":["import { ElevenLabsClient } from '@elevenlabs/elevenlabs-js'\nimport {\n  getApiKeyFromEnv,\n  generateId as sharedGenerateId,\n} from '@tanstack/ai-utils'\nimport type { ElevenLabsOutputFormat } from '../model-meta'\n\n/**\n * Configuration for any ElevenLabs adapter. When `apiKey` is omitted we read\n * `ELEVENLABS_API_KEY` from `process.env` / `window.env` to match the\n * pattern the realtime adapters already use.\n */\nexport interface ElevenLabsClientConfig {\n  apiKey?: string\n  /** Override the API base URL — handy for tests + self-hosted proxies. */\n  baseUrl?: string\n  /** Per-request timeout passed through to the SDK (ms). */\n  timeoutInSeconds?: number\n  /** Override the number of SDK-level retries. */\n  maxRetries?: number\n  /** Extra headers attached to every request (e.g. test multiplexing). */\n  headers?: Record<string, string>\n}\n\ninterface EnvObject {\n  ELEVENLABS_API_KEY?: string\n  ELEVENLABS_AGENT_ID?: string\n}\n\ninterface WindowWithEnv {\n  env?: EnvObject\n}\n\nfunction getEnvironment(): EnvObject | undefined {\n  if (typeof globalThis !== 'undefined') {\n    const win = (globalThis as { window?: WindowWithEnv }).window\n    if (win?.env) return win.env\n  }\n  if (typeof process !== 'undefined') {\n    return process.env as EnvObject\n  }\n  return undefined\n}\n\nexport function getElevenLabsApiKeyFromEnv(): string {\n  return getApiKeyFromEnv('ELEVENLABS_API_KEY')\n}\n\nexport function getElevenLabsAgentIdFromEnv(): string {\n  const id = getEnvironment()?.ELEVENLABS_AGENT_ID\n  if (!id) {\n    throw new Error(\n      'ELEVENLABS_AGENT_ID is required. Please set it in your environment ' +\n        'variables or pass `agentId` explicitly to elevenlabsRealtimeToken().',\n    )\n  }\n  return id\n}\n\n/**\n * Build an `ElevenLabsClient` with env-based or explicit credentials.\n * Each adapter calls this once at construction time so unit tests can\n * pass in an explicit key without needing `process.env`.\n */\nexport function createElevenLabsClient(\n  config?: ElevenLabsClientConfig,\n): ElevenLabsClient {\n  const apiKey = config?.apiKey ?? getElevenLabsApiKeyFromEnv()\n  return new ElevenLabsClient({\n    apiKey,\n    ...(config?.baseUrl ? { baseUrl: config.baseUrl } : {}),\n    ...(config?.timeoutInSeconds != null\n      ? { timeoutInSeconds: config.timeoutInSeconds }\n      : {}),\n    ...(config?.maxRetries != null ? { maxRetries: config.maxRetries } : {}),\n    ...(config?.headers ? { headers: config.headers } : {}),\n  })\n}\n\n// Re-exported from `@tanstack/ai-utils` so existing callers keep working\n// while the implementation stays deduped across provider packages.\nexport const generateId = sharedGenerateId\n\n/**\n * Convert an ArrayBuffer to base64 in a cross-runtime way.\n *\n * The naive `btoa(String.fromCharCode(...bytes))` form blows up V8's argument\n * limit (~65k) on realistic audio payloads, so we either use `Buffer`\n * (Node / Bun) or walk the byte array in a single loop (browser).\n */\nexport function arrayBufferToBase64(buffer: ArrayBuffer): string {\n  if (typeof Buffer !== 'undefined' && typeof Buffer.from === 'function') {\n    return Buffer.from(buffer).toString('base64')\n  }\n  const view = new Uint8Array(buffer)\n  let binary = ''\n  for (let i = 0; i < view.byteLength; i += 1) {\n    binary += String.fromCharCode(view[i]!)\n  }\n  return btoa(binary)\n}\n\n/**\n * Drain a `ReadableStream<Uint8Array>` (what the ElevenLabs SDK returns for\n * audio endpoints) into a single `Uint8Array`, then expose it as an\n * `ArrayBuffer` slice. We concatenate ourselves rather than going through\n * `new Response(stream).arrayBuffer()` so we stay runtime-agnostic.\n */\nexport async function readStreamToArrayBuffer(\n  stream: ReadableStream<Uint8Array>,\n): Promise<ArrayBuffer> {\n  const reader = stream.getReader()\n  const chunks: Array<Uint8Array> = []\n  let total = 0\n  try {\n    let result = await reader.read()\n    while (!result.done) {\n      chunks.push(result.value)\n      total += result.value.byteLength\n      result = await reader.read()\n    }\n  } finally {\n    reader.releaseLock()\n  }\n  const merged = new Uint8Array(total)\n  let offset = 0\n  for (const chunk of chunks) {\n    merged.set(chunk, offset)\n    offset += chunk.byteLength\n  }\n  return merged.buffer.slice(\n    merged.byteOffset,\n    merged.byteOffset + merged.byteLength,\n  )\n}\n\n/**\n * Decode a `data:` URL into a Blob for upload to the ElevenLabs STT API\n * (which only accepts multipart files or https URLs). Supports base64 and\n * URL-encoded payloads. Returns `undefined` for non-data-URL strings so the\n * caller can fall through to treating the input as an https URL.\n */\nexport function dataUrlToBlob(value: string): Blob | undefined {\n  if (!value.startsWith('data:')) return undefined\n  const commaIndex = value.indexOf(',')\n  if (commaIndex === -1) return undefined\n\n  const header = value.slice(5, commaIndex)\n  const payload = value.slice(commaIndex + 1)\n  const isBase64 = /;base64$/i.test(header)\n  const mimeType = header.split(';')[0] || 'application/octet-stream'\n\n  if (isBase64) {\n    const binary = atob(payload)\n    const bytes = new Uint8Array(binary.length)\n    for (let i = 0; i < binary.length; i += 1) {\n      bytes[i] = binary.charCodeAt(i)\n    }\n    return new Blob([bytes], { type: mimeType })\n  }\n\n  return new Blob([decodeURIComponent(payload)], { type: mimeType })\n}\n\n/**\n * Break an ElevenLabs `output_format` string (`mp3_44100_128`,\n * `pcm_24000`, `opus_48000_64`, `ulaw_8000`, ...) into a file extension and\n * content-type suitable for `TTSResult` / `AudioGenerationResult` consumers.\n *\n * Unknown codecs fall back to `mp3` / `audio/mpeg` because the ElevenLabs\n * default is `mp3_44100_128` — mispredicting on an exotic format is safer\n * than throwing in the adapter.\n */\nexport function parseOutputFormat(fmt: ElevenLabsOutputFormat | undefined): {\n  format: string\n  contentType: string\n} {\n  const codec = (fmt || 'mp3_44100_128').split('_')[0]?.toLowerCase()\n  switch (codec) {\n    case 'mp3':\n      return { format: 'mp3', contentType: 'audio/mpeg' }\n    case 'pcm':\n      return { format: 'pcm', contentType: 'audio/pcm' }\n    case 'opus':\n      return { format: 'opus', contentType: 'audio/opus' }\n    case 'ulaw':\n      return { format: 'ulaw', contentType: 'audio/basic' }\n    case 'alaw':\n      return { format: 'alaw', contentType: 'audio/x-alaw-basic' }\n    default:\n      return { format: 'mp3', contentType: 'audio/mpeg' }\n  }\n}\n"],"names":["sharedGenerateId"],"mappings":";;AAiCA,SAAS,iBAAwC;AAC/C,MAAI,OAAO,eAAe,aAAa;AACrC,UAAM,MAAO,WAA0C;AACvD,QAAI,KAAK,IAAK,QAAO,IAAI;AAAA,EAC3B;AACA,MAAI,OAAO,YAAY,aAAa;AAClC,WAAO,QAAQ;AAAA,EACjB;AACA,SAAO;AACT;AAEO,SAAS,6BAAqC;AACnD,SAAO,iBAAiB,oBAAoB;AAC9C;AAEO,SAAS,8BAAsC;AACpD,QAAM,KAAK,kBAAkB;AAC7B,MAAI,CAAC,IAAI;AACP,UAAM,IAAI;AAAA,MACR;AAAA,IAAA;AAAA,EAGJ;AACA,SAAO;AACT;AAOO,SAAS,uBACd,QACkB;AAClB,QAAM,SAAS,QAAQ,UAAU,2BAAA;AACjC,SAAO,IAAI,iBAAiB;AAAA,IAC1B;AAAA,IACA,GAAI,QAAQ,UAAU,EAAE,SAAS,OAAO,QAAA,IAAY,CAAA;AAAA,IACpD,GAAI,QAAQ,oBAAoB,OAC5B,EAAE,kBAAkB,OAAO,iBAAA,IAC3B,CAAA;AAAA,IACJ,GAAI,QAAQ,cAAc,OAAO,EAAE,YAAY,OAAO,WAAA,IAAe,CAAA;AAAA,IACrE,GAAI,QAAQ,UAAU,EAAE,SAAS,OAAO,QAAA,IAAY,CAAA;AAAA,EAAC,CACtD;AACH;AAIO,MAAM,aAAaA;AASnB,SAAS,oBAAoB,QAA6B;AAC/D,MAAI,OAAO,WAAW,eAAe,OAAO,OAAO,SAAS,YAAY;AACtE,WAAO,OAAO,KAAK,MAAM,EAAE,SAAS,QAAQ;AAAA,EAC9C;AACA,QAAM,OAAO,IAAI,WAAW,MAAM;AAClC,MAAI,SAAS;AACb,WAAS,IAAI,GAAG,IAAI,KAAK,YAAY,KAAK,GAAG;AAC3C,cAAU,OAAO,aAAa,KAAK,CAAC,CAAE;AAAA,EACxC;AACA,SAAO,KAAK,MAAM;AACpB;AAQA,eAAsB,wBACpB,QACsB;AACtB,QAAM,SAAS,OAAO,UAAA;AACtB,QAAM,SAA4B,CAAA;AAClC,MAAI,QAAQ;AACZ,MAAI;AACF,QAAI,SAAS,MAAM,OAAO,KAAA;AAC1B,WAAO,CAAC,OAAO,MAAM;AACnB,aAAO,KAAK,OAAO,KAAK;AACxB,eAAS,OAAO,MAAM;AACtB,eAAS,MAAM,OAAO,KAAA;AAAA,IACxB;AAAA,EACF,UAAA;AACE,WAAO,YAAA;AAAA,EACT;AACA,QAAM,SAAS,IAAI,WAAW,KAAK;AACnC,MAAI,SAAS;AACb,aAAW,SAAS,QAAQ;AAC1B,WAAO,IAAI,OAAO,MAAM;AACxB,cAAU,MAAM;AAAA,EAClB;AACA,SAAO,OAAO,OAAO;AAAA,IACnB,OAAO;AAAA,IACP,OAAO,aAAa,OAAO;AAAA,EAAA;AAE/B;AAQO,SAAS,cAAc,OAAiC;AAC7D,MAAI,CAAC,MAAM,WAAW,OAAO,EAAG,QAAO;AACvC,QAAM,aAAa,MAAM,QAAQ,GAAG;AACpC,MAAI,eAAe,GAAI,QAAO;AAE9B,QAAM,SAAS,MAAM,MAAM,GAAG,UAAU;AACxC,QAAM,UAAU,MAAM,MAAM,aAAa,CAAC;AAC1C,QAAM,WAAW,YAAY,KAAK,MAAM;AACxC,QAAM,WAAW,OAAO,MAAM,GAAG,EAAE,CAAC,KAAK;AAEzC,MAAI,UAAU;AACZ,UAAM,SAAS,KAAK,OAAO;AAC3B,UAAM,QAAQ,IAAI,WAAW,OAAO,MAAM;AAC1C,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK,GAAG;AACzC,YAAM,CAAC,IAAI,OAAO,WAAW,CAAC;AAAA,IAChC;AACA,WAAO,IAAI,KAAK,CAAC,KAAK,GAAG,EAAE,MAAM,UAAU;AAAA,EAC7C;AAEA,SAAO,IAAI,KAAK,CAAC,mBAAmB,OAAO,CAAC,GAAG,EAAE,MAAM,UAAU;AACnE;AAWO,SAAS,kBAAkB,KAGhC;AACA,QAAM,SAAS,OAAO,iBAAiB,MAAM,GAAG,EAAE,CAAC,GAAG,YAAA;AACtD,UAAQ,OAAA;AAAA,IACN,KAAK;AACH,aAAO,EAAE,QAAQ,OAAO,aAAa,aAAA;AAAA,IACvC,KAAK;AACH,aAAO,EAAE,QAAQ,OAAO,aAAa,YAAA;AAAA,IACvC,KAAK;AACH,aAAO,EAAE,QAAQ,QAAQ,aAAa,aAAA;AAAA,IACxC,KAAK;AACH,aAAO,EAAE,QAAQ,QAAQ,aAAa,cAAA;AAAA,IACxC,KAAK;AACH,aAAO,EAAE,QAAQ,QAAQ,aAAa,qBAAA;AAAA,IACxC;AACE,aAAO,EAAE,QAAQ,OAAO,aAAa,aAAA;AAAA,EAAa;AAExD;"}
+{"version":3,"file":"client.js","sources":["../../../src/utils/client.ts"],"sourcesContent":["import { ElevenLabsClient } from '@elevenlabs/elevenlabs-js'\nimport {\n  getApiKeyFromEnv,\n  generateId as sharedGenerateId,\n} from '@tanstack/ai-utils'\nimport type { ElevenLabsOutputFormat } from '../model-meta'\n\n/**\n * Configuration for any ElevenLabs adapter. When `apiKey` is omitted we read\n * `ELEVENLABS_API_KEY` from `process.env` / `window.env` to match the\n * pattern the realtime adapters already use.\n */\nexport interface ElevenLabsClientConfig {\n  apiKey?: string\n  /** Override the API base URL — handy for tests + self-hosted proxies. */\n  baseUrl?: string\n  /** Per-request timeout passed through to the SDK (ms). */\n  timeoutInSeconds?: number\n  /** Override the number of SDK-level retries. */\n  maxRetries?: number\n  /** Extra headers attached to every request (e.g. test multiplexing). */\n  headers?: Record<string, string>\n}\n\ninterface EnvObject {\n  ELEVENLABS_API_KEY?: string\n  ELEVENLABS_AGENT_ID?: string\n}\n\ninterface WindowWithEnv {\n  env?: EnvObject\n}\n\nfunction getEnvironment(): EnvObject | undefined {\n  if (typeof globalThis !== 'undefined') {\n    const win = (globalThis as { window?: WindowWithEnv }).window\n    if (win?.env) return win.env\n  }\n  if (typeof process !== 'undefined') {\n    return process.env as EnvObject\n  }\n  return undefined\n}\n\nexport function getElevenLabsApiKeyFromEnv(): string {\n  return getApiKeyFromEnv('ELEVENLABS_API_KEY')\n}\n\nexport function getElevenLabsAgentIdFromEnv(): string {\n  const id = getEnvironment()?.ELEVENLABS_AGENT_ID\n  if (!id) {\n    throw new Error(\n      'ELEVENLABS_AGENT_ID is required. Please set it in your environment ' +\n        'variables or pass `agentId` explicitly to elevenlabsRealtimeToken().',\n    )\n  }\n  return id\n}\n\n/**\n * Build an `ElevenLabsClient` with env-based or explicit credentials.\n * Each adapter calls this once at construction time so unit tests can\n * pass in an explicit key without needing `process.env`.\n */\nexport function createElevenLabsClient(\n  config?: ElevenLabsClientConfig,\n): ElevenLabsClient {\n  const apiKey = config?.apiKey ?? getElevenLabsApiKeyFromEnv()\n  return new ElevenLabsClient({\n    apiKey,\n    ...(config?.baseUrl ? { baseUrl: config.baseUrl } : {}),\n    ...(config?.timeoutInSeconds != null\n      ? { timeoutInSeconds: config.timeoutInSeconds }\n      : {}),\n    ...(config?.maxRetries != null ? { maxRetries: config.maxRetries } : {}),\n    ...(config?.headers ? { headers: config.headers } : {}),\n  })\n}\n\n// Re-exported from `@tanstack/ai-utils` so existing callers keep working\n// while the implementation stays deduped across provider packages.\nexport const generateId = sharedGenerateId\n\n/**\n * Convert an ArrayBuffer to base64 in a cross-runtime way.\n *\n * The naive `btoa(String.fromCharCode(...bytes))` form blows up V8's argument\n * limit (~65k) on realistic audio payloads, so we either use `Buffer`\n * (Node / Bun) or walk the byte array in a single loop (browser).\n */\nexport function arrayBufferToBase64(buffer: ArrayBuffer): string {\n  if (typeof Buffer !== 'undefined' && typeof Buffer.from === 'function') {\n    return Buffer.from(buffer).toString('base64')\n  }\n  const view = new Uint8Array(buffer)\n  let binary = ''\n  for (const byte of view) {\n    binary += String.fromCharCode(byte)\n  }\n  return btoa(binary)\n}\n\n/**\n * Drain a `ReadableStream<Uint8Array>` (what the ElevenLabs SDK returns for\n * audio endpoints) into a single `Uint8Array`, then expose it as an\n * `ArrayBuffer` slice. We concatenate ourselves rather than going through\n * `new Response(stream).arrayBuffer()` so we stay runtime-agnostic.\n */\nexport async function readStreamToArrayBuffer(\n  stream: ReadableStream<Uint8Array>,\n): Promise<ArrayBuffer> {\n  const reader = stream.getReader()\n  const chunks: Array<Uint8Array> = []\n  let total = 0\n  try {\n    let result = await reader.read()\n    while (!result.done) {\n      chunks.push(result.value)\n      total += result.value.byteLength\n      result = await reader.read()\n    }\n  } finally {\n    reader.releaseLock()\n  }\n  const merged = new Uint8Array(total)\n  let offset = 0\n  for (const chunk of chunks) {\n    merged.set(chunk, offset)\n    offset += chunk.byteLength\n  }\n  return merged.buffer.slice(\n    merged.byteOffset,\n    merged.byteOffset + merged.byteLength,\n  )\n}\n\n/**\n * Decode a `data:` URL into a Blob for upload to the ElevenLabs STT API\n * (which only accepts multipart files or https URLs). Supports base64 and\n * URL-encoded payloads. Returns `undefined` for non-data-URL strings so the\n * caller can fall through to treating the input as an https URL.\n */\nexport function dataUrlToBlob(value: string): Blob | undefined {\n  if (!value.startsWith('data:')) return undefined\n  const commaIndex = value.indexOf(',')\n  if (commaIndex === -1) return undefined\n\n  const header = value.slice(5, commaIndex)\n  const payload = value.slice(commaIndex + 1)\n  const isBase64 = /;base64$/i.test(header)\n  const mimeType = header.split(';')[0] || 'application/octet-stream'\n\n  if (isBase64) {\n    const binary = atob(payload)\n    const bytes = new Uint8Array(binary.length)\n    for (let i = 0; i < binary.length; i += 1) {\n      bytes[i] = binary.charCodeAt(i)\n    }\n    return new Blob([bytes], { type: mimeType })\n  }\n\n  return new Blob([decodeURIComponent(payload)], { type: mimeType })\n}\n\n/**\n * Break an ElevenLabs `output_format` string (`mp3_44100_128`,\n * `pcm_24000`, `opus_48000_64`, `ulaw_8000`, ...) into a file extension and\n * content-type suitable for `TTSResult` / `AudioGenerationResult` consumers.\n *\n * Unknown codecs fall back to `mp3` / `audio/mpeg` because the ElevenLabs\n * default is `mp3_44100_128` — mispredicting on an exotic format is safer\n * than throwing in the adapter.\n */\nexport function parseOutputFormat(fmt: ElevenLabsOutputFormat | undefined): {\n  format: string\n  contentType: string\n} {\n  const codec = (fmt || 'mp3_44100_128').split('_')[0]?.toLowerCase()\n  switch (codec) {\n    case 'mp3':\n      return { format: 'mp3', contentType: 'audio/mpeg' }\n    case 'pcm':\n      return { format: 'pcm', contentType: 'audio/pcm' }\n    case 'opus':\n      return { format: 'opus', contentType: 'audio/opus' }\n    case 'ulaw':\n      return { format: 'ulaw', contentType: 'audio/basic' }\n    case 'alaw':\n      return { format: 'alaw', contentType: 'audio/x-alaw-basic' }\n    case undefined:\n    default:\n      // `split('_')[0]?.toLowerCase()` can be `undefined` only if `fmt` is\n      // an empty string; any other unknown codec also falls back to mp3.\n      return { format: 'mp3', contentType: 'audio/mpeg' }\n  }\n}\n"],"names":["sharedGenerateId"],"mappings":";;AAiCA,SAAS,iBAAwC;AAC/C,MAAI,OAAO,eAAe,aAAa;AACrC,UAAM,MAAO,WAA0C;AACvD,QAAI,KAAK,IAAK,QAAO,IAAI;AAAA,EAC3B;AACA,MAAI,OAAO,YAAY,aAAa;AAClC,WAAO,QAAQ;AAAA,EACjB;AACA,SAAO;AACT;AAEO,SAAS,6BAAqC;AACnD,SAAO,iBAAiB,oBAAoB;AAC9C;AAEO,SAAS,8BAAsC;AACpD,QAAM,KAAK,kBAAkB;AAC7B,MAAI,CAAC,IAAI;AACP,UAAM,IAAI;AAAA,MACR;AAAA,IAAA;AAAA,EAGJ;AACA,SAAO;AACT;AAOO,SAAS,uBACd,QACkB;AAClB,QAAM,SAAS,QAAQ,UAAU,2BAAA;AACjC,SAAO,IAAI,iBAAiB;AAAA,IAC1B;AAAA,IACA,GAAI,QAAQ,UAAU,EAAE,SAAS,OAAO,QAAA,IAAY,CAAA;AAAA,IACpD,GAAI,QAAQ,oBAAoB,OAC5B,EAAE,kBAAkB,OAAO,iBAAA,IAC3B,CAAA;AAAA,IACJ,GAAI,QAAQ,cAAc,OAAO,EAAE,YAAY,OAAO,WAAA,IAAe,CAAA;AAAA,IACrE,GAAI,QAAQ,UAAU,EAAE,SAAS,OAAO,QAAA,IAAY,CAAA;AAAA,EAAC,CACtD;AACH;AAIO,MAAM,aAAaA;AASnB,SAAS,oBAAoB,QAA6B;AAC/D,MAAI,OAAO,WAAW,eAAe,OAAO,OAAO,SAAS,YAAY;AACtE,WAAO,OAAO,KAAK,MAAM,EAAE,SAAS,QAAQ;AAAA,EAC9C;AACA,QAAM,OAAO,IAAI,WAAW,MAAM;AAClC,MAAI,SAAS;AACb,aAAW,QAAQ,MAAM;AACvB,cAAU,OAAO,aAAa,IAAI;AAAA,EACpC;AACA,SAAO,KAAK,MAAM;AACpB;AAQA,eAAsB,wBACpB,QACsB;AACtB,QAAM,SAAS,OAAO,UAAA;AACtB,QAAM,SAA4B,CAAA;AAClC,MAAI,QAAQ;AACZ,MAAI;AACF,QAAI,SAAS,MAAM,OAAO,KAAA;AAC1B,WAAO,CAAC,OAAO,MAAM;AACnB,aAAO,KAAK,OAAO,KAAK;AACxB,eAAS,OAAO,MAAM;AACtB,eAAS,MAAM,OAAO,KAAA;AAAA,IACxB;AAAA,EACF,UAAA;AACE,WAAO,YAAA;AAAA,EACT;AACA,QAAM,SAAS,IAAI,WAAW,KAAK;AACnC,MAAI,SAAS;AACb,aAAW,SAAS,QAAQ;AAC1B,WAAO,IAAI,OAAO,MAAM;AACxB,cAAU,MAAM;AAAA,EAClB;AACA,SAAO,OAAO,OAAO;AAAA,IACnB,OAAO;AAAA,IACP,OAAO,aAAa,OAAO;AAAA,EAAA;AAE/B;AAQO,SAAS,cAAc,OAAiC;AAC7D,MAAI,CAAC,MAAM,WAAW,OAAO,EAAG,QAAO;AACvC,QAAM,aAAa,MAAM,QAAQ,GAAG;AACpC,MAAI,eAAe,GAAI,QAAO;AAE9B,QAAM,SAAS,MAAM,MAAM,GAAG,UAAU;AACxC,QAAM,UAAU,MAAM,MAAM,aAAa,CAAC;AAC1C,QAAM,WAAW,YAAY,KAAK,MAAM;AACxC,QAAM,WAAW,OAAO,MAAM,GAAG,EAAE,CAAC,KAAK;AAEzC,MAAI,UAAU;AACZ,UAAM,SAAS,KAAK,OAAO;AAC3B,UAAM,QAAQ,IAAI,WAAW,OAAO,MAAM;AAC1C,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK,GAAG;AACzC,YAAM,CAAC,IAAI,OAAO,WAAW,CAAC;AAAA,IAChC;AACA,WAAO,IAAI,KAAK,CAAC,KAAK,GAAG,EAAE,MAAM,UAAU;AAAA,EAC7C;AAEA,SAAO,IAAI,KAAK,CAAC,mBAAmB,OAAO,CAAC,GAAG,EAAE,MAAM,UAAU;AACnE;AAWO,SAAS,kBAAkB,KAGhC;AACA,QAAM,SAAS,OAAO,iBAAiB,MAAM,GAAG,EAAE,CAAC,GAAG,YAAA;AACtD,UAAQ,OAAA;AAAA,IACN,KAAK;AACH,aAAO,EAAE,QAAQ,OAAO,aAAa,aAAA;AAAA,IACvC,KAAK;AACH,aAAO,EAAE,QAAQ,OAAO,aAAa,YAAA;AAAA,IACvC,KAAK;AACH,aAAO,EAAE,QAAQ,QAAQ,aAAa,aAAA;AAAA,IACxC,KAAK;AACH,aAAO,EAAE,QAAQ,QAAQ,aAAa,cAAA;AAAA,IACxC,KAAK;AACH,aAAO,EAAE,QAAQ,QAAQ,aAAa,qBAAA;AAAA,IACxC,KAAK;AAAA,IACL;AAGE,aAAO,EAAE,QAAQ,OAAO,aAAa,aAAA;AAAA,EAAa;AAExD;"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tanstack/ai-elevenlabs",
-  "version": "0.2.7",
+  "version": "0.2.9",
   "description": "ElevenLabs adapter for TanStack AI realtime voice",
   "author": "",
   "license": "MIT",
@@ -43,13 +43,13 @@
     "@tanstack/ai-utils": "0.2.0"
   },
   "peerDependencies": {
-    "@tanstack/ai": "^0.20.0",
-    "@tanstack/ai-client": "^0.11.2"
+    "@tanstack/ai": "^0.21.0",
+    "@tanstack/ai-client": "^0.11.4"
   },
   "devDependencies": {
     "@vitest/coverage-v8": "4.0.14",
-    "@tanstack/ai": "0.20.0",
-    "@tanstack/ai-client": "0.11.2"
+    "@tanstack/ai": "0.21.0",
+    "@tanstack/ai-client": "0.11.4"
   },
   "scripts": {
     "build": "vite build",

package/src/adapters/audio.ts

@@ -106,7 +106,7 @@ export class ElevenLabsAudioAdapter<
 > extends BaseAudioAdapter<TModel, ElevenLabsAudioProviderOptions> {
   readonly name = 'elevenlabs' as const
 
-  private client: ElevenLabsClient
+  private readonly client: ElevenLabsClient
 
   constructor(model: TModel, config?: ElevenLabsClientConfig) {
     super(model, config ?? {})

package/src/adapters/speech.ts

@@ -82,7 +82,7 @@ export class ElevenLabsSpeechAdapter<
 > extends BaseTTSAdapter<TModel, ElevenLabsSpeechProviderOptions> {
   readonly name = 'elevenlabs' as const
 
-  private client: ElevenLabsClient
+  private readonly client: ElevenLabsClient
 
   constructor(model: TModel, config?: ElevenLabsClientConfig) {
     super(model, config ?? {})
@@ -211,6 +211,9 @@ function inferOutputFormatFromResponseFormat(
       return 'opus_48000_128'
     case undefined:
       return undefined
+    case 'aac':
+    case 'flac':
+    case 'wav':
     default:
       // `aac` / `flac` / `wav` are not native ElevenLabs formats —
       // fall back to mp3 rather than blowing up mid-request.

package/src/adapters/transcription.ts

@@ -82,7 +82,7 @@ export class ElevenLabsTranscriptionAdapter<
 > {
   readonly name = 'elevenlabs' as const
 
-  private client: ElevenLabsClient
+  private readonly client: ElevenLabsClient
 
   constructor(model: TModel, config?: ElevenLabsClientConfig) {
     super(model, config ?? {})
@@ -165,6 +165,7 @@ export class ElevenLabsTranscriptionAdapter<
     // The SDK types this as a union of single- and multi-channel responses.
     // We treat multi-channel as "join the channel transcripts" — consumers
     // who care about per-channel detail can re-parse from `modelOptions`.
+    // 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
     const data = response as unknown as {
       text?: string
       languageCode?: string
@@ -261,7 +262,7 @@ function buildWordsAndSegments(
   segments?: Array<TranscriptionSegment>
 } {
   const timedWords = words.filter(
-    (w) =>
+    (w): w is typeof w & { start: number; end: number } =>
       typeof w.start === 'number' &&
       typeof w.end === 'number' &&
       w.type !== 'spacing',
@@ -270,8 +271,8 @@ function buildWordsAndSegments(
 
   const outWords: Array<TranscriptionWord> = timedWords.map((w) => ({
     word: w.text,
-    start: w.start!,
-    end: w.end!,
+    start: w.start,
+    end: w.end,
   }))
 
   // Group contiguous words that share a speaker into segments. If no speaker
@@ -287,8 +288,8 @@ function buildWordsAndSegments(
   for (const w of timedWords) {
     if (!current) {
       current = {
-        start: w.start!,
-        end: w.end!,
+        start: w.start,
+        end: w.end,
         text: w.text,
         ...(w.speakerId ? { speaker: w.speakerId } : {}),
       }
@@ -297,14 +298,14 @@ function buildWordsAndSegments(
     if (w.speakerId && current.speaker !== w.speakerId) {
       segments.push({ id: segments.length, ...current })
       current = {
-        start: w.start!,
-        end: w.end!,
+        start: w.start,
+        end: w.end,
         text: w.text,
         speaker: w.speakerId,
       }
       continue
     }
-    current.end = w.end!
+    current.end = w.end
     current.text = current.text ? `${current.text} ${w.text}` : w.text
   }
   if (current) segments.push({ id: segments.length, ...current })

package/src/realtime/adapter.ts

@@ -8,7 +8,6 @@ import type {
   RealtimeMessage,
   RealtimeMode,
   RealtimeSessionConfig,
-  RealtimeStatus,
   RealtimeToken,
 } from '@tanstack/ai'
 import type { InternalLogger } from '@tanstack/ai/adapter-internals'
@@ -119,7 +118,7 @@ async function createElevenLabsConnection(
       logger.provider(`provider=elevenlabs direction=in type=connect`, {
         frame: { type: 'connect' },
       })
-      emit('status_change', { status: 'connected' as RealtimeStatus })
+      emit('status_change', { status: 'connected' })
       emit('mode_change', { mode: 'listening' })
     },
 
@@ -127,7 +126,7 @@ async function createElevenLabsConnection(
       logger.provider(`provider=elevenlabs direction=in type=disconnect`, {
         frame: { type: 'disconnect' },
       })
-      emit('status_change', { status: 'idle' as RealtimeStatus })
+      emit('status_change', { status: 'idle' })
       emit('mode_change', { mode: 'idle' })
     },
 
@@ -205,7 +204,7 @@ async function createElevenLabsConnection(
         await conversation.endSession()
         conversation = null
       }
-      emit('status_change', { status: 'idle' as RealtimeStatus })
+      emit('status_change', { status: 'idle' })
     },
 
     async startAudioCapture() {
@@ -257,10 +256,12 @@ async function createElevenLabsConnection(
       event: TEvent,
       handler: RealtimeEventHandler<TEvent>,
     ): () => void {
-      if (!eventHandlers.has(event)) {
-        eventHandlers.set(event, new Set())
+      let handlers = eventHandlers.get(event)
+      if (!handlers) {
+        handlers = new Set()
+        eventHandlers.set(event, handlers)
       }
-      eventHandlers.get(event)!.add(handler)
+      handlers.add(handler)
 
       return () => {
         eventHandlers.get(event)?.delete(handler)

package/src/realtime/token.ts

@@ -49,17 +49,28 @@ export function elevenlabsRealtimeToken(
       // Signed URLs are valid for 30 minutes
       const expiresAt = Date.now() + 30 * 60 * 1000
 
+      // Conditionally spread optional fields — the vendor target type
+      // (`RealtimeSessionConfig` in `@tanstack/ai`) uses strict
+      // `field?: T` semantics under `exactOptionalPropertyTypes`, so we
+      // omit keys entirely when the override is absent rather than
+      // setting them to `undefined`.
       return {
         provider: 'elevenlabs',
         token: response.signedUrl,
         expiresAt,
         config: {
-          voice: overrides?.voiceId,
-          instructions: overrides?.systemPrompt,
+          ...(overrides?.voiceId !== undefined && { voice: overrides.voiceId }),
+          ...(overrides?.systemPrompt !== undefined && {
+            instructions: overrides.systemPrompt,
+          }),
           providerOptions: {
             agentId,
-            firstMessage: overrides?.firstMessage,
-            language: overrides?.language,
+            ...(overrides?.firstMessage !== undefined && {
+              firstMessage: overrides.firstMessage,
+            }),
+            ...(overrides?.language !== undefined && {
+              language: overrides.language,
+            }),
           },
         },
       }

package/src/utils/client.ts

@@ -94,8 +94,8 @@ export function arrayBufferToBase64(buffer: ArrayBuffer): string {
   }
   const view = new Uint8Array(buffer)
   let binary = ''
-  for (let i = 0; i < view.byteLength; i += 1) {
-    binary += String.fromCharCode(view[i]!)
+  for (const byte of view) {
+    binary += String.fromCharCode(byte)
   }
   return btoa(binary)
 }
@@ -187,7 +187,10 @@ export function parseOutputFormat(fmt: ElevenLabsOutputFormat | undefined): {
       return { format: 'ulaw', contentType: 'audio/basic' }
     case 'alaw':
       return { format: 'alaw', contentType: 'audio/x-alaw-basic' }
+    case undefined:
     default:
+      // `split('_')[0]?.toLowerCase()` can be `undefined` only if `fmt` is
+      // an empty string; any other unknown codec also falls back to mp3.
       return { format: 'mp3', contentType: 'audio/mpeg' }
   }
 }