@tanstack/ai-elevenlabs 0.2.33 → 0.2.35

package/dist/esm/adapters/transcription.js

@@ -1,143 +1,176 @@
+import { createElevenLabsClient, dataUrlToBlob, generateId } from "../utils/client.js";
 import { BaseTranscriptionAdapter } from "@tanstack/ai/adapters";
-import { createElevenLabsClient, generateId, dataUrlToBlob } from "../utils/client.js";
-class ElevenLabsTranscriptionAdapter extends BaseTranscriptionAdapter {
-  name = "elevenlabs";
-  client;
-  constructor(model, config) {
-    super(model, config ?? {});
-    this.client = createElevenLabsClient(config);
-  }
-  async transcribe(options) {
-    const { logger } = options;
-    logger.request(
-      `activity=generateTranscription provider=elevenlabs model=${this.model}`,
-      { provider: "elevenlabs", model: this.model }
-    );
-    try {
-      const modelOpts = options.modelOptions ?? {};
-      const audioInput = normalizeAudioInput(options.audio);
-      const response = await this.client.speechToText.convert({
-        modelId: this.model,
-        ...audioInput.kind === "file" ? { file: audioInput.value } : { cloudStorageUrl: audioInput.value },
-        ...options.language ? { languageCode: options.language } : {},
-        ...modelOpts.tagAudioEvents != null ? { tagAudioEvents: modelOpts.tagAudioEvents } : {},
-        ...modelOpts.numSpeakers != null ? { numSpeakers: modelOpts.numSpeakers } : {},
-        ...modelOpts.timestampsGranularity ? { timestampsGranularity: modelOpts.timestampsGranularity } : {},
-        ...modelOpts.diarize != null ? { diarize: modelOpts.diarize } : {},
-        ...modelOpts.diarizationThreshold != null ? { diarizationThreshold: modelOpts.diarizationThreshold } : {},
-        ...modelOpts.detectSpeakerRoles != null ? { detectSpeakerRoles: modelOpts.detectSpeakerRoles } : {},
-        ...modelOpts.keyterms ? { keyterms: modelOpts.keyterms } : {},
-        ...modelOpts.entityDetection ? { entityDetection: modelOpts.entityDetection } : {},
-        ...modelOpts.entityRedaction ? { entityRedaction: modelOpts.entityRedaction } : {},
-        ...modelOpts.entityRedactionMode ? { entityRedactionMode: modelOpts.entityRedactionMode } : {},
-        ...modelOpts.noVerbatim != null ? { noVerbatim: modelOpts.noVerbatim } : {},
-        ...modelOpts.temperature != null ? { temperature: modelOpts.temperature } : {},
-        ...modelOpts.seed != null ? { seed: modelOpts.seed } : {},
-        ...modelOpts.enableLogging != null ? { enableLogging: modelOpts.enableLogging } : {},
-        ...modelOpts.useMultiChannel != null ? { useMultiChannel: modelOpts.useMultiChannel } : {},
-        ...modelOpts.fileFormat ? { fileFormat: modelOpts.fileFormat } : {}
-      });
-      return this.transformResponse(response);
-    } catch (error) {
-      logger.errors("elevenlabs.generateTranscription fatal", {
-        error,
-        source: "elevenlabs.generateTranscription"
-      });
-      throw error;
-    }
-  }
-  transformResponse(response) {
-    const data = response;
-    if (data.transcripts) {
-      const joinedText = data.transcripts.map((t) => t.text ?? "").filter(Boolean).join("\n");
-      const joinedWords = data.transcripts.flatMap((t) => t.words ?? []);
-      const duration = data.transcripts.reduce(
-        (max, t) => Math.max(max, t.audioDurationSecs ?? 0),
-        0
-      );
-      const firstLang = data.transcripts.find(
-        (t) => t.languageCode
-      )?.languageCode;
-      return {
-        id: generateId(this.name),
-        model: this.model,
-        text: joinedText,
-        ...firstLang ? { language: firstLang } : {},
-        ...duration ? { duration } : {},
-        ...buildWordsAndSegments(joinedWords)
-      };
-    }
-    return {
-      id: generateId(this.name),
-      model: this.model,
-      text: data.text ?? "",
-      ...data.languageCode ? { language: data.languageCode } : {},
-      ...data.audioDurationSecs ? { duration: data.audioDurationSecs } : {},
-      ...buildWordsAndSegments(data.words ?? [])
-    };
-  }
-  generateId() {
-    return generateId(this.name);
-  }
-}
+//#region src/adapters/transcription.ts
+/**
+* ElevenLabs speech-to-text adapter built on the official SDK's Scribe family.
+*
+* @example
+* ```ts
+* const adapter = elevenlabsTranscription('scribe_v1')
+* const result = await generateTranscription({
+*   adapter,
+*   audio: fileInput,
+*   language: 'en',
+* })
+* ```
+*/
+var ElevenLabsTranscriptionAdapter = class extends BaseTranscriptionAdapter {
+	name = "elevenlabs";
+	client;
+	constructor(model, config) {
+		super(model, config ?? {});
+		this.client = createElevenLabsClient(config);
+	}
+	async transcribe(options) {
+		const { logger } = options;
+		logger.request(`activity=generateTranscription provider=elevenlabs model=${this.model}`, {
+			provider: "elevenlabs",
+			model: this.model
+		});
+		try {
+			const modelOpts = options.modelOptions ?? {};
+			const audioInput = normalizeAudioInput(options.audio);
+			const response = await this.client.speechToText.convert({
+				modelId: this.model,
+				...audioInput.kind === "file" ? { file: audioInput.value } : { cloudStorageUrl: audioInput.value },
+				...options.language ? { languageCode: options.language } : {},
+				...modelOpts.tagAudioEvents != null ? { tagAudioEvents: modelOpts.tagAudioEvents } : {},
+				...modelOpts.numSpeakers != null ? { numSpeakers: modelOpts.numSpeakers } : {},
+				...modelOpts.timestampsGranularity ? { timestampsGranularity: modelOpts.timestampsGranularity } : {},
+				...modelOpts.diarize != null ? { diarize: modelOpts.diarize } : {},
+				...modelOpts.diarizationThreshold != null ? { diarizationThreshold: modelOpts.diarizationThreshold } : {},
+				...modelOpts.detectSpeakerRoles != null ? { detectSpeakerRoles: modelOpts.detectSpeakerRoles } : {},
+				...modelOpts.keyterms ? { keyterms: modelOpts.keyterms } : {},
+				...modelOpts.entityDetection ? { entityDetection: modelOpts.entityDetection } : {},
+				...modelOpts.entityRedaction ? { entityRedaction: modelOpts.entityRedaction } : {},
+				...modelOpts.entityRedactionMode ? { entityRedactionMode: modelOpts.entityRedactionMode } : {},
+				...modelOpts.noVerbatim != null ? { noVerbatim: modelOpts.noVerbatim } : {},
+				...modelOpts.temperature != null ? { temperature: modelOpts.temperature } : {},
+				...modelOpts.seed != null ? { seed: modelOpts.seed } : {},
+				...modelOpts.enableLogging != null ? { enableLogging: modelOpts.enableLogging } : {},
+				...modelOpts.useMultiChannel != null ? { useMultiChannel: modelOpts.useMultiChannel } : {},
+				...modelOpts.fileFormat ? { fileFormat: modelOpts.fileFormat } : {}
+			});
+			return this.transformResponse(response);
+		} catch (error) {
+			logger.errors("elevenlabs.generateTranscription fatal", {
+				error,
+				source: "elevenlabs.generateTranscription"
+			});
+			throw error;
+		}
+	}
+	transformResponse(response) {
+		const data = response;
+		if (data.transcripts) {
+			const joinedText = data.transcripts.map((t) => t.text ?? "").filter(Boolean).join("\n");
+			const joinedWords = data.transcripts.flatMap((t) => t.words ?? []);
+			const duration = data.transcripts.reduce((max, t) => Math.max(max, t.audioDurationSecs ?? 0), 0);
+			const firstLang = data.transcripts.find((t) => t.languageCode)?.languageCode;
+			return {
+				id: generateId(this.name),
+				model: this.model,
+				text: joinedText,
+				...firstLang ? { language: firstLang } : {},
+				...duration ? { duration } : {},
+				...buildWordsAndSegments(joinedWords)
+			};
+		}
+		return {
+			id: generateId(this.name),
+			model: this.model,
+			text: data.text ?? "",
+			...data.languageCode ? { language: data.languageCode } : {},
+			...data.audioDurationSecs ? { duration: data.audioDurationSecs } : {},
+			...buildWordsAndSegments(data.words ?? [])
+		};
+	}
+	generateId() {
+		return generateId(this.name);
+	}
+};
 function normalizeAudioInput(audio) {
-  if (audio instanceof ArrayBuffer) {
-    return { kind: "file", value: new Blob([audio]) };
-  }
-  if (typeof audio === "string") {
-    const blob = dataUrlToBlob(audio);
-    if (blob) return { kind: "file", value: blob };
-    return { kind: "url", value: audio };
-  }
-  return { kind: "file", value: audio };
+	if (audio instanceof ArrayBuffer) return {
+		kind: "file",
+		value: new Blob([audio])
+	};
+	if (typeof audio === "string") {
+		const blob = dataUrlToBlob(audio);
+		if (blob) return {
+			kind: "file",
+			value: blob
+		};
+		return {
+			kind: "url",
+			value: audio
+		};
+	}
+	return {
+		kind: "file",
+		value: audio
+	};
 }
 function buildWordsAndSegments(words) {
-  const timedWords = words.filter(
-    (w) => typeof w.start === "number" && typeof w.end === "number" && w.type !== "spacing"
-  );
-  if (timedWords.length === 0) return {};
-  const outWords = timedWords.map((w) => ({
-    word: w.text,
-    start: w.start,
-    end: w.end
-  }));
-  const segments = [];
-  let current = null;
-  for (const w of timedWords) {
-    if (!current) {
-      current = {
-        start: w.start,
-        end: w.end,
-        text: w.text,
-        ...w.speakerId ? { speaker: w.speakerId } : {}
-      };
-      continue;
-    }
-    if (w.speakerId && current.speaker !== w.speakerId) {
-      segments.push({ id: segments.length, ...current });
-      current = {
-        start: w.start,
-        end: w.end,
-        text: w.text,
-        speaker: w.speakerId
-      };
-      continue;
-    }
-    current.end = w.end;
-    current.text = current.text ? `${current.text} ${w.text}` : w.text;
-  }
-  if (current) segments.push({ id: segments.length, ...current });
-  return { words: outWords, segments };
+	const timedWords = words.filter((w) => typeof w.start === "number" && typeof w.end === "number" && w.type !== "spacing");
+	if (timedWords.length === 0) return {};
+	const outWords = timedWords.map((w) => ({
+		word: w.text,
+		start: w.start,
+		end: w.end
+	}));
+	const segments = [];
+	let current = null;
+	for (const w of timedWords) {
+		if (!current) {
+			current = {
+				start: w.start,
+				end: w.end,
+				text: w.text,
+				...w.speakerId ? { speaker: w.speakerId } : {}
+			};
+			continue;
+		}
+		if (w.speakerId && current.speaker !== w.speakerId) {
+			segments.push({
+				id: segments.length,
+				...current
+			});
+			current = {
+				start: w.start,
+				end: w.end,
+				text: w.text,
+				speaker: w.speakerId
+			};
+			continue;
+		}
+		current.end = w.end;
+		current.text = current.text ? `${current.text} ${w.text}` : w.text;
+	}
+	if (current) segments.push({
+		id: segments.length,
+		...current
+	});
+	return {
+		words: outWords,
+		segments
+	};
 }
+/**
+* Create an ElevenLabs transcription adapter using `ELEVENLABS_API_KEY` from env.
+*/
 function elevenlabsTranscription(model, config) {
-  return new ElevenLabsTranscriptionAdapter(model, config);
+	return new ElevenLabsTranscriptionAdapter(model, config);
 }
+/**
+* Create an ElevenLabs transcription adapter with an explicit API key.
+*/
 function createElevenLabsTranscription(model, apiKey, config) {
-  return new ElevenLabsTranscriptionAdapter(model, { apiKey, ...config });
+	return new ElevenLabsTranscriptionAdapter(model, {
+		apiKey,
+		...config
+	});
 }
-export {
-  ElevenLabsTranscriptionAdapter,
-  createElevenLabsTranscription,
-  elevenlabsTranscription
-};
-//# sourceMappingURL=transcription.js.map
+//#endregion
+export { ElevenLabsTranscriptionAdapter, createElevenLabsTranscription, elevenlabsTranscription };
+
+//# sourceMappingURL=transcription.js.map

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 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;"}
+{"version":3,"file":"transcription.js","names":[],"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    // oxlint-disable-next-line eslint-js/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"],"mappings":";;;;;;;;;;;;;;;;AA4EA,IAAa,iCAAb,cAEU,yBAGR;CACA,OAAgB;CAEhB;CAEA,YAAY,OAAe,QAAiC;EAC1D,MAAM,OAAO,UAAU,CAAC,CAAC;EACzB,KAAK,SAAS,uBAAuB,MAAM;CAC7C;CAEA,MAAM,WACJ,SAC8B;EAC9B,MAAM,EAAE,WAAW;EACnB,OAAO,QACL,4DAA4D,KAAK,SACjE;GAAE,UAAU;GAAc,OAAO,KAAK;EAAM,CAC9C;EACA,IAAI;GACF,MAAM,YAAY,QAAQ,gBAAgB,CAAC;GAC3C,MAAM,aAAa,oBAAoB,QAAQ,KAAK;GAEpD,MAAM,WAAW,MAAM,KAAK,OAAO,aAAa,QAAQ;IACtD,SAAS,KAAK;IACd,GAAI,WAAW,SAAS,SACpB,EAAE,MAAM,WAAW,MAAM,IACzB,EAAE,iBAAiB,WAAW,MAAM;IACxC,GAAI,QAAQ,WAAW,EAAE,cAAc,QAAQ,SAAS,IAAI,CAAC;IAC7D,GAAI,UAAU,kBAAkB,OAC5B,EAAE,gBAAgB,UAAU,eAAe,IAC3C,CAAC;IACL,GAAI,UAAU,eAAe,OACzB,EAAE,aAAa,UAAU,YAAY,IACrC,CAAC;IACL,GAAI,UAAU,wBACV,EAAE,uBAAuB,UAAU,sBAAsB,IACzD,CAAC;IACL,GAAI,UAAU,WAAW,OAAO,EAAE,SAAS,UAAU,QAAQ,IAAI,CAAC;IAClE,GAAI,UAAU,wBAAwB,OAClC,EAAE,sBAAsB,UAAU,qBAAqB,IACvD,CAAC;IACL,GAAI,UAAU,sBAAsB,OAChC,EAAE,oBAAoB,UAAU,mBAAmB,IACnD,CAAC;IACL,GAAI,UAAU,WAAW,EAAE,UAAU,UAAU,SAAS,IAAI,CAAC;IAC7D,GAAI,UAAU,kBACV,EAAE,iBAAiB,UAAU,gBAAgB,IAC7C,CAAC;IACL,GAAI,UAAU,kBACV,EAAE,iBAAiB,UAAU,gBAAgB,IAC7C,CAAC;IACL,GAAI,UAAU,sBACV,EAAE,qBAAqB,UAAU,oBAAoB,IACrD,CAAC;IACL,GAAI,UAAU,cAAc,OACxB,EAAE,YAAY,UAAU,WAAW,IACnC,CAAC;IACL,GAAI,UAAU,eAAe,OACzB,EAAE,aAAa,UAAU,YAAY,IACrC,CAAC;IACL,GAAI,UAAU,QAAQ,OAAO,EAAE,MAAM,UAAU,KAAK,IAAI,CAAC;IACzD,GAAI,UAAU,iBAAiB,OAC3B,EAAE,eAAe,UAAU,cAAc,IACzC,CAAC;IACL,GAAI,UAAU,mBAAmB,OAC7B,EAAE,iBAAiB,UAAU,gBAAgB,IAC7C,CAAC;IACL,GAAI,UAAU,aAAa,EAAE,YAAY,UAAU,WAAW,IAAI,CAAC;GACrE,CAA+D;GAE/D,OAAO,KAAK,kBAAkB,QAAQ;EACxC,SAAS,OAAO;GACd,OAAO,OAAO,0CAA0C;IACtD;IACA,QAAQ;GACV,CAAC;GACD,MAAM;EACR;CACF;CAEA,kBACE,UACqB;EAKrB,MAAM,OAAO;EA0Bb,IAAI,KAAK,aAAa;GACpB,MAAM,aAAa,KAAK,YACrB,KAAK,MAAM,EAAE,QAAQ,EAAE,CAAC,CACxB,OAAO,OAAO,CAAC,CACf,KAAK,IAAI;GACZ,MAAM,cAAc,KAAK,YAAY,SAAS,MAAM,EAAE,SAAS,CAAC,CAAC;GACjE,MAAM,WAAW,KAAK,YAAY,QAC/B,KAAK,MAAM,KAAK,IAAI,KAAK,EAAE,qBAAqB,CAAC,GAClD,CACF;GACA,MAAM,YAAY,KAAK,YAAY,MAChC,MAAM,EAAE,YACX,CAAC,EAAE;GACH,OAAO;IACL,IAAI,WAAW,KAAK,IAAI;IACxB,OAAO,KAAK;IACZ,MAAM;IACN,GAAI,YAAY,EAAE,UAAU,UAAU,IAAI,CAAC;IAC3C,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;IAC/B,GAAG,sBAAsB,WAAW;GACtC;EACF;EAEA,OAAO;GACL,IAAI,WAAW,KAAK,IAAI;GACxB,OAAO,KAAK;GACZ,MAAM,KAAK,QAAQ;GACnB,GAAI,KAAK,eAAe,EAAE,UAAU,KAAK,aAAa,IAAI,CAAC;GAC3D,GAAI,KAAK,oBAAoB,EAAE,UAAU,KAAK,kBAAkB,IAAI,CAAC;GACrE,GAAG,sBAAsB,KAAK,SAAS,CAAC,CAAC;EAC3C;CACF;CAEA,aAAwC;EACtC,OAAO,WAAW,KAAK,IAAI;CAC7B;AACF;AAMA,SAAS,oBACP,OACiB;CACjB,IAAI,iBAAiB,aACnB,OAAO;EAAE,MAAM;EAAQ,OAAO,IAAI,KAAK,CAAC,KAAK,CAAC;CAAE;CAElD,IAAI,OAAO,UAAU,UAAU;EAC7B,MAAM,OAAO,cAAc,KAAK;EAChC,IAAI,MAAM,OAAO;GAAE,MAAM;GAAQ,OAAO;EAAK;EAC7C,OAAO;GAAE,MAAM;GAAO,OAAO;EAAM;CACrC;CAEA,OAAO;EAAE,MAAM;EAAQ,OAAO;CAAM;AACtC;AAEA,SAAS,sBACP,OAUA;CACA,MAAM,aAAa,MAAM,QACtB,MACC,OAAO,EAAE,UAAU,YACnB,OAAO,EAAE,QAAQ,YACjB,EAAE,SAAS,SACf;CACA,IAAI,WAAW,WAAW,GAAG,OAAO,CAAC;CAErC,MAAM,WAAqC,WAAW,KAAK,OAAO;EAChE,MAAM,EAAE;EACR,OAAO,EAAE;EACT,KAAK,EAAE;CACT,EAAE;CAIF,MAAM,WAAwC,CAAC;CAC/C,IAAI,UAKO;CAEX,KAAK,MAAM,KAAK,YAAY;EAC1B,IAAI,CAAC,SAAS;GACZ,UAAU;IACR,OAAO,EAAE;IACT,KAAK,EAAE;IACP,MAAM,EAAE;IACR,GAAI,EAAE,YAAY,EAAE,SAAS,EAAE,UAAU,IAAI,CAAC;GAChD;GACA;EACF;EACA,IAAI,EAAE,aAAa,QAAQ,YAAY,EAAE,WAAW;GAClD,SAAS,KAAK;IAAE,IAAI,SAAS;IAAQ,GAAG;GAAQ,CAAC;GACjD,UAAU;IACR,OAAO,EAAE;IACT,KAAK,EAAE;IACP,MAAM,EAAE;IACR,SAAS,EAAE;GACb;GACA;EACF;EACA,QAAQ,MAAM,EAAE;EAChB,QAAQ,OAAO,QAAQ,OAAO,GAAG,QAAQ,KAAK,GAAG,EAAE,SAAS,EAAE;CAChE;CACA,IAAI,SAAS,SAAS,KAAK;EAAE,IAAI,SAAS;EAAQ,GAAG;CAAQ,CAAC;CAE9D,OAAO;EAAE,OAAO;EAAU;CAAS;AACrC;;;;AAKA,SAAgB,wBAGd,OACA,QACwC;CACxC,OAAO,IAAI,+BAA+B,OAAO,MAAM;AACzD;;;;AAKA,SAAgB,8BAGd,OACA,QACA,QACwC;CACxC,OAAO,IAAI,+BAA+B,OAAO;EAAE;EAAQ,GAAG;CAAO,CAAC;AACxE"}

package/dist/esm/index.js

@@ -1,27 +1,10 @@
+import { getElevenLabsApiKeyFromEnv } from "./utils/client.js";
 import { elevenlabsRealtimeToken } from "./realtime/token.js";
 import { elevenlabsRealtime } from "./realtime/adapter.js";
+import "./realtime/index.js";
 import { ElevenLabsSpeechAdapter, createElevenLabsSpeech, elevenlabsSpeech } from "./adapters/speech.js";
+import { ELEVENLABS_AUDIO_MODELS, ELEVENLABS_TRANSCRIPTION_MODELS, ELEVENLABS_TTS_MODELS, isElevenLabsMusicModel, isElevenLabsSoundEffectsModel } from "./model-meta.js";
 import { ElevenLabsAudioAdapter, createElevenLabsAudio, elevenlabsAudio } from "./adapters/audio.js";
 import { ElevenLabsTranscriptionAdapter, createElevenLabsTranscription, elevenlabsTranscription } from "./adapters/transcription.js";
-import { ELEVENLABS_AUDIO_MODELS, ELEVENLABS_TRANSCRIPTION_MODELS, ELEVENLABS_TTS_MODELS, isElevenLabsMusicModel, isElevenLabsSoundEffectsModel } from "./model-meta.js";
-import { getElevenLabsApiKeyFromEnv } from "./utils/client.js";
-export {
-  ELEVENLABS_AUDIO_MODELS,
-  ELEVENLABS_TRANSCRIPTION_MODELS,
-  ELEVENLABS_TTS_MODELS,
-  ElevenLabsAudioAdapter,
-  ElevenLabsSpeechAdapter,
-  ElevenLabsTranscriptionAdapter,
-  createElevenLabsAudio,
-  createElevenLabsSpeech,
-  createElevenLabsTranscription,
-  elevenlabsAudio,
-  elevenlabsRealtime,
-  elevenlabsRealtimeToken,
-  elevenlabsSpeech,
-  elevenlabsTranscription,
-  getElevenLabsApiKeyFromEnv,
-  isElevenLabsMusicModel,
-  isElevenLabsSoundEffectsModel
-};
-//# sourceMappingURL=index.js.map
+import "./utils/index.js";
+export { ELEVENLABS_AUDIO_MODELS, ELEVENLABS_TRANSCRIPTION_MODELS, ELEVENLABS_TTS_MODELS, ElevenLabsAudioAdapter, ElevenLabsSpeechAdapter, ElevenLabsTranscriptionAdapter, createElevenLabsAudio, createElevenLabsSpeech, createElevenLabsTranscription, elevenlabsAudio, elevenlabsRealtime, elevenlabsRealtimeToken, elevenlabsSpeech, elevenlabsTranscription, getElevenLabsApiKeyFromEnv, isElevenLabsMusicModel, isElevenLabsSoundEffectsModel };

package/dist/esm/model-meta.js

@@ -1,32 +1,47 @@
-const ELEVENLABS_TTS_MODELS = [
-  "eleven_v3",
-  "eleven_multilingual_v2",
-  "eleven_flash_v2_5",
-  "eleven_flash_v2",
-  "eleven_turbo_v2_5",
-  "eleven_turbo_v2",
-  "eleven_monolingual_v1"
+//#region src/model-meta.ts
+/**
+* ElevenLabs model identifiers. The lists below are the source of truth —
+* callers are blocked from passing unknown model IDs. Keep them in sync with
+* the ElevenLabs SDK via the automated update pipeline.
+*/
+/**
+* Text-to-speech models.
+* @see https://elevenlabs.io/docs/models
+*/
+var ELEVENLABS_TTS_MODELS = [
+	"eleven_v3",
+	"eleven_multilingual_v2",
+	"eleven_flash_v2_5",
+	"eleven_flash_v2",
+	"eleven_turbo_v2_5",
+	"eleven_turbo_v2",
+	"eleven_monolingual_v1"
 ];
-const ELEVENLABS_AUDIO_MODELS = [
-  "music_v1",
-  "eleven_text_to_sound_v2",
-  "eleven_text_to_sound_v1"
+/**
+* Audio generation models — music (`music_v1`) + sound effects
+* (`eleven_text_to_sound_v*`) share one `generateAudio` adapter.
+* The adapter dispatches by model id so callers pick behavior via the model.
+*
+* @see https://elevenlabs.io/docs/overview/capabilities/music
+* @see https://elevenlabs.io/docs/overview/capabilities/sound-effects
+*/
+var ELEVENLABS_AUDIO_MODELS = [
+	"music_v1",
+	"eleven_text_to_sound_v2",
+	"eleven_text_to_sound_v1"
 ];
 function isElevenLabsMusicModel(model) {
-  return model === "music_v1";
+	return model === "music_v1";
 }
 function isElevenLabsSoundEffectsModel(model) {
-  return model.startsWith("eleven_text_to_sound_");
+	return model.startsWith("eleven_text_to_sound_");
 }
-const ELEVENLABS_TRANSCRIPTION_MODELS = [
-  "scribe_v2",
-  "scribe_v1"
-];
-export {
-  ELEVENLABS_AUDIO_MODELS,
-  ELEVENLABS_TRANSCRIPTION_MODELS,
-  ELEVENLABS_TTS_MODELS,
-  isElevenLabsMusicModel,
-  isElevenLabsSoundEffectsModel
-};
-//# sourceMappingURL=model-meta.js.map
+/**
+* Speech-to-text (transcription) models — Scribe family.
+* @see https://elevenlabs.io/docs/overview/capabilities/speech-to-text
+*/
+var ELEVENLABS_TRANSCRIPTION_MODELS = ["scribe_v2", "scribe_v1"];
+//#endregion
+export { ELEVENLABS_AUDIO_MODELS, ELEVENLABS_TRANSCRIPTION_MODELS, ELEVENLABS_TTS_MODELS, isElevenLabsMusicModel, isElevenLabsSoundEffectsModel };
+
+//# sourceMappingURL=model-meta.js.map

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

@@ -1 +1 @@
-{"version":3,"file":"model-meta.js","sources":["../../src/model-meta.ts"],"sourcesContent":["import type { ElevenLabs } from '@elevenlabs/elevenlabs-js'\n\n/**\n * ElevenLabs model identifiers. The lists below are the source of truth —\n * callers are blocked from passing unknown model IDs. Keep them in sync with\n * the ElevenLabs SDK via the automated update pipeline.\n */\n\n/**\n * Text-to-speech models.\n * @see https://elevenlabs.io/docs/models\n */\nexport const ELEVENLABS_TTS_MODELS = [\n  'eleven_v3',\n  'eleven_multilingual_v2',\n  'eleven_flash_v2_5',\n  'eleven_flash_v2',\n  'eleven_turbo_v2_5',\n  'eleven_turbo_v2',\n  'eleven_monolingual_v1',\n] as const\n\nexport type ElevenLabsTTSModel = (typeof ELEVENLABS_TTS_MODELS)[number]\n\n/**\n * Audio generation models — music (`music_v1`) + sound effects\n * (`eleven_text_to_sound_v*`) share one `generateAudio` adapter.\n * The adapter dispatches by model id so callers pick behavior via the model.\n *\n * @see https://elevenlabs.io/docs/overview/capabilities/music\n * @see https://elevenlabs.io/docs/overview/capabilities/sound-effects\n */\nexport const ELEVENLABS_AUDIO_MODELS = [\n  'music_v1',\n  'eleven_text_to_sound_v2',\n  'eleven_text_to_sound_v1',\n] as const\n\nexport type ElevenLabsAudioModel = (typeof ELEVENLABS_AUDIO_MODELS)[number]\n\n/** Music models within the audio family. */\nexport type ElevenLabsMusicModel = 'music_v1'\n/** SFX models within the audio family. */\nexport type ElevenLabsSoundEffectsModel =\n  | 'eleven_text_to_sound_v2'\n  | 'eleven_text_to_sound_v1'\n\nexport function isElevenLabsMusicModel(\n  model: string,\n): model is ElevenLabsMusicModel {\n  return model === 'music_v1'\n}\n\nexport function isElevenLabsSoundEffectsModel(\n  model: string,\n): model is ElevenLabsSoundEffectsModel {\n  return model.startsWith('eleven_text_to_sound_')\n}\n\n/**\n * Speech-to-text (transcription) models — Scribe family.\n * @see https://elevenlabs.io/docs/overview/capabilities/speech-to-text\n */\nexport const ELEVENLABS_TRANSCRIPTION_MODELS = [\n  'scribe_v2',\n  'scribe_v1',\n] as const\n\nexport type ElevenLabsTranscriptionModel =\n  (typeof ELEVENLABS_TRANSCRIPTION_MODELS)[number]\n\n/**\n * Supported `output_format` strings, encoded as `codec_samplerate[_bitrate]`.\n * Aliased to the SDK's `AllowedOutputFormats` so the list stays in sync\n * automatically whenever the `@elevenlabs/elevenlabs-js` dependency is bumped.\n *\n * @see https://elevenlabs.io/docs/api-reference/text-to-speech/convert\n */\nexport type ElevenLabsOutputFormat = ElevenLabs.AllowedOutputFormats\n"],"names":[],"mappings":"AAYO,MAAM,wBAAwB;AAAA,EACnC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAYO,MAAM,0BAA0B;AAAA,EACrC;AAAA,EACA;AAAA,EACA;AACF;AAWO,SAAS,uBACd,OAC+B;AAC/B,SAAO,UAAU;AACnB;AAEO,SAAS,8BACd,OACsC;AACtC,SAAO,MAAM,WAAW,uBAAuB;AACjD;AAMO,MAAM,kCAAkC;AAAA,EAC7C;AAAA,EACA;AACF;"}
+{"version":3,"file":"model-meta.js","names":[],"sources":["../../src/model-meta.ts"],"sourcesContent":["import type { ElevenLabs } from '@elevenlabs/elevenlabs-js'\n\n/**\n * ElevenLabs model identifiers. The lists below are the source of truth —\n * callers are blocked from passing unknown model IDs. Keep them in sync with\n * the ElevenLabs SDK via the automated update pipeline.\n */\n\n/**\n * Text-to-speech models.\n * @see https://elevenlabs.io/docs/models\n */\nexport const ELEVENLABS_TTS_MODELS = [\n  'eleven_v3',\n  'eleven_multilingual_v2',\n  'eleven_flash_v2_5',\n  'eleven_flash_v2',\n  'eleven_turbo_v2_5',\n  'eleven_turbo_v2',\n  'eleven_monolingual_v1',\n] as const\n\nexport type ElevenLabsTTSModel = (typeof ELEVENLABS_TTS_MODELS)[number]\n\n/**\n * Audio generation models — music (`music_v1`) + sound effects\n * (`eleven_text_to_sound_v*`) share one `generateAudio` adapter.\n * The adapter dispatches by model id so callers pick behavior via the model.\n *\n * @see https://elevenlabs.io/docs/overview/capabilities/music\n * @see https://elevenlabs.io/docs/overview/capabilities/sound-effects\n */\nexport const ELEVENLABS_AUDIO_MODELS = [\n  'music_v1',\n  'eleven_text_to_sound_v2',\n  'eleven_text_to_sound_v1',\n] as const\n\nexport type ElevenLabsAudioModel = (typeof ELEVENLABS_AUDIO_MODELS)[number]\n\n/** Music models within the audio family. */\nexport type ElevenLabsMusicModel = 'music_v1'\n/** SFX models within the audio family. */\nexport type ElevenLabsSoundEffectsModel =\n  | 'eleven_text_to_sound_v2'\n  | 'eleven_text_to_sound_v1'\n\nexport function isElevenLabsMusicModel(\n  model: string,\n): model is ElevenLabsMusicModel {\n  return model === 'music_v1'\n}\n\nexport function isElevenLabsSoundEffectsModel(\n  model: string,\n): model is ElevenLabsSoundEffectsModel {\n  return model.startsWith('eleven_text_to_sound_')\n}\n\n/**\n * Speech-to-text (transcription) models — Scribe family.\n * @see https://elevenlabs.io/docs/overview/capabilities/speech-to-text\n */\nexport const ELEVENLABS_TRANSCRIPTION_MODELS = [\n  'scribe_v2',\n  'scribe_v1',\n] as const\n\nexport type ElevenLabsTranscriptionModel =\n  (typeof ELEVENLABS_TRANSCRIPTION_MODELS)[number]\n\n/**\n * Supported `output_format` strings, encoded as `codec_samplerate[_bitrate]`.\n * Aliased to the SDK's `AllowedOutputFormats` so the list stays in sync\n * automatically whenever the `@elevenlabs/elevenlabs-js` dependency is bumped.\n *\n * @see https://elevenlabs.io/docs/api-reference/text-to-speech/convert\n */\nexport type ElevenLabsOutputFormat = ElevenLabs.AllowedOutputFormats\n"],"mappings":";;;;;;;;;;AAYA,IAAa,wBAAwB;CACnC;CACA;CACA;CACA;CACA;CACA;CACA;AACF;;;;;;;;;AAYA,IAAa,0BAA0B;CACrC;CACA;CACA;AACF;AAWA,SAAgB,uBACd,OAC+B;CAC/B,OAAO,UAAU;AACnB;AAEA,SAAgB,8BACd,OACsC;CACtC,OAAO,MAAM,WAAW,uBAAuB;AACjD;;;;;AAMA,IAAa,kCAAkC,CAC7C,aACA,WACF"}