@tanstack/ai-elevenlabs 0.2.36 → 0.2.37
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/esm/adapters/audio.js +2 -1
- package/dist/esm/adapters/audio.js.map +1 -1
- package/dist/esm/adapters/speech.js +4 -2
- package/dist/esm/adapters/speech.js.map +1 -1
- package/dist/esm/realtime/token.js +1 -1
- package/dist/esm/realtime/token.js.map +1 -1
- package/dist/esm/utils/client.js +2 -1
- package/dist/esm/utils/client.js.map +1 -1
- package/package.json +4 -4
|
@@ -72,7 +72,8 @@ var ElevenLabsAudioAdapter = class extends BaseAudioAdapter {
|
|
|
72
72
|
return this.finalize(stream, outputFormat, options.duration);
|
|
73
73
|
}
|
|
74
74
|
async finalize(stream, outputFormat, duration) {
|
|
75
|
-
const
|
|
75
|
+
const buffer = await readStreamToArrayBuffer(stream);
|
|
76
|
+
const base64 = arrayBufferToBase64(buffer);
|
|
76
77
|
const { contentType } = parseOutputFormat(outputFormat);
|
|
77
78
|
return {
|
|
78
79
|
id: generateId(this.name),
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"audio.js","names":[],"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"],"mappings":";;;;;;;;;;;;;;;;;;AAuGA,IAAa,yBAAb,cAEU,iBAAyD;CACjE,OAAgB;CAEhB;CAEA,YAAY,OAAe,QAAiC;EAC1D,MAAM,OAAO,UAAU,CAAC,CAAC;EACzB,KAAK,SAAS,uBAAuB,MAAM;CAC7C;CAEA,MAAM,cACJ,SACgC;EAChC,MAAM,EAAE,WAAW;EACnB,OAAO,QACL,oDAAoD,KAAK,SACzD;GAAE,UAAU;GAAc,OAAO,KAAK;EAAM,CAC9C;EACA,IAAI;GACF,IAAI,uBAAuB,KAAK,KAAK,GACnC,OAAO,MAAM,KAAK,SAAS,OAAO;GAEpC,IAAI,8BAA8B,KAAK,KAAK,GAC1C,OAAO,MAAM,KAAK,gBAAgB,OAAO;GAE3C,MAAM,IAAI,MACR,uCAAuC,KAAK,MAAM,gFACpD;EACF,SAAS,OAAO;GACd,OAAO,OAAO,kCAAkC;IAC9C;IACA,QAAQ;GACV,CAAC;GACD,MAAM;EACR;CACF;CAEA,MAAc,SACZ,SACgC;EAEhC,MAAM,UAAU,KAAK;EACrB,MAAM,QAAS,QAAQ,gBAAgB,CAAC;EACxC,MAAM,eAAe,MAAM;EAE3B,MAAM,SAAS,MAAM,KAAK,OAAO,MAAM,QAAQ;GAC7C;GACA,GAAI,QAAQ,UAAU,CAAC,MAAM,kBACzB,EAAE,QAAQ,QAAQ,OAAO,IACzB,CAAC;GACL,GAAI,MAAM,kBACN,EAAE,iBAAiB,cAAc,MAAM,eAAe,EAAE,IACxD,CAAC;GACL,GAAI,QAAQ,YAAY,QAAQ,CAAC,MAAM,kBACnC,EAAE,eAAe,KAAK,MAAM,QAAQ,WAAW,GAAI,EAAE,IACrD,CAAC;GACL,GAAI,eAAe,EAAE,aAAa,IAAI,CAAC;GACvC,GAAI,MAAM,QAAQ,OAAO,EAAE,MAAM,MAAM,KAAK,IAAI,CAAC;GACjD,GAAI,MAAM,qBAAqB,OAC3B,EAAE,mBAAmB,MAAM,kBAAkB,IAC7C,CAAC;GACL,GAAI,MAAM,4BAA4B,OAClC,EAAE,0BAA0B,MAAM,yBAAyB,IAC3D,CAAC;EACP,CAAC;EAED,OAAO,KAAK,SAAS,QAAQ,cAAc,QAAQ,QAAQ;CAC7D;CAEA,MAAc,gBACZ,SACgC;EAEhC,MAAM,UAAU,KAAK;EACrB,MAAM,MAAO,QAAQ,gBACnB,CAAC;EACH,MAAM,eAAe,IAAI;EAEzB,MAAM,SAAS,MAAM,KAAK,OAAO,mBAAmB,QAAQ;GAC1D,MAAM,QAAQ;GACd;GACA,GAAI,QAAQ,YAAY,OACpB,EAAE,iBAAiB,QAAQ,SAAS,IACpC,CAAC;GACL,GAAI,eAAe,EAAE,aAAa,IAAI,CAAC;GACvC,GAAI,IAAI,mBAAmB,OACvB,EAAE,iBAAiB,IAAI,gBAAgB,IACvC,CAAC;GACL,GAAI,IAAI,QAAQ,OAAO,EAAE,MAAM,IAAI,KAAK,IAAI,CAAC;EAC/C,CAAC;EAED,OAAO,KAAK,SAAS,QAAQ,cAAc,QAAQ,QAAQ;CAC7D;CAEA,MAAc,SACZ,QACA,cACA,UACgC;EAEhC,MAAM,SAAS,oBAAoB,MADd,wBAAwB,MAAM,CACV;EACzC,MAAM,EAAE,gBAAgB,kBAAkB,YAAY;EACtD,OAAO;GACL,IAAI,WAAW,KAAK,IAAI;GACxB,OAAO,KAAK;GACZ,OAAO;IACL,SAAS;IACT;IACA,GAAI,YAAY,OAAO,EAAE,SAAS,IAAI,CAAC;GACzC;EACF;CACF;CAEA,aAAwC;EACtC,OAAO,WAAW,KAAK,IAAI;CAC7B;AACF;AAEA,SAAS,cAAc,MAAsC;CAC3D,OAAO;EACL,sBAAsB,KAAK,wBAAwB,CAAC;EACpD,sBAAsB,KAAK,wBAAwB,CAAC;EACpD,WAAW,KAAK,YAAY,CAAC,EAAA,CAAG,KAAK,aAAa;GAChD,aAAa,QAAQ;GACrB,qBAAqB,QAAQ,uBAAuB,CAAC;GACrD,qBAAqB,QAAQ,uBAAuB,CAAC;GACrD,YAAY,QAAQ,cAAc;GAClC,OAAO,QAAQ,SAAS,CAAC;EAC3B,EAAE;CACJ;AACF;;;;AAKA,SAAgB,gBACd,OACA,QACgC;CAChC,OAAO,IAAI,uBAAuB,OAAO,MAAM;AACjD;;;;AAKA,SAAgB,sBACd,OACA,QACA,QACgC;CAChC,OAAO,IAAI,uBAAuB,OAAO;EAAE;EAAQ,GAAG;CAAO,CAAC;AAChE"}
|
|
1
|
+
{"version":3,"file":"audio.js","names":[],"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"],"mappings":";;;;;;;;;;;;;;;;;;AAuGA,IAAa,yBAAb,cAEU,iBAAyD;CACjE,OAAgB;CAEhB;CAEA,YAAY,OAAe,QAAiC;EAC1D,MAAM,OAAO,UAAU,CAAC,CAAC;EACzB,KAAK,SAAS,uBAAuB,MAAM;CAC7C;CAEA,MAAM,cACJ,SACgC;EAChC,MAAM,EAAE,WAAW;EACnB,OAAO,QACL,oDAAoD,KAAK,SACzD;GAAE,UAAU;GAAc,OAAO,KAAK;EAAM,CAC9C;EACA,IAAI;GACF,IAAI,uBAAuB,KAAK,KAAK,GACnC,OAAO,MAAM,KAAK,SAAS,OAAO;GAEpC,IAAI,8BAA8B,KAAK,KAAK,GAC1C,OAAO,MAAM,KAAK,gBAAgB,OAAO;GAE3C,MAAM,IAAI,MACR,uCAAuC,KAAK,MAAM,gFACpD;EACF,SAAS,OAAO;GACd,OAAO,OAAO,kCAAkC;IAC9C;IACA,QAAQ;GACV,CAAC;GACD,MAAM;EACR;CACF;CAEA,MAAc,SACZ,SACgC;EAEhC,MAAM,UAAU,KAAK;EACrB,MAAM,QAAS,QAAQ,gBAAgB,CAAC;EACxC,MAAM,eAAe,MAAM;EAE3B,MAAM,SAAS,MAAM,KAAK,OAAO,MAAM,QAAQ;GAC7C;GACA,GAAI,QAAQ,UAAU,CAAC,MAAM,kBACzB,EAAE,QAAQ,QAAQ,OAAO,IACzB,CAAC;GACL,GAAI,MAAM,kBACN,EAAE,iBAAiB,cAAc,MAAM,eAAe,EAAE,IACxD,CAAC;GACL,GAAI,QAAQ,YAAY,QAAQ,CAAC,MAAM,kBACnC,EAAE,eAAe,KAAK,MAAM,QAAQ,WAAW,GAAI,EAAE,IACrD,CAAC;GACL,GAAI,eAAe,EAAE,aAAa,IAAI,CAAC;GACvC,GAAI,MAAM,QAAQ,OAAO,EAAE,MAAM,MAAM,KAAK,IAAI,CAAC;GACjD,GAAI,MAAM,qBAAqB,OAC3B,EAAE,mBAAmB,MAAM,kBAAkB,IAC7C,CAAC;GACL,GAAI,MAAM,4BAA4B,OAClC,EAAE,0BAA0B,MAAM,yBAAyB,IAC3D,CAAC;EACP,CAAC;EAED,OAAO,KAAK,SAAS,QAAQ,cAAc,QAAQ,QAAQ;CAC7D;CAEA,MAAc,gBACZ,SACgC;EAEhC,MAAM,UAAU,KAAK;EACrB,MAAM,MAAO,QAAQ,gBACnB,CAAC;EACH,MAAM,eAAe,IAAI;EAEzB,MAAM,SAAS,MAAM,KAAK,OAAO,mBAAmB,QAAQ;GAC1D,MAAM,QAAQ;GACd;GACA,GAAI,QAAQ,YAAY,OACpB,EAAE,iBAAiB,QAAQ,SAAS,IACpC,CAAC;GACL,GAAI,eAAe,EAAE,aAAa,IAAI,CAAC;GACvC,GAAI,IAAI,mBAAmB,OACvB,EAAE,iBAAiB,IAAI,gBAAgB,IACvC,CAAC;GACL,GAAI,IAAI,QAAQ,OAAO,EAAE,MAAM,IAAI,KAAK,IAAI,CAAC;EAC/C,CAAC;EAED,OAAO,KAAK,SAAS,QAAQ,cAAc,QAAQ,QAAQ;CAC7D;CAEA,MAAc,SACZ,QACA,cACA,UACgC;EAChC,MAAM,SAAS,MAAM,wBAAwB,MAAM;EACnD,MAAM,SAAS,oBAAoB,MAAM;EACzC,MAAM,EAAE,gBAAgB,kBAAkB,YAAY;EACtD,OAAO;GACL,IAAI,WAAW,KAAK,IAAI;GACxB,OAAO,KAAK;GACZ,OAAO;IACL,SAAS;IACT;IACA,GAAI,YAAY,OAAO,EAAE,SAAS,IAAI,CAAC;GACzC;EACF;CACF;CAEA,aAAwC;EACtC,OAAO,WAAW,KAAK,IAAI;CAC7B;AACF;AAEA,SAAS,cAAc,MAAsC;CAC3D,OAAO;EACL,sBAAsB,KAAK,wBAAwB,CAAC;EACpD,sBAAsB,KAAK,wBAAwB,CAAC;EACpD,WAAW,KAAK,YAAY,CAAC,EAAA,CAAG,KAAK,aAAa;GAChD,aAAa,QAAQ;GACrB,qBAAqB,QAAQ,uBAAuB,CAAC;GACrD,qBAAqB,QAAQ,uBAAuB,CAAC;GACrD,YAAY,QAAQ,cAAc;GAClC,OAAO,QAAQ,SAAS,CAAC;EAC3B,EAAE;CACJ;AACF;;;;AAKA,SAAgB,gBACd,OACA,QACgC;CAChC,OAAO,IAAI,uBAAuB,OAAO,MAAM;AACjD;;;;AAKA,SAAgB,sBACd,OACA,QACA,QACgC;CAChC,OAAO,IAAI,uBAAuB,OAAO;EAAE;EAAQ,GAAG;CAAO,CAAC;AAChE"}
|
|
@@ -33,7 +33,7 @@ var ElevenLabsSpeechAdapter = class extends BaseTTSAdapter {
|
|
|
33
33
|
if (!voiceId) throw new Error("ElevenLabs TTS requires a voice. Pass `voice` on generateSpeech() or `voiceId` in modelOptions.");
|
|
34
34
|
const { outputFormat, voiceSettings, languageCode, seed, previousText, nextText, previousRequestIds, nextRequestIds, applyTextNormalization, applyLanguageTextNormalization, optimizeStreamingLatency, enableLogging } = options.modelOptions ?? {};
|
|
35
35
|
const effectiveOutputFormat = outputFormat ?? inferOutputFormatFromResponseFormat(options.format);
|
|
36
|
-
const
|
|
36
|
+
const stream = await this.client.textToSpeech.convert(voiceId, {
|
|
37
37
|
text: options.text,
|
|
38
38
|
modelId: this.model,
|
|
39
39
|
...effectiveOutputFormat ? { outputFormat: effectiveOutputFormat } : {},
|
|
@@ -48,7 +48,9 @@ var ElevenLabsSpeechAdapter = class extends BaseTTSAdapter {
|
|
|
48
48
|
...applyLanguageTextNormalization != null ? { applyLanguageTextNormalization } : {},
|
|
49
49
|
...optimizeStreamingLatency != null ? { optimizeStreamingLatency } : {},
|
|
50
50
|
...enableLogging != null ? { enableLogging } : {}
|
|
51
|
-
})
|
|
51
|
+
});
|
|
52
|
+
const buffer = await readStreamToArrayBuffer(stream);
|
|
53
|
+
const base64 = arrayBufferToBase64(buffer);
|
|
52
54
|
const { format, contentType } = parseOutputFormat(effectiveOutputFormat);
|
|
53
55
|
return {
|
|
54
56
|
id: generateId(this.name),
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"speech.js","names":[],"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"],"mappings":";;;;;;;;;;;;;;;;;AA+EA,IAAa,0BAAb,cAEU,eAAwD;CAChE,OAAgB;CAEhB;CAEA,YAAY,OAAe,QAAiC;EAC1D,MAAM,OAAO,UAAU,CAAC,CAAC;EACzB,KAAK,SAAS,uBAAuB,MAAM;CAC7C;CAEA,MAAM,eACJ,SACoB;EACpB,MAAM,EAAE,WAAW;EACnB,OAAO,QACL,qDAAqD,KAAK,SAC1D;GAAE,UAAU;GAAc,OAAO,KAAK;EAAM,CAC9C;EACA,IAAI;GACF,MAAM,UAAU,QAAQ,SAAS,QAAQ,cAAc;GACvD,IAAI,CAAC,SACH,MAAM,IAAI,MACR,iGACF;GAEF,MAAM,EACJ,cACA,eACA,cACA,MACA,cACA,UACA,oBACA,gBACA,wBACA,gCACA,0BACA,kBACE,QAAQ,gBAAgB,CAAC;GAC7B,MAAM,wBACJ,gBAAgB,oCAAoC,QAAQ,MAAM;GA8BpE,MAAM,SAAS,oBAAoB,MADd,wBAAwB,MA3BxB,KAAK,OAAO,aAAa,QAAQ,SAAS;IAC7D,MAAM,QAAQ;IACd,SAAS,KAAK;IACd,GAAI,wBACA,EAAE,cAAc,sBAAsB,IACtC,CAAC;IACL,GAAI,gBACA,EAAE,eAAe,iBAAiB,eAAe,QAAQ,KAAK,EAAE,IAChE,QAAQ,SAAS,OACf,EAAE,eAAe,EAAE,OAAO,QAAQ,MAAM,EAAE,IAC1C,CAAC;IACP,GAAI,eAAe,EAAE,aAAa,IAAI,CAAC;IACvC,GAAI,QAAQ,OAAO,EAAE,KAAK,IAAI,CAAC;IAC/B,GAAI,eAAe,EAAE,aAAa,IAAI,CAAC;IACvC,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;IAC/B,GAAI,qBAAqB,EAAE,mBAAmB,IAAI,CAAC;IACnD,GAAI,iBAAiB,EAAE,eAAe,IAAI,CAAC;IAC3C,GAAI,yBAAyB,EAAE,uBAAuB,IAAI,CAAC;IAC3D,GAAI,kCAAkC,OAClC,EAAE,+BAA+B,IACjC,CAAC;IACL,GAAI,4BAA4B,OAC5B,EAAE,yBAAyB,IAC3B,CAAC;IACL,GAAI,iBAAiB,OAAO,EAAE,cAAc,IAAI,CAAC;GACnD,CAAC,CAEkD,CACV;GACzC,MAAM,EAAE,QAAQ,gBAAgB,kBAAkB,qBAAqB;GAEvE,OAAO;IACL,IAAI,WAAW,KAAK,IAAI;IACxB,OAAO,KAAK;IACZ,OAAO;IACP;IACA;GACF;EACF,SAAS,OAAO;GACd,OAAO,OAAO,mCAAmC;IAC/C;IACA,QAAQ;GACV,CAAC;GACD,MAAM;EACR;CACF;CAEA,aAAwC;EACtC,OAAO,WAAW,KAAK,IAAI;CAC7B;AACF;AAEA,SAAS,iBACP,UACA,eACyB;CACzB,OAAO;EACL,GAAI,SAAS,aAAa,OAAO,EAAE,WAAW,SAAS,UAAU,IAAI,CAAC;EACtE,GAAI,SAAS,mBAAmB,OAC5B,EAAE,iBAAiB,SAAS,gBAAgB,IAC5C,CAAC;EACL,GAAI,SAAS,SAAS,OAAO,EAAE,OAAO,SAAS,MAAM,IAAI,CAAC;EAC1D,GAAI,iBAAiB,OACjB,EAAE,OAAO,cAAc,IACvB,SAAS,SAAS,OAChB,EAAE,OAAO,SAAS,MAAM,IACxB,CAAC;EACP,GAAI,SAAS,mBAAmB,OAC5B,EAAE,iBAAiB,SAAS,gBAAgB,IAC5C,CAAC;CACP;AACF;;;;;;AAOA,SAAS,oCACP,QACoC;CACpC,QAAQ,QAAR;EACE,KAAK,OACH,OAAO;EACT,KAAK,OACH,OAAO;EACT,KAAK,QACH,OAAO;EACT,KAAK,KAAA,GACH;EAIF,SAGE,OAAO;CACX;AACF;;;;AAKA,SAAgB,iBACd,OACA,QACiC;CACjC,OAAO,IAAI,wBAAwB,OAAO,MAAM;AAClD;;;;AAKA,SAAgB,uBACd,OACA,QACA,QACiC;CACjC,OAAO,IAAI,wBAAwB,OAAO;EAAE;EAAQ,GAAG;CAAO,CAAC;AACjE"}
|
|
1
|
+
{"version":3,"file":"speech.js","names":[],"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"],"mappings":";;;;;;;;;;;;;;;;;AA+EA,IAAa,0BAAb,cAEU,eAAwD;CAChE,OAAgB;CAEhB;CAEA,YAAY,OAAe,QAAiC;EAC1D,MAAM,OAAO,UAAU,CAAC,CAAC;EACzB,KAAK,SAAS,uBAAuB,MAAM;CAC7C;CAEA,MAAM,eACJ,SACoB;EACpB,MAAM,EAAE,WAAW;EACnB,OAAO,QACL,qDAAqD,KAAK,SAC1D;GAAE,UAAU;GAAc,OAAO,KAAK;EAAM,CAC9C;EACA,IAAI;GACF,MAAM,UAAU,QAAQ,SAAS,QAAQ,cAAc;GACvD,IAAI,CAAC,SACH,MAAM,IAAI,MACR,iGACF;GAEF,MAAM,EACJ,cACA,eACA,cACA,MACA,cACA,UACA,oBACA,gBACA,wBACA,gCACA,0BACA,kBACE,QAAQ,gBAAgB,CAAC;GAC7B,MAAM,wBACJ,gBAAgB,oCAAoC,QAAQ,MAAM;GAEpE,MAAM,SAAS,MAAM,KAAK,OAAO,aAAa,QAAQ,SAAS;IAC7D,MAAM,QAAQ;IACd,SAAS,KAAK;IACd,GAAI,wBACA,EAAE,cAAc,sBAAsB,IACtC,CAAC;IACL,GAAI,gBACA,EAAE,eAAe,iBAAiB,eAAe,QAAQ,KAAK,EAAE,IAChE,QAAQ,SAAS,OACf,EAAE,eAAe,EAAE,OAAO,QAAQ,MAAM,EAAE,IAC1C,CAAC;IACP,GAAI,eAAe,EAAE,aAAa,IAAI,CAAC;IACvC,GAAI,QAAQ,OAAO,EAAE,KAAK,IAAI,CAAC;IAC/B,GAAI,eAAe,EAAE,aAAa,IAAI,CAAC;IACvC,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;IAC/B,GAAI,qBAAqB,EAAE,mBAAmB,IAAI,CAAC;IACnD,GAAI,iBAAiB,EAAE,eAAe,IAAI,CAAC;IAC3C,GAAI,yBAAyB,EAAE,uBAAuB,IAAI,CAAC;IAC3D,GAAI,kCAAkC,OAClC,EAAE,+BAA+B,IACjC,CAAC;IACL,GAAI,4BAA4B,OAC5B,EAAE,yBAAyB,IAC3B,CAAC;IACL,GAAI,iBAAiB,OAAO,EAAE,cAAc,IAAI,CAAC;GACnD,CAAC;GAED,MAAM,SAAS,MAAM,wBAAwB,MAAM;GACnD,MAAM,SAAS,oBAAoB,MAAM;GACzC,MAAM,EAAE,QAAQ,gBAAgB,kBAAkB,qBAAqB;GAEvE,OAAO;IACL,IAAI,WAAW,KAAK,IAAI;IACxB,OAAO,KAAK;IACZ,OAAO;IACP;IACA;GACF;EACF,SAAS,OAAO;GACd,OAAO,OAAO,mCAAmC;IAC/C;IACA,QAAQ;GACV,CAAC;GACD,MAAM;EACR;CACF;CAEA,aAAwC;EACtC,OAAO,WAAW,KAAK,IAAI;CAC7B;AACF;AAEA,SAAS,iBACP,UACA,eACyB;CACzB,OAAO;EACL,GAAI,SAAS,aAAa,OAAO,EAAE,WAAW,SAAS,UAAU,IAAI,CAAC;EACtE,GAAI,SAAS,mBAAmB,OAC5B,EAAE,iBAAiB,SAAS,gBAAgB,IAC5C,CAAC;EACL,GAAI,SAAS,SAAS,OAAO,EAAE,OAAO,SAAS,MAAM,IAAI,CAAC;EAC1D,GAAI,iBAAiB,OACjB,EAAE,OAAO,cAAc,IACvB,SAAS,SAAS,OAChB,EAAE,OAAO,SAAS,MAAM,IACxB,CAAC;EACP,GAAI,SAAS,mBAAmB,OAC5B,EAAE,iBAAiB,SAAS,gBAAgB,IAC5C,CAAC;CACP;AACF;;;;;;AAOA,SAAS,oCACP,QACoC;CACpC,QAAQ,QAAR;EACE,KAAK,OACH,OAAO;EACT,KAAK,OACH,OAAO;EACT,KAAK,QACH,OAAO;EACT,KAAK,KAAA,GACH;EAIF,SAGE,OAAO;CACX;AACF;;;;AAKA,SAAgB,iBACd,OACA,QACiC;CACjC,OAAO,IAAI,wBAAwB,OAAO,MAAM;AAClD;;;;AAKA,SAAgB,uBACd,OACA,QACA,QACiC;CACjC,OAAO,IAAI,wBAAwB,OAAO;EAAE;EAAQ,GAAG;CAAO,CAAC;AACjE"}
|
|
@@ -33,7 +33,7 @@ function elevenlabsRealtimeToken(options = {}) {
|
|
|
33
33
|
const { overrides } = options;
|
|
34
34
|
const agentId = options.agentId ?? getElevenLabsAgentIdFromEnv();
|
|
35
35
|
const response = await client.conversationalAi.conversations.getSignedUrl({ agentId });
|
|
36
|
-
const expiresAt = Date.now() +
|
|
36
|
+
const expiresAt = Date.now() + 18e5;
|
|
37
37
|
return {
|
|
38
38
|
provider: "elevenlabs",
|
|
39
39
|
token: response.signedUrl,
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"token.js","names":[],"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"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;AAgCA,SAAgB,wBACd,UAA0C,CAAC,GACrB;CACtB,MAAM,SAAS,uBAAuB;CAEtC,OAAO;EACL,UAAU;EAEV,MAAM,gBAAwC;GAC5C,MAAM,EAAE,cAAc;GACtB,MAAM,UAAU,QAAQ,WAAW,4BAA4B;GAE/D,MAAM,WAAW,MAAM,OAAO,iBAAiB,cAAc,aAC3D,EAAE,QAAQ,CACZ;GAGA,MAAM,YAAY,KAAK,IAAI,IAAI
|
|
1
|
+
{"version":3,"file":"token.js","names":[],"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"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;AAgCA,SAAgB,wBACd,UAA0C,CAAC,GACrB;CACtB,MAAM,SAAS,uBAAuB;CAEtC,OAAO;EACL,UAAU;EAEV,MAAM,gBAAwC;GAC5C,MAAM,EAAE,cAAc;GACtB,MAAM,UAAU,QAAQ,WAAW,4BAA4B;GAE/D,MAAM,WAAW,MAAM,OAAO,iBAAiB,cAAc,aAC3D,EAAE,QAAQ,CACZ;GAGA,MAAM,YAAY,KAAK,IAAI,IAAI;GAO/B,OAAO;IACL,UAAU;IACV,OAAO,SAAS;IAChB;IACA,QAAQ;KACN,GAAI,WAAW,YAAY,KAAA,KAAa,EAAE,OAAO,UAAU,QAAQ;KACnE,GAAI,WAAW,iBAAiB,KAAA,KAAa,EAC3C,cAAc,UAAU,aAC1B;KACA,iBAAiB;MACf;MACA,GAAI,WAAW,iBAAiB,KAAA,KAAa,EAC3C,cAAc,UAAU,aAC1B;MACA,GAAI,WAAW,aAAa,KAAA,KAAa,EACvC,UAAU,UAAU,SACtB;KACF;IACF;GACF;EACF;CACF;AACF"}
|
package/dist/esm/utils/client.js
CHANGED
|
@@ -22,8 +22,9 @@ function getElevenLabsAgentIdFromEnv() {
|
|
|
22
22
|
* pass in an explicit key without needing `process.env`.
|
|
23
23
|
*/
|
|
24
24
|
function createElevenLabsClient(config) {
|
|
25
|
+
const apiKey = config?.apiKey ?? getElevenLabsApiKeyFromEnv();
|
|
25
26
|
return new ElevenLabsClient({
|
|
26
|
-
apiKey
|
|
27
|
+
apiKey,
|
|
27
28
|
...config?.baseUrl ? { baseUrl: config.baseUrl } : {},
|
|
28
29
|
...config?.timeoutInSeconds != null ? { timeoutInSeconds: config.timeoutInSeconds } : {},
|
|
29
30
|
...config?.maxRetries != null ? { maxRetries: config.maxRetries } : {},
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"client.js","names":[],"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"],"mappings":";;;AAiCA,SAAS,iBAAwC;CAC/C,IAAI,OAAO,eAAe,aAAa;EACrC,MAAM,MAAO,WAA0C;EACvD,IAAI,KAAK,KAAK,OAAO,IAAI;CAC3B;CACA,IAAI,OAAO,YAAY,aACrB,OAAO,QAAQ;AAGnB;AAEA,SAAgB,6BAAqC;CACnD,OAAO,iBAAiB,oBAAoB;AAC9C;AAEA,SAAgB,8BAAsC;CACpD,MAAM,KAAK,eAAe,CAAC,EAAE;CAC7B,IAAI,CAAC,IACH,MAAM,IAAI,MACR,yIAEF;CAEF,OAAO;AACT;;;;;;AAOA,SAAgB,uBACd,QACkB;
|
|
1
|
+
{"version":3,"file":"client.js","names":[],"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"],"mappings":";;;AAiCA,SAAS,iBAAwC;CAC/C,IAAI,OAAO,eAAe,aAAa;EACrC,MAAM,MAAO,WAA0C;EACvD,IAAI,KAAK,KAAK,OAAO,IAAI;CAC3B;CACA,IAAI,OAAO,YAAY,aACrB,OAAO,QAAQ;AAGnB;AAEA,SAAgB,6BAAqC;CACnD,OAAO,iBAAiB,oBAAoB;AAC9C;AAEA,SAAgB,8BAAsC;CACpD,MAAM,KAAK,eAAe,CAAC,EAAE;CAC7B,IAAI,CAAC,IACH,MAAM,IAAI,MACR,yIAEF;CAEF,OAAO;AACT;;;;;;AAOA,SAAgB,uBACd,QACkB;CAClB,MAAM,SAAS,QAAQ,UAAU,2BAA2B;CAC5D,OAAO,IAAI,iBAAiB;EAC1B;EACA,GAAI,QAAQ,UAAU,EAAE,SAAS,OAAO,QAAQ,IAAI,CAAC;EACrD,GAAI,QAAQ,oBAAoB,OAC5B,EAAE,kBAAkB,OAAO,iBAAiB,IAC5C,CAAC;EACL,GAAI,QAAQ,cAAc,OAAO,EAAE,YAAY,OAAO,WAAW,IAAI,CAAC;EACtE,GAAI,QAAQ,UAAU,EAAE,SAAS,OAAO,QAAQ,IAAI,CAAC;CACvD,CAAC;AACH;AAIA,IAAa,eAAa;;;;;;;;AAS1B,SAAgB,oBAAoB,QAA6B;CAC/D,IAAI,OAAO,WAAW,eAAe,OAAO,OAAO,SAAS,YAC1D,OAAO,OAAO,KAAK,MAAM,CAAC,CAAC,SAAS,QAAQ;CAE9C,MAAM,OAAO,IAAI,WAAW,MAAM;CAClC,IAAI,SAAS;CACb,KAAK,MAAM,QAAQ,MACjB,UAAU,OAAO,aAAa,IAAI;CAEpC,OAAO,KAAK,MAAM;AACpB;;;;;;;AAQA,eAAsB,wBACpB,QACsB;CACtB,MAAM,SAAS,OAAO,UAAU;CAChC,MAAM,SAA4B,CAAC;CACnC,IAAI,QAAQ;CACZ,IAAI;EACF,IAAI,SAAS,MAAM,OAAO,KAAK;EAC/B,OAAO,CAAC,OAAO,MAAM;GACnB,OAAO,KAAK,OAAO,KAAK;GACxB,SAAS,OAAO,MAAM;GACtB,SAAS,MAAM,OAAO,KAAK;EAC7B;CACF,UAAU;EACR,OAAO,YAAY;CACrB;CACA,MAAM,SAAS,IAAI,WAAW,KAAK;CACnC,IAAI,SAAS;CACb,KAAK,MAAM,SAAS,QAAQ;EAC1B,OAAO,IAAI,OAAO,MAAM;EACxB,UAAU,MAAM;CAClB;CACA,OAAO,OAAO,OAAO,MACnB,OAAO,YACP,OAAO,aAAa,OAAO,UAC7B;AACF;;;;;;;AAQA,SAAgB,cAAc,OAAiC;CAC7D,IAAI,CAAC,MAAM,WAAW,OAAO,GAAG,OAAO,KAAA;CACvC,MAAM,aAAa,MAAM,QAAQ,GAAG;CACpC,IAAI,eAAe,IAAI,OAAO,KAAA;CAE9B,MAAM,SAAS,MAAM,MAAM,GAAG,UAAU;CACxC,MAAM,UAAU,MAAM,MAAM,aAAa,CAAC;CAC1C,MAAM,WAAW,YAAY,KAAK,MAAM;CACxC,MAAM,WAAW,OAAO,MAAM,GAAG,CAAC,CAAC,MAAM;CAEzC,IAAI,UAAU;EACZ,MAAM,SAAS,KAAK,OAAO;EAC3B,MAAM,QAAQ,IAAI,WAAW,OAAO,MAAM;EAC1C,KAAK,IAAI,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK,GACtC,MAAM,KAAK,OAAO,WAAW,CAAC;EAEhC,OAAO,IAAI,KAAK,CAAC,KAAK,GAAG,EAAE,MAAM,SAAS,CAAC;CAC7C;CAEA,OAAO,IAAI,KAAK,CAAC,mBAAmB,OAAO,CAAC,GAAG,EAAE,MAAM,SAAS,CAAC;AACnE;;;;;;;;;;AAWA,SAAgB,kBAAkB,KAGhC;CAEA,SADe,OAAO,gBAAA,CAAiB,MAAM,GAAG,CAAC,CAAC,EAAE,EAAE,YAAY,GAClE;EACE,KAAK,OACH,OAAO;GAAE,QAAQ;GAAO,aAAa;EAAa;EACpD,KAAK,OACH,OAAO;GAAE,QAAQ;GAAO,aAAa;EAAY;EACnD,KAAK,QACH,OAAO;GAAE,QAAQ;GAAQ,aAAa;EAAa;EACrD,KAAK,QACH,OAAO;GAAE,QAAQ;GAAQ,aAAa;EAAc;EACtD,KAAK,QACH,OAAO;GAAE,QAAQ;GAAQ,aAAa;EAAqB;EAC7D,KAAK,KAAA;EACL,SAGE,OAAO;GAAE,QAAQ;GAAO,aAAa;EAAa;CACtD;AACF"}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@tanstack/ai-elevenlabs",
|
|
3
|
-
"version": "0.2.
|
|
3
|
+
"version": "0.2.37",
|
|
4
4
|
"description": "ElevenLabs adapter for TanStack AI realtime voice, text-to-speech, transcription, music, and sound effects.",
|
|
5
5
|
"author": "Tanner Linsley",
|
|
6
6
|
"license": "MIT",
|
|
@@ -53,11 +53,11 @@
|
|
|
53
53
|
"@tanstack/ai-utils": "^0.4.0"
|
|
54
54
|
},
|
|
55
55
|
"peerDependencies": {
|
|
56
|
-
"@tanstack/ai": "^0.
|
|
56
|
+
"@tanstack/ai": "^0.45.0"
|
|
57
57
|
},
|
|
58
58
|
"devDependencies": {
|
|
59
|
-
"@vitest/coverage-v8": "4.
|
|
60
|
-
"@tanstack/ai": "0.
|
|
59
|
+
"@vitest/coverage-v8": "4.1.10",
|
|
60
|
+
"@tanstack/ai": "0.45.0"
|
|
61
61
|
},
|
|
62
62
|
"scripts": {
|
|
63
63
|
"build": "vite build",
|