@tanstack/ai-grok 0.9.2 → 0.10.1

package/dist/esm/adapters/image.js

@@ -1,6 +1,7 @@
 import OpenAI from "openai";
 import { BaseImageAdapter } from "@tanstack/ai/adapters";
 import { toRunErrorPayload } from "@tanstack/ai/adapter-internals";
+import { buildImagesUsage } from "@tanstack/openai-base";
 import { generateId } from "@tanstack/ai-utils";
 import { withGrokDefaults, getGrokApiKeyFromEnv } from "../utils/client.js";
 import { validatePrompt, validateImageSize, validateNumberOfImages } from "../image/image-provider-options.js";
@@ -54,17 +55,12 @@ class GrokImageAdapter extends BaseImageAdapter {
           return [];
         }
       );
+      const usage = buildImagesUsage(response.usage);
       return {
         id: generateId(this.name),
         model,
         images,
-        ...response.usage && {
-          usage: {
-            inputTokens: response.usage.input_tokens,
-            outputTokens: response.usage.output_tokens,
-            totalTokens: response.usage.total_tokens
-          }
-        }
+        ...usage ? { usage } : {}
       };
     } catch (error) {
       options.logger.errors(`${this.name}.generateImages fatal`, {

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

@@ -1 +1 @@
-{"version":3,"file":"image.js","sources":["../../../src/adapters/image.ts"],"sourcesContent":["import OpenAI from 'openai'\nimport { BaseImageAdapter } from '@tanstack/ai/adapters'\nimport { toRunErrorPayload } from '@tanstack/ai/adapter-internals'\nimport { generateId } from '@tanstack/ai-utils'\nimport { getGrokApiKeyFromEnv, withGrokDefaults } from '../utils/client'\nimport {\n  validateImageSize,\n  validateNumberOfImages,\n  validatePrompt,\n} from '../image/image-provider-options'\nimport type {\n  GeneratedImage,\n  ImageGenerationOptions,\n  ImageGenerationResult,\n} from '@tanstack/ai'\nimport type OpenAI_SDK from 'openai'\nimport type { GrokImageModel } from '../model-meta'\nimport type {\n  GrokImageModelProviderOptionsByName,\n  GrokImageModelSizeByName,\n  GrokImageProviderOptions,\n} from '../image/image-provider-options'\nimport type { GrokClientConfig } from '../utils'\n\n/**\n * Configuration for Grok image adapter\n */\nexport interface GrokImageConfig extends GrokClientConfig {}\n\n/**\n * Grok Image Generation Adapter\n *\n * Tree-shakeable adapter for Grok image generation functionality.\n * Supports grok-2-image-1212 model.\n *\n * Features:\n * - Model-specific type-safe provider options\n * - Size validation per model\n * - Number of images validation\n */\nexport class GrokImageAdapter<\n  TModel extends GrokImageModel,\n> extends BaseImageAdapter<\n  TModel,\n  GrokImageProviderOptions,\n  GrokImageModelProviderOptionsByName,\n  GrokImageModelSizeByName\n> {\n  override readonly kind = 'image' as const\n  readonly name = 'grok' as const\n\n  protected client: OpenAI\n\n  constructor(config: GrokImageConfig, model: TModel) {\n    super(model, {})\n    this.client = new OpenAI(withGrokDefaults(config))\n  }\n\n  async generateImages(\n    options: ImageGenerationOptions<GrokImageProviderOptions>,\n  ): Promise<ImageGenerationResult> {\n    const { model, prompt, numberOfImages, size, modelOptions } = options\n\n    validatePrompt({ prompt, model })\n    validateImageSize(model, size)\n    validateNumberOfImages(model, numberOfImages)\n\n    const resolvedSize = size as OpenAI_SDK.Images.ImageGenerateParams['size']\n    const request: OpenAI_SDK.Images.ImageGenerateParamsNonStreaming = {\n      model,\n      prompt,\n      n: numberOfImages ?? 1,\n      ...(resolvedSize !== undefined && { size: resolvedSize }),\n      stream: false,\n      ...modelOptions,\n    }\n\n    try {\n      options.logger.request(\n        `activity=image provider=${this.name} model=${model} n=${request.n ?? 1} size=${request.size ?? 'default'}`,\n        { provider: this.name, model },\n      )\n      const response = await this.client.images.generate(request)\n\n      const images: Array<GeneratedImage> = (response.data ?? []).flatMap(\n        (item): Array<GeneratedImage> => {\n          const revisedPrompt = item.revised_prompt\n          if (item.b64_json) {\n            return [\n              {\n                b64Json: item.b64_json,\n                ...(revisedPrompt !== undefined && { revisedPrompt }),\n              },\n            ]\n          }\n          if (item.url) {\n            return [\n              {\n                url: item.url,\n                ...(revisedPrompt !== undefined && { revisedPrompt }),\n              },\n            ]\n          }\n          return []\n        },\n      )\n\n      return {\n        id: generateId(this.name),\n        model,\n        images,\n        ...(response.usage && {\n          usage: {\n            inputTokens: response.usage.input_tokens,\n            outputTokens: response.usage.output_tokens,\n            totalTokens: response.usage.total_tokens,\n          },\n        }),\n      }\n    } catch (error: unknown) {\n      options.logger.errors(`${this.name}.generateImages fatal`, {\n        error: toRunErrorPayload(error, `${this.name}.generateImages failed`),\n        source: `${this.name}.generateImages`,\n      })\n      throw error\n    }\n  }\n}\n\n/**\n * Creates a Grok image adapter with explicit API key.\n * Type resolution happens here at the call site.\n *\n * @param model - The model name (e.g., 'grok-2-image-1212')\n * @param apiKey - Your xAI API key\n * @param config - Optional additional configuration\n * @returns Configured Grok image adapter instance with resolved types\n *\n * @example\n * ```typescript\n * const adapter = createGrokImage('grok-2-image-1212', \"xai-...\");\n *\n * const result = await generateImage({\n *   adapter,\n *   prompt: 'A cute baby sea otter'\n * });\n * ```\n */\nexport function createGrokImage<TModel extends GrokImageModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<GrokImageConfig, 'apiKey'>,\n): GrokImageAdapter<TModel> {\n  return new GrokImageAdapter({ apiKey, ...config }, model)\n}\n\n/**\n * Creates a Grok image adapter with automatic API key detection from environment variables.\n * Type resolution happens here at the call site.\n *\n * Looks for `XAI_API_KEY` in:\n * - `process.env` (Node.js)\n * - `window.env` (Browser with injected env)\n *\n * @param model - The model name (e.g., 'grok-2-image-1212')\n * @param config - Optional configuration (excluding apiKey which is auto-detected)\n * @returns Configured Grok image adapter instance with resolved types\n * @throws Error if XAI_API_KEY is not found in environment\n *\n * @example\n * ```typescript\n * // Automatically uses XAI_API_KEY from environment\n * const adapter = grokImage('grok-2-image-1212');\n *\n * const result = await generateImage({\n *   adapter,\n *   prompt: 'A beautiful sunset over mountains'\n * });\n * ```\n */\nexport function grokImage<TModel extends GrokImageModel>(\n  model: TModel,\n  config?: Omit<GrokImageConfig, 'apiKey'>,\n): GrokImageAdapter<TModel> {\n  const apiKey = getGrokApiKeyFromEnv()\n  return createGrokImage(model, apiKey, config)\n}\n"],"names":[],"mappings":";;;;;;AAwCO,MAAM,yBAEH,iBAKR;AAAA,EACkB,OAAO;AAAA,EAChB,OAAO;AAAA,EAEN;AAAA,EAEV,YAAY,QAAyB,OAAe;AAClD,UAAM,OAAO,EAAE;AACf,SAAK,SAAS,IAAI,OAAO,iBAAiB,MAAM,CAAC;AAAA,EACnD;AAAA,EAEA,MAAM,eACJ,SACgC;AAChC,UAAM,EAAE,OAAO,QAAQ,gBAAgB,MAAM,iBAAiB;AAE9D,mBAAe,EAAE,OAAc,CAAC;AAChC,sBAAkB,OAAO,IAAI;AAC7B,2BAAuB,OAAO,cAAc;AAE5C,UAAM,eAAe;AACrB,UAAM,UAA6D;AAAA,MACjE;AAAA,MACA;AAAA,MACA,GAAG,kBAAkB;AAAA,MACrB,GAAI,iBAAiB,UAAa,EAAE,MAAM,aAAA;AAAA,MAC1C,QAAQ;AAAA,MACR,GAAG;AAAA,IAAA;AAGL,QAAI;AACF,cAAQ,OAAO;AAAA,QACb,2BAA2B,KAAK,IAAI,UAAU,KAAK,MAAM,QAAQ,KAAK,CAAC,SAAS,QAAQ,QAAQ,SAAS;AAAA,QACzG,EAAE,UAAU,KAAK,MAAM,MAAA;AAAA,MAAM;AAE/B,YAAM,WAAW,MAAM,KAAK,OAAO,OAAO,SAAS,OAAO;AAE1D,YAAM,UAAiC,SAAS,QAAQ,CAAA,GAAI;AAAA,QAC1D,CAAC,SAAgC;AAC/B,gBAAM,gBAAgB,KAAK;AAC3B,cAAI,KAAK,UAAU;AACjB,mBAAO;AAAA,cACL;AAAA,gBACE,SAAS,KAAK;AAAA,gBACd,GAAI,kBAAkB,UAAa,EAAE,cAAA;AAAA,cAAc;AAAA,YACrD;AAAA,UAEJ;AACA,cAAI,KAAK,KAAK;AACZ,mBAAO;AAAA,cACL;AAAA,gBACE,KAAK,KAAK;AAAA,gBACV,GAAI,kBAAkB,UAAa,EAAE,cAAA;AAAA,cAAc;AAAA,YACrD;AAAA,UAEJ;AACA,iBAAO,CAAA;AAAA,QACT;AAAA,MAAA;AAGF,aAAO;AAAA,QACL,IAAI,WAAW,KAAK,IAAI;AAAA,QACxB;AAAA,QACA;AAAA,QACA,GAAI,SAAS,SAAS;AAAA,UACpB,OAAO;AAAA,YACL,aAAa,SAAS,MAAM;AAAA,YAC5B,cAAc,SAAS,MAAM;AAAA,YAC7B,aAAa,SAAS,MAAM;AAAA,UAAA;AAAA,QAC9B;AAAA,MACF;AAAA,IAEJ,SAAS,OAAgB;AACvB,cAAQ,OAAO,OAAO,GAAG,KAAK,IAAI,yBAAyB;AAAA,QACzD,OAAO,kBAAkB,OAAO,GAAG,KAAK,IAAI,wBAAwB;AAAA,QACpE,QAAQ,GAAG,KAAK,IAAI;AAAA,MAAA,CACrB;AACD,YAAM;AAAA,IACR;AAAA,EACF;AACF;AAqBO,SAAS,gBACd,OACA,QACA,QAC0B;AAC1B,SAAO,IAAI,iBAAiB,EAAE,QAAQ,GAAG,OAAA,GAAU,KAAK;AAC1D;AA0BO,SAAS,UACd,OACA,QAC0B;AAC1B,QAAM,SAAS,qBAAA;AACf,SAAO,gBAAgB,OAAO,QAAQ,MAAM;AAC9C;"}
+{"version":3,"file":"image.js","sources":["../../../src/adapters/image.ts"],"sourcesContent":["import OpenAI from 'openai'\nimport { BaseImageAdapter } from '@tanstack/ai/adapters'\nimport { toRunErrorPayload } from '@tanstack/ai/adapter-internals'\nimport { buildImagesUsage } from '@tanstack/openai-base'\nimport { generateId } from '@tanstack/ai-utils'\nimport { getGrokApiKeyFromEnv, withGrokDefaults } from '../utils/client'\nimport {\n  validateImageSize,\n  validateNumberOfImages,\n  validatePrompt,\n} from '../image/image-provider-options'\nimport type {\n  GeneratedImage,\n  ImageGenerationOptions,\n  ImageGenerationResult,\n} from '@tanstack/ai'\nimport type OpenAI_SDK from 'openai'\nimport type { GrokImageModel } from '../model-meta'\nimport type {\n  GrokImageModelProviderOptionsByName,\n  GrokImageModelSizeByName,\n  GrokImageProviderOptions,\n} from '../image/image-provider-options'\nimport type { GrokClientConfig } from '../utils'\n\n/**\n * Configuration for Grok image adapter\n */\nexport interface GrokImageConfig extends GrokClientConfig {}\n\n/**\n * Grok Image Generation Adapter\n *\n * Tree-shakeable adapter for Grok image generation functionality.\n * Supports grok-2-image-1212 model.\n *\n * Features:\n * - Model-specific type-safe provider options\n * - Size validation per model\n * - Number of images validation\n */\nexport class GrokImageAdapter<\n  TModel extends GrokImageModel,\n> extends BaseImageAdapter<\n  TModel,\n  GrokImageProviderOptions,\n  GrokImageModelProviderOptionsByName,\n  GrokImageModelSizeByName\n> {\n  override readonly kind = 'image' as const\n  readonly name = 'grok' as const\n\n  protected client: OpenAI\n\n  constructor(config: GrokImageConfig, model: TModel) {\n    super(model, {})\n    this.client = new OpenAI(withGrokDefaults(config))\n  }\n\n  async generateImages(\n    options: ImageGenerationOptions<GrokImageProviderOptions>,\n  ): Promise<ImageGenerationResult> {\n    const { model, prompt, numberOfImages, size, modelOptions } = options\n\n    validatePrompt({ prompt, model })\n    validateImageSize(model, size)\n    validateNumberOfImages(model, numberOfImages)\n\n    const resolvedSize = size as OpenAI_SDK.Images.ImageGenerateParams['size']\n    const request: OpenAI_SDK.Images.ImageGenerateParamsNonStreaming = {\n      model,\n      prompt,\n      n: numberOfImages ?? 1,\n      ...(resolvedSize !== undefined && { size: resolvedSize }),\n      stream: false,\n      ...modelOptions,\n    }\n\n    try {\n      options.logger.request(\n        `activity=image provider=${this.name} model=${model} n=${request.n ?? 1} size=${request.size ?? 'default'}`,\n        { provider: this.name, model },\n      )\n      const response = await this.client.images.generate(request)\n\n      const images: Array<GeneratedImage> = (response.data ?? []).flatMap(\n        (item): Array<GeneratedImage> => {\n          const revisedPrompt = item.revised_prompt\n          if (item.b64_json) {\n            return [\n              {\n                b64Json: item.b64_json,\n                ...(revisedPrompt !== undefined && { revisedPrompt }),\n              },\n            ]\n          }\n          if (item.url) {\n            return [\n              {\n                url: item.url,\n                ...(revisedPrompt !== undefined && { revisedPrompt }),\n              },\n            ]\n          }\n          return []\n        },\n      )\n\n      const usage = buildImagesUsage(response.usage)\n\n      return {\n        id: generateId(this.name),\n        model,\n        images,\n        ...(usage ? { usage } : {}),\n      }\n    } catch (error: unknown) {\n      options.logger.errors(`${this.name}.generateImages fatal`, {\n        error: toRunErrorPayload(error, `${this.name}.generateImages failed`),\n        source: `${this.name}.generateImages`,\n      })\n      throw error\n    }\n  }\n}\n\n/**\n * Creates a Grok image adapter with explicit API key.\n * Type resolution happens here at the call site.\n *\n * @param model - The model name (e.g., 'grok-2-image-1212')\n * @param apiKey - Your xAI API key\n * @param config - Optional additional configuration\n * @returns Configured Grok image adapter instance with resolved types\n *\n * @example\n * ```typescript\n * const adapter = createGrokImage('grok-2-image-1212', \"xai-...\");\n *\n * const result = await generateImage({\n *   adapter,\n *   prompt: 'A cute baby sea otter'\n * });\n * ```\n */\nexport function createGrokImage<TModel extends GrokImageModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<GrokImageConfig, 'apiKey'>,\n): GrokImageAdapter<TModel> {\n  return new GrokImageAdapter({ apiKey, ...config }, model)\n}\n\n/**\n * Creates a Grok image adapter with automatic API key detection from environment variables.\n * Type resolution happens here at the call site.\n *\n * Looks for `XAI_API_KEY` in:\n * - `process.env` (Node.js)\n * - `window.env` (Browser with injected env)\n *\n * @param model - The model name (e.g., 'grok-2-image-1212')\n * @param config - Optional configuration (excluding apiKey which is auto-detected)\n * @returns Configured Grok image adapter instance with resolved types\n * @throws Error if XAI_API_KEY is not found in environment\n *\n * @example\n * ```typescript\n * // Automatically uses XAI_API_KEY from environment\n * const adapter = grokImage('grok-2-image-1212');\n *\n * const result = await generateImage({\n *   adapter,\n *   prompt: 'A beautiful sunset over mountains'\n * });\n * ```\n */\nexport function grokImage<TModel extends GrokImageModel>(\n  model: TModel,\n  config?: Omit<GrokImageConfig, 'apiKey'>,\n): GrokImageAdapter<TModel> {\n  const apiKey = getGrokApiKeyFromEnv()\n  return createGrokImage(model, apiKey, config)\n}\n"],"names":[],"mappings":";;;;;;;AAyCO,MAAM,yBAEH,iBAKR;AAAA,EACkB,OAAO;AAAA,EAChB,OAAO;AAAA,EAEN;AAAA,EAEV,YAAY,QAAyB,OAAe;AAClD,UAAM,OAAO,EAAE;AACf,SAAK,SAAS,IAAI,OAAO,iBAAiB,MAAM,CAAC;AAAA,EACnD;AAAA,EAEA,MAAM,eACJ,SACgC;AAChC,UAAM,EAAE,OAAO,QAAQ,gBAAgB,MAAM,iBAAiB;AAE9D,mBAAe,EAAE,OAAc,CAAC;AAChC,sBAAkB,OAAO,IAAI;AAC7B,2BAAuB,OAAO,cAAc;AAE5C,UAAM,eAAe;AACrB,UAAM,UAA6D;AAAA,MACjE;AAAA,MACA;AAAA,MACA,GAAG,kBAAkB;AAAA,MACrB,GAAI,iBAAiB,UAAa,EAAE,MAAM,aAAA;AAAA,MAC1C,QAAQ;AAAA,MACR,GAAG;AAAA,IAAA;AAGL,QAAI;AACF,cAAQ,OAAO;AAAA,QACb,2BAA2B,KAAK,IAAI,UAAU,KAAK,MAAM,QAAQ,KAAK,CAAC,SAAS,QAAQ,QAAQ,SAAS;AAAA,QACzG,EAAE,UAAU,KAAK,MAAM,MAAA;AAAA,MAAM;AAE/B,YAAM,WAAW,MAAM,KAAK,OAAO,OAAO,SAAS,OAAO;AAE1D,YAAM,UAAiC,SAAS,QAAQ,CAAA,GAAI;AAAA,QAC1D,CAAC,SAAgC;AAC/B,gBAAM,gBAAgB,KAAK;AAC3B,cAAI,KAAK,UAAU;AACjB,mBAAO;AAAA,cACL;AAAA,gBACE,SAAS,KAAK;AAAA,gBACd,GAAI,kBAAkB,UAAa,EAAE,cAAA;AAAA,cAAc;AAAA,YACrD;AAAA,UAEJ;AACA,cAAI,KAAK,KAAK;AACZ,mBAAO;AAAA,cACL;AAAA,gBACE,KAAK,KAAK;AAAA,gBACV,GAAI,kBAAkB,UAAa,EAAE,cAAA;AAAA,cAAc;AAAA,YACrD;AAAA,UAEJ;AACA,iBAAO,CAAA;AAAA,QACT;AAAA,MAAA;AAGF,YAAM,QAAQ,iBAAiB,SAAS,KAAK;AAE7C,aAAO;AAAA,QACL,IAAI,WAAW,KAAK,IAAI;AAAA,QACxB;AAAA,QACA;AAAA,QACA,GAAI,QAAQ,EAAE,UAAU,CAAA;AAAA,MAAC;AAAA,IAE7B,SAAS,OAAgB;AACvB,cAAQ,OAAO,OAAO,GAAG,KAAK,IAAI,yBAAyB;AAAA,QACzD,OAAO,kBAAkB,OAAO,GAAG,KAAK,IAAI,wBAAwB;AAAA,QACpE,QAAQ,GAAG,KAAK,IAAI;AAAA,MAAA,CACrB;AACD,YAAM;AAAA,IACR;AAAA,EACF;AACF;AAqBO,SAAS,gBACd,OACA,QACA,QAC0B;AAC1B,SAAO,IAAI,iBAAiB,EAAE,QAAQ,GAAG,OAAA,GAAU,KAAK;AAC1D;AA0BO,SAAS,UACd,OACA,QAC0B;AAC1B,QAAM,SAAS,qBAAA;AACf,SAAO,gBAAgB,OAAO,QAAQ,MAAM;AAC9C;"}

package/dist/esm/adapters/transcription.js

@@ -54,13 +54,20 @@ class GrokTranscriptionAdapter extends BaseTranscriptionAdapter {
         }
       );
       const resolvedLanguage = data.language ?? language;
+      const usage = data.duration !== void 0 && data.duration > 0 ? {
+        promptTokens: 0,
+        completionTokens: 0,
+        totalTokens: 0,
+        durationSeconds: data.duration
+      } : void 0;
       return {
         id: generateId(this.name),
         model,
         text: data.text,
         ...resolvedLanguage !== void 0 && { language: resolvedLanguage },
         duration: data.duration,
-        ...words !== void 0 && { words }
+        ...words !== void 0 && { words },
+        ...usage !== void 0 && { usage }
       };
     } catch (error) {
       logger.errors("grok.transcribe fatal", {

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 { generateId, getGrokApiKeyFromEnv, toAudioFile } from '../utils'\nimport type {\n  TranscriptionOptions,\n  TranscriptionResult,\n  TranscriptionWord,\n} from '@tanstack/ai'\nimport type { GrokTranscriptionModel } from '../model-meta'\nimport type { GrokTranscriptionProviderOptions } from '../audio/transcription-provider-options'\n\n/**\n * Grok-specific extension of `TranscriptionWord` that surfaces the extra\n * fields xAI returns when diarization / confidence are enabled. The base\n * cross-provider `TranscriptionWord` contract doesn't include these, so\n * callers who know they're using Grok can narrow with:\n *\n * ```ts\n * const words = result.words as Array<GrokTranscriptionWord> | undefined\n * ```\n */\nexport interface GrokTranscriptionWord extends TranscriptionWord {\n  /** Model confidence for the word, when xAI returns one. */\n  confidence?: number\n  /** Speaker index, populated when `modelOptions.diarize === true`. */\n  speaker?: number\n}\n\nconst DEFAULT_GROK_BASE_URL = 'https://api.x.ai/v1'\n\n/**\n * Configuration for the Grok transcription adapter.\n *\n * Uses direct `fetch` rather than the OpenAI SDK because xAI's `/v1/stt`\n * endpoint is not OpenAI-compatible.\n */\nexport interface GrokTranscriptionConfig {\n  apiKey: string\n  baseURL?: string\n  /** Additional headers to merge into every request (e.g., test IDs). */\n  defaultHeaders?: Record<string, string>\n}\n\n/**\n * xAI STT response shape from `POST /v1/stt`.\n * Grok returns word-level timestamps only; no segment array.\n */\ninterface GrokSTTWord {\n  text: string\n  start: number\n  end: number\n  confidence?: number\n  speaker?: number\n}\n\ninterface GrokSTTResponse {\n  text: string\n  language?: string\n  duration?: number\n  words?: Array<GrokSTTWord>\n  channels?: Array<unknown>\n}\n\n/**\n * Grok Speech-to-Text Adapter.\n *\n * Talks to `POST {baseURL}/stt` per\n * https://docs.x.ai/developers/rest-api-reference/inference/voice\n */\nexport class GrokTranscriptionAdapter<\n  TModel extends GrokTranscriptionModel,\n> extends BaseTranscriptionAdapter<TModel, GrokTranscriptionProviderOptions> {\n  readonly name = 'grok' as const\n\n  private readonly apiKey: string\n  private readonly baseURL: string\n  private readonly defaultHeaders: Record<string, string>\n\n  constructor(config: GrokTranscriptionConfig, model: TModel) {\n    super(model, config)\n    this.apiKey = config.apiKey\n    this.baseURL = (config.baseURL ?? DEFAULT_GROK_BASE_URL).replace(/\\/+$/, '')\n    this.defaultHeaders = config.defaultHeaders ?? {}\n  }\n\n  async transcribe(\n    options: TranscriptionOptions<GrokTranscriptionProviderOptions>,\n  ): Promise<TranscriptionResult> {\n    const { logger } = options\n    const { model, audio, language, modelOptions } = options\n\n    logger.request(\n      `activity=generateTranscription provider=grok model=${model}`,\n      { provider: 'grok', model },\n    )\n\n    const file = toAudioFile(audio, modelOptions?.audio_format)\n    const form = buildTranscriptionFormData({ file, language, modelOptions })\n\n    try {\n      const response = await fetch(`${this.baseURL}/stt`, {\n        method: 'POST',\n        headers: {\n          // `defaultHeaders` first so Authorization always wins.\n          ...this.defaultHeaders,\n          Authorization: `Bearer ${this.apiKey}`,\n        },\n        body: form,\n      })\n\n      if (!response.ok) {\n        const errorText = await response.text()\n        throw new Error(\n          `Grok transcription request failed: ${response.status} ${errorText}`,\n        )\n      }\n\n      const data = (await response.json()) as GrokSTTResponse\n\n      const words: Array<TranscriptionWord> | undefined = data.words?.map(\n        (w) => {\n          // Construct a GrokTranscriptionWord so that `confidence` and\n          // `speaker` (when xAI returns them under `diarize` / confidence\n          // mode) are preserved on the result. The returned array is typed\n          // as `Array<TranscriptionWord>` per the cross-provider contract;\n          // callers who want the extras narrow via `as Array<GrokTranscriptionWord>`.\n          const tw: GrokTranscriptionWord = {\n            word: w.text,\n            start: w.start,\n            end: w.end,\n          }\n          if (w.confidence !== undefined) tw.confidence = w.confidence\n          if (w.speaker !== undefined) tw.speaker = w.speaker\n          return tw\n        },\n      )\n\n      const resolvedLanguage = data.language ?? language\n      return {\n        id: generateId(this.name),\n        model,\n        text: data.text,\n        ...(resolvedLanguage !== undefined && { language: resolvedLanguage }),\n        duration: data.duration,\n        ...(words !== undefined && { words }),\n      }\n    } catch (error) {\n      logger.errors('grok.transcribe fatal', {\n        error,\n        source: 'grok.transcribe',\n      })\n      throw error\n    }\n  }\n}\n\n/**\n * Build the multipart/form-data body for `POST /v1/stt`, coercing SDK-level\n * model options into xAI's wire format (booleans as `'true'`/`'false'`\n * strings, numeric fields stringified, etc.).\n *\n * Wire-field mapping:\n *   - `modelOptions.inverse_text_normalization` → `format` (xAI's chosen\n *     wire-field name for the ITN boolean; the SDK surfaces it under the\n *     clearer `inverse_text_normalization` key).\n *   - `modelOptions.audio_format`, `sample_rate`, `multichannel`, `channels`,\n *     `diarize` map to same-named form fields.\n */\nexport function buildTranscriptionFormData(options: {\n  file: File\n  language: string | undefined\n  modelOptions: GrokTranscriptionProviderOptions | undefined\n}): FormData {\n  const { file, language, modelOptions } = options\n  const form = new FormData()\n  form.set('file', file)\n  if (language) form.set('language', language)\n  if (modelOptions?.audio_format !== undefined) {\n    form.set('audio_format', modelOptions.audio_format)\n  }\n  if (modelOptions?.sample_rate !== undefined) {\n    form.set('sample_rate', String(modelOptions.sample_rate))\n  }\n  if (modelOptions?.inverse_text_normalization !== undefined) {\n    form.set(\n      'format',\n      modelOptions.inverse_text_normalization ? 'true' : 'false',\n    )\n  }\n  if (modelOptions?.multichannel !== undefined) {\n    form.set('multichannel', modelOptions.multichannel ? 'true' : 'false')\n  }\n  if (modelOptions?.channels !== undefined) {\n    form.set('channels', String(modelOptions.channels))\n  }\n  if (modelOptions?.diarize !== undefined) {\n    form.set('diarize', modelOptions.diarize ? 'true' : 'false')\n  }\n  return form\n}\n\n/**\n * Creates a Grok transcription adapter with an explicit API key.\n *\n * @example\n * ```typescript\n * const adapter = createGrokTranscription('grok-stt', 'xai-...')\n * const result = await generateTranscription({\n *   adapter,\n *   audio: audioFile,\n *   language: 'en',\n * })\n * ```\n */\nexport function createGrokTranscription<TModel extends GrokTranscriptionModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<GrokTranscriptionConfig, 'apiKey'>,\n): GrokTranscriptionAdapter<TModel> {\n  return new GrokTranscriptionAdapter({ apiKey, ...config }, model)\n}\n\n/**\n * Creates a Grok transcription adapter, reading the API key from\n * `XAI_API_KEY` in the environment.\n *\n * @throws Error if `XAI_API_KEY` is not set.\n */\nexport function grokTranscription<TModel extends GrokTranscriptionModel>(\n  model: TModel,\n  config?: Omit<GrokTranscriptionConfig, 'apiKey'>,\n): GrokTranscriptionAdapter<TModel> {\n  const apiKey = getGrokApiKeyFromEnv()\n  return createGrokTranscription(model, apiKey, config)\n}\n"],"names":[],"mappings":";;;;;AA2BA,MAAM,wBAAwB;AAyCvB,MAAM,iCAEH,yBAAmE;AAAA,EAClE,OAAO;AAAA,EAEC;AAAA,EACA;AAAA,EACA;AAAA,EAEjB,YAAY,QAAiC,OAAe;AAC1D,UAAM,OAAO,MAAM;AACnB,SAAK,SAAS,OAAO;AACrB,SAAK,WAAW,OAAO,WAAW,uBAAuB,QAAQ,QAAQ,EAAE;AAC3E,SAAK,iBAAiB,OAAO,kBAAkB,CAAA;AAAA,EACjD;AAAA,EAEA,MAAM,WACJ,SAC8B;AAC9B,UAAM,EAAE,WAAW;AACnB,UAAM,EAAE,OAAO,OAAO,UAAU,iBAAiB;AAEjD,WAAO;AAAA,MACL,sDAAsD,KAAK;AAAA,MAC3D,EAAE,UAAU,QAAQ,MAAA;AAAA,IAAM;AAG5B,UAAM,OAAO,YAAY,OAAO,cAAc,YAAY;AAC1D,UAAM,OAAO,2BAA2B,EAAE,MAAM,UAAU,cAAc;AAExE,QAAI;AACF,YAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,QAAQ;AAAA,QAClD,QAAQ;AAAA,QACR,SAAS;AAAA;AAAA,UAEP,GAAG,KAAK;AAAA,UACR,eAAe,UAAU,KAAK,MAAM;AAAA,QAAA;AAAA,QAEtC,MAAM;AAAA,MAAA,CACP;AAED,UAAI,CAAC,SAAS,IAAI;AAChB,cAAM,YAAY,MAAM,SAAS,KAAA;AACjC,cAAM,IAAI;AAAA,UACR,sCAAsC,SAAS,MAAM,IAAI,SAAS;AAAA,QAAA;AAAA,MAEtE;AAEA,YAAM,OAAQ,MAAM,SAAS,KAAA;AAE7B,YAAM,QAA8C,KAAK,OAAO;AAAA,QAC9D,CAAC,MAAM;AAML,gBAAM,KAA4B;AAAA,YAChC,MAAM,EAAE;AAAA,YACR,OAAO,EAAE;AAAA,YACT,KAAK,EAAE;AAAA,UAAA;AAET,cAAI,EAAE,eAAe,OAAW,IAAG,aAAa,EAAE;AAClD,cAAI,EAAE,YAAY,OAAW,IAAG,UAAU,EAAE;AAC5C,iBAAO;AAAA,QACT;AAAA,MAAA;AAGF,YAAM,mBAAmB,KAAK,YAAY;AAC1C,aAAO;AAAA,QACL,IAAI,WAAW,KAAK,IAAI;AAAA,QACxB;AAAA,QACA,MAAM,KAAK;AAAA,QACX,GAAI,qBAAqB,UAAa,EAAE,UAAU,iBAAA;AAAA,QAClD,UAAU,KAAK;AAAA,QACf,GAAI,UAAU,UAAa,EAAE,MAAA;AAAA,MAAM;AAAA,IAEvC,SAAS,OAAO;AACd,aAAO,OAAO,yBAAyB;AAAA,QACrC;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,YAAM;AAAA,IACR;AAAA,EACF;AACF;AAcO,SAAS,2BAA2B,SAI9B;AACX,QAAM,EAAE,MAAM,UAAU,aAAA,IAAiB;AACzC,QAAM,OAAO,IAAI,SAAA;AACjB,OAAK,IAAI,QAAQ,IAAI;AACrB,MAAI,SAAU,MAAK,IAAI,YAAY,QAAQ;AAC3C,MAAI,cAAc,iBAAiB,QAAW;AAC5C,SAAK,IAAI,gBAAgB,aAAa,YAAY;AAAA,EACpD;AACA,MAAI,cAAc,gBAAgB,QAAW;AAC3C,SAAK,IAAI,eAAe,OAAO,aAAa,WAAW,CAAC;AAAA,EAC1D;AACA,MAAI,cAAc,+BAA+B,QAAW;AAC1D,SAAK;AAAA,MACH;AAAA,MACA,aAAa,6BAA6B,SAAS;AAAA,IAAA;AAAA,EAEvD;AACA,MAAI,cAAc,iBAAiB,QAAW;AAC5C,SAAK,IAAI,gBAAgB,aAAa,eAAe,SAAS,OAAO;AAAA,EACvE;AACA,MAAI,cAAc,aAAa,QAAW;AACxC,SAAK,IAAI,YAAY,OAAO,aAAa,QAAQ,CAAC;AAAA,EACpD;AACA,MAAI,cAAc,YAAY,QAAW;AACvC,SAAK,IAAI,WAAW,aAAa,UAAU,SAAS,OAAO;AAAA,EAC7D;AACA,SAAO;AACT;AAeO,SAAS,wBACd,OACA,QACA,QACkC;AAClC,SAAO,IAAI,yBAAyB,EAAE,QAAQ,GAAG,OAAA,GAAU,KAAK;AAClE;AAQO,SAAS,kBACd,OACA,QACkC;AAClC,QAAM,SAAS,qBAAA;AACf,SAAO,wBAAwB,OAAO,QAAQ,MAAM;AACtD;"}
+{"version":3,"file":"transcription.js","sources":["../../../src/adapters/transcription.ts"],"sourcesContent":["import { BaseTranscriptionAdapter } from '@tanstack/ai/adapters'\nimport { generateId, getGrokApiKeyFromEnv, toAudioFile } from '../utils'\nimport type {\n  TokenUsage,\n  TranscriptionOptions,\n  TranscriptionResult,\n  TranscriptionWord,\n} from '@tanstack/ai'\nimport type { GrokTranscriptionModel } from '../model-meta'\nimport type { GrokTranscriptionProviderOptions } from '../audio/transcription-provider-options'\n\n/**\n * Grok-specific extension of `TranscriptionWord` that surfaces the extra\n * fields xAI returns when diarization / confidence are enabled. The base\n * cross-provider `TranscriptionWord` contract doesn't include these, so\n * callers who know they're using Grok can narrow with:\n *\n * ```ts\n * const words = result.words as Array<GrokTranscriptionWord> | undefined\n * ```\n */\nexport interface GrokTranscriptionWord extends TranscriptionWord {\n  /** Model confidence for the word, when xAI returns one. */\n  confidence?: number\n  /** Speaker index, populated when `modelOptions.diarize === true`. */\n  speaker?: number\n}\n\nconst DEFAULT_GROK_BASE_URL = 'https://api.x.ai/v1'\n\n/**\n * Configuration for the Grok transcription adapter.\n *\n * Uses direct `fetch` rather than the OpenAI SDK because xAI's `/v1/stt`\n * endpoint is not OpenAI-compatible.\n */\nexport interface GrokTranscriptionConfig {\n  apiKey: string\n  baseURL?: string\n  /** Additional headers to merge into every request (e.g., test IDs). */\n  defaultHeaders?: Record<string, string>\n}\n\n/**\n * xAI STT response shape from `POST /v1/stt`.\n * Grok returns word-level timestamps only; no segment array.\n */\ninterface GrokSTTWord {\n  text: string\n  start: number\n  end: number\n  confidence?: number\n  speaker?: number\n}\n\ninterface GrokSTTResponse {\n  text: string\n  language?: string\n  duration?: number\n  words?: Array<GrokSTTWord>\n  channels?: Array<unknown>\n}\n\n/**\n * Grok Speech-to-Text Adapter.\n *\n * Talks to `POST {baseURL}/stt` per\n * https://docs.x.ai/developers/rest-api-reference/inference/voice\n */\nexport class GrokTranscriptionAdapter<\n  TModel extends GrokTranscriptionModel,\n> extends BaseTranscriptionAdapter<TModel, GrokTranscriptionProviderOptions> {\n  readonly name = 'grok' as const\n\n  private readonly apiKey: string\n  private readonly baseURL: string\n  private readonly defaultHeaders: Record<string, string>\n\n  constructor(config: GrokTranscriptionConfig, model: TModel) {\n    super(model, config)\n    this.apiKey = config.apiKey\n    this.baseURL = (config.baseURL ?? DEFAULT_GROK_BASE_URL).replace(/\\/+$/, '')\n    this.defaultHeaders = config.defaultHeaders ?? {}\n  }\n\n  async transcribe(\n    options: TranscriptionOptions<GrokTranscriptionProviderOptions>,\n  ): Promise<TranscriptionResult> {\n    const { logger } = options\n    const { model, audio, language, modelOptions } = options\n\n    logger.request(\n      `activity=generateTranscription provider=grok model=${model}`,\n      { provider: 'grok', model },\n    )\n\n    const file = toAudioFile(audio, modelOptions?.audio_format)\n    const form = buildTranscriptionFormData({ file, language, modelOptions })\n\n    try {\n      const response = await fetch(`${this.baseURL}/stt`, {\n        method: 'POST',\n        headers: {\n          // `defaultHeaders` first so Authorization always wins.\n          ...this.defaultHeaders,\n          Authorization: `Bearer ${this.apiKey}`,\n        },\n        body: form,\n      })\n\n      if (!response.ok) {\n        const errorText = await response.text()\n        throw new Error(\n          `Grok transcription request failed: ${response.status} ${errorText}`,\n        )\n      }\n\n      const data = (await response.json()) as GrokSTTResponse\n\n      const words: Array<TranscriptionWord> | undefined = data.words?.map(\n        (w) => {\n          // Construct a GrokTranscriptionWord so that `confidence` and\n          // `speaker` (when xAI returns them under `diarize` / confidence\n          // mode) are preserved on the result. The returned array is typed\n          // as `Array<TranscriptionWord>` per the cross-provider contract;\n          // callers who want the extras narrow via `as Array<GrokTranscriptionWord>`.\n          const tw: GrokTranscriptionWord = {\n            word: w.text,\n            start: w.start,\n            end: w.end,\n          }\n          if (w.confidence !== undefined) tw.confidence = w.confidence\n          if (w.speaker !== undefined) tw.speaker = w.speaker\n          return tw\n        },\n      )\n\n      const resolvedLanguage = data.language ?? language\n      // xAI's /v1/stt response carries no token counts — STT is duration-billed —\n      // so surface the audio duration as `durationSeconds`, mirroring the\n      // whisper-1 path in the OpenAI transcription adapter.\n      const usage: TokenUsage | undefined =\n        data.duration !== undefined && data.duration > 0\n          ? {\n              promptTokens: 0,\n              completionTokens: 0,\n              totalTokens: 0,\n              durationSeconds: data.duration,\n            }\n          : undefined\n      return {\n        id: generateId(this.name),\n        model,\n        text: data.text,\n        ...(resolvedLanguage !== undefined && { language: resolvedLanguage }),\n        duration: data.duration,\n        ...(words !== undefined && { words }),\n        ...(usage !== undefined && { usage }),\n      }\n    } catch (error) {\n      logger.errors('grok.transcribe fatal', {\n        error,\n        source: 'grok.transcribe',\n      })\n      throw error\n    }\n  }\n}\n\n/**\n * Build the multipart/form-data body for `POST /v1/stt`, coercing SDK-level\n * model options into xAI's wire format (booleans as `'true'`/`'false'`\n * strings, numeric fields stringified, etc.).\n *\n * Wire-field mapping:\n *   - `modelOptions.inverse_text_normalization` → `format` (xAI's chosen\n *     wire-field name for the ITN boolean; the SDK surfaces it under the\n *     clearer `inverse_text_normalization` key).\n *   - `modelOptions.audio_format`, `sample_rate`, `multichannel`, `channels`,\n *     `diarize` map to same-named form fields.\n */\nexport function buildTranscriptionFormData(options: {\n  file: File\n  language: string | undefined\n  modelOptions: GrokTranscriptionProviderOptions | undefined\n}): FormData {\n  const { file, language, modelOptions } = options\n  const form = new FormData()\n  form.set('file', file)\n  if (language) form.set('language', language)\n  if (modelOptions?.audio_format !== undefined) {\n    form.set('audio_format', modelOptions.audio_format)\n  }\n  if (modelOptions?.sample_rate !== undefined) {\n    form.set('sample_rate', String(modelOptions.sample_rate))\n  }\n  if (modelOptions?.inverse_text_normalization !== undefined) {\n    form.set(\n      'format',\n      modelOptions.inverse_text_normalization ? 'true' : 'false',\n    )\n  }\n  if (modelOptions?.multichannel !== undefined) {\n    form.set('multichannel', modelOptions.multichannel ? 'true' : 'false')\n  }\n  if (modelOptions?.channels !== undefined) {\n    form.set('channels', String(modelOptions.channels))\n  }\n  if (modelOptions?.diarize !== undefined) {\n    form.set('diarize', modelOptions.diarize ? 'true' : 'false')\n  }\n  return form\n}\n\n/**\n * Creates a Grok transcription adapter with an explicit API key.\n *\n * @example\n * ```typescript\n * const adapter = createGrokTranscription('grok-stt', 'xai-...')\n * const result = await generateTranscription({\n *   adapter,\n *   audio: audioFile,\n *   language: 'en',\n * })\n * ```\n */\nexport function createGrokTranscription<TModel extends GrokTranscriptionModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<GrokTranscriptionConfig, 'apiKey'>,\n): GrokTranscriptionAdapter<TModel> {\n  return new GrokTranscriptionAdapter({ apiKey, ...config }, model)\n}\n\n/**\n * Creates a Grok transcription adapter, reading the API key from\n * `XAI_API_KEY` in the environment.\n *\n * @throws Error if `XAI_API_KEY` is not set.\n */\nexport function grokTranscription<TModel extends GrokTranscriptionModel>(\n  model: TModel,\n  config?: Omit<GrokTranscriptionConfig, 'apiKey'>,\n): GrokTranscriptionAdapter<TModel> {\n  const apiKey = getGrokApiKeyFromEnv()\n  return createGrokTranscription(model, apiKey, config)\n}\n"],"names":[],"mappings":";;;;;AA4BA,MAAM,wBAAwB;AAyCvB,MAAM,iCAEH,yBAAmE;AAAA,EAClE,OAAO;AAAA,EAEC;AAAA,EACA;AAAA,EACA;AAAA,EAEjB,YAAY,QAAiC,OAAe;AAC1D,UAAM,OAAO,MAAM;AACnB,SAAK,SAAS,OAAO;AACrB,SAAK,WAAW,OAAO,WAAW,uBAAuB,QAAQ,QAAQ,EAAE;AAC3E,SAAK,iBAAiB,OAAO,kBAAkB,CAAA;AAAA,EACjD;AAAA,EAEA,MAAM,WACJ,SAC8B;AAC9B,UAAM,EAAE,WAAW;AACnB,UAAM,EAAE,OAAO,OAAO,UAAU,iBAAiB;AAEjD,WAAO;AAAA,MACL,sDAAsD,KAAK;AAAA,MAC3D,EAAE,UAAU,QAAQ,MAAA;AAAA,IAAM;AAG5B,UAAM,OAAO,YAAY,OAAO,cAAc,YAAY;AAC1D,UAAM,OAAO,2BAA2B,EAAE,MAAM,UAAU,cAAc;AAExE,QAAI;AACF,YAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,QAAQ;AAAA,QAClD,QAAQ;AAAA,QACR,SAAS;AAAA;AAAA,UAEP,GAAG,KAAK;AAAA,UACR,eAAe,UAAU,KAAK,MAAM;AAAA,QAAA;AAAA,QAEtC,MAAM;AAAA,MAAA,CACP;AAED,UAAI,CAAC,SAAS,IAAI;AAChB,cAAM,YAAY,MAAM,SAAS,KAAA;AACjC,cAAM,IAAI;AAAA,UACR,sCAAsC,SAAS,MAAM,IAAI,SAAS;AAAA,QAAA;AAAA,MAEtE;AAEA,YAAM,OAAQ,MAAM,SAAS,KAAA;AAE7B,YAAM,QAA8C,KAAK,OAAO;AAAA,QAC9D,CAAC,MAAM;AAML,gBAAM,KAA4B;AAAA,YAChC,MAAM,EAAE;AAAA,YACR,OAAO,EAAE;AAAA,YACT,KAAK,EAAE;AAAA,UAAA;AAET,cAAI,EAAE,eAAe,OAAW,IAAG,aAAa,EAAE;AAClD,cAAI,EAAE,YAAY,OAAW,IAAG,UAAU,EAAE;AAC5C,iBAAO;AAAA,QACT;AAAA,MAAA;AAGF,YAAM,mBAAmB,KAAK,YAAY;AAI1C,YAAM,QACJ,KAAK,aAAa,UAAa,KAAK,WAAW,IAC3C;AAAA,QACE,cAAc;AAAA,QACd,kBAAkB;AAAA,QAClB,aAAa;AAAA,QACb,iBAAiB,KAAK;AAAA,MAAA,IAExB;AACN,aAAO;AAAA,QACL,IAAI,WAAW,KAAK,IAAI;AAAA,QACxB;AAAA,QACA,MAAM,KAAK;AAAA,QACX,GAAI,qBAAqB,UAAa,EAAE,UAAU,iBAAA;AAAA,QAClD,UAAU,KAAK;AAAA,QACf,GAAI,UAAU,UAAa,EAAE,MAAA;AAAA,QAC7B,GAAI,UAAU,UAAa,EAAE,MAAA;AAAA,MAAM;AAAA,IAEvC,SAAS,OAAO;AACd,aAAO,OAAO,yBAAyB;AAAA,QACrC;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,YAAM;AAAA,IACR;AAAA,EACF;AACF;AAcO,SAAS,2BAA2B,SAI9B;AACX,QAAM,EAAE,MAAM,UAAU,aAAA,IAAiB;AACzC,QAAM,OAAO,IAAI,SAAA;AACjB,OAAK,IAAI,QAAQ,IAAI;AACrB,MAAI,SAAU,MAAK,IAAI,YAAY,QAAQ;AAC3C,MAAI,cAAc,iBAAiB,QAAW;AAC5C,SAAK,IAAI,gBAAgB,aAAa,YAAY;AAAA,EACpD;AACA,MAAI,cAAc,gBAAgB,QAAW;AAC3C,SAAK,IAAI,eAAe,OAAO,aAAa,WAAW,CAAC;AAAA,EAC1D;AACA,MAAI,cAAc,+BAA+B,QAAW;AAC1D,SAAK;AAAA,MACH;AAAA,MACA,aAAa,6BAA6B,SAAS;AAAA,IAAA;AAAA,EAEvD;AACA,MAAI,cAAc,iBAAiB,QAAW;AAC5C,SAAK,IAAI,gBAAgB,aAAa,eAAe,SAAS,OAAO;AAAA,EACvE;AACA,MAAI,cAAc,aAAa,QAAW;AACxC,SAAK,IAAI,YAAY,OAAO,aAAa,QAAQ,CAAC;AAAA,EACpD;AACA,MAAI,cAAc,YAAY,QAAW;AACvC,SAAK,IAAI,WAAW,aAAa,UAAU,SAAS,OAAO;AAAA,EAC7D;AACA,SAAO;AACT;AAeO,SAAS,wBACd,OACA,QACA,QACkC;AAClC,SAAO,IAAI,yBAAyB,EAAE,QAAQ,GAAG,OAAA,GAAU,KAAK;AAClE;AAQO,SAAS,kBACd,OACA,QACkC;AAClC,QAAM,SAAS,qBAAA;AACf,SAAO,wBAAwB,OAAO,QAAQ,MAAM;AACtD;"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tanstack/ai-grok",
-  "version": "0.9.2",
+  "version": "0.10.1",
   "description": "xAI Grok adapter for TanStack AI chat, image generation, realtime, and structured outputs.",
   "author": "",
   "license": "MIT",
@@ -43,17 +43,17 @@
   "dependencies": {
     "openai": "^6.9.1",
     "@tanstack/ai-utils": "0.2.1",
-    "@tanstack/openai-base": "0.5.0"
+    "@tanstack/openai-base": "0.6.1"
   },
   "devDependencies": {
     "@vitest/coverage-v8": "4.0.14",
     "vite": "^7.3.3",
-    "@tanstack/ai": "0.24.0",
-    "@tanstack/ai-client": "0.15.0"
+    "@tanstack/ai": "0.26.0",
+    "@tanstack/ai-client": "0.15.2"
   },
   "peerDependencies": {
     "zod": "^4.0.0",
-    "@tanstack/ai": "^0.24.0"
+    "@tanstack/ai": "^0.26.0"
   },
   "scripts": {
     "build": "vite build",

package/src/adapters/image.ts

@@ -1,6 +1,7 @@
 import OpenAI from 'openai'
 import { BaseImageAdapter } from '@tanstack/ai/adapters'
 import { toRunErrorPayload } from '@tanstack/ai/adapter-internals'
+import { buildImagesUsage } from '@tanstack/openai-base'
 import { generateId } from '@tanstack/ai-utils'
 import { getGrokApiKeyFromEnv, withGrokDefaults } from '../utils/client'
 import {
@@ -105,17 +106,13 @@ export class GrokImageAdapter<
         },
       )
 
+      const usage = buildImagesUsage(response.usage)
+
       return {
         id: generateId(this.name),
         model,
         images,
-        ...(response.usage && {
-          usage: {
-            inputTokens: response.usage.input_tokens,
-            outputTokens: response.usage.output_tokens,
-            totalTokens: response.usage.total_tokens,
-          },
-        }),
+        ...(usage ? { usage } : {}),
       }
     } catch (error: unknown) {
       options.logger.errors(`${this.name}.generateImages fatal`, {

package/src/adapters/transcription.ts

@@ -1,6 +1,7 @@
 import { BaseTranscriptionAdapter } from '@tanstack/ai/adapters'
 import { generateId, getGrokApiKeyFromEnv, toAudioFile } from '../utils'
 import type {
+  TokenUsage,
   TranscriptionOptions,
   TranscriptionResult,
   TranscriptionWord,
@@ -135,6 +136,18 @@ export class GrokTranscriptionAdapter<
       )
 
       const resolvedLanguage = data.language ?? language
+      // xAI's /v1/stt response carries no token counts — STT is duration-billed —
+      // so surface the audio duration as `durationSeconds`, mirroring the
+      // whisper-1 path in the OpenAI transcription adapter.
+      const usage: TokenUsage | undefined =
+        data.duration !== undefined && data.duration > 0
+          ? {
+              promptTokens: 0,
+              completionTokens: 0,
+              totalTokens: 0,
+              durationSeconds: data.duration,
+            }
+          : undefined
       return {
         id: generateId(this.name),
         model,
@@ -142,6 +155,7 @@ export class GrokTranscriptionAdapter<
         ...(resolvedLanguage !== undefined && { language: resolvedLanguage }),
         duration: data.duration,
         ...(words !== undefined && { words }),
+        ...(usage !== undefined && { usage }),
       }
     } catch (error) {
       logger.errors('grok.transcribe fatal', {