@tanstack/ai-gemini 0.11.0 → 0.12.1

package/dist/esm/adapters/image.js

@@ -93,7 +93,17 @@ class GeminiImageAdapter extends BaseImageAdapter {
     return {
       id: generateId(this.name),
       model,
-      images
+      images,
+      // Surface token usage when the model reports it (e.g. Nano Banana via
+      // generateContent). Conditionally spread to satisfy
+      // exactOptionalPropertyTypes — only include usage when present. See #330.
+      ...response.usageMetadata && {
+        usage: {
+          inputTokens: response.usageMetadata.promptTokenCount ?? 0,
+          outputTokens: response.usageMetadata.candidatesTokenCount ?? 0,
+          totalTokens: response.usageMetadata.totalTokenCount ?? 0
+        }
+      }
     };
   }
   buildImagenConfig(options) {

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

@@ -1 +1 @@
-{"version":3,"file":"image.js","sources":["../../../src/adapters/image.ts"],"sourcesContent":["import { BaseImageAdapter } from '@tanstack/ai/adapters'\nimport {\n  createGeminiClient,\n  generateId,\n  getGeminiApiKeyFromEnv,\n} from '../utils'\nimport {\n  parseNativeImageSize,\n  sizeToAspectRatio,\n  validateImageSize,\n  validateNumberOfImages,\n  validatePrompt,\n} from '../image/image-provider-options'\nimport type { GEMINI_IMAGE_MODELS } from '../model-meta'\nimport type {\n  GeminiImageModelProviderOptionsByName,\n  GeminiImageModelSizeByName,\n  GeminiImageProviderOptions,\n} from '../image/image-provider-options'\nimport type {\n  GeneratedImage,\n  ImageGenerationOptions,\n  ImageGenerationResult,\n} from '@tanstack/ai'\nimport type {\n  GenerateContentConfig,\n  GenerateContentResponse,\n  GenerateImagesConfig,\n  GenerateImagesResponse,\n  GoogleGenAI,\n} from '@google/genai'\nimport type { GeminiClientConfig } from '../utils'\n\n/**\n * Configuration for Gemini image adapter\n */\nexport interface GeminiImageConfig extends GeminiClientConfig {}\n\n/** Model type for Gemini Image */\nexport type GeminiImageModel = (typeof GEMINI_IMAGE_MODELS)[number]\n\n/**\n * Gemini Image Generation Adapter\n *\n * Tree-shakeable adapter for Gemini image generation functionality.\n * Supports Imagen 3/4 models (via generateImages API) and Gemini native\n * image models like Nano Banana 2 (via generateContent API).\n *\n * Features:\n * - Aspect ratio-based image sizing\n * - Person generation controls\n * - Safety filtering\n * - Watermark options\n * - Extended resolution tiers (Nano Banana 2)\n */\nexport class GeminiImageAdapter<\n  TModel extends GeminiImageModel,\n> extends BaseImageAdapter<\n  TModel,\n  GeminiImageProviderOptions,\n  GeminiImageModelProviderOptionsByName,\n  GeminiImageModelSizeByName\n> {\n  override readonly kind = 'image' as const\n  readonly name = 'gemini' as const\n\n  // Type-only property - never assigned at runtime\n  declare '~types': {\n    providerOptions: GeminiImageProviderOptions\n    modelProviderOptionsByName: GeminiImageModelProviderOptionsByName\n    modelSizeByName: GeminiImageModelSizeByName\n  }\n\n  private readonly client: GoogleGenAI\n\n  constructor(config: GeminiImageConfig, model: TModel) {\n    super(model, config)\n    this.client = createGeminiClient(config)\n  }\n\n  async generateImages(\n    options: ImageGenerationOptions<GeminiImageProviderOptions>,\n  ): Promise<ImageGenerationResult> {\n    const { model, prompt, logger } = options\n\n    logger.request(\n      `activity=generateImage provider=gemini model=${this.model}`,\n      {\n        provider: 'gemini',\n        model: this.model,\n      },\n    )\n\n    try {\n      validatePrompt({ prompt, model })\n\n      if (this.isGeminiImageModel(model)) {\n        return await this.generateWithGeminiApi(options)\n      }\n\n      // Imagen models path (generateImages API)\n      validateImageSize(model, options.size)\n      validateNumberOfImages(model, options.numberOfImages)\n\n      const config = this.buildImagenConfig(options)\n\n      const response = await this.client.models.generateImages({\n        model,\n        prompt,\n        config,\n      })\n\n      return this.transformImagenResponse(model, response)\n    } catch (error) {\n      logger.errors('gemini.generateImage fatal', {\n        error,\n        source: 'gemini.generateImage',\n      })\n      throw error\n    }\n  }\n\n  private isGeminiImageModel(model: string): boolean {\n    return model.startsWith('gemini-')\n  }\n\n  private async generateWithGeminiApi(\n    options: ImageGenerationOptions<GeminiImageProviderOptions>,\n  ): Promise<ImageGenerationResult> {\n    const { model, prompt, size, numberOfImages, modelOptions } = options\n\n    const parsedSize = size ? parseNativeImageSize(size) : undefined\n\n    // The generateContent API has no numberOfImages parameter.\n    // Instead, augment the prompt to request multiple images when needed.\n    const augmentedPrompt =\n      numberOfImages && numberOfImages > 1\n        ? `${prompt} Generate ${numberOfImages} distinct images.`\n        : prompt\n\n    // GeminiImageProviderOptions is Imagen-shaped — most fields\n    // (personGeneration, safetyFilterLevel, addWatermark, outputMimeType,\n    // outputCompressionQuality, guidanceScale, enhancePrompt,\n    // includeSafetyAttributes, includeRaiReason, outputGcsUri, labels,\n    // negativePrompt, language) are only valid on GenerateImagesConfig and\n    // would be rejected by the Gemini-native generateContent path. Pick only\n    // the fields that are valid on GenerateContentConfig instead of spreading\n    // the whole options object.\n    const nativeConfig: GenerateContentConfig = {}\n    if (modelOptions?.seed !== undefined) {\n      nativeConfig.seed = modelOptions.seed\n    }\n\n    const config: GenerateContentConfig = {\n      ...nativeConfig,\n      // Include TEXT so the model can interleave descriptions between images.\n      // IMPORTANT: responseModalities is a protected default — set it AFTER\n      // nativeConfig so nothing can silently disable image output.\n      responseModalities: ['TEXT', 'IMAGE'],\n      ...(parsedSize && {\n        imageConfig: {\n          ...(parsedSize.aspectRatio && {\n            aspectRatio: parsedSize.aspectRatio,\n          }),\n          ...(parsedSize.resolution && {\n            imageSize: parsedSize.resolution,\n          }),\n        },\n      }),\n    }\n\n    const response = await this.client.models.generateContent({\n      model,\n      contents: augmentedPrompt,\n      config,\n    })\n\n    return this.transformGeminiResponse(model, response)\n  }\n\n  private transformGeminiResponse(\n    model: string,\n    response: GenerateContentResponse,\n  ): ImageGenerationResult {\n    const images: Array<GeneratedImage> = []\n    const textParts: Array<string> = []\n    const parts = response.candidates?.[0]?.content?.parts ?? []\n\n    for (const part of parts) {\n      if (\n        part.inlineData?.data &&\n        typeof part.inlineData.data === 'string' &&\n        part.inlineData.data.length > 0\n      ) {\n        images.push({ b64Json: part.inlineData.data })\n      } else if (typeof part.text === 'string' && part.text.length > 0) {\n        textParts.push(part.text)\n      }\n    }\n\n    // If the model returned only text parts (for example a safety refusal\n    // or a \"can't do that\" message), surface the text instead of silently\n    // resolving to an empty images array — otherwise callers can't tell a\n    // generation failure apart from a genuine empty response.\n    if (images.length === 0) {\n      const reason =\n        textParts.length > 0\n          ? `: ${textParts.join(' ').trim()}`\n          : ' (no inline image or text parts were returned).'\n      throw new Error(`Gemini ${model} returned no images${reason}`)\n    }\n\n    return {\n      id: generateId(this.name),\n      model,\n      images,\n    }\n  }\n\n  private buildImagenConfig(\n    options: ImageGenerationOptions<GeminiImageProviderOptions>,\n  ): GenerateImagesConfig {\n    const { size, numberOfImages, modelOptions } = options\n\n    // Build with conditional spreads — under exactOptionalPropertyTypes the\n    // vendor `GenerateImagesConfig` fields are `field?: T` (no `| undefined`),\n    // so we can only assign the property when we actually have a value.\n    const sizeAspectRatio = size ? sizeToAspectRatio(size) : undefined\n    return {\n      numberOfImages: numberOfImages ?? 1,\n      // Map size to aspect ratio if provided (modelOptions.aspectRatio will override)\n      ...(sizeAspectRatio !== undefined && { aspectRatio: sizeAspectRatio }),\n      ...modelOptions,\n    }\n  }\n\n  private transformImagenResponse(\n    model: string,\n    response: GenerateImagesResponse,\n  ): ImageGenerationResult {\n    const entries = response.generatedImages ?? []\n    const images: Array<GeneratedImage> = []\n    const filterReasons: Array<string> = []\n\n    for (const item of entries) {\n      const b64Json = item.image?.imageBytes\n      if (b64Json) {\n        images.push({\n          b64Json,\n          ...(item.enhancedPrompt !== undefined && {\n            revisedPrompt: item.enhancedPrompt,\n          }),\n        })\n        continue\n      }\n      // Imagen can drop individual entries with a raiFilteredReason when\n      // Responsible-AI filters fire. Preserve the reason so callers can\n      // surface it instead of silently getting back fewer images.\n      const reason = (item as { raiFilteredReason?: string }).raiFilteredReason\n      if (reason) {\n        filterReasons.push(reason)\n      }\n    }\n\n    // Every entry was filtered — no usable images to return. Throw rather\n    // than resolve to an empty array so the caller is forced to handle the\n    // failure mode explicitly.\n    if (entries.length > 0 && images.length === 0) {\n      const joined = filterReasons.length > 0 ? filterReasons.join('; ') : ''\n      throw new Error(\n        `Imagen ${model} returned no images: all ${entries.length} generated image(s) were filtered by Responsible-AI${joined ? ` (${joined})` : ''}.`,\n      )\n    }\n\n    // Partial filter: surface via console.warn since ImageGenerationResult\n    // has no warnings field. Callers that care can still inspect the count\n    // mismatch between requested and returned images.\n    if (filterReasons.length > 0 && typeof console !== 'undefined') {\n      console.warn(\n        `[gemini-image] ${filterReasons.length} of ${entries.length} images from ${model} were filtered by Responsible-AI: ${filterReasons.join('; ')}`,\n      )\n    }\n\n    return {\n      id: generateId(this.name),\n      model,\n      images,\n    }\n  }\n}\n\n/**\n * Creates a Gemini image adapter with explicit API key.\n * Type resolution happens here at the call site.\n *\n * @param model - The model name (e.g., 'imagen-3.0-generate-002')\n * @param apiKey - Your Google API key\n * @param config - Optional additional configuration\n * @returns Configured Gemini image adapter instance with resolved types\n *\n * @example\n * ```typescript\n * const adapter = createGeminiImage('imagen-3.0-generate-002', \"your-api-key\");\n *\n * const result = await generateImage({\n *   adapter,\n *   prompt: 'A cute baby sea otter'\n * });\n * ```\n */\nexport function createGeminiImage<TModel extends GeminiImageModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<GeminiImageConfig, 'apiKey'>,\n): GeminiImageAdapter<TModel> {\n  return new GeminiImageAdapter({ apiKey, ...config }, model)\n}\n\n/**\n * Creates a Gemini image adapter with automatic API key detection from environment variables.\n * Type resolution happens here at the call site.\n *\n * Looks for `GOOGLE_API_KEY` or `GEMINI_API_KEY` in:\n * - `process.env` (Node.js)\n * - `window.env` (Browser with injected env)\n *\n * @param model - The model name (e.g., 'imagen-4.0-generate-001')\n * @param config - Optional configuration (excluding apiKey which is auto-detected)\n * @returns Configured Gemini image adapter instance with resolved types\n * @throws Error if GOOGLE_API_KEY or GEMINI_API_KEY is not found in environment\n *\n * @example\n * ```typescript\n * // Automatically uses GOOGLE_API_KEY from environment\n * const adapter = geminiImage('imagen-4.0-generate-001');\n *\n * const result = await generateImage({\n *   adapter,\n *   prompt: 'A beautiful sunset over mountains'\n * });\n * ```\n */\nexport function geminiImage<TModel extends GeminiImageModel>(\n  model: TModel,\n  config?: Omit<GeminiImageConfig, 'apiKey'>,\n): GeminiImageAdapter<TModel> {\n  const apiKey = getGeminiApiKeyFromEnv()\n  return createGeminiImage(model, apiKey, config)\n}\n"],"names":[],"mappings":";;;AAuDO,MAAM,2BAEH,iBAKR;AAAA,EACkB,OAAO;AAAA,EAChB,OAAO;AAAA,EASC;AAAA,EAEjB,YAAY,QAA2B,OAAe;AACpD,UAAM,OAAO,MAAM;AACnB,SAAK,SAAS,mBAAmB,MAAM;AAAA,EACzC;AAAA,EAEA,MAAM,eACJ,SACgC;AAChC,UAAM,EAAE,OAAO,QAAQ,OAAA,IAAW;AAElC,WAAO;AAAA,MACL,gDAAgD,KAAK,KAAK;AAAA,MAC1D;AAAA,QACE,UAAU;AAAA,QACV,OAAO,KAAK;AAAA,MAAA;AAAA,IACd;AAGF,QAAI;AACF,qBAAe,EAAE,QAAQ,OAAO;AAEhC,UAAI,KAAK,mBAAmB,KAAK,GAAG;AAClC,eAAO,MAAM,KAAK,sBAAsB,OAAO;AAAA,MACjD;AAGA,wBAAkB,OAAO,QAAQ,IAAI;AACrC,6BAAuB,OAAO,QAAQ,cAAc;AAEpD,YAAM,SAAS,KAAK,kBAAkB,OAAO;AAE7C,YAAM,WAAW,MAAM,KAAK,OAAO,OAAO,eAAe;AAAA,QACvD;AAAA,QACA;AAAA,QACA;AAAA,MAAA,CACD;AAED,aAAO,KAAK,wBAAwB,OAAO,QAAQ;AAAA,IACrD,SAAS,OAAO;AACd,aAAO,OAAO,8BAA8B;AAAA,QAC1C;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEQ,mBAAmB,OAAwB;AACjD,WAAO,MAAM,WAAW,SAAS;AAAA,EACnC;AAAA,EAEA,MAAc,sBACZ,SACgC;AAChC,UAAM,EAAE,OAAO,QAAQ,MAAM,gBAAgB,iBAAiB;AAE9D,UAAM,aAAa,OAAO,qBAAqB,IAAI,IAAI;AAIvD,UAAM,kBACJ,kBAAkB,iBAAiB,IAC/B,GAAG,MAAM,aAAa,cAAc,sBACpC;AAUN,UAAM,eAAsC,CAAA;AAC5C,QAAI,cAAc,SAAS,QAAW;AACpC,mBAAa,OAAO,aAAa;AAAA,IACnC;AAEA,UAAM,SAAgC;AAAA,MACpC,GAAG;AAAA;AAAA;AAAA;AAAA,MAIH,oBAAoB,CAAC,QAAQ,OAAO;AAAA,MACpC,GAAI,cAAc;AAAA,QAChB,aAAa;AAAA,UACX,GAAI,WAAW,eAAe;AAAA,YAC5B,aAAa,WAAW;AAAA,UAAA;AAAA,UAE1B,GAAI,WAAW,cAAc;AAAA,YAC3B,WAAW,WAAW;AAAA,UAAA;AAAA,QACxB;AAAA,MACF;AAAA,IACF;AAGF,UAAM,WAAW,MAAM,KAAK,OAAO,OAAO,gBAAgB;AAAA,MACxD;AAAA,MACA,UAAU;AAAA,MACV;AAAA,IAAA,CACD;AAED,WAAO,KAAK,wBAAwB,OAAO,QAAQ;AAAA,EACrD;AAAA,EAEQ,wBACN,OACA,UACuB;AACvB,UAAM,SAAgC,CAAA;AACtC,UAAM,YAA2B,CAAA;AACjC,UAAM,QAAQ,SAAS,aAAa,CAAC,GAAG,SAAS,SAAS,CAAA;AAE1D,eAAW,QAAQ,OAAO;AACxB,UACE,KAAK,YAAY,QACjB,OAAO,KAAK,WAAW,SAAS,YAChC,KAAK,WAAW,KAAK,SAAS,GAC9B;AACA,eAAO,KAAK,EAAE,SAAS,KAAK,WAAW,MAAM;AAAA,MAC/C,WAAW,OAAO,KAAK,SAAS,YAAY,KAAK,KAAK,SAAS,GAAG;AAChE,kBAAU,KAAK,KAAK,IAAI;AAAA,MAC1B;AAAA,IACF;AAMA,QAAI,OAAO,WAAW,GAAG;AACvB,YAAM,SACJ,UAAU,SAAS,IACf,KAAK,UAAU,KAAK,GAAG,EAAE,KAAA,CAAM,KAC/B;AACN,YAAM,IAAI,MAAM,UAAU,KAAK,sBAAsB,MAAM,EAAE;AAAA,IAC/D;AAEA,WAAO;AAAA,MACL,IAAI,WAAW,KAAK,IAAI;AAAA,MACxB;AAAA,MACA;AAAA,IAAA;AAAA,EAEJ;AAAA,EAEQ,kBACN,SACsB;AACtB,UAAM,EAAE,MAAM,gBAAgB,aAAA,IAAiB;AAK/C,UAAM,kBAAkB,OAAO,kBAAkB,IAAI,IAAI;AACzD,WAAO;AAAA,MACL,gBAAgB,kBAAkB;AAAA;AAAA,MAElC,GAAI,oBAAoB,UAAa,EAAE,aAAa,gBAAA;AAAA,MACpD,GAAG;AAAA,IAAA;AAAA,EAEP;AAAA,EAEQ,wBACN,OACA,UACuB;AACvB,UAAM,UAAU,SAAS,mBAAmB,CAAA;AAC5C,UAAM,SAAgC,CAAA;AACtC,UAAM,gBAA+B,CAAA;AAErC,eAAW,QAAQ,SAAS;AAC1B,YAAM,UAAU,KAAK,OAAO;AAC5B,UAAI,SAAS;AACX,eAAO,KAAK;AAAA,UACV;AAAA,UACA,GAAI,KAAK,mBAAmB,UAAa;AAAA,YACvC,eAAe,KAAK;AAAA,UAAA;AAAA,QACtB,CACD;AACD;AAAA,MACF;AAIA,YAAM,SAAU,KAAwC;AACxD,UAAI,QAAQ;AACV,sBAAc,KAAK,MAAM;AAAA,MAC3B;AAAA,IACF;AAKA,QAAI,QAAQ,SAAS,KAAK,OAAO,WAAW,GAAG;AAC7C,YAAM,SAAS,cAAc,SAAS,IAAI,cAAc,KAAK,IAAI,IAAI;AACrE,YAAM,IAAI;AAAA,QACR,UAAU,KAAK,4BAA4B,QAAQ,MAAM,sDAAsD,SAAS,KAAK,MAAM,MAAM,EAAE;AAAA,MAAA;AAAA,IAE/I;AAKA,QAAI,cAAc,SAAS,KAAK,OAAO,YAAY,aAAa;AAC9D,cAAQ;AAAA,QACN,kBAAkB,cAAc,MAAM,OAAO,QAAQ,MAAM,gBAAgB,KAAK,qCAAqC,cAAc,KAAK,IAAI,CAAC;AAAA,MAAA;AAAA,IAEjJ;AAEA,WAAO;AAAA,MACL,IAAI,WAAW,KAAK,IAAI;AAAA,MACxB;AAAA,MACA;AAAA,IAAA;AAAA,EAEJ;AACF;AAqBO,SAAS,kBACd,OACA,QACA,QAC4B;AAC5B,SAAO,IAAI,mBAAmB,EAAE,QAAQ,GAAG,OAAA,GAAU,KAAK;AAC5D;AA0BO,SAAS,YACd,OACA,QAC4B;AAC5B,QAAM,SAAS,uBAAA;AACf,SAAO,kBAAkB,OAAO,QAAQ,MAAM;AAChD;"}
+{"version":3,"file":"image.js","sources":["../../../src/adapters/image.ts"],"sourcesContent":["import { BaseImageAdapter } from '@tanstack/ai/adapters'\nimport {\n  createGeminiClient,\n  generateId,\n  getGeminiApiKeyFromEnv,\n} from '../utils'\nimport {\n  parseNativeImageSize,\n  sizeToAspectRatio,\n  validateImageSize,\n  validateNumberOfImages,\n  validatePrompt,\n} from '../image/image-provider-options'\nimport type { GEMINI_IMAGE_MODELS } from '../model-meta'\nimport type {\n  GeminiImageModelProviderOptionsByName,\n  GeminiImageModelSizeByName,\n  GeminiImageProviderOptions,\n} from '../image/image-provider-options'\nimport type {\n  GeneratedImage,\n  ImageGenerationOptions,\n  ImageGenerationResult,\n} from '@tanstack/ai'\nimport type {\n  GenerateContentConfig,\n  GenerateContentResponse,\n  GenerateImagesConfig,\n  GenerateImagesResponse,\n  GoogleGenAI,\n} from '@google/genai'\nimport type { GeminiClientConfig } from '../utils'\n\n/**\n * Configuration for Gemini image adapter\n */\nexport interface GeminiImageConfig extends GeminiClientConfig {}\n\n/** Model type for Gemini Image */\nexport type GeminiImageModel = (typeof GEMINI_IMAGE_MODELS)[number]\n\n/**\n * Gemini Image Generation Adapter\n *\n * Tree-shakeable adapter for Gemini image generation functionality.\n * Supports Imagen 3/4 models (via generateImages API) and Gemini native\n * image models like Nano Banana 2 (via generateContent API).\n *\n * Features:\n * - Aspect ratio-based image sizing\n * - Person generation controls\n * - Safety filtering\n * - Watermark options\n * - Extended resolution tiers (Nano Banana 2)\n */\nexport class GeminiImageAdapter<\n  TModel extends GeminiImageModel,\n> extends BaseImageAdapter<\n  TModel,\n  GeminiImageProviderOptions,\n  GeminiImageModelProviderOptionsByName,\n  GeminiImageModelSizeByName\n> {\n  override readonly kind = 'image' as const\n  readonly name = 'gemini' as const\n\n  // Type-only property - never assigned at runtime\n  declare '~types': {\n    providerOptions: GeminiImageProviderOptions\n    modelProviderOptionsByName: GeminiImageModelProviderOptionsByName\n    modelSizeByName: GeminiImageModelSizeByName\n  }\n\n  private readonly client: GoogleGenAI\n\n  constructor(config: GeminiImageConfig, model: TModel) {\n    super(model, config)\n    this.client = createGeminiClient(config)\n  }\n\n  async generateImages(\n    options: ImageGenerationOptions<GeminiImageProviderOptions>,\n  ): Promise<ImageGenerationResult> {\n    const { model, prompt, logger } = options\n\n    logger.request(\n      `activity=generateImage provider=gemini model=${this.model}`,\n      {\n        provider: 'gemini',\n        model: this.model,\n      },\n    )\n\n    try {\n      validatePrompt({ prompt, model })\n\n      if (this.isGeminiImageModel(model)) {\n        return await this.generateWithGeminiApi(options)\n      }\n\n      // Imagen models path (generateImages API)\n      validateImageSize(model, options.size)\n      validateNumberOfImages(model, options.numberOfImages)\n\n      const config = this.buildImagenConfig(options)\n\n      const response = await this.client.models.generateImages({\n        model,\n        prompt,\n        config,\n      })\n\n      return this.transformImagenResponse(model, response)\n    } catch (error) {\n      logger.errors('gemini.generateImage fatal', {\n        error,\n        source: 'gemini.generateImage',\n      })\n      throw error\n    }\n  }\n\n  private isGeminiImageModel(model: string): boolean {\n    return model.startsWith('gemini-')\n  }\n\n  private async generateWithGeminiApi(\n    options: ImageGenerationOptions<GeminiImageProviderOptions>,\n  ): Promise<ImageGenerationResult> {\n    const { model, prompt, size, numberOfImages, modelOptions } = options\n\n    const parsedSize = size ? parseNativeImageSize(size) : undefined\n\n    // The generateContent API has no numberOfImages parameter.\n    // Instead, augment the prompt to request multiple images when needed.\n    const augmentedPrompt =\n      numberOfImages && numberOfImages > 1\n        ? `${prompt} Generate ${numberOfImages} distinct images.`\n        : prompt\n\n    // GeminiImageProviderOptions is Imagen-shaped — most fields\n    // (personGeneration, safetyFilterLevel, addWatermark, outputMimeType,\n    // outputCompressionQuality, guidanceScale, enhancePrompt,\n    // includeSafetyAttributes, includeRaiReason, outputGcsUri, labels,\n    // negativePrompt, language) are only valid on GenerateImagesConfig and\n    // would be rejected by the Gemini-native generateContent path. Pick only\n    // the fields that are valid on GenerateContentConfig instead of spreading\n    // the whole options object.\n    const nativeConfig: GenerateContentConfig = {}\n    if (modelOptions?.seed !== undefined) {\n      nativeConfig.seed = modelOptions.seed\n    }\n\n    const config: GenerateContentConfig = {\n      ...nativeConfig,\n      // Include TEXT so the model can interleave descriptions between images.\n      // IMPORTANT: responseModalities is a protected default — set it AFTER\n      // nativeConfig so nothing can silently disable image output.\n      responseModalities: ['TEXT', 'IMAGE'],\n      ...(parsedSize && {\n        imageConfig: {\n          ...(parsedSize.aspectRatio && {\n            aspectRatio: parsedSize.aspectRatio,\n          }),\n          ...(parsedSize.resolution && {\n            imageSize: parsedSize.resolution,\n          }),\n        },\n      }),\n    }\n\n    const response = await this.client.models.generateContent({\n      model,\n      contents: augmentedPrompt,\n      config,\n    })\n\n    return this.transformGeminiResponse(model, response)\n  }\n\n  private transformGeminiResponse(\n    model: string,\n    response: GenerateContentResponse,\n  ): ImageGenerationResult {\n    const images: Array<GeneratedImage> = []\n    const textParts: Array<string> = []\n    const parts = response.candidates?.[0]?.content?.parts ?? []\n\n    for (const part of parts) {\n      if (\n        part.inlineData?.data &&\n        typeof part.inlineData.data === 'string' &&\n        part.inlineData.data.length > 0\n      ) {\n        images.push({ b64Json: part.inlineData.data })\n      } else if (typeof part.text === 'string' && part.text.length > 0) {\n        textParts.push(part.text)\n      }\n    }\n\n    // If the model returned only text parts (for example a safety refusal\n    // or a \"can't do that\" message), surface the text instead of silently\n    // resolving to an empty images array — otherwise callers can't tell a\n    // generation failure apart from a genuine empty response.\n    if (images.length === 0) {\n      const reason =\n        textParts.length > 0\n          ? `: ${textParts.join(' ').trim()}`\n          : ' (no inline image or text parts were returned).'\n      throw new Error(`Gemini ${model} returned no images${reason}`)\n    }\n\n    return {\n      id: generateId(this.name),\n      model,\n      images,\n      // Surface token usage when the model reports it (e.g. Nano Banana via\n      // generateContent). Conditionally spread to satisfy\n      // exactOptionalPropertyTypes — only include usage when present. See #330.\n      ...(response.usageMetadata && {\n        usage: {\n          inputTokens: response.usageMetadata.promptTokenCount ?? 0,\n          outputTokens: response.usageMetadata.candidatesTokenCount ?? 0,\n          totalTokens: response.usageMetadata.totalTokenCount ?? 0,\n        },\n      }),\n    }\n  }\n\n  private buildImagenConfig(\n    options: ImageGenerationOptions<GeminiImageProviderOptions>,\n  ): GenerateImagesConfig {\n    const { size, numberOfImages, modelOptions } = options\n\n    // Build with conditional spreads — under exactOptionalPropertyTypes the\n    // vendor `GenerateImagesConfig` fields are `field?: T` (no `| undefined`),\n    // so we can only assign the property when we actually have a value.\n    const sizeAspectRatio = size ? sizeToAspectRatio(size) : undefined\n    return {\n      numberOfImages: numberOfImages ?? 1,\n      // Map size to aspect ratio if provided (modelOptions.aspectRatio will override)\n      ...(sizeAspectRatio !== undefined && { aspectRatio: sizeAspectRatio }),\n      ...modelOptions,\n    }\n  }\n\n  private transformImagenResponse(\n    model: string,\n    response: GenerateImagesResponse,\n  ): ImageGenerationResult {\n    const entries = response.generatedImages ?? []\n    const images: Array<GeneratedImage> = []\n    const filterReasons: Array<string> = []\n\n    for (const item of entries) {\n      const b64Json = item.image?.imageBytes\n      if (b64Json) {\n        images.push({\n          b64Json,\n          ...(item.enhancedPrompt !== undefined && {\n            revisedPrompt: item.enhancedPrompt,\n          }),\n        })\n        continue\n      }\n      // Imagen can drop individual entries with a raiFilteredReason when\n      // Responsible-AI filters fire. Preserve the reason so callers can\n      // surface it instead of silently getting back fewer images.\n      const reason = (item as { raiFilteredReason?: string }).raiFilteredReason\n      if (reason) {\n        filterReasons.push(reason)\n      }\n    }\n\n    // Every entry was filtered — no usable images to return. Throw rather\n    // than resolve to an empty array so the caller is forced to handle the\n    // failure mode explicitly.\n    if (entries.length > 0 && images.length === 0) {\n      const joined = filterReasons.length > 0 ? filterReasons.join('; ') : ''\n      throw new Error(\n        `Imagen ${model} returned no images: all ${entries.length} generated image(s) were filtered by Responsible-AI${joined ? ` (${joined})` : ''}.`,\n      )\n    }\n\n    // Partial filter: surface via console.warn since ImageGenerationResult\n    // has no warnings field. Callers that care can still inspect the count\n    // mismatch between requested and returned images.\n    if (filterReasons.length > 0 && typeof console !== 'undefined') {\n      console.warn(\n        `[gemini-image] ${filterReasons.length} of ${entries.length} images from ${model} were filtered by Responsible-AI: ${filterReasons.join('; ')}`,\n      )\n    }\n\n    return {\n      id: generateId(this.name),\n      model,\n      images,\n    }\n  }\n}\n\n/**\n * Creates a Gemini image adapter with explicit API key.\n * Type resolution happens here at the call site.\n *\n * @param model - The model name (e.g., 'imagen-3.0-generate-002')\n * @param apiKey - Your Google API key\n * @param config - Optional additional configuration\n * @returns Configured Gemini image adapter instance with resolved types\n *\n * @example\n * ```typescript\n * const adapter = createGeminiImage('imagen-3.0-generate-002', \"your-api-key\");\n *\n * const result = await generateImage({\n *   adapter,\n *   prompt: 'A cute baby sea otter'\n * });\n * ```\n */\nexport function createGeminiImage<TModel extends GeminiImageModel>(\n  model: TModel,\n  apiKey: string,\n  config?: Omit<GeminiImageConfig, 'apiKey'>,\n): GeminiImageAdapter<TModel> {\n  return new GeminiImageAdapter({ apiKey, ...config }, model)\n}\n\n/**\n * Creates a Gemini image adapter with automatic API key detection from environment variables.\n * Type resolution happens here at the call site.\n *\n * Looks for `GOOGLE_API_KEY` or `GEMINI_API_KEY` in:\n * - `process.env` (Node.js)\n * - `window.env` (Browser with injected env)\n *\n * @param model - The model name (e.g., 'imagen-4.0-generate-001')\n * @param config - Optional configuration (excluding apiKey which is auto-detected)\n * @returns Configured Gemini image adapter instance with resolved types\n * @throws Error if GOOGLE_API_KEY or GEMINI_API_KEY is not found in environment\n *\n * @example\n * ```typescript\n * // Automatically uses GOOGLE_API_KEY from environment\n * const adapter = geminiImage('imagen-4.0-generate-001');\n *\n * const result = await generateImage({\n *   adapter,\n *   prompt: 'A beautiful sunset over mountains'\n * });\n * ```\n */\nexport function geminiImage<TModel extends GeminiImageModel>(\n  model: TModel,\n  config?: Omit<GeminiImageConfig, 'apiKey'>,\n): GeminiImageAdapter<TModel> {\n  const apiKey = getGeminiApiKeyFromEnv()\n  return createGeminiImage(model, apiKey, config)\n}\n"],"names":[],"mappings":";;;AAuDO,MAAM,2BAEH,iBAKR;AAAA,EACkB,OAAO;AAAA,EAChB,OAAO;AAAA,EASC;AAAA,EAEjB,YAAY,QAA2B,OAAe;AACpD,UAAM,OAAO,MAAM;AACnB,SAAK,SAAS,mBAAmB,MAAM;AAAA,EACzC;AAAA,EAEA,MAAM,eACJ,SACgC;AAChC,UAAM,EAAE,OAAO,QAAQ,OAAA,IAAW;AAElC,WAAO;AAAA,MACL,gDAAgD,KAAK,KAAK;AAAA,MAC1D;AAAA,QACE,UAAU;AAAA,QACV,OAAO,KAAK;AAAA,MAAA;AAAA,IACd;AAGF,QAAI;AACF,qBAAe,EAAE,QAAQ,OAAO;AAEhC,UAAI,KAAK,mBAAmB,KAAK,GAAG;AAClC,eAAO,MAAM,KAAK,sBAAsB,OAAO;AAAA,MACjD;AAGA,wBAAkB,OAAO,QAAQ,IAAI;AACrC,6BAAuB,OAAO,QAAQ,cAAc;AAEpD,YAAM,SAAS,KAAK,kBAAkB,OAAO;AAE7C,YAAM,WAAW,MAAM,KAAK,OAAO,OAAO,eAAe;AAAA,QACvD;AAAA,QACA;AAAA,QACA;AAAA,MAAA,CACD;AAED,aAAO,KAAK,wBAAwB,OAAO,QAAQ;AAAA,IACrD,SAAS,OAAO;AACd,aAAO,OAAO,8BAA8B;AAAA,QAC1C;AAAA,QACA,QAAQ;AAAA,MAAA,CACT;AACD,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEQ,mBAAmB,OAAwB;AACjD,WAAO,MAAM,WAAW,SAAS;AAAA,EACnC;AAAA,EAEA,MAAc,sBACZ,SACgC;AAChC,UAAM,EAAE,OAAO,QAAQ,MAAM,gBAAgB,iBAAiB;AAE9D,UAAM,aAAa,OAAO,qBAAqB,IAAI,IAAI;AAIvD,UAAM,kBACJ,kBAAkB,iBAAiB,IAC/B,GAAG,MAAM,aAAa,cAAc,sBACpC;AAUN,UAAM,eAAsC,CAAA;AAC5C,QAAI,cAAc,SAAS,QAAW;AACpC,mBAAa,OAAO,aAAa;AAAA,IACnC;AAEA,UAAM,SAAgC;AAAA,MACpC,GAAG;AAAA;AAAA;AAAA;AAAA,MAIH,oBAAoB,CAAC,QAAQ,OAAO;AAAA,MACpC,GAAI,cAAc;AAAA,QAChB,aAAa;AAAA,UACX,GAAI,WAAW,eAAe;AAAA,YAC5B,aAAa,WAAW;AAAA,UAAA;AAAA,UAE1B,GAAI,WAAW,cAAc;AAAA,YAC3B,WAAW,WAAW;AAAA,UAAA;AAAA,QACxB;AAAA,MACF;AAAA,IACF;AAGF,UAAM,WAAW,MAAM,KAAK,OAAO,OAAO,gBAAgB;AAAA,MACxD;AAAA,MACA,UAAU;AAAA,MACV;AAAA,IAAA,CACD;AAED,WAAO,KAAK,wBAAwB,OAAO,QAAQ;AAAA,EACrD;AAAA,EAEQ,wBACN,OACA,UACuB;AACvB,UAAM,SAAgC,CAAA;AACtC,UAAM,YAA2B,CAAA;AACjC,UAAM,QAAQ,SAAS,aAAa,CAAC,GAAG,SAAS,SAAS,CAAA;AAE1D,eAAW,QAAQ,OAAO;AACxB,UACE,KAAK,YAAY,QACjB,OAAO,KAAK,WAAW,SAAS,YAChC,KAAK,WAAW,KAAK,SAAS,GAC9B;AACA,eAAO,KAAK,EAAE,SAAS,KAAK,WAAW,MAAM;AAAA,MAC/C,WAAW,OAAO,KAAK,SAAS,YAAY,KAAK,KAAK,SAAS,GAAG;AAChE,kBAAU,KAAK,KAAK,IAAI;AAAA,MAC1B;AAAA,IACF;AAMA,QAAI,OAAO,WAAW,GAAG;AACvB,YAAM,SACJ,UAAU,SAAS,IACf,KAAK,UAAU,KAAK,GAAG,EAAE,KAAA,CAAM,KAC/B;AACN,YAAM,IAAI,MAAM,UAAU,KAAK,sBAAsB,MAAM,EAAE;AAAA,IAC/D;AAEA,WAAO;AAAA,MACL,IAAI,WAAW,KAAK,IAAI;AAAA,MACxB;AAAA,MACA;AAAA;AAAA;AAAA;AAAA,MAIA,GAAI,SAAS,iBAAiB;AAAA,QAC5B,OAAO;AAAA,UACL,aAAa,SAAS,cAAc,oBAAoB;AAAA,UACxD,cAAc,SAAS,cAAc,wBAAwB;AAAA,UAC7D,aAAa,SAAS,cAAc,mBAAmB;AAAA,QAAA;AAAA,MACzD;AAAA,IACF;AAAA,EAEJ;AAAA,EAEQ,kBACN,SACsB;AACtB,UAAM,EAAE,MAAM,gBAAgB,aAAA,IAAiB;AAK/C,UAAM,kBAAkB,OAAO,kBAAkB,IAAI,IAAI;AACzD,WAAO;AAAA,MACL,gBAAgB,kBAAkB;AAAA;AAAA,MAElC,GAAI,oBAAoB,UAAa,EAAE,aAAa,gBAAA;AAAA,MACpD,GAAG;AAAA,IAAA;AAAA,EAEP;AAAA,EAEQ,wBACN,OACA,UACuB;AACvB,UAAM,UAAU,SAAS,mBAAmB,CAAA;AAC5C,UAAM,SAAgC,CAAA;AACtC,UAAM,gBAA+B,CAAA;AAErC,eAAW,QAAQ,SAAS;AAC1B,YAAM,UAAU,KAAK,OAAO;AAC5B,UAAI,SAAS;AACX,eAAO,KAAK;AAAA,UACV;AAAA,UACA,GAAI,KAAK,mBAAmB,UAAa;AAAA,YACvC,eAAe,KAAK;AAAA,UAAA;AAAA,QACtB,CACD;AACD;AAAA,MACF;AAIA,YAAM,SAAU,KAAwC;AACxD,UAAI,QAAQ;AACV,sBAAc,KAAK,MAAM;AAAA,MAC3B;AAAA,IACF;AAKA,QAAI,QAAQ,SAAS,KAAK,OAAO,WAAW,GAAG;AAC7C,YAAM,SAAS,cAAc,SAAS,IAAI,cAAc,KAAK,IAAI,IAAI;AACrE,YAAM,IAAI;AAAA,QACR,UAAU,KAAK,4BAA4B,QAAQ,MAAM,sDAAsD,SAAS,KAAK,MAAM,MAAM,EAAE;AAAA,MAAA;AAAA,IAE/I;AAKA,QAAI,cAAc,SAAS,KAAK,OAAO,YAAY,aAAa;AAC9D,cAAQ;AAAA,QACN,kBAAkB,cAAc,MAAM,OAAO,QAAQ,MAAM,gBAAgB,KAAK,qCAAqC,cAAc,KAAK,IAAI,CAAC;AAAA,MAAA;AAAA,IAEjJ;AAEA,WAAO;AAAA,MACL,IAAI,WAAW,KAAK,IAAI;AAAA,MACxB;AAAA,MACA;AAAA,IAAA;AAAA,EAEJ;AACF;AAqBO,SAAS,kBACd,OACA,QACA,QAC4B;AAC5B,SAAO,IAAI,mBAAmB,EAAE,QAAQ,GAAG,OAAA,GAAU,KAAK;AAC5D;AA0BO,SAAS,YACd,OACA,QAC4B;AAC5B,QAAM,SAAS,uBAAA;AACf,SAAO,kBAAkB,OAAO,QAAQ,MAAM;AAChD;"}

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

@@ -0,0 +1,6 @@
+/**
+ * @experimental Gemini's Interactions API is in Beta per Google. Shapes and
+ * behavior may change between minor releases of `@tanstack/ai-gemini`.
+ */
+export { GeminiTextInteractionsAdapter, createGeminiTextInteractions, geminiTextInteractions, type GeminiTextInteractionsConfig, type GeminiTextInteractionsProviderOptions, type GeminiInteractionsStream, } from './text-interactions/adapter.js';
+export type { GeminiInteractionsCustomEvent, GeminiInteractionsCustomEventName, GeminiInteractionsCustomEventValue, } from './text-interactions/events.js';

package/dist/esm/experimental/index.js

@@ -0,0 +1,7 @@
+import { GeminiTextInteractionsAdapter, createGeminiTextInteractions, geminiTextInteractions } from "./text-interactions/adapter.js";
+export {
+  GeminiTextInteractionsAdapter,
+  createGeminiTextInteractions,
+  geminiTextInteractions
+};
+//# sourceMappingURL=index.js.map

package/dist/esm/experimental/index.js.map

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

package/dist/esm/experimental/text-interactions/adapter.d.ts

@@ -0,0 +1,75 @@
+import { BaseTextAdapter, StructuredOutputOptions, StructuredOutputResult } from '@tanstack/ai/adapters';
+import { GeminiChatModelToolCapabilitiesByName, GeminiModelInputModalitiesByName, GeminiModels } from '../../model-meta.js';
+import { Modality, StreamChunk, TextOptions } from '@tanstack/ai';
+import { GeminiInteractionsStream } from './events.js';
+import { ExternalTextInteractionsProviderOptions } from './provider-options.js';
+import { GeminiMessageMetadataByModality } from '../../message-types.js';
+import { GeminiClientConfig } from '../../utils.js';
+export type GeminiTextInteractionsConfig = GeminiClientConfig;
+export type GeminiTextInteractionsProviderOptions = ExternalTextInteractionsProviderOptions;
+/**
+ * Resolve provider options for a specific model. The Interactions API's
+ * request shape is the same across all chat-capable Gemini models — the
+ * SDK doesn't expose a per-model param union — so this currently falls
+ * through to the flat `GeminiTextInteractionsProviderOptions` for every
+ * model. The alias exists for parity with `GeminiTextAdapter`, so a
+ * per-model map can be slotted in later without changing the adapter
+ * signature.
+ */
+type ResolveProviderOptions = GeminiTextInteractionsProviderOptions;
+/**
+ * Resolve input modalities for a specific model. Reuses the chat-model
+ * modality map from `model-meta.ts`: passing a `document` content block
+ * to a model that doesn't support it is a compile error, matching the
+ * sibling `GeminiTextAdapter`.
+ */
+type ResolveInputModalities<TModel extends string> = TModel extends keyof GeminiModelInputModalitiesByName ? GeminiModelInputModalitiesByName[TModel] : readonly ['text', 'image', 'audio', 'video', 'document'];
+/**
+ * Resolve tool capabilities for a specific model. Reuses the chat-model
+ * capability map: `google_maps` / `google_search_retrieval` /
+ * `mcp_server` are rejected at runtime by `convertToolsToInteractionsFormat`,
+ * but per-model gating happens here at compile time.
+ */
+type ResolveToolCapabilities<TModel extends string> = TModel extends keyof GeminiChatModelToolCapabilitiesByName ? NonNullable<GeminiChatModelToolCapabilitiesByName[TModel]> : readonly [];
+/**
+ * Tree-shakeable adapter for Gemini's stateful Interactions API. Routes
+ * through `client.interactions.create` and surfaces the server-assigned
+ * `interactionId` via an AG-UI `CUSTOM` event with
+ * `name: 'gemini.interactionId'` emitted just before `RUN_FINISHED`; pass
+ * that id back on the next turn via `modelOptions.previous_interaction_id`
+ * to continue the conversation without resending history.
+ *
+ * The Interactions API does NOT support stateless multi-turn replay —
+ * passing more than one message in `messages` without a
+ * `previous_interaction_id` throws. For a chat UI that maintains local
+ * history (e.g. `useChat`), see the "Wiring with `useChat`" section of
+ * `docs/adapters/gemini.md` for the canonical client/server pattern.
+ *
+ * Supports user-defined function tools and the built-in tools
+ * `google_search`, `code_execution`, `url_context`, `file_search`, and
+ * `computer_use`. Built-in tool *activity* for the four search/exec
+ * variants is surfaced via `CUSTOM` events
+ * (`gemini.googleSearchCall` / `gemini.googleSearchResult` and the
+ * corresponding per-tool variants) carrying the raw Interactions delta;
+ * see {@link GeminiInteractionsCustomEvent}. `computer_use` is accepted
+ * in the request but the Interactions API does not currently stream
+ * per-delta CUSTOM events for it. `google_search_retrieval`,
+ * `google_maps`, and `mcp_server` are not supported on this adapter.
+ *
+ * @experimental Interactions API is in Beta per Google; shapes may change.
+ * @see https://ai.google.dev/gemini-api/docs/interactions
+ */
+export declare class GeminiTextInteractionsAdapter<TModel extends GeminiModels, TProviderOptions extends Record<string, any> = ResolveProviderOptions, TInputModalities extends ReadonlyArray<Modality> = ResolveInputModalities<TModel>, TToolCapabilities extends ReadonlyArray<string> = ResolveToolCapabilities<TModel>> extends BaseTextAdapter<TModel, TProviderOptions, TInputModalities, GeminiMessageMetadataByModality, TToolCapabilities> {
+    readonly kind: "text";
+    readonly name: "gemini-text-interactions";
+    private readonly client;
+    private readonly interactionIdByThread;
+    constructor(config: GeminiTextInteractionsConfig, model: TModel);
+    chatStream(options: TextOptions<GeminiTextInteractionsProviderOptions>): AsyncIterable<StreamChunk>;
+    structuredOutput(options: StructuredOutputOptions<GeminiTextInteractionsProviderOptions>): Promise<StructuredOutputResult<unknown>>;
+}
+/** @experimental Interactions API is in Beta. */
+export declare function createGeminiTextInteractions<TModel extends GeminiModels>(model: TModel, apiKey: string, config?: Omit<GeminiTextInteractionsConfig, 'apiKey'>): GeminiTextInteractionsAdapter<TModel, ResolveProviderOptions, ResolveInputModalities<TModel>, ResolveToolCapabilities<TModel>>;
+/** @experimental Interactions API is in Beta. */
+export declare function geminiTextInteractions<TModel extends GeminiModels>(model: TModel, config?: Omit<GeminiTextInteractionsConfig, 'apiKey'>): GeminiTextInteractionsAdapter<TModel, ResolveProviderOptions, ResolveInputModalities<TModel>, ResolveToolCapabilities<TModel>>;
+export type { GeminiInteractionsStream };