@ai-sdk/google - npm Package Compare versions

Comparing version 3.0.25 to 3.0.26

CHANGELOG.md

 # @ai-sdk/google
 
+## 3.0.26
+
+### Patch Changes
+
+- 4c27179: feat(google): allow using Gemini image models with `generateImage`
+
 ## 3.0.25
@@ -4,0 +10,0 @@

dist/index.d.mts

@@ -163,3 +163,3 @@ import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 
-type GoogleGenerativeAIImageModelId = 'imagen-4.0-generate-001' | 'imagen-4.0-ultra-generate-001' | 'imagen-4.0-fast-generate-001' | (string & {});
+type GoogleGenerativeAIImageModelId = 'imagen-4.0-generate-001' | 'imagen-4.0-ultra-generate-001' | 'imagen-4.0-fast-generate-001' | 'gemini-2.5-flash-image' | 'gemini-3-pro-image-preview' | (string & {});
 interface GoogleGenerativeAIImageSettings {
@@ -166,0 +166,0 @@ /**

dist/index.d.ts

@@ -163,3 +163,3 @@ import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 
-type GoogleGenerativeAIImageModelId = 'imagen-4.0-generate-001' | 'imagen-4.0-ultra-generate-001' | 'imagen-4.0-fast-generate-001' | (string & {});
+type GoogleGenerativeAIImageModelId = 'imagen-4.0-generate-001' | 'imagen-4.0-ultra-generate-001' | 'imagen-4.0-fast-generate-001' | 'gemini-2.5-flash-image' | 'gemini-3-pro-image-preview' | (string & {});
 interface GoogleGenerativeAIImageSettings {
@@ -166,0 +166,0 @@ /**

package.json

 {
   "name": "@ai-sdk/google",
-  "version": "3.0.25",
+  "version": "3.0.26",
   "license": "Apache-2.0",
@@ -39,4 +39,4 @@ "sideEffects": false,
   "dependencies": {
-    "@ai-sdk/provider": "3.0.8",
-    "@ai-sdk/provider-utils": "4.0.14"
+    "@ai-sdk/provider-utils": "4.0.14",
+    "@ai-sdk/provider": "3.0.8"
   },
@@ -43,0 +43,0 @@ "devDependencies": {

src/google-generative-ai-image-model.ts

@@ -1,5 +0,12 @@
-import { ImageModelV3, SharedV3Warning } from '@ai-sdk/provider';
 import {
+  ImageModelV3,
+  LanguageModelV3Prompt,
+  SharedV3Warning,
+} from '@ai-sdk/provider';
+import {
   combineHeaders,
+  convertToBase64,
   createJsonResponseHandler,
+  FetchFunction,
+  generateId as defaultGenerateId,
   type InferSchema,
@@ -9,2 +16,3 @@ lazySchema,
   postJsonToApi,
+  Resolvable,
   resolve,
@@ -19,3 +27,3 @@ zodSchema,
 } from './google-generative-ai-image-settings';
-import { FetchFunction, Resolvable } from '@ai-sdk/provider-utils';
+import { GoogleGenerativeAILanguageModel } from './google-generative-ai-language-model';
 
@@ -37,4 +45,11 @@ interface GoogleGenerativeAIImageModelConfig {
   get maxImagesPerCall(): number {
+    if (this.settings.maxImagesPerCall != null) {
+      return this.settings.maxImagesPerCall;
+    }
+    // https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/2-5-flash-image
+    if (isGeminiModel(this.modelId)) {
+      return 10;
+    }
     // https://ai.google.dev/gemini-api/docs/imagen#imagen-model
-    return this.settings.maxImagesPerCall ?? 4;
+    return 4;
   }
@@ -55,2 +70,12 @@
   ): Promise<Awaited<ReturnType<ImageModelV3['doGenerate']>>> {
+    // Gemini image models use the language model API internally
+    if (isGeminiModel(this.modelId)) {
+      return this.doGenerateGemini(options);
+    }
+    return this.doGenerateImagen(options);
+  }
+
+  private async doGenerateImagen(
+    options: Parameters<ImageModelV3['doGenerate']>[0],
+  ): Promise<Awaited<ReturnType<ImageModelV3['doGenerate']>>> {
     const {
@@ -70,6 +95,6 @@ prompt,
 
-    // Google Generative AI does not support image editing
+    // Imagen API endpoints do not support image editing
     if (files != null && files.length > 0) {
       throw new Error(
-        'Google Generative AI does not support image editing. ' +
+        'Google Generative AI does not support image editing with Imagen models. ' +
           'Use Google Vertex AI (@ai-sdk/google-vertex) for image editing capabilities.',
@@ -146,6 +171,6 @@ );
       ),
-      warnings: warnings ?? [],
+      warnings,
       providerMetadata: {
         google: {
-          images: response.predictions.map(prediction => ({
+          images: response.predictions.map(() => ({
             // Add any prediction-specific metadata here
@@ -162,4 +187,144 @@ })),
   }
+
+  private async doGenerateGemini(
+    options: Parameters<ImageModelV3['doGenerate']>[0],
+  ): Promise<Awaited<ReturnType<ImageModelV3['doGenerate']>>> {
+    const {
+      prompt,
+      n,
+      size,
+      aspectRatio,
+      seed,
+      providerOptions,
+      headers,
+      abortSignal,
+      files,
+      mask,
+    } = options;
+    const warnings: Array<SharedV3Warning> = [];
+
+    // Gemini does not support mask-based inpainting
+    if (mask != null) {
+      throw new Error(
+        'Gemini image models do not support mask-based image editing.',
+      );
+    }
+
+    // Gemini does not support generating multiple images per call via n parameter
+    if (n != null && n > 1) {
+      throw new Error(
+        'Gemini image models do not support generating a set number of images per call. Use n=1 or omit the n parameter.',
+      );
+    }
+
+    if (size != null) {
+      warnings.push({
+        type: 'unsupported',
+        feature: 'size',
+        details:
+          'This model does not support the `size` option. Use `aspectRatio` instead.',
+      });
+    }
+
+    // Build user message content for language model
+    const userContent: Array<
+      | { type: 'text'; text: string }
+      | { type: 'file'; data: string | Uint8Array | URL; mediaType: string }
+    > = [];
+
+    // Add text prompt
+    if (prompt != null) {
+      userContent.push({ type: 'text', text: prompt });
+    }
+
+    // Add input images for editing
+    if (files != null && files.length > 0) {
+      for (const file of files) {
+        if (file.type === 'url') {
+          userContent.push({
+            type: 'file',
+            data: new URL(file.url),
+            mediaType: 'image/*',
+          });
+        } else {
+          userContent.push({
+            type: 'file',
+            data:
+              typeof file.data === 'string'
+                ? file.data
+                : new Uint8Array(file.data),
+            mediaType: file.mediaType,
+          });
+        }
+      }
+    }
+
+    const languageModelPrompt: LanguageModelV3Prompt = [
+      { role: 'user', content: userContent },
+    ];
+
+    // Instantiate language model
+    const languageModel = new GoogleGenerativeAILanguageModel(this.modelId, {
+      provider: this.config.provider,
+      baseURL: this.config.baseURL,
+      headers: this.config.headers ?? {},
+      fetch: this.config.fetch,
+      generateId: this.config.generateId ?? defaultGenerateId,
+    });
+
+    // Call language model with image-only response modality
+    const result = await languageModel.doGenerate({
+      prompt: languageModelPrompt,
+      seed,
+      providerOptions: {
+        google: {
+          responseModalities: ['IMAGE'],
+          imageConfig: aspectRatio ? { aspectRatio } : undefined,
+          ...((providerOptions?.google as Record<string, unknown>) ?? {}),
+        },
+      },
+      headers,
+      abortSignal,
+    });
+
+    const currentDate = this.config._internal?.currentDate?.() ?? new Date();
+
+    // Extract images from language model response
+    const images: string[] = [];
+    for (const part of result.content) {
+      if (part.type === 'file' && part.mediaType.startsWith('image/')) {
+        images.push(convertToBase64(part.data));
+      }
+    }
+
+    return {
+      images,
+      warnings,
+      providerMetadata: {
+        google: {
+          images: images.map(() => ({})),
+        },
+      },
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: result.response?.headers,
+      },
+      usage: result.usage
+        ? {
+            inputTokens: result.usage.inputTokens.total,
+            outputTokens: result.usage.outputTokens.total,
+            totalTokens:
+              (result.usage.inputTokens.total ?? 0) +
+              (result.usage.outputTokens.total ?? 0),
+          }
+        : undefined,
+    };
+  }
 }
 
+function isGeminiModel(modelId: string): boolean {
+  return modelId.startsWith('gemini-');
+}
+
 // minimal version of the schema
@@ -166,0 +331,0 @@ const googleImageResponseSchema = lazySchema(() =>

src/google-generative-ai-image-settings.ts

 export type GoogleGenerativeAIImageModelId =
+  // Imagen models (use :predict API)
   | 'imagen-4.0-generate-001'
   | 'imagen-4.0-ultra-generate-001'
   | 'imagen-4.0-fast-generate-001'
+  // Gemini image models (technically multimodal output language models, use :generateContent API)
+  | 'gemini-2.5-flash-image'
+  | 'gemini-3-pro-image-preview'
   | (string & {});
@@ -6,0 +10,0 @@

New alerts

Long strings

Supply chain risk

Contains long string literals, which may be a sign of obfuscated or packed code.

Found 1 instance in 1 package

URL strings

Supply chain risk

Package contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.

Found 1 instance in 1 package

Fixed alerts

Long strings

Supply chain risk

Contains long string literals, which may be a sign of obfuscated or packed code.

Found 1 instance in 1 package

URL strings

Supply chain risk

Package contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.

Found 1 instance in 1 package

Improved metrics

Total package byte prevSize

1056146

2.9%

Lines of code

11741

3.44%

No dependency changes