@umituz/react-native-ai-gemini-provider 1.8.2 → 1.8.3

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@umituz/react-native-ai-gemini-provider",
-  "version": "1.8.2",
+  "version": "1.8.3",
   "description": "Google Gemini AI provider for React Native applications",
   "main": "src/index.ts",
   "types": "src/index.ts",

package/src/domain/entities/models.ts

@@ -14,11 +14,16 @@ export const GEMINI_MODELS = {
     PRO: "gemini-1.5-pro",
   },
 
-  // Image generation models (Imagen 4.0)
-  IMAGE: {
+  // Text-to-Image models (Imagen 4.0) - generates images from text only
+  TEXT_TO_IMAGE: {
     DEFAULT: "imagen-4.0-generate-001",
   },
 
+  // Image editing models (Gemini) - transforms/edits images with input image + prompt
+  IMAGE_EDIT: {
+    DEFAULT: "gemini-2.0-flash-exp-image-generation",
+  },
+
   // Video understanding models
   VIDEO: {
     FLASH: "gemini-2.0-flash",
@@ -30,7 +35,8 @@ export const GEMINI_MODELS = {
  */
 export const DEFAULT_MODELS = {
   TEXT: GEMINI_MODELS.TEXT.FLASH,
-  IMAGE: GEMINI_MODELS.IMAGE.DEFAULT,
+  TEXT_TO_IMAGE: GEMINI_MODELS.TEXT_TO_IMAGE.DEFAULT,
+  IMAGE_EDIT: GEMINI_MODELS.IMAGE_EDIT.DEFAULT,
   VIDEO: GEMINI_MODELS.VIDEO.FLASH,
 } as const;
 

package/src/infrastructure/services/gemini-client.service.ts

@@ -23,7 +23,7 @@ const DEFAULT_CONFIG: Partial<GeminiConfig> = {
   maxDelay: 10000,
   defaultTimeoutMs: 60000,
   defaultModel: DEFAULT_MODELS.TEXT,
-  imageModel: DEFAULT_MODELS.IMAGE,
+  imageModel: DEFAULT_MODELS.TEXT_TO_IMAGE,
 };
 
 const RETRYABLE_ERROR_PATTERNS = [
@@ -264,7 +264,7 @@ class GeminiClientService {
   ): Promise<GeminiImageGenerationResult> {
     this.validateInitialization();
 
-    const imageModel = this.config?.imageModel || DEFAULT_MODELS.IMAGE;
+    const imageModel = this.config?.imageModel || DEFAULT_MODELS.TEXT_TO_IMAGE;
     const apiKey = this.config?.apiKey;
 
     if (typeof __DEV__ !== "undefined" && __DEV__) {
@@ -339,6 +339,109 @@ class GeminiClientService {
     return result;
   }
 
+  /**
+   * Edit/transform image using Gemini generateContent API
+   * Takes input image + prompt and generates new image
+   */
+  async editImage(
+    prompt: string,
+    images: Array<{ base64: string; mimeType: string }>,
+  ): Promise<GeminiImageGenerationResult> {
+    this.validateInitialization();
+
+    const editModel = DEFAULT_MODELS.IMAGE_EDIT;
+    const apiKey = this.config?.apiKey;
+
+    if (typeof __DEV__ !== "undefined" && __DEV__) {
+      // eslint-disable-next-line no-console
+      console.log("[GeminiClient] editImage() called", {
+        model: editModel,
+        promptLength: prompt.length,
+        imagesCount: images.length,
+      });
+    }
+
+    const url = `https://generativelanguage.googleapis.com/v1beta/models/${editModel}:generateContent`;
+
+    const parts: Array<Record<string, unknown>> = [];
+
+    for (const image of images) {
+      parts.push({
+        inlineData: {
+          mimeType: image.mimeType,
+          data: extractBase64Data(image.base64),
+        },
+      });
+    }
+
+    parts.push({ text: prompt });
+
+    const requestBody = {
+      contents: [{ parts }],
+      generationConfig: {
+        responseModalities: ["TEXT", "IMAGE"],
+      },
+    };
+
+    if (typeof __DEV__ !== "undefined" && __DEV__) {
+      // eslint-disable-next-line no-console
+      console.log("[GeminiClient] editImage() request", {
+        url,
+        partsCount: parts.length,
+      });
+    }
+
+    const response = await this.executeWithRetry(async () => {
+      const res = await fetch(url, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          "x-goog-api-key": apiKey!,
+        },
+        body: JSON.stringify(requestBody),
+      });
+
+      if (!res.ok) {
+        const errorText = await res.text();
+        throw new Error(`Image edit API error (${res.status}): ${errorText}`);
+      }
+
+      return res.json();
+    });
+
+    const result: GeminiImageGenerationResult = {
+      text: undefined,
+      imageUrl: undefined,
+      imageBase64: undefined,
+      mimeType: "image/png",
+    };
+
+    const candidate = response.candidates?.[0];
+    const responseParts = candidate?.content?.parts || [];
+
+    for (const part of responseParts) {
+      if (part.text) {
+        result.text = part.text;
+      }
+      if (part.inlineData) {
+        result.imageBase64 = part.inlineData.data;
+        result.mimeType = part.inlineData.mimeType || "image/png";
+        result.imageUrl = `data:${result.mimeType};base64,${result.imageBase64}`;
+      }
+    }
+
+    if (typeof __DEV__ !== "undefined" && __DEV__) {
+      // eslint-disable-next-line no-console
+      console.log("[GeminiClient] editImage() completed", {
+        hasImage: !!result.imageBase64,
+        hasText: !!result.text,
+        imageDataLength: result.imageBase64?.length ?? 0,
+      });
+    }
+
+    return result;
+  }
+
   /**
    * Stream content generation
    */

package/src/infrastructure/services/gemini-provider.service.ts

@@ -201,14 +201,24 @@ class GeminiProviderService {
   }
 
   /**
-   * Generate image with optional input images
-   * Convenience method for image generation
+   * Generate image from text only (Imagen API)
+   * Use for text-to-image generation without input images
    */
   async generateImage(
     prompt: string,
-    images?: GeminiImageInput[],
   ): Promise<GeminiImageGenerationResult> {
-    return geminiClientService.generateImage(prompt, images);
+    return geminiClientService.generateImage(prompt);
+  }
+
+  /**
+   * Edit/transform image using input image + prompt (Gemini API)
+   * Use for image editing, transformation, style transfer
+   */
+  async editImage(
+    prompt: string,
+    images: GeminiImageInput[],
+  ): Promise<GeminiImageGenerationResult> {
+    return geminiClientService.editImage(prompt, images);
   }
 
   /**