Version 1.0 Release

src/services/geminiService.ts

@@ -0,0 +1,169 @@
+import { GoogleGenAI } from '@google/genai';
+
+// Note: In production, this should be handled via a backend proxy
+const API_KEY = import.meta.env.VITE_GEMINI_API_KEY || 'demo-key';
+const genAI = new GoogleGenAI({ apiKey: API_KEY });
+
+export interface GenerationRequest {
+  prompt: string;
+  referenceImages?: string[]; // base64 array
+  temperature?: number;
+  seed?: number;
+}
+
+export interface EditRequest {
+  instruction: string;
+  originalImage: string; // base64
+  referenceImages?: string[]; // base64 array
+  maskImage?: string; // base64
+  temperature?: number;
+  seed?: number;
+}
+
+export interface SegmentationRequest {
+  image: string; // base64
+  query: string; // "the object at pixel (x,y)" or "the red car"
+}
+
+export class GeminiService {
+  async generateImage(request: GenerationRequest): Promise<string[]> {
+    try {
+      const contents: any[] = [{ text: request.prompt }];
+      
+      // Add reference images if provided
+      if (request.referenceImages && request.referenceImages.length > 0) {
+        request.referenceImages.forEach(image => {
+          contents.push({
+            inlineData: {
+              mimeType: "image/png",
+              data: image,
+            },
+          });
+        });
+      }
+
+      const response = await genAI.models.generateContent({
+        model: "gemini-2.5-flash-image-preview",
+        contents,
+      });
+
+      const images: string[] = [];
+
+      for (const part of response.candidates[0].content.parts) {
+        if (part.inlineData) {
+          images.push(part.inlineData.data);
+        }
+      }
+
+      return images;
+    } catch (error) {
+      console.error('Error generating image:', error);
+      throw new Error('Failed to generate image. Please try again.');
+    }
+  }
+
+  async editImage(request: EditRequest): Promise<string[]> {
+    try {
+      const contents = [
+        { text: this.buildEditPrompt(request) },
+        {
+          inlineData: {
+            mimeType: "image/png",
+            data: request.originalImage,
+          },
+        },
+      ];
+
+      // Add reference images if provided
+      if (request.referenceImages && request.referenceImages.length > 0) {
+        request.referenceImages.forEach(image => {
+          contents.push({
+            inlineData: {
+              mimeType: "image/png",
+              data: image,
+            },
+          });
+        });
+      }
+
+      if (request.maskImage) {
+        contents.push({
+          inlineData: {
+            mimeType: "image/png",
+            data: request.maskImage,
+          },
+        });
+      }
+
+      const response = await genAI.models.generateContent({
+        model: "gemini-2.5-flash-image-preview",
+        contents,
+      });
+
+      const images: string[] = [];
+
+      for (const part of response.candidates[0].content.parts) {
+        if (part.inlineData) {
+          images.push(part.inlineData.data);
+        }
+      }
+
+      return images;
+    } catch (error) {
+      console.error('Error editing image:', error);
+      throw new Error('Failed to edit image. Please try again.');
+    }
+  }
+
+  async segmentImage(request: SegmentationRequest): Promise<any> {
+    try {
+      const prompt = [
+        { text: `Analyze this image and create a segmentation mask for: ${request.query}
+
+Return a JSON object with this exact structure:
+{
+  "masks": [
+    {
+      "label": "description of the segmented object",
+      "box_2d": [x, y, width, height],
+      "mask": "base64-encoded binary mask image"
+    }
+  ]
+}
+
+Only segment the specific object or region requested. The mask should be a binary PNG where white pixels (255) indicate the selected region and black pixels (0) indicate the background.` },
+        {
+          inlineData: {
+            mimeType: "image/png",
+            data: request.image,
+          },
+        },
+      ];
+
+      const response = await genAI.models.generateContent({
+        model: "gemini-2.5-flash-image-preview",
+        contents: prompt,
+      });
+
+      const responseText = response.candidates[0].content.parts[0].text;
+      return JSON.parse(responseText);
+    } catch (error) {
+      console.error('Error segmenting image:', error);
+      throw new Error('Failed to segment image. Please try again.');
+    }
+  }
+
+  private buildEditPrompt(request: EditRequest): string {
+    const maskInstruction = request.maskImage 
+      ? "\n\nIMPORTANT: Apply changes ONLY where the mask image shows white pixels (value 255). Leave all other areas completely unchanged. Respect the mask boundaries precisely and maintain seamless blending at the edges."
+      : "";
+
+    return `Edit this image according to the following instruction: ${request.instruction}
+
+Maintain the original image's lighting, perspective, and overall composition. Make the changes look natural and seamlessly integrated.${maskInstruction}
+
+Preserve image quality and ensure the edit looks professional and realistic.`;
+  }
+}
+
+export const geminiService = new GeminiService();