A research-grade image analysis tool powered by Google's Gemini 3 Pro model. This application allows users to upload images, select specific analysis modes and target audiences, and receive structured, expert insights ranging from simple explanations to academic research comparisons.

• 📸 Multimodal Analysis: Upload a single image for deep analysis or Compare two images side-by-side to find differences and similarities.
• 🧠 Intelligent Modes:
  • Explain It: Clear, direct answers about what is in the image.
  • Research View: Academic tone with technical specifications and detailed bullet points.
  • Teach Me: Educational format with "Key Takeaways" and a "Pop Quiz".
• 👥 Adaptive Audience: Tailors the complexity of the response for:
  • 5-Year-Olds
  • High School Students
  • Undergraduates
  • Industry Experts
• 🌗 Dark Mode: Fully responsive dark/light theme toggle.
• 💾 History & Auth: Mock authentication system that saves analysis history to the browser's local storage.
• 🔒 Secure Architecture: API keys are securely managed on a backend server, ensuring no credentials are exposed to the client.

This project uses a Split Architecture to ensure security and scalability:

• Tech Stack: React, Vite, Tailwind CSS, Lucide React.
• Deployment: Optimized for static hosting (e.g., GitHub Pages, Vercel, Netlify).
• Security: Contains NO API keys. It sends images to the backend via HTTP POST.

• Tech Stack: Node.js, Express, Multer (file handling), @google/genai SDK.
• Deployment: Optimized for containerized serverless hosting (e.g., Google Cloud Run).
• Security: Stores the API_KEY in environment variables. Handles CORS to allow requests only from your frontend.

• Node.js (v20+)
• A Google Cloud Project with the Gemini API enabled.
• An API Key from Google AI Studio.

The backend must be running for the app to work.

cd backend

# Install dependencies
npm install

# Configure Environment
# Create a .env file based on the example
cp .env.example .env

# Open .env and paste your GEMINI_API_KEY
# API_KEY=AIzaSy...

# Run the server (Defaults to port 8080)
npm start

cd frontend

# Install dependencies
npm install

# Run the development server
# By default, it connects to http://localhost:8080 defined in services/api.ts
npm run dev

1. Containerize: Build the Docker image using the provided Dockerfile.

   gcloud builds submit --tag gcr.io/YOUR_PROJECT_ID/ai-camera-backend

2. Deploy:

   gcloud run deploy ai-camera-backend \
     --image gcr.io/YOUR_PROJECT_ID/ai-camera-backend \
     --platform managed \
     --region us-central1 \
     --allow-unauthenticated \
     --set-env-vars API_KEY=YOUR_ACTUAL_API_KEY

3. Copy URL: Note the Service URL (e.g., https://ai-camera-backend-xyz.a.run.app).

1. Configure URL: In your local environment or build pipeline, set the VITE_API_URL environment variable to your Cloud Run URL.

   Or update frontend/src/services/api.ts directly if not using env vars during build.

2. Update Base Path: Open frontend/vite.config.ts and set the base property to your repository name:

   base: '/your-repo-name/',

3. Build & Deploy:

   npm run build
   # Upload the contents of the /dist folder to your gh-pages branch

This project has been audited and secured against API key exposure.

• ✅ No API keys in frontend code - API keys stay securely on the backend
• ✅ Environment file protection - All .env* files excluded via .gitignore
• ✅ Secure architecture - Frontend calls backend, backend calls Gemini API
• ✅ Build verification - No secrets exposed in compiled JavaScript

• 📋 SECURITY.md - Comprehensive security guide and best practices
• 🚀 DEPLOYMENT.md - Secure deployment instructions
• 📊 SECURITY_AUDIT_SUMMARY.md - Latest security audit results

• Use backend/.env for API keys (never commit this file)
• Use frontend/ directory for production (not root App.tsx)
• Set VITE_API_URL environment variable in frontend deployments
• Configure CORS to allow only your frontend domain in production
• Verify no API keys in built files: grep -r "AIza" frontend/dist/

For detailed security information, see SECURITY.md

This project is open source and available under the MIT License.