AI-powered furniture discovery. Upload a photo of your room, describe what you want, and swipe through real products from real stores — placed directly into your space.

1. You upload a room photo
2. You describe what you're looking for ("Add a sofa", "Make it cozy", etc.)
3. Gemini AI analyzes your room — style, colors, dimensions, lighting
4. Products from our database are matched, scored, and ranked
5. Each product is composited into your room photo (background removal + placement)
6. You swipe through results Tinder-style — right to save, left to pass

Products link directly to real stores (Blueprint Home, Elte, IKEA, CB2, etc.) so you can buy what you like.

• React 18 + TypeScript + Vite
• Tailwind CSS + shadcn/ui + Framer Motion
• Google Gemini 2.5 (room analysis, intent extraction, product ranking)
• Supabase (product catalog, stores, attributes, dimensions)
• Netlify Functions (serverless API)
• @imgly/background-removal (client-side bg removal for compositing)

• Node.js 18+
• A Google AI Studio API key (Gemini)
• A Supabase project with the furniture catalog schema

npm install
cp .env.example .env

Fill in .env:

GEMINI_API_KEY=AIzaSy...
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_SERVICE_ROLE_KEY=eyJ...
VITE_SUPABASE_URL=https://your-project.supabase.co
VITE_SUPABASE_ANON_KEY=eyJ...

You need two processes — the Vite frontend and the local API server:

# Terminal 1: API server (port 8888)
npm run dev:api

# Terminal 2: Frontend (port 8080)
npm run dev

Open http://localhost:8080

Apply the migration to set up tables:

# Using Supabase CLI
supabase db push

# Or manually run:
# supabase/migrations/20240001_furniture_catalog.sql

Seed data is in supabase/seed/mock_ontario_furniture.sql.

src/
├── components/           # UI components
│   ├── SwipeResultsPage  # Main swipe interface (AI results)
│   ├── RoomUploadPage    # Room photo + prompt input
│   ├── TrenderSwipeScreen # Tinder-style card deck (sample data)
│   ├── LandingPage       # Hero + entry points
│   └── ui/               # shadcn primitives
├── lib/
│   ├── gemini/           # Gemini client, prompts, parsers
│   ├── room-overlay/     # Pipeline: analyze → query → score → rank
│   ├── compositor/       # Background removal + image stitching
│   └── ingestion/        # Product data ingestion providers
├── services/             # Legacy API services
├── types/                # TypeScript interfaces
└── data/                 # Static sample furniture data
server/
└── index.ts              # Local dev API server (mirrors Netlify Functions)
netlify/functions/        # Serverless endpoints (production)
scripts/                  # Scraping, ingestion, DB utilities

User prompt → extractIntent → translateUserPrompt → analyzeRoomWithGemini
  → buildProductCandidates (Supabase query)
  → scoreAndRankCandidates (structured scoring)
  → rankProductsWithGemini (Gemini rerank, optional)
  → return top candidates + placement guidance

Product image → background removal (@imgly) → scale + position → composite onto room photo

The app is configured for Netlify with netlify.toml. Functions in netlify/functions/ are deployed automatically.

Set environment variables in the Netlify dashboard:

• GEMINI_API_KEY
• SUPABASE_URL
• SUPABASE_SERVICE_ROLE_KEY

npm run build
# Output: dist/

MIT