🍽️ RecipeAI — Snap, Scan & Cook

Upload a food photo or paste any recipe text. AI structures it, analyzes it, and makes it interactive.

---

What is RecipeAI?

RecipeAI is a full-stack AI-powered recipe assistant that lets users:

• Upload a food photo — AI identifies the dish and fetches a structured recipe
• Paste any recipe text — even messy, emoji-heavy WhatsApp-style formats — and AI parses and structures it automatically
• Scale servings — ingredient quantities adjust in real time
• Localize recipes — adapt dishes to local/regional ingredients (e.g. a Kenyan twist on a European dish)
• Substitute ingredients — "I don't have thyme, what can I use?"
• Get nutritional info — calories, macros, and dietary flags per serving

---

Tech Stack

Layer	Technology
Frontend	Next.js 15 (App Router, Tailwind CSS)
Backend / API	FastAPI (Python)
AI — Text & Reasoning	Google Gemini API (free tier)
AI — Image Recognition	Google Cloud Vision API (free tier)
Auth	Supabase Auth
Database	Supabase (PostgreSQL)
Frontend Deployment	Vercel
Backend Deployment	Render

---

Project Structure

recipeai/
├── frontend/                  # Next.js app
│   ├── app/
│   │   ├── (auth)/            # Login / signup pages
│   │   ├── dashboard/         # User saved recipes
│   │   ├── recipe/[id]/       # Single recipe view
│   │   └── upload/            # Upload & paste flow
│   ├── components/
│   │   ├── RecipeCard.tsx
│   │   ├── IngredientList.tsx
│   │   ├── ServingScaler.tsx
│   │   └── NutritionBadge.tsx
│   ├── lib/
│   │   ├── supabase.ts        # Supabase client
│   │   └── api.ts             # FastAPI client helpers
│   └── public/
│
├── backend/                   # FastAPI app
│   ├── main.py
│   ├── routers/
│   │   ├── recipes.py         # Parse, structure, save
│   │   ├── vision.py          # Image recognition
│   │   ├── ai.py              # Gemini AI calls
│   │   └── auth.py            # Supabase JWT verification
│   ├── models/
│   │   └── recipe.py          # Pydantic models
│   ├── services/
│   │   ├── gemini_service.py
│   │   └── vision_service.py
│   └── requirements.txt
│
├── supabase/
│   └── migrations/            # DB schema SQL files
│
├── .env.example
├── .gitignore
└── README.md

---

Getting Started

Prerequisites

• Node.js >= 18
• Python >= 3.10
• A Supabase project
• A Google AI Studio API key (Gemini — free)
• A Google Cloud project with Vision API enabled (free tier)

---

1. Clone the repo

git clone https://github.com/your-username/recipeai.git
cd recipeai

---

2. Set up environment variables

Copy the example env file and fill in your keys:

cp .env.example .env

# Supabase
NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key
SUPABASE_SERVICE_ROLE_KEY=your_service_role_key

# Google Gemini
GEMINI_API_KEY=your_gemini_api_key

# Google Cloud Vision
GOOGLE_APPLICATION_CREDENTIALS=path/to/your/credentials.json

# FastAPI
BACKEND_URL=http://localhost:8000

---

3. Frontend setup

cd frontend
npm install
npm run dev

Frontend runs at http://localhost:3000

---

4. Backend setup

cd backend
python -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r requirements.txt
uvicorn main:app --reload

Backend runs at http://localhost:8000

API docs available at http://localhost:8000/docs

---

5. Supabase DB setup

Run the migration in supabase/migrations/ via the Supabase dashboard SQL editor, or use the Supabase CLI:

supabase db push

---

Database Schema (Supabase)

-- Users handled by Supabase Auth

create table recipes (
  id uuid primary key default gen_random_uuid(),
  user_id uuid references auth.users(id) on delete cascade,
  title text not null,
  description text,
  image_url text,
  source_text text,             -- original pasted/uploaded input
  structured_data jsonb,        -- parsed ingredients, steps, etc.
  nutrition jsonb,
  servings integer default 4,
  created_at timestamptz default now()
);

create table saved_recipes (
  id uuid primary key default gen_random_uuid(),
  user_id uuid references auth.users(id) on delete cascade,
  recipe_id uuid references recipes(id) on delete cascade,
  created_at timestamptz default now()
);

---

API Endpoints

Method	Endpoint	Description
POST	/api/recipes/parse	Parse raw recipe text using Gemini AI
POST	/api/vision/identify	Identify dish from uploaded image
GET	/api/recipes/{id}	Fetch a single saved recipe
POST	/api/recipes/save	Save structured recipe to Supabase
GET	/api/recipes/user/{user_id}	Get all recipes for a user
POST	/api/recipes/substitute	Get ingredient substitutions
POST	/api/recipes/localize	Localize recipe to a region

---

AI Features In Detail

Recipe Parser (Gemini)

Sends raw text (including emojis, informal formatting) to Gemini with a structured prompt. Returns clean JSON with title, ingredients[], steps[], servings, and tags.

Image Recognition (Google Vision)

Uploads image to Vision API → gets label annotations → uses top labels as search terms → Gemini generates or fetches a matching recipe.

Ingredient Substitution (Gemini)

Sends ingredient + dish context → returns 2–3 substitution options with notes on flavor impact.

Localization (Gemini)

Sends full recipe + target region → returns adapted recipe with locally available ingredient equivalents.

---

Deployment

Frontend → Vercel

# Push to GitHub, connect repo on vercel.com
# Add all NEXT_PUBLIC_* env vars in Vercel dashboard

Backend → Render

# Push backend/ to GitHub
# Create a new Web Service on render.com
# Build command: pip install -r requirements.txt
# Start command: uvicorn main:app --host 0.0.0.0 --port 10000
# Add all env vars in Render dashboard

---

Screenshots

---

---

Contributing

Pull requests are welcome. For major changes, please open an issue first.

---

License

MIT

---

Author

Built by Daniel Giathi — a Software Engineer passionate about building AI-powered tools for everyday life.