NutriNani - Decoding Labels for a Healthier You

🌐 Live demo: nutrinani.vercel.app

A healthcare & wellness GenAI app that helps you scan food labels, get personalized verdicts, safe recipes, and trusted health advice — all tailored to your family's dietary needs, allergies, and health conditions.

Features

• 🔍 Food Label Scanner - Scan ingredients and nutrition labels, get an instant safe/unsafe verdict based on the active profile's health conditions and allergies
• 👥 Family Profiles - Netflix-style multiple profiles so every family member gets recommendations tuned to their own diet, allergies, and diseases
• 🍽️ Personalized Recipes - Recipes filtered by what's in your pantry and safe for the active profile
• 📦 Smart Inventory - Track pantry items with expiry dates, one-click "cook with what I have"
• 💬 Nani Chatbot - Conversational health assistant with full chat history (last 5 sessions, autosaved)
• 🎙️ Voice Bot - Ask Nani questions hands-free
• 👨‍👩‍👧 Community - See what other families are cooking and sharing

Tech Stack

Frontend

• React 18 + TypeScript + Vite
• Tailwind CSS + Radix UI + shadcn/ui
• React Router, TanStack Query, React Hook Form + Zod
• react-markdown for chatbot replies
• @zxing/browser + tesseract.js for barcode/OCR scanning

Backend (AWS, serverless)

• AWS Lambda (Node.js 20.x) behind API Gateway HTTP API v2
• DynamoDB for inventory, profiles, and chat sessions
• AWS Cognito (User Pool + Hosted UI) for auth, with Google OAuth
• Amazon Bedrock + Gemini for the chatbot
• AWS SAM for infra-as-code

Hosting

• Frontend: Vercel
• Backend: AWS (ap-southeast-2)

Getting Started

Prerequisites

• Node.js 20+ & npm
• An AWS account (only if you want to redeploy the backend — the hosted backend is already live)

Local development

git clone https://github.com/SirjanSingh/nutrinani.git
cd nutrinani/nutrinani
npm install
cp .env.example .env       # fill in your Cognito + API values
npm run dev

The app runs at http://localhost:8080.

Environment variables

See .env.example for the full list. The minimum required set:

Variable	Purpose
VITE_COGNITO_REGION	AWS region of your Cognito User Pool
VITE_COGNITO_USER_POOL_ID	Cognito User Pool ID
VITE_COGNITO_USER_POOL_CLIENT_ID	Cognito App Client ID (no secret)
VITE_COGNITO_DOMAIN	Hosted UI domain (needed for Google sign-in)
VITE_GOOGLE_CLIENT_ID	Google OAuth Web Client ID
VITE_API_BASE_URL	API Gateway base URL for the deployed Lambdas

VITE_COGNITO_REDIRECT_SIGN_IN / _SIGN_OUT should be left unset in production — the app falls back to window.location.origin, which works for every Vercel preview and production URL automatically.

Deploying

Frontend → Vercel

This repo is Vercel-ready:

• vercel.json configures the Vite build and adds an SPA rewrite (/* → /index.html) so React Router deep links don't 404
• Root Directory in Vercel project settings: nutrinani
• Add the env vars from .env.example under Settings → Environment Variables
• Cognito callback URLs and Google OAuth authorized origins must include your Vercel URL

Backend → AWS SAM

cd nutrinani/backend/inventory-api
sam build
sam deploy --guided

The chat-sessions Lambda (backend/chat-sessions-api/) is deployed manually via AWS CLI. It attaches to the same API Gateway (/chat/sessions, /chat/sessions/{id}) and uses the same Cognito JWT authorizer as the inventory API.

Project Structure

nutrinani/
├── nutrinani/                    # frontend (Vite app, Vercel root)
│   ├── src/
│   │   ├── components/           # UI components
│   │   │   ├── ChatBot.tsx       # Nani chat with markdown + history sidebar
│   │   │   ├── Scanner.tsx       # label scanner (barcode + OCR)
│   │   │   ├── Dashboard.tsx
│   │   │   ├── Inventory.tsx
│   │   │   ├── Recipes.tsx
│   │   │   └── profiles/         # family profile management
│   │   ├── contexts/             # AuthContext, ProfileContext
│   │   ├── lib/                  # amplify config, API clients
│   │   ├── pages/                # route entry points
│   │   └── types/                # shared TS types
│   ├── public/                   # static assets
│   ├── vercel.json               # Vercel SPA config
│   └── .env.example
└── backend/
    ├── inventory-api/            # pantry CRUD Lambda
    └── chat-sessions-api/        # chat history CRUD Lambda

Available Scripts

From nutrinani/nutrinani/:

• npm run dev – start dev server on :8080
• npm run build – production build to dist/
• npm run preview – preview the production build locally
• npm run lint – ESLint

Family Profiles

NutriNani is built around a Netflix-style profile picker:

• Up to 6 profiles per account
• Each profile stores its own age, diet type, allergies, diseases, and preferences
• Switching profiles instantly re-filters recipes, scanner verdicts, and chatbot answers
• Profile data is scoped per-user in DynamoDB (key = Cognito sub)

Author

Sirjan Singh — @SirjanSingh

License

MIT — see LICENSE for details.

Support

For support, email support@nutrinani.com or open an issue on this repo.