MysteryBag

Save food. Save money.
A Flutter app for buying discounted surplus-food "mystery bags" from local restaurants — with an on-device TensorFlow Lite food classifier and a Gemini-powered food assistant.

_{Home · Bag details · AI food assistant}

---

📖 About

MysteryBag is the final project of the CiC program. Restaurants put their unsold, still-good food into discounted mystery bags; users reserve a bag and pick it up — less food waste, cheaper meals. On top of a full e-commerce flow (auth, cart, checkout, reviews, points), the app adds two AI features: a TensorFlow Lite model that recognises the dish from a photo on-device, and a Gemini vision check that looks for signs of spoilage, plus a food-only AI chat assistant.

✨ Features

• 🔐 Authentication — email/password, Google and Facebook sign-in (Firebase Auth)
• 👋 Onboarding and splash flow
• 🍱 Restaurants & mystery bags — available bags, bags left, old vs. new price (EGP), bag details, "reserve for pickup"
• ⭐ Reviews — streamed live from Firestore, with average rating and review analytics
• 🛒 Cart & checkout — cash on delivery, PayPal, or Visa card (flutter_credit_card)
• 🎁 Points & rewards — points balance with a rewards & discounts page
• 📸 Food scan — on-device TFLite classifier (2,000+ dish labels, 224×224 input), then a Gemini vision analysis in Arabic: verdict, visible signs and a safety tip
• 💬 AI food chat — Gemini chat with keyword rules that keep it on food topics (calories, healthy choices, suggestions)
• ❤️ Favourites, avatar picker, profile
• 🌍 English / Arabic localization and 🌙 light / dark theme

🛠 Tech stack

Concern	Choice
Framework	Flutter, Dart 3
State management	bloc / flutter_bloc (Cubits) + provider
Architecture	Feature-first Clean Architecture (data / domain / presentation), get_it DI, dartz Either errors
Backend	Firebase Auth, Cloud Firestore, Cloud Functions
On-device ML	tflite_flutter, image (model in assets/tflite/)
Generative AI	Gemini via a proxy — Firebase Function (functions/) or Cloudflare Worker (worker.js) — so the API key stays server-side
Payments	flutter_paypal_payment, flutter_credit_card
Localization	flutter_localizations + ARB files (gen-l10n)
Other	shared_preferences, flutter_phoenix, dots_indicator, image_picker, crypto

🏗 Project structure

lib/
├── core/
│   ├── cubits/         # locale, theme, avatar
│   ├── entities/ models/ repos/   # products, restaurants, bags, reviews, orders
│   ├── services/       # Firebase auth, Firestore, prefs, get_it
│   └── widgets/ utils/ helper_functions/
├── features/
│   ├── auth/  onBoarding/  splash/
│   ├── home/           # restaurants, bags, cart, profile, points, reviews
│   ├── check_out/      # cash / PayPal / Visa
│   ├── food_scan/      # TFLite classifier + Gemini vision
│   └── ai_chat/        # Gemini food assistant
└── l10n/               # app_en.arb, app_ar.arb
assets/tflite/          # food_model.tflite + labels.txt
functions/              # Express API: POST /chat, POST /vision
worker.js               # Cloudflare Worker version of the same proxy

🚀 Getting started

Prerequisites: Flutter SDK (Dart ^3.10), a Firebase project, and a Gemini API key for the AI features.

git clone https://github.com/youssef336/CiC_final_project.git
cd CiC_final_project
flutter pub get
flutter run

Configuration

1. Firebase — run flutterfire configure to use your own project.
2. Gemini proxy — either deploy worker.js to Cloudflare Workers with a GEMINI_API_KEY secret, or deploy functions/ after setting firebase functions:config:set gemini.key=.... Then update _proxyBaseUrl in lib/features/ai_chat/presentation/views/ai_chat_view.dart and lib/features/food_scan/presentation/views/food_scan_view.dart.
3. PayPal — use your own sandbox credentials in lib/core/utils/app_key.dart.

👤 Author

Youssef Belal Zakaria · GitHub @youssef336

⭐ If you like this project, give it a star!