Nutrimate is an AI-powered web application that recommends recipes based on ingredients, nutritional preferences, and user input. It leverages advanced machine learning techniques, GPT-3.5, Gemini Vision, and a modular architecture to deliver fast, accurate, and personalized recipe recommendations.
-
🔍 Multi-Input Search
- Text-based search
- Image-based ingredient recognition (Gemini Vision)
- Nutrient-based filtering and diet customization
-
🧠 AI-Powered Intelligence
- Recipe parsing with OpenAI GPT-3.5
- Ingredient detection from images using Gemini Vision API
- Hybrid ML model (TF-IDF + FastText + CNN + SHAP)
-
📊 Nutritional Insights
- Custom dietary filters (vegan, keto, etc.)
- Macronutrient and calorie tracking
-
🧾 User Search History
- Personalized experience with search logs
- Managed securely using Appwrite Authentication
-
📱 Modern Frontend
- Built with Next.js and Tailwind CSS
- Mobile responsive and accessible UI
| Layer | Technology |
|---|---|
| Frontend | Next.js, Tailwind CSS, SWR |
| Backend | Flask (Python), REST API |
| AI Services | OpenAI GPT-3.5, Gemini Vision API |
| Authentication | Appwrite (JWT, OAuth) |
| Database | MongoDB Atlas |
| Deployment | Vercel (Frontend), Hugging Face Spaces (Backend) |
The project follows a modular microservices architecture:
- Frontend: React components organized into pages, layouts, features
- Backend: Flask app divided into services (recommendation, analysis, user)
- AI Adapters: GPT-3.5 for text parsing, Gemini Vision for image analysis
- Data Layer: MongoDB with collections for recipes, users, feedback, and history
| Endpoint | Description |
|---|---|
/api/recommend |
Main recommendation engine |
/api/analyze-food-image |
Ingredient recognition via image upload |
/api/extract-recipe-attributes |
Recipe parsing with GPT-3.5 |
/api/user/history |
Fetch or update search history |
/api/user/preferences |
Manage dietary filters and preferences |