Nutrimate: AI-Powered Nutrition Guidance with RAG

🌟 Overview

Nutrimate is an AI-driven web application that utilizes Retrieval-Augmented Generation (RAG) to provide personalized nutritional insights. Built with Next.js and TypeScript, Nutrimate leverages OpenAI's GPT-4-turbo and text-embedding-3-small model to enhance nutrition-based recommendations. The application retrieves relevant nutritional data from a curated knowledge base stored in Astra DB, making it an intelligent assistant for dietary and health-conscious users.

🚀 Key Features

• AI-Powered Nutrition Insights: Uses RAG to provide contextual and evidence-based dietary recommendations.
• Personalized Suggestions: Processes user medical history, allergies, and preferences for customized advice.
• Efficient Data Retrieval: Utilizes Astra DB as a vector database for fast and relevant query results.
• Interactive Chatbot: Allows users to query food-related concerns and receive AI-generated responses.
• Seamless Web Experience: Developed with Next.js and React 19, ensuring optimal performance and user experience.

🛠️ Tech Stack

• Frontend: Next.js (15), TypeScript, React (19)
• Backend: Node.js, Express (if applicable), Serverless API
• AI Models:
  • OpenAI GPT-4-turbo for response generation
  • text-embedding-3-small for vector embeddings
• Database:
  • Astra DB for vector search
  • Custom dataset (data/data.json) for nutritional and medical insights
• APIs & Libraries:
  • langchain for RAG implementation
  • @datastax/astra-db-ts for database connectivity
  • puppeteer for potential web scraping (if used)

📊 How It Works

1. User Query Processing:

   • User enters a nutrition-related query.
   • The query is embedded using OpenAI's text-embedding-3-small.

2. Retrieval via Astra DB:

   • The vector database finds the most relevant documents from the stored knowledge base.
   • The retrieved context is fed into the LLM for response generation.

3. AI-Generated Response:

   • GPT-4-turbo generates a personalized answer using both retrieved data and prior knowledge.
   • The response is displayed in an intuitive chat-based interface.

📂 Dataset & Sources

• Custom Dataset (data/data.json): Contains structured medical and dietary information.
• Vectorized Content Storage: Implemented using Astra DB for high-speed retrieval.
• Potential External Integrations (Future Scope):
  • USDA FoodData Central
  • Open Food Facts
  • NutritionIX API

🎯 Why Nutrimate?

• Minimal Effort, Maximum Impact: Users get quick, AI-driven nutrition insights without manually searching databases.
• Personalized & Context-Aware: Unlike generic nutrition apps, Nutrimate tailors suggestions based on user history.
• Efficient RAG Pipeline: Enhances accuracy and relevance by retrieving real-world data before generating responses.

📌 Future Enhancements

• Integration with Public Nutrition APIs (USDA, Open Food Facts)
• Mobile App Version using React Native
• Voice Assistant Capabilities
• Advanced User Profiling for hyper-personalized insights

🔗 Links & References

• Project Repository: https://github.com/UtkarshK95/nutrimate
• Model Details:
  • GPT-4-turbo
  • text-embedding-3-small
  • Astra DB

---