🧠 NutriGen: RAG-Based Nutritional Chatbot

NutriGen is a Retrieval-Augmented Generation (RAG) system designed to intelligently answer questions about nutrition and dietary science using contextual retrieval, embeddings, and LLM-based augmentation.
It combines Jupyter-based data preparation, Supabase vector search, and a Next.js + OpenAI frontend for real-time conversational responses.

---

🚀 Overview

NutriGen leverages the power of:

• LLM embeddings & retrieval for understanding nutrition documents
• Supabase pgvector for fast similarity search
• OpenAI GPT models (and optionally Gemma 7B locally) for natural language responses
• Next.js frontend for an elegant chat-based user interface

The system reads and embeds research documents, stores them as vectors in Supabase, and then uses OpenAI to generate context-aware responses.

---

🧩 System Architecture

📁 NutriGen/ ├── 📓 main_notebook.ipynb ← Jupyter notebook for manual chunking & embedding │ ├─ Handles 5 chunking strategies │ ├─ Embedding via all-mpnet-base-v2 and local Gemma 7B │ └─ Performs RAG pipeline testing locally │ ├── 📁 Backend API/ │ ├── ingest.py ← Automated ingestion + embedding pipeline │ │ ├─ Uses OpenAI text-embedding-3-small │ │ ├─ Stores embeddings in Supabase (pgvector) │ │ └─ Connects PostgreSQL for vector queries │ └── ... other scripts │ ├── 📁 app/ │ ├── api/chat/route.ts ← Next.js API route for RAG query + OpenAI augmentation │ └── page.tsx ← Frontend chat interface │ ├── 📄 .env ← Environment variables (OpenAI & Supabase keys) ├── 📄 package.json ├── 📄 README.md └── 📁 public/ └── (images for visualization)

yaml Copy code

---

🧠 Workflow

1️⃣ Data Preparation (in Jupyter)

• Manually chunked 5 types of text splitting for different structural analysis.
• Used Sentence Transformers (all-MiniLM-L6-v2, all-mpnet-base-v2) to embed text.
• Tested Gemma 7B locally for comparison in generation quality.
• Stored embeddings temporarily for inspection and visualization.

2️⃣ Backend Ingestion (Supabase + pgvector)

• ingest.py automates embedding creation and uploading.
• Uses OpenAI’s text-embedding-3-small model.
• Embeddings are stored in Supabase (PostgreSQL + pgvector).
• A RPC function match_documents in Supabase retrieves top-k matches by cosine similarity.

3️⃣ RAG API (Next.js Route)

• route.ts receives a query → embeds it → retrieves relevant chunks via Supabase RPC.
• If results found: constructs a contextual prompt and sends to OpenAI Chat Completion API.
• If not: returns a helpful fallback response.

4️⃣ Frontend (Next.js + Tailwind + Framer Motion)

• Built in /app/page.tsx
• Features a minimal sleek chat interface inspired by OpenAI Chat UI.
• Displays:
  • User + Assistant messages
  • Animated typing indicator (three dots while generating)
  • Clickable citations showing contextual snippets from source pages
  • Dynamic sources section with similarity scores
• Styled in the theme of arcprize.org for a refined scientific aesthetic.

---

🧰 Tech Stack

Layer	Tools / Frameworks
Frontend	Next.js 15, React, Tailwind CSS, Framer Motion
Backend API	Node.js, OpenAI API
Database / Vector Store	Supabase (PostgreSQL + pgvector)
Embedding Models	OpenAI text-embedding-3-small, Sentence Transformers
Generation Models	OpenAI GPT-4o / GPT-3.5, Gemma 7B (local)
Data Prep Notebook	Python, torch, sentence-transformers, numpy, pandas

---

⚙️ Environment Variables

Create a .env file in the root:

OPENAI_API_KEY=your_openai_key
SUPABASE_URL=your_supabase_project_url
SUPABASE_SERVICE_ROLE_KEY=your_supabase_service_role_key