🔮 GraphRAG — Knowledge Graph-Powered RAG

Retrieval Augmented Generation enhanced with Knowledge Graphs for deeper, context-aware AI answers.

GraphRAG combines vector similarity search (FAISS) with knowledge graph traversal (NetworkX) to provide richer, more connected answers than traditional RAG systems. Upload documents, watch entities and relationships emerge as an interactive graph, and query your knowledge base with graph-augmented context.

---

✨ Features

• 📄 Document Upload — Drag & drop PDF or TXT files for automatic processing
• 🧠 Entity & Relation Extraction — LLM-powered extraction of entities and their relationships
• 🕸️ Knowledge Graph — Automatic construction of a directed knowledge graph from documents
• 🔍 Hybrid Retrieval — Combines FAISS vector search with graph traversal for comprehensive context
• 📊 Interactive D3.js Visualization — Explore your knowledge graph with zoom, pan, drag, and hover tooltips
• 💬 Chat Interface — Ask questions with a real-time pipeline visualization showing each RAG step
• 🎨 Glassmorphism UI — Premium dark-themed interface with animations and particle effects

---

🏗️ Architecture

┌─────────────┐     ┌──────────────────────────────────────────┐
│  Upload PDF │────▶│           GraphRAG Pipeline              │
│  or TXT     │     │                                          │
└─────────────┘     │  1. Chunk Text (500 words, 100 overlap)  │
                    │  2. Extract Entities & Relations (LLM)   │
                    │  3. Build Knowledge Graph (NetworkX)     │
                    │  4. Create Embeddings (Sentence-BERT)    │
                    │  5. Index in FAISS                       │
                    └──────────────────────────────────────────┘

┌─────────────┐     ┌──────────────────────────────────────────┐
│  Ask a      │────▶│           Query Pipeline                 │
│  Question   │     │                                          │
└─────────────┘     │  1. Vector Search (FAISS, top-3 chunks)  │
                    │  2. Extract Query Entities (LLM)         │
                    │  3. Graph Traversal (2-hop neighbors)    │
                    │  4. Combined Context → LLM Answer        │
                    └──────────────────────────────────────────┘

---

🚀 Getting Started

Prerequisites

• Python 3.9+
• Groq API Key (free tier available)

Installation

1. Clone the repository

   git clone https://github.com/<your-username>/graph-rag.git
   cd graph-rag

2. Create a virtual environment

   python -m venv venv
   source venv/bin/activate  # macOS/Linux
   # or
   venv\Scripts\activate     # Windows

3. Install dependencies

   pip install -r requirements.txt

4. Configure environment variables

   cp .env.example .env

   Edit .env and add your Groq API key:

   GROQ_API_KEY=your_groq_api_key_here
   

5. Run the application

   python app.py

6. Open in browser

   http://localhost:5001
   

---

📡 API Endpoints

Method	Endpoint	Description
GET	/	Serve the frontend UI
POST	/upload	Upload and process a document (TXT/PDF)
POST	/query	Query the GraphRAG system
GET	/graph	Get knowledge graph data (D3.js format)
GET	/stats	Get system statistics
POST	/reset	Reset the entire system

---

🛠️ Tech Stack

Component	Technology
Backend	Flask, Python
Knowledge Graph	NetworkX (Directed Graph)
Vector Search	FAISS (Facebook AI Similarity Search)
Embeddings	Sentence-Transformers (all-MiniLM-L6-v2)
LLM	Groq (llama-3.1-8b-instant)
Frontend	Vanilla JS, D3.js, CSS
PDF Parsing	PyPDF2

---

📁 Project Structure

graph-rag/
├── app.py              # Flask server & REST API
├── graph_rag.py        # Core GraphRAG engine
├── requirements.txt    # Python dependencies
├── .env                # Environment variables (not committed)
├── sample_doc.txt      # Sample document for testing
└── static/
    ├── index.html      # Frontend UI
    ├── style.css       # Glassmorphism styles & animations
    └── app.js          # D3.js graph viz, chat, & upload logic

---

⚠️ Rate Limits

The free tier of Groq has a 6,000 tokens per minute limit. The engine includes built-in retry logic with exponential backoff to handle rate limiting gracefully. For heavier usage, consider upgrading to the Groq Dev Tier.

---

📝 License

MIT