NodeRAG: Structuring Graph-based RAG with Heterogeneous Nodes

📢 News

• [2025-03-18] 🚀 NodeRAG v0.1.0 Released! The first stable version is now available on PyPI. Install it with pip install NodeRAG.

• [2025-03-18] 🌐 Official Website Launched! Visit NodeRAG_web for comprehensive documentation, tutorials, and examples.

---

🚀 NodeRAG is a heterogeneous graph-based generation and retrieval RAG system that you can install and use in multiple ways. 🖥️ We also provide a user interface (local deployment) and convenient tools for visualization generation. You can read our paper 📄 to learn more. For experimental discussions, check out our blog posts 📝.

---

🚀 Quick Start

📖 View our official website for comprehensive documentation and tutorials:
👉 NodeRAG_web

🧩 Workflow

---

NodeRAG

Conda Setup

Create and activate a virtual environment for NodeRAG:

conda create -n NodeRAG python=3.10
conda activate NodeRAG

---

Install uv (Optional: Faster Package Installation)

To speed up package installation, use uv:

pip install uv

---

Install NodeRAG

Install NodeRAG using uv for optimized performance:

uv pip install NodeRAG

Next

For indexing and answering processes, please refer to our website: Indexing and Answering

✨ Features

🔗 Enhancing Graph Structure for RAG

NodeRAG introduces a heterogeneous graph structure that strengthens the foundation of graph-based Retrieval-Augmented Generation (RAG).

🔍 Fine-Grained and Explainable Retrieval

NodeRAG leverages HeteroGraphs to enable functionally distinct nodes, ensuring precise and context-aware retrieval while improving interpretability.

🧱 A Unified Information Retrieval

Instead of treating extracted insights and raw data as separate layers, NodeRAG integrates them as interconnected nodes, creating a seamless and adaptable retrieval system.

⚡ Optimized Performance and Speed

NodeRAG achieves faster graph construction and retrieval speeds through unified algorithms and optimized implementations.

🔄 Incremental Graph Updates

NodeRAG supports incremental updates within heterogeneous graphs using graph connectivity mechanisms.

📊 Visualization and User Interface

NodeRAG offers a user-friendly visualization system. Coupled with a fully developed Web UI, users can explore, analyze, and manage the graph structure with ease.

⚙️ Performance

📊 Benchmark Performance

NodeRAG demonstrates strong performance across multiple benchmark tasks, showcasing efficiency and retrieval quality.

---

🖥️ System Performance

Optimized for speed and scalability, NodeRAG achieves fast indexing and query response times even on large datasets.