🧠 Knowledge Graph RAG

A Retrieval-Augmented Generation (RAG) system powered by Neo4J Knowledge Graphs and LangChain — combining structured graph reasoning with LLM-based question answering.

---

🎯 Motivation

Traditional RAG systems retrieve flat text chunks from vector databases. This project explores a graph-enhanced approach: entities and their relationships are stored in a Neo4J Knowledge Graph, enabling the LLM to reason over structured connections — not just similar text.

This is particularly powerful for domains where relationships matter: education, healthcare, e-commerce, and enterprise knowledge management.

---

🏗️ Architecture

User Query
    │
    ▼
┌─────────────────────┐
│   Query Processor   │  → Entity extraction (spaCy / LLM)
└─────────┬───────────┘
          │
    ┌─────┴──────┐
    │            │
    ▼            ▼
┌────────┐  ┌──────────────┐
│ Neo4J  │  │ Vector Store │  (Chroma / Qdrant)
│  KG    │  │  Embeddings  │
└────┬───┘  └──────┬───────┘
     │              │
     └──────┬───────┘
            │  Context Fusion
            ▼
    ┌───────────────┐
    │  LLM (Claude  │
    │  / GPT-4)     │
    └───────┬───────┘
            │
            ▼
       Final Answer

---

✨ Features

• 🔗 Graph-based retrieval via Neo4J Cypher queries
• 🧩 Hybrid search — combines graph traversal + semantic vector search
• 🤖 LangChain integration with GraphCypherQAChain
• 📊 Knowledge Graph builder — ingest raw text and extract entities/relations automatically
• 🖥️ FastAPI endpoint for easy integration
• 📓 Jupyter notebooks for exploration and demos

---

Screenshot

---

🚀 Quickstart

Prerequisites

• Python 3.11+
• Neo4J Desktop or Docker
• OpenAI or Anthropic API key

1. Clone & install

git clone https://github.com/YOUR_USERNAME/knowledge-graph-rag.git
cd knowledge-graph-rag
pip install -r requirements.txt

2. Start Neo4J

docker-compose up -d

Neo4J Browser: http://localhost:7474 (user: neo4j, password: password)

3. Configure environment

cp .env.example .env
# Edit .env with your API keys and Neo4J credentials

4. Build the Knowledge Graph

python src/graph/kg_builder.py --input data/raw/sample_docs.txt

5. Run a query

python src/main.py --query "What are the prerequisites for machine learning?"

6. Start the Web-UI

uvicorn src.api.app:app --reload
# Web-UI: http://localhost:8000

7. Start the API

uvicorn src.api.app:app --reload
# API docs: http://localhost:8000/docs

---

📁 Project Structure

knowledge-graph-rag/
├── data/
│   ├── raw/                  # Input documents (txt, pdf, json)
│   └── processed/            # Extracted entities & relations
├── src/
│   ├── graph/
│   │   ├── kg_builder.py     # Entity/relation extraction → Neo4J
│   │   ├── neo4j_client.py   # Neo4J connection & Cypher queries
│   │   └── schema.py         # Graph schema definition
│   ├── rag/
│   │   ├── retriever.py      # Hybrid graph + vector retriever
│   │   ├── embeddings.py     # Embedding model wrapper
│   │   └── chain.py          # LangChain RAG chain
│   ├── api/
│   │   └── app.py            # FastAPI application
│   └── main.py               # CLI entrypoint
├── notebooks/
│   ├── 01_kg_exploration.ipynb    # Explore the knowledge graph
│   ├── 02_rag_pipeline.ipynb      # RAG pipeline walkthrough
│   └── 03_evaluation.ipynb        # Evaluate retrieval quality
├── tests/
│   ├── test_kg_builder.py
│   ├── test_retriever.py
│   └── test_api.py
├── docs/
│   └── architecture.md
├── docker-compose.yml
├── requirements.txt
├── .env.example
└── README.md

---

🧪 Example

from src.rag.chain import GraphRAGChain

chain = GraphRAGChain()
response = chain.query("Which topics are related to neural networks?")
print(response)
# → "Neural networks are related to: deep learning, backpropagation,
#    activation functions, and gradient descent. In the knowledge graph,
#    they connect directly to CNNs, RNNs, and Transformers..."

---

📓 Notebooks

Notebook	Description
01_kg_exploration.ipynb	Visualize the graph, run Cypher queries
02_rag_pipeline.ipynb	End-to-end RAG demo step by step
03_evaluation.ipynb	Compare graph RAG vs. standard RAG

---

🛣️ Roadmap

• Support PDF ingestion
• Add GraphRAG (Microsoft) comparison
• Streamlit demo UI
• Multi-hop reasoning over graph paths
• Evaluation with RAGAS framework

---

🤝 Related Work

• DeepLearning.AI – Knowledge Graphs for RAG
• LangChain GraphCypherQAChain
• LightRAG – Knowledge Graph + RAG

---

👤 Author

Paul Meteng — Software Engineer & AI Engineer
LinkedIn · GitHub

---

📄 License

MIT License — see LICENSE