https://github.com/dianaow/products-knowledge-graph

Implementing Retrieval-Augmented Generation (RAG) with constructed Knowledge Graph
https://github.com/dianaow/products-knowledge-graph

knowledge-graph langchain langchain-python neo4j rag vector-search

Last synced: about 1 month ago
JSON representation

Implementing Retrieval-Augmented Generation (RAG) with constructed Knowledge Graph

• Host: GitHub
• URL: https://github.com/dianaow/products-knowledge-graph
• Owner: dianaow
• Created: 2024-02-05T13:49:27.000Z (11 months ago)
• Default Branch: main
• Last Pushed: 2024-02-05T16:15:44.000Z (11 months ago)
• Last Synced: 2024-02-05T17:34:57.675Z (11 months ago)
• Topics: knowledge-graph, langchain, langchain-python, neo4j, rag, vector-search
• Language: HTML
• Homepage: https://dianaow.com/blog/knowledge-graph-rag/
• Size: 2.11 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Implementing Retrieval-Augmented Generation (RAG) with a Knowledge Graph (KG)

### Constructing a KG with skincare products found on an e-commerce website

### Product similarity search with vector embeddings

Tools: Python, OpenAI (free tier), LangChain, Neo4j

### Set up

1. Create a Python virtual environment where the dependencies for this project will be installed.

```

cd server

python3 -m venv venv

```

2. Activate the environment and install all the packages available in the requirement.txt file.

```

source venv/bin/activate

pip install -r ./requirements.txt

```

3. If a  `.env` file is not present in the server folder, create one to store the private OpenAI API key, which is required to use the LLMs. 

```

OPENAI_API_KEY=XXXXXX

NEO4J_PW=XXX

```

4. Run the Python script to web scrape product information found online.

```

python3 web_scrape.py

```

5. The `llm_kg.ipynb` file explains:

- How to use LLM to extract new relations from product descriptions and construct Knowledge Graph

- Methods to query the graph database (based on embeddings, LLM-generated entities in prompt, Cypher)

![Image of Knowledge Graph](https://github.com/dianaow/products-knowledge-graph/blob/main/graph_overview.png)