This project is a practical application of the concepts discussed in Enhancing RAG-based application accuracy by constructing and leveraging knowledge graphs*. It aims to advance the original implementation by integrating its logic into a basic application framework.

Features

  • Incorporation of Knowledge Graphs into Agent RAG for enhanced performance.
  • Use of the Tavily search engine as an auxiliary tool.
  • Simple deployment via LangServe.

Technical Details

  • Graph Database: We employ Neo4J for our graph database needs.
  • Model: We utilize two models, gpt-4-0125-preview for generative tasks, and text-embedding-3-small for embeddings.
  • App Deployment: The application is hosted on a t3.xlarge EC2 instance, ensuring robust performance.
  • App Construction: Development was carried out using LangServe, facilitating a streamlined build process.

Code Walkthrough

Below are the main components of our application, along with links to their implementation:

Component Link
Application serve.py
Step by Step Implementation graph_retrieval_playground.ipynb

Getting Started

Interact with the application via the following URL: http://ec2-54-80-98-106.compute-1.amazonaws.com:8001/agent/playground/. To get a feel for its capabilities, try out these queries:

  • "What are the Generative AI Practice Requirements?" - This query demonstrates the use of the Enhanced RAG with Knowledge Graphs.
  • "Who won the 2001 Copa America?" - This query showcases the integration with TavilySearchResults.