/GRAPHRAG_Exploration

This repository explores Microsoft GRAPHRAG working by ingesting a PDF document.

Primary LanguageJupyter Notebook

Repository Name: RAG Exploration

Description

Basic RAG method is based on similarity search of vectors and hence most of the times fails to capture the overall syntax of the text presented to it, especially understanding it at higher level. To solve this problem, Knowledge Graph based RAG has been introduced, and Microsoft's GRAPHRAG seems to be a gamechanger, which takes it even further by capturing the entities and their relationship at a very deeperlevel.

This Repository consists of 2 notebooks which compare the results of knowledge graphs generated by GRAPHRAG and simple LLM based model.

Installation

Please run the following command to install required dependencies:

pip install -m requirements.txt

NOTE: Open-AI subscription will be required to run GRAPHRAG project. I have used my Azure based subscription.

Usage

Some of the commands specific to GRAPHRAG in this repo has to be run in terminal window and has been mentioned explictly.

Contributing

Explain how others can contribute to the repository. Include guidelines for submitting issues or pull requests.

Acknowledgments

https://microsoft.github.io/graphrag/posts/get_started/ Prompt Engineering YouTube Channel