Show HN: A Simple GraphRAG Implementation

#1

Post by gusye » Fri, Aug 16, 2024, 3:48 AM UTC

I made a simple GraphRAG call nano-graphrag. The reason is I try to hack the official implementation released by Microsoft but that version is very hard to read/hack. This algorithm should not be implemented that annoying, I think. So I made a simpler one.

It's about 800-900 lines of Python, and it's portable.

Show HN: A Simple GraphRAG Implementation
github.com

Re: Show HN: A Simple GraphRAG Implementation

#2

Post by vintagedave » Fri, Aug 16, 2024, 11:54 AM UTC

This is really neat. It seems clear and usable… I may try it out in one of my projects and I hope you see lots of interest / success with it.

Re: Show HN: A Simple GraphRAG Implementation

#3

Post by 383toast » Fri, Aug 16, 2024, 9:14 PM UTC

I love that you had an "issues" section https://github.com/gusye1234/nano-graphrag?tab=readme-ov-fil...

but how does that affect the performance, theoretically?

Re: Show HN: A Simple GraphRAG Implementation

#4

Post by 383toast » Fri, Aug 16, 2024, 9:14 PM UTC

383toast wrote: post #3

I love that you had an "issues" section https://github.com/gusye1234/nano-graphrag?tab=readme-ov-fil... but how does that affect the performance, theoretically?

performance vs GraphRAG in terms of accuracy, etc.

Re: Show HN: A Simple GraphRAG Implementation

#5

Post by gusye » Sat, Aug 17, 2024, 5:10 AM UTC

383toast wrote: post #4

383toast wrote: post #3

I love that you had an "issues" section https://github.com/gusye1234/nano-graphrag?tab=readme-ov-fil... but how does that affect the performance, theoretically?

performance vs GraphRAG in terms of accuracy, etc.

Yeah, that's something I definitely love to add in next few days. I plan to use one of the Test Dataset from GraphRAG paper.

Re: Show HN: A Simple GraphRAG Implementation

#6

Post by gkorland » Sun, Aug 18, 2024, 9:03 AM UTC

[dead]