Intro to GraphRAG: Session Resources

[multispark]

This is the resources and discussion thread for this live stream: https://aka.ms/raghack/graph
It will be updated with links after the session.

Recording: https://www.youtube.com/watch?v=f6pUqDeMiG0

Slides:
PDF
PPTX

John's Code Samples: https://github.com/ms-johnalex/intro-to-graphrag

Please ask any follow-up questions here, and the speakers will see them.

[Yash-Kavaiya]

How can we apply Graph RAG using Github Models ?

[moapurva]

Can you be more specific about what you mean by GitHub Models?

[pamelafox]

I'm guessing Yash is referring to https://github.com/marketplace/models
They are freely usable models, but they have very low rate limits, so I doubt they can be used for GraphRAG. I only use them when testing single calls to the APIs, I never use them for multiple call situations like evaluation or processing.

[moapurva]

Thanks Pamela. @Yash-Kavaiya, we recommend using only GPT-4 or newer models, and while GPT-4o-mini is available, the rate limits are too low. https://docs.github.com/en/github-models/prototyping-with-ai-models#rate-limits.

[shyshin]

As a follow up to your answer on using single domain input for graph RAG, how do we go about graph updation? when GraphRAG app eventually goes into production. Do we index everything again?

[moapurva]

Incremental update functionality is on the roadmap. Until then, start with smaller datasets to validate proper graph creation before adding more data.