Working with Knowledge Graph (.ttl) for Retrieved Augmented Generation purpose

Hi everyone,

This topic is more a general discussion, as i'm looking for experiences or opinions.

Have you ever used a knowledge graph hosted on Stardog to set up a RAG. More specifically, as a external data source on which the LLM relies to answer a question?

My question concerns the interpretability of a Knowledge Graph into something that the LLM can understand (a collection of triplets, for example).

If you're talking RAG and Stardog you should start with their new Voicebox feature and what they're calling Safety RAG. Safety RAG: Improving AI Safety by Extending AI's Data Reach | Stardog

If you're taking about backing a RAG based LLM solution in general take a look at the work done by Microsoft Research on GraphRAG GraphRAG: Unlocking LLM discovery on narrative private data - Microsoft Research

On a somewhat related topic that I just came across this, llama3 in a single Java file! GitHub - mukel/llama3.java: Practical Llama 3 inference in Java

1 Like

Your answer is very informative, it's great.

The link to Github is great, that's the kind of thing I was looking for.

Thanks a lot!

You're very welcome. I think Stardog is a much overlooked platform for not only using AI but for AIOps. I'm going to hop in my Time Machine and head back to this SimplePhotoAnnotation - W3C Wiki Check out the time stamp at the bottom. 2005!