Build a simple Graph RAG workflow with LlamaIndex and Kuzu

[@jasoncole3253]

Thanks for the video. Unfortunately I find the text2cypher to be quite off, in my case to start the graph traversal it works better to store graph entities in a vector db and match them with closest entities from the query

[@KuzuDB]

Hi Jason, yes, you're right - in many cases the text2Cypher depends on the underlying LLM and the associated prompt (it takes a bit of prompt engineering to get better Cypher). There are also some fine-tuned Cypher generation LLMs these days, so that might be a good option too. The demo shown here was purely for demonstrating a workflow - you can look at the following repo for some ideas on custom prompts for Cypher generation: github.com/kuzudb/graph-rag-workshop