Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
Graph RAG: Combining the power of vectors and graphs for better retrieval
Learn how to combine graph and vector search for improved RAG, extract structured data from unstructured text, and build an updatable on-disk knowledge graph.
In this demo, I’ll showcase how to bring together the power of graph and vector search to provide LLMs with better context when generating responses for RAG. Because constructing a knowledge graph is typically the biggest bottleneck towards using them in real applications, I’ll also quickly showcase some new frameworks and tools to prompt LLMs to extract structured data from unstructured text, and how to easily store them as an on-disk graph that can be improved over time.
Graph RAG experiments utilize Kzu embedded graph database for persistent knowledge retrieval.
Constructs knowledge graphs using NetworkX, LanceDB, and gensim for entity embeddings and NER.
Compose Email
Loading recent emails...