Embeddings databases are a union of vector indexes (sparse and dense), graph networks and relational databases. This enables vector search with SQL, t

Advanced RAG with graph path traversal

submited by

Style Pass

2024-07-04 11:30:04

Embeddings databases are a union of vector indexes (sparse and dense), graph networks and relational databases. This enables vector search with SQL, topic modeling, retrieval augmented generation (RAG) and more.

A standard RAG process typically runs a single vector search query and returns the closest matches. Those matches are then passed into a LLM prompt and used to limit the context and help ensure more factually correct answers are generated. This works well with most simple cases. More complex use cases, require a more advanced approach.

This article will demonstrate how semantic graphs can be used to build a more comprehensive context for LLM generation. It will cover an example of writing a short book on English history from the fall of the Roman Empire to the Norman conquest.

The first step we'll take in writing this book is collecting information. We'll use the txtai-wikipedia database, which stores all Wikipedia article abstracts as of January 2024. This database is a great way to explore a wide variety of topics. It also has the number of page views integrated in, which enables pulling frequently viewed or popular articles on a topic.