Retrieval Augmented Generation with Elastic

In this chapter, we will continue on our journey through the world of Elasticsearch and take a deep dive into one of the most advanced and exciting features that Elastic has to offer: retrieval augmented generation (RAG) search experiences. If you followed along with our previous chapters, you’ll be familiar with the Elastic Learned Sparse EncodeR (ELSER) pre-trained model and the concept of a vector search, which offers you the luxury of semantic understanding in search results. You also learned about the power of reciprocal rank fusion (RRF) in combining the strengths of lexical and vector searches.

In this chapter, we will take it up a notch by integrating these concepts into a comprehensive pipeline, culminating in the expansion of a large language model (LLM) with RAG. The goal here is to combine the strengths of lexical, vector, and contextual search to offer the most relevant search results, enhancing the user experience...