Building AI Agents with LLMs, RAG, and Knowledge Graphs

6

Advanced RAG Techniques for Information Retrieval and Augmentation

In the previous chapter, we discussed RAG and how this paradigm has evolved to solve some shortcomings of LLMs. However, even naïve RAG (the basic form of this paradigm) is not without its challenges and problems. Naïve RAG consists of a few simple components: an embedder, a vector database for retrieval, and an LLM for generation. As mentioned in the previous chapter, naïve RAG involves a collection of text being embedded in a database; once a query from a user arrives, text chunks that are relevant to the query are searched for and provided to the LLM to generate a response. These components allow us to respond effectively to user queries; but as we shall see, we can add ...