So far, we have worked with Large Language Models that operate on their internal, pre-trained knowledge. This approach has a clear limitation: the model is unaware of any data created after its training cut-off date and has no access to private or domain-specific information. This chapter introduces a method to overcome this by connecting an LLM to external data sources.

This method is known as Retrieval Augmented Generation (RAG). The core idea is to first retrieve relevant documents from an external knowledge base and then provide them to the LLM as context for generating a response. This grounds the model's output in factual, timely, and specific information.

Throughout this chapter, we will construct a RAG system piece by piece. You will learn how to:

• Load Documents: Use LangChain’s DocumentLoaders to ingest data from various file formats like PDFs and text files.
• Process Text: Split large documents into smaller, manageable chunks using TextSplitters.
• Create Embeddings: Convert text chunks into numerical vectors, which allows for semantic comparison.
• Store and Retrieve: Use a VectorStore to index these embeddings for efficient similarity searches, where relevance is determined by the distance between a query vector $v_q$ and document vectors $v_d$.
• Build a Q&A Chain: Combine these components with a retriever and an LLM to build a complete question-answering application that can reason over your own data.