Most RAG systems are designed for single-turn queries: a user asks a question, the system retrieves relevant documents, and the model generates an answer. But real interactions are rarely that simple. Users often ask follow-up questions, refer to earlier context, or refine their queries over multiple turns.

This is where Conversational RAG comes in. Instead of treating every question independently, the system maintains conversation history and uses it to better understand the user’s intent. It may rewrite follow-up questions, track context across turns, and retrieve documents that are relevant to the entire conversation.

Handling conversations introduces new challenges. The system must manage context windows, memory, query rewriting, and retrieval across multiple turns while still returning accurate and grounded answers.

In this chapter, we will explore how to design RAG systems that support multi-turn conversations, allowing users to interact naturally while maintaining relevant and context-aware responses.