Overview

This chapter introduced agentic behavior in RAG.

The main idea is to let a system choose between different search tools instead of always following one fixed retrieval path. This builds on recursive RAG, where the system searches, looks at the results, and continues searching.

This chapter was mostly conceptual. We focused on how agentic search could work rather than building a complete agentic RAG system.

Key Concepts

AI Agents

An agent can be thought of as a loop:

1. Choose an action.
2. Use a tool.
3. Look at the result.
4. Decide what to do next.

The next action depends on what the agent has learned so far.

Agentic Search

Agentic search applies this idea to retrieval. Instead of using the same search process every time, the system can choose the search tool that best fits the query and the results it has already found.

Search Tools

The chapter discussed several possible search tools:

• Keyword search finds exact words or phrases.
• Semantic search finds related meanings or concepts.
• Regex search finds specific text patterns.
• Genre search filters results by genre.