RAG-Driven Generative AI, Second Edition | RAG-Driven Generative AI

Retrieval, Reimagined for Real Systems 🧠

Retrieval-augmented generation has moved past the demo stage. The interesting problems now live in architecture: how to coordinate multiple retrievers, how to ground answers in graphs and video, and how to keep the whole pipeline fast enough for production. Denis Rothman’s RAG-Driven Generative AI, Second Edition takes on that shift. The book’s subtitle names the territory directly: MAS-RAG, DualRAG, GraphRAG, multimodal video pipelines, and Oracle Database 23ai.

This is a working developer’s book. It does not stop at explaining what RAG is. It shows how different retrieval strategies fit together, where they break, and what it takes to build systems that can be inspected, evaluated, and improved.

What the Second Edition Brings

First published in September 2024, this edition arrives in April 2026 with a sharper focus on multi-agent and graph-aware retrieval, multimodal pipelines, and Oracle Database 23ai. If you already understand basic vector search, the material here pushes into the design decisions that separate a prototype from a maintainable AI service.

Inside the Chapters ⚙️

The opening chapters build a foundation. Chapter 1 asks why RAG matters and compares naive, advanced, and modular configurations. It breaks the pipeline into retriever, generator, evaluator, and trainer roles, then shows how those pieces behave in practice.

Chapter 2 moves into embeddings and Oracle vector stores. It walks through Oracle 23ai setup, DBA-side management, secure connections, data ingestion, chunking, batch embedding, and semantic search. Chapter 3 applies the stack to a live recruiter agent, covering schema creation, ETL, vectorization, and the orchestration logic behind an LLM-augmented query.

From Vector Search to Multi-Agent Retrieval

The book treats retrieval as an engineering discipline, not a single query. MAS-RAG and DualRAG suggest multiple agents or retrieval paths working together. GraphRAG adds structure and relationships that flat vector search can miss. Multimodal video pipelines extend the same thinking to non-text data. Oracle Database 23ai serves as the persistent, enterprise-ready backing store.

Who This Book Is For

AI engineers, full-stack developers, ML practitioners, and technical architects will get the most from this book. It assumes you are comfortable reading code and working with modern AI tooling. Beginners may want a gentler introduction to embeddings and LLMs first. The reward for experienced readers is a concrete path from RAG concepts to production-oriented architecture.

Why It Belongs on Your Shelf

RAG is evolving quickly, and the gap between simple retrieval and reliable AI systems is where real projects succeed or fail. This second edition captures that gap. It is a practical, code-informed guide for anyone building retrieval systems that need to handle more than text, more than one agent, and more than a single database. Add it to your Digital Delights library and keep the source close while you build.