book

RAG-Driven Generative AI - Second Edition

Name: RAG-Driven Generative AI - Second Edition
Author: Denis Rothman
ISBN: 9781807424954

by Denis Rothman

April 2026

Intermediate

430 pages

9h 49m

English

Packt Publishing

Read now

Unlock full access

Preface
Free benefits with your book
Chapter 1: Why Retrieval-Augmented Generation?
What is RAG?Naïve, advanced, and modular RAG configurationsRAG versus fine-tuningThe RAG ecosystemRetriever (D)Generator (G)Evaluator (E)Trainer (T)Implementing naïve, advanced, and modular RAGThe environmentThe GeneratorThe dataThe queryNaïve RAGKeyword search and matchingAugmented inputGenerationAdvanced RAGVector searchAugmented input and generationIndex-based searchModular RAGBuilding the retrieval componentModular RAG strategies in actionSummaryQuestionsReferencesOther resources
Chapter 2: RAG Embeddings in Oracle Vector Stores
The architecture of symbiosisThe agnostic advantageThe converged ecosystemGetting started with Oracle 23aiPhase 1: DBA Oracle managementInstallation and setupGlobal configuration flagsOracle environment setup and wallet extractionInstall Oracle database driverEstablish secure database connectionCreating the tablesData maintenancePhase 2: Data ingestion, OracleInstallation and setupEstablish secure database connectionData preparation: The context library (procedural RAG)Initialize staging directoryData acquisition: Secure remote ingestionThe context library (the instructions)Updating the data loading and processing logicHelper functions for chunking and embeddingRobust text chunkingBatch embedding generationProcessing and uploading the data to OracleProcessing and uploading the context library to OracleProcessing and uploading knowledge base to OracleFinal verificationSemantic search testFinal table verificationPhase 3: LLM-augmented queryDefining the Oracle vector search functionHandling Oracle LOB dataRobust text extraction from database rowsOrchestrating the RAG pipelineStep 1: The retrieval timer and dual searchStep 2: Prompt assembly and generationStep 3: Reporting and outputSummaryQuestionsReferencesFurther reading
Chapter 3: Building a Live Recruiter Agent
Architecture of the Recruiter agentPhase 1: DBA Oracle management (the infrastructure)Global configuration flagsSecure database connectionInitialize the HR recruitment schemaVerify the converged schemaData maintenance and hygieneTruncate tables (non-destructive reset)Validate data removalDestructive maintenance: Drop schemaPhase 2: Data ingestion (the ETL pipeline)Environment and connection setupData acquisition and stagingVectorization utilitiesDual-stream data ingestionVector enrichmentHybrid query verificationPhase 3: LLM-augmented hybrid queryEnvironment and connection setupDynamic context retrieval: hiring personasHybrid search engineRecruiter agentLive simulation: Recruiter agent in actionSummaryQuestionsReferencesFurther reading
Chapter 4: Building Sovereign Enterprise Agents
System architecture and framework designWhat's new in this chapter and what isn'tPhase 1: Infrastructure layer (oracle_lib.py)Imports and logging configurationSingleton class structureInitialization logicContext managerUtilities and cleanupPhase 2: Unstructured data agent (agent_archivist.py)Imports and agent signaturesInput parsing and vectorizationExecution of the vector searchFormatting the output with MCPPhase 3: Structured data agent (agent_recruiter.py)Imports and agent signaturesParsing hybrid constraintsExecution of the hybrid queryFormatting the structured outputPhase 4: Verification suite (Unified_Agents_Unit_Test.ipynb)Setup and initializationImporting the modular agentsTest case A: Verifying the ArchivistTest case B: Verifying the RecruiterCleanupSummaryQuestionsReferencesFurther reading
Chapter 5: Building a Universal Context Engine
Architecture of the Universal Context EngineWhat's new in this chapterStep 1: Update registry.pyUpdating the global toolkit registryArchitectural prompt engineering for the PlannerStep 2: Configure the hybrid runtimeInstalling universal dependenciesInitializing cloud clientsEstablishing the enterprise connectionStep 3: Deploy convergence scenariosThe execution wrapperScenario A: Searching for experienced Python developersScenario B: Multi-step reasoning with structured constraintsSystem cleanup and finalizationSummaryQuestionsReferencesFurther reading
Chapter 6: Operationalizing the Universal Context Engine
Architecture of the operational layersWhat's new in this chapterPart 1: Assembling the Universal Context EngineStep 1: Initialize session stateStep 2: Build chat layoutImporting widgets and defining stylesCreating the interface componentsAssembling the application containerStep 3: Deploy the interactive loopEvent handler logicBinding the interactionDefining main application loopTroubleshooting during the development phaseHardening the RecruiterImproving UI responsivenessWhen a record isn't foundPart II: Operationalizing the Universal Context EngineBuilding the interactive control deckExecution and configuration in the engine roomAnalyzing the traceStep 1: RecruiterStep 2: WriterTroubleshooting during the deployment phaseUnstable network connectionsNon-compliant promptsSummaryQuestionsReferencesFurther reading
Chapter 7: Empowering AI Models by Fine-Tuning RAG Data
Fine-tuning architecture for static RAG dataFine-tuning within the RAG ecosystemInstalling the environmentPreparing the dataset for fine-tuningDownloading and visualizing the datasetPreparing the dataset for fine-tuningFine-tuning the modelMonitoring fine-tuning jobsThe success emailThe failure emailUsing the fine-tuned OpenAI modelAnalyzing fine-tuned jobs and metricsSummaryQuestionsReferencesFurther reading
Chapter 8: Boosting RAG Performance with Human Feedback
Architecture of adaptive RAGBuilding hybrid-adaptive RAG in PythonRetrieverInstalling the environmentRetrieving data from user queriesGeneratorIntegrating human feedbackInputSimulating the ranking mechanismInstalling the generative AI environmentGenerating contentEvaluatorResponse timeCosine similarity scoreHuman user ratingHuman-expert evaluationA hybrid-adaptive RAG guideStep 1: Ask the questionStep 2: Choose your modeStep 3: Grade the resultStep 4: The "Fix It" Loop (expert feedback)Summary checklistSummaryQuestionsReferencesFurther reading
Chapter 9: Building a Conversational RAG Agent
Architecture of multimodal RAGControl plane (infrastructure)Semantic plane (metadata)Visual plane (assets)System flow and user interactionUniversal Oracle DBA consoleSchema registryDefining the scopeSchema operations engineSchema status verificationInteractive DBA consoleVideo dataset visualization consoleEnvironment setup and token retrievalVideo retrievalVideo display utilitiesProcessing the introduction videoBatch processing the video datasetImplementing the Video Store agentStep 1: Crawler (ingesting metadata and URLs)Breaking down the provisioning logicInitializing the ingestion pipelineStep 2: Video Store agent (class definition and instantiation)SummaryQuestionsReferencesFurther reading

Chapter 10: Building an Agent with Spatial-RAG and GraphRAG
The converged architectureVector search in Oracle 23AISQL property graphsPython application layerExtending the universal Oracle DBA console for spatial and graphUpdating the schema registryUpdating the table registryUpdating the interactive console UICreating the spatial and graph schemaVerifying the schema stateBuilding the hyper-contextual Recruiter agentEnvironment setupThe spatial dimension: GeolocationThe social dimension: SQL property graphsThe hyper-query (combining vector, spatial, and graph search)Step 1: Defining the constraintsStep 2: Generating the embeddingStep 3: Executing the converged queryStep 4: Running the cursorStep 5: Displaying the resultsGenerating the Recruiter recommendationSummaryQuestionsReferencesFurther reading
Chapter 11: Scaling AI Workloads with Oracle Exadata
Scalable AI architecture with Oracle Cloud and ExadataThe execution substrate: Oracle Cloud InfrastructureThe database engine: Oracle ExadataChapter journey from prototype to productionExtending the universal Oracle DBA console for AI scalingUpdating the schema registryUpdating the console table registryUpdating the console UICreating the benchmark schemaVerifying the schema stateScaling the workflowEnvironment setupIdempotent data ingestionEstablishing the performance baselineScaling the infrastructureResults and conclusionCPU load with random and meaningful dataSummaryQuestionsReferencesFurther reading
Chapter 12: The Autonomous Database Architect
Evolution of data interactionPhase 1: Era of the SQL priesthood (1980s – 2010s)Phase 2: Era of application abstraction (2010s – 2023)Phase 3: Era of probabilistic retrieval (2023 – 2025)Phase 4: Era of the autonomous architect (2026 and beyond)Architecture of the Autonomous Database Architect systemArchitect agent (the designer)ETL Worker agent (the builder)Governance layer (the safety net)Implementing the Autonomous Database ArchitectAutonomous ArchitectETL WorkerScenario A: IT operations and incident log simulationPersona constraint (who)Format constraint (how)Data typing heuristics (reasoning)Safety constraints (what not to do)Scenario B: Supply chain crisisUnstructured input (the chaos)Perception and planningExecution (the act)Business resultThree laws of autonomous safetyLaw 1: Schema isolation (the sandbox)Law 2: Forensic auditability (the "why" behind the "what")Law 3: Lifecycle reversibility (the clean slate)Day-2 operations: Self-healing schemasEvolution loop (catch, analyze, evolve)Limit of evolution (the immutable core)SummaryQuestionsReferencesFurther reading
Appendix A: Answers
Chapter 1, Why Retrieval-Augmented Generation?Chapter 2, RAG Embeddings in Oracle Vector StoresChapter 3, Building a Live Recruiter AgentChapter 4, Building Sovereign Enterprise AgentsChapter 5, Building a Universal Context EngineChapter 6, Operationalizing the Universal Context EngineChapter 7, Empowering AI Models by Fine-Tuning RAG DataChapter 8, Boosting RAG Performance with Human FeedbackChapter 9, Building a Conversational RAG AgentChapter 10, Building an Agent with Spatial-RAG and GraphRAGChapter 11, Scaling AI Workloads with Oracle ExadataChapter 12, The Autonomous Database Architect
Appendix B: Unlock Your Exclusive Benefits
Unlock this Book's Free Benefits in 3 Easy Steps
Index

Content preview from RAG-Driven Generative AI - Second Edition

Appendix A

Answers

Use this appendix to cross-check your responses after attempting the questions included at the end of each chapter. Where possible, revisit the relevant chapter sections to understand the reasoning behind each answer rather than treating this as a simple correctness check.

Chapter 1, Why Retrieval-Augmented Generation?

No. RAG is designed to augment Generative AI models by providing them with external, up-to-date knowledge they were not trained on.
Yes. Traditional architecture often involves ETL pipelines that move data to specialized external vector databases, creating potential security risks.
No. The revolutionary shift is bringing AI to the data, where vector search and retrieval happen inside the secure database environment. ...

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

Generative AI with LangChain - Second Edition

Publisher Resources

ISBN: 9781807424954

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design