Hands-On RAG for Production. Design, Develop, and Deploy Production-Ready RAG Applications Ofer Mendelevitch, Forrest Sheng Bao

Dostępność produktu

Produkt nie został jeszcze oceniony pod kątem ułatwień dostępu lub nie podano żadnych informacji o ułatwieniach dostępu lub są one niewystarczające. Prawdopodobnie Wydawca/Dostawca jeszcze nie umożliwił dokonania walidacji produktu lub nie przekazał odpowiednich informacji na temat jego dostępności.

Spis treści książki

• Foreword by Sharon Zhou
• Foreword by Jim Dowling
• Preface
  • What This Book Is About
  • Who This Book Is For
  • Who This Book Is Not For
  • Using Code Examples
  • Prerequisites
  • Navigating This Book
  • Conventions Used in This Book
  • OReilly Online Learning
  • How to Contact Us
  • Acknowledgments
• 1. Introduction to Retrieval-Augmented Generation (RAG)
  • How Does RAG Work?
  • The Blueprint of a RAG Stack
    • The Ingestion Flow
    • The Query Flow
  • Example: RAG with LangChain
  • RAG Versus Other Approaches
    • RAG Versus Chat with PDF
    • RAG Versus Fine-Tuning
      • Expertise gap
      • Cost of fine-tuning
      • Access controls and permissions
  • Key Benefits of RAG
    • RAG Is Scalable and Efficient
    • RAG Helps Reduce Hallucinations
    • RAG Enables Explainability
    • Near-Instant Addition and Removal of Knowledge
    • Access Controls and Security
  • RAG Use Cases
    • Virtual Assistants and AI Chatbots
    • Enterprise Knowledge Management and Internal Search
    • Automated Content Creation and Document Summarization
    • Generating Attractive and Effective Personalized Ads
    • Question-Answering Systems
    • Medical and Healthcare Applications
    • Legal and Compliance Research
  • Advanced RAG
    • Agentic RAG
    • Multimodal RAG
    • RAG with Knowledge Graphs
  • Conclusion
• 2. The Base RAG Stack
  • RAG Stack Flows
    • The Ingestion Flow
    • The Query Flow
  • Document Parsing
    • Extracting Text from Various File Formats
    • Document Parsing with VisionLanguage Models
    • Code Example: Parsing Files
      • Working with PDF files in PyMuPDF
      • Working with DOCX
      • Parsing PDF files using an LLM
  • Text Chunking
    • Chunking Strategies
    • Code Example: Chunking in Python
  • Embedding Models
    • What Is (an) Embedding?
    • Selection Criteria for Embedding Models
    • Practical Tips and Considerations
    • Code Example: Generating Embeddings with Sentence Transformers
  • Vector Databases and Vector Search
    • Understanding Vector-Based Similarity Search
    • Approximate Nearest Neighbor Algorithms
    • Vector Databases
    • Parameters to Consider When Using Vector Search
    • Code Example: Storing and Retrieving Vectors Using pgvector
  • Generative LLMs
    • LLMs
    • RAG Prompt Engineering
    • Evaluating LLMs and Prompt Templates
    • Code Example: Using Anthropic Claude to Generate Responses in RAG
  • Conclusion
• 3. Scaling Your RAG Stack
  • RAG at Scale
    • Volume and Complexity of Documents
    • Index Freshness
    • Cost Management and Optimization
  • Advanced Data Ingestion
    • Handling a Large Volume of Documents
    • Dealing with Inconsistent Data Quality
    • Handling Large Documents
    • Example: Splitting a Large PDF File
    • Managing Document Updates and Refresh
  • Advanced Retrieval
    • The Two-Stage Retrieval Pipeline
    • Hybrid Search
    • Reranking
      • Relevance reranking
      • Example: Using the bge-reranker-v2 model
      • Maximum Marginal Relevance Reranking
      • Custom reranking
  • Implementing Guardrails
    • Guardrails for AI Safety
      • Addressing bias and safety in RAG
      • Example: Using ShieldGemma to implement guardrails
    • Preventing Prompt Injection Attacks
      • Input sanitization
      • Instruction defense
  • Controlling Hallucinations in RAG
    • Defining Hallucinations in RAG
    • LLM Hallucinations Versus RAG Hallucinations
    • Hallucination Detection
      • LMM-as-a-judge
      • Hallucination evaluation model
      • Example: HHEM evaluation
    • Hallucination Correction
  • Building a Great RAG User Experience
    • Considerations for RAG User Experience
      • Capture of user input
      • Presentation of results
      • User control and feedback
    • Multimodal User Interfaces
    • Tools and Reference Implementations
      • Assistant-ui
      • Streamlit and Gradio
      • Vectara-answer
  • Conclusion
• 4. Deploying RAG to Production
  • Challenges with RAG in Production
    • Response Quality and Reduced Hallucinations
      • Reason 1: No relevant data
      • Reason 2: Weak retrieval pipeline
      • Reason 3: LLM hallucinations
      • Reason 4: Prompt engineering
    • High Latency
      • Parallelization and auto-scaling
      • Using alternative LLMs
      • Software or hardware acceleration
      • Efficient data indexing
      • Caching
    • Data Security and Privacy
      • Ingestion layer security
      • Data store safeguards
      • Preventing data leaks
      • LLM generation guardrails
    • Vendor Chaos and Integration Woes
    • Team and Expertise
    • Total Cost of Ownership
      • Direct costs
      • Indirect and ongoing costs
      • Additional total cost of ownership considerations
      • Cost monitoring
    • RAG Evaluation
  • A Reference Production Architecture
  • Successful Transition from Proof of Concept to Production
    • Summarize What You Learned in the Proof of Concept
    • Define Goals and Requirements
  • Ensuring Continued RAG Success
  • Conclusion
• 5. The RAG Platform
  • DIY Versus Platform RAG
    • Core RAG Capabilities
      • Embedding models
      • Vector databases
      • Advanced retrieval
      • Prompt engineering
      • Support for multiple LLMs
      • Hallucination detection and correction
    • Data Sources
    • RAG Sprawl and Centralized Governance
    • Cost and Upkeep
    • Deployment Options
      • DIY RAG deployment options
      • RAG platform deployment options
  • Example RAG Platform: Vectara
    • Getting Started
    • Ingesting Data into Vectara
      • File upload
      • Direct data ingestion
    • Running Queries
    • Hallucination Correction
    • Other RAG Admin API Endpoints
  • Conclusion
• 6. Evaluating Your RAG Application
  • How Does RAG Fail?
    • Retrieval Failures
    • Generation Failures
      • Faithfulness failure
      • Context utilization failure
      • Answer relevance failure
    • Failures Due to Inadequate Data Ingestion
      • Structural parsing error
      • Content staleness
    • Summary of RAG Failures
  • Using LLMs for Evaluation: LLM-as-a-Judge
    • What Is LLM-as-a-Judge?
      • The nuances of performance
      • Inherent biases
      • Operational hurdles: Cost and latency
      • The challenge of stochastic stability
    • How LLM-as-a-Judge Works
  • RAG Evaluation Metrics
    • Retrieval Metrics
      • Basic retrieval metrics: Precision, recall, and F1-score
      • Rank-aware metrics: Mean reciprocal rank, mean average precision, and normalized discounted cumulative gain
        • Mean reciprocal rank (MRR)
        • Mean average precision (MAP)
        • Normalized discounted cumulative gain (nDCG)
      • UMBRELA scores
    • Generation Metrics
      • Context utilization
      • Answer accuracy
      • Faithfulness
      • Citation accuracy
      • Response consistency
    • Bias and Safety
  • RAG Evaluation Offerings
    • Open RAG Eval
    • Retrieval-Augmented Generation Assessment
    • DeepEval
    • Amazon Bedrock
  • Human Feedback
  • Integrating RAG Evaluation in Production
    • Using LLM-as-a-Judge in Production
    • Offline RAG Evaluation
    • Online RAG Evaluation
  • System Metrics: Latency and Uptime
    • Latency and Throughput
    • Reliability and Uptime
    • Cost and Resource Efficiency
  • Conclusion
• 7. From RAG to AI Agents
  • What Is an AI Agent?
    • The Agentic Stack
    • Single-Agent Versus Multi-Agent Systems
      • Single-agent systems
      • Multi-agent systems
  • Agentic Use Cases
    • Agents in Customer Service
    • Agents in Financial Services
    • Agentic AI in Healthcare
    • AI Coding Agents
  • The Agentic Loop
    • Tool Calling
    • Model Context Protocol
    • Model Context Protocol Architecture
    • MCP in Enterprise Agentic AI
    • Agent-to-Agent Communication
  • Hands-On with Agentic AI Frameworks
    • AI Chatbots Using LangChain
    • Document Generation Agent with LlamaIndex
    • Building an Agent with Vectara
    • Building a Multi-Agent System with CrewAI
  • Agentic Memory
    • Short-Term Versus Long-Term Memory
    • Implementation Memory with Agentic RAG
    • Enterprise Guardrails: Privacy and Integrity
  • Evaluation and Observability with AI Agents
    • Agentic Observability
    • Tracing an Agent
    • Agentic Observability Metrics
    • Tools for Agentic Observability
  • Conclusion
• 8. Multimodal RAG
  • Documents with Embedded Tables
    • Why Are Embedded Tables Important?
    • Extracting Tables from Documents
    • Why Naive Chunking Fails for Tables
    • Processing Tables for RAG
    • Dealing with Multi-Page Tables
  • Documents with Embedded Images
    • The Image Summarization Approach
    • Multimodal Retrieval with a Shared Embedding Space
  • Audio and Video in RAG
    • The Baseline: High-Fidelity Transcription
    • Visual Semantics: The Red Button Problem
      • Fixed-interval frame extraction
      • Temporal segment extraction
      • Keyframe extraction
      • Summary of video extraction techniques
  • Production Considerations
    • Computational Economics and Latency
    • Modality Alignment
    • The Interface Layer: Visual Citations
    • Security, Privacy, and Governance
    • Deep Observability, Tracing, and Security at Scale
  • Hallucinations and Evaluation in Multimodal RAG
    • Detecting Multimodal Hallucinations
    • Evaluating Multimodal Retrieval and Generation
  • Conclusion
• 9. Knowledge-Enhanced RAG
  • Knowledge Graphs: An Overview
    • How Do You Search a Knowledge Graph?
      • Cypher
      • SPARQL
    • Ontologies Versus Schemas
  • Using Knowledge Graphs in RAG
    • Building a Knowledge Graph for Movies
    • Using the Knowledge Graph at Query Time
      • Chunk enrichment
      • Hybrid-graph retrieval
    • Choosing Between Enrichment and Hybrid Retrieval
  • Building Knowledge Graphs
    • Automating Knowledge Graph Construction
    • Leveraging Standard Ontologies and Knowledge Graphs
    • GraphRAG
    • The Graph Database Infrastructure
    • Graph Update Patterns and Evolution
    • The Accuracy/Cost Trade-off
  • Conclusion
• 10. The Future of RAG
  • The Evolution of Retrieval
  • The Shift to Agentic RAG
  • The Reality of Data Gravity and Federated Retrieval
  • The Impact of Longer Context
    • From Prompt Engineering to Context Engineering
    • The Shift from Reactive to Proactive RAG
  • RAG at the Edge: Small Language Models
  • Governance and Compliance at Scale
  • Conclusion: The Living Knowledge Base
• Index

pokaż cały spis treści

ukryj recenzje