Vector Databases. A Practical Introduction Nitin Borwankar

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

• Preface
  • Whats in This Book
  • Who This Book Is For
  • How to Use This Book
    • Software, Environment, and Resource Requirements
  • Conventions Used in This Book
  • Using Code Examples
  • OReilly Online Learning
  • How to Contact Us
  • Acknowledgments
• 1. Introduction to Vector Databases
  • Why Do You Need Vector Databases?
    • A New Data Type: Vector
    • Similarity Search
      • Similarity search example
        • Keyword search (literal word matching)
        • Semantic search (meaning matching)
        • One-sentence summary
      • An additional query
      • Embedding
    • Whats Different About the Vector Type?
      • Similarity comparison
      • Nearest-neighbor search
      • Vector addition and subtraction
    • Where Do You Use Vector Databases?
  • SQL Versus Vector Databases
    • The Foundation of Business Math: Accounting Arithmetic
    • Vector Representation in a Relational Database Management System
    • The Need for Vector-Specific Capabilities
  • NoSQL Versus Vector Databases
    • NoSQL Databases and Vector Storage
    • Limitations of Vector Extensions in NoSQL Databases
    • When to Choose NoSQL with Vector Extensions
  • Hybrid Approaches: Combining Structured and Vector Data
    • The Need for Both Vector Data and Metadata
    • Limitations of Pure Vector Storage
    • Hybrid Database Architecture
    • Example of a Hybrid Query
    • Benefits of the Hybrid Approach
  • Conclusion
• 2. Embeddings
  • Understanding Vector Embeddings: Why We Need Them
  • Word2Vec: The Breakthrough That Changed Everything
  • Doc2Vec: From Words to Documents
  • From Embeddings to Modern Language Models: The Transformer Connection
    • Encoder-Only Transformers (BERT and Its Variants)
    • Decoder-Only Transformers (GPT Family)
    • Encoder-Decoder Transformers (T5, BART)
  • Embedding Models: The Specialized Vector Generators
    • Distinction from Traditional Models
    • Role in Modern LLM Applications
    • Practical Applications and Use Cases
    • Simple RAG Pipeline
    • The sentence-transformers Library: The Swiss Army Knife of Text Embeddings
      • Most popular models and their characteristics
      • Training approach
      • all-MiniLM-L6-v2 for RAG applications
      • Best practices when using sentence-transformers
    • Best Practices for Using SentenceTransformers: A Detailed Guide
      • Batch processing
      • Length handling
      • Normalization
      • GPU utilization
      • Memory management
      • Error handling
  • The Embedding Layer: The Gateway to Zero-Shot Learning
    • Anatomy of Transformer Embeddings
    • Connection to Zero-Shot Learning
    • Key Characteristics That Enable Zero-Shot Learning
    • Limitations and Considerations
      • Quality of embeddings
      • Computational costs
      • Bias and fairness
    • Latest Developments and Trends
  • Vector Arithmetic with Word2Vec: A Hands-On Guide
    • Step 1: Setup and Installation
    • Step 2: Load Pretrained Word2Vec Model
    • Step 3: Implement Vector Arithmetic Functions
    • Step 4: Classic KingQueen Analogy
    • Step 5: More Interesting Analogies
    • Step 6: Interactive Exploration Tool
    • Final Words on Vector Arithmetic
  • Conclusion
• 3. Similarity Search with FAISS
  • Foundations
    • Vector Representations
    • Distance Metrics
      • L2 (Euclidean) distance
      • Inner-product distance
      • Cosine similarity
    • Selection Heuristics
  • FAISS Indexes
    • Flat Indexes (Brute Force)
    • IVF-Based Indexes
    • LSH-Based Indexes
    • HNSW-Based Indexes
    • Other Specialized Indexes
    • Composite and Transformative Indexes
    • Choosing the Right Index
  • Quantization
    • SQ
      • Benefits of SQ in FAISS
      • Trade-offs of SQ in FAISS
    • PQ
      • The problem
      • Walk-through of PQ
      • Details of PQ in FAISS
      • Key advantages of PQ in FAISS
      • When is PQ a good choice?
  • The ANN Problem
    • The Problem
    • Avoid Computational Cost
    • Key ANN Techniques in FAISS
    • Choosing an Index in FAISS
    • Code Example
  • Understanding HNSW Indexes
    • What Is HNSW?
    • How HNSW Works
      • Building the index
      • Searching the index
    • Key Parameters Explained
      • M (graph degree parameter)
      • efConstruction (construction factor)
      • efSearch (search factor)
    • Practical Example: Building a Similarity Search System
    • Performance Characteristics
    • Best Practices
  • FAISS Architecture and Components
    • Foundation
    • Core Concepts
    • Key Components
    • Common Workflow
    • Illustrative Example
    • Key Takeaways
    • Further Exploration
  • Conclusion
• 4. Semantic Search with SQLite3
  • Understanding the SQLite Vector Similarity Search Extension
    • Core Capabilities
    • Architecture Overview
    • Limitations
  • Setting Up the Development Environment
    • Installing Dependencies
    • Verifying the Installation
    • Operational Pragmas
  • Designing the Database Schema
    • Schema Requirements
    • Table Definitions
    • Schema Design Decisions
  • Connecting to Reddit with the Python Reddit API Wrapper
    • Creating Reddit API Credentials
    • PRAW Client Implementation
    • Usage Example
  • Content Extraction and Preprocessing
    • Text Cleaning Pipeline
    • Quality Filtering
  • Generating and Storing Embeddings
    • Embedding Generator
    • Database Storage
    • Batch Processing Pipeline
  • Building the Vector Index
    • Understanding VSS Indexing
    • Index Management
  • Implementing Semantic Search
    • Search Result Container
    • Search Engine
  • Putting It All Together
    • Workflow Example
    • Example Output
  • Extension: Incremental Indexing
  • Conclusion
• 5. Building an ArXiv Paper Search System with PostgreSQL pgvector
  • The Challenge of Searching Scientific Literature
    • Why ArXiv Makes an Ideal Data Source
    • Real-World Use Cases
    • Technology Stack Rationale
  • Architecture Overview
    • System Components
    • Data Flow
    • Design Philosophy
  • Environment Setup and Dependencies
    • PostgreSQL and pgvector Installation
    • Python Environment Configuration
    • Directory Structure and Configuration
    • Verification and Testing
  • Database Design for Scientific Papers
    • Schema Design Principles
    • Core Tables Structure
    • Vector Storage Strategy
    • Indexing Strategy
  • ArXiv Integration and PDF Management
    • ArXiv API Client Implementation
    • PDF Download Pipeline
    • Batch Processing System
  • PDF Text Extraction and Processing
    • PDF Extraction Challenges
    • Intelligent Text Chunking
  • Embedding Generation and Storage
    • Embedding Model Strategy
    • Batch Processing Pipeline
  • Similarity Search Implementation
  • Interactive Application and UI
  • Docker Packaging for Local Deployment
    • Container Architecture
    • Docker Compose Configuration
    • Database Initialization Scripts
    • Development Workflow
    • Cloud-Ready Design
  • Basic Performance Tuning
    • Index Configuration
    • Query Performance
    • Resource Management
  • Next Steps
    • Current Limitations
    • Enhancement Ideas
  • What We Did
    • System Achievements
    • Technical Skills Gained
    • Practical Research Tool
    • Foundation for Advanced Systems
    • Future Potential
  • Conclusion
• 6. Building a Retrieval-Augmented Generation System with SQLite VSS and Ollama
  • System Architecture Overview
  • Database Foundation with Vector Support
    • Setting Up the Vector-Enabled Database
    • Schema Design for RAG
    • Creating Search Indexes
  • Text Processing and Embedding Generation
    • Embedding Model Management
    • Intelligent Text Chunking
    • Storing Content with Embeddings
  • Hybrid Search Implementation
    • Hybrid Search Algorithm
    • Semantic Search Component
    • Keyword Search Component
    • Score Fusion and Ranking
  • LLM Integration with Ollama
    • Ollama API Client
    • Health Check Function
  • The RAG Pipeline
    • Context Formatting
    • Question-Answering Pipeline
  • Demonstration and Testing
    • Sample Data Loading
    • Main Demonstration Function
    • Interactive Q&A Interface
    • Quick Testing Utility
  • Next Steps: Extending the System
    • Missing Reddit Data Features
    • Performance Optimizations
    • Production Considerations
    • Advanced RAG Patterns
  • Conclusion
• 7. Building a Scientific RAG System with PostgreSQL and pgvector
  • System Goals and Capabilities
  • Architecture Overview
  • Database Foundation with pgvector
    • Database Configuration and Setup
    • Schema Design for Scientific Papers
    • High-Performance Vector Indexes
  • Embedding Generation Strategy
  • ArXiv Integration and PDF Processing
    • Paper Discovery with ArXiv API
    • Intelligent PDF Text Extraction
  • Advanced Text Chunking
  • Storage Pipeline with Embeddings
  • Multilevel Semantic Search
    • Abstract-Level Search
    • Section-Level Search
  • The RAG Pipeline: Deep Dive
    • Local LLM Integration with Ollama
    • Health Check and Model Discovery
    • Intelligent Context Retrieval
    • Scientific Prompt Engineering
    • Complete RAG Execution Pipeline
  • Demonstration and Interactive Interface
    • Main Demonstration Flow
    • Search Demonstrations
    • RAG Demonstration
    • Interactive Search Interface
    • Entry Point with Mode Selection
    • Technical Note on HNSW
    • How to Evaluate Your Results
  • Next Steps: Extending the Scientific RAG System
  • Conclusion
• 8. Building a Complete Conversation Search and RAG System
  • System Goals and Capabilities
  • System Architecture Overview
  • What Well Build Together
  • Database Foundation for Conversation Storage
    • Designing the Conversation Schema
    • Three-Table Architecture for Optimal Performance
    • High-Performance Vector Indexing
  • Conversation Import and Data Processing Pipeline
    • Robust JSON Import with Error Handling
    • Atomic Transaction Processing
    • Timestamp Handling and Data Validation
    • Error Recovery and Logging
  • Efficient Embedding Generation and Batch Processing
    • Singleton Pattern for Model Management
    • Incremental Processing Strategy
    • Batch Processing for Optimal Performance
    • Database Insertion with Conflict Handling
  • Contextual Search with Conversational Understanding
    • Semantic Similarity Search
    • Multitable Joins for Rich Context
    • Result Formatting and Structure
    • Conversation Context Retrieval
    • Context Window Calculation
  • RAG Integration for Conversation History
    • Structured Context Management
    • Local LLM Integration with Ollama
    • Health Monitoring and Model Discovery
    • Context Retrieval and Assembly
    • Conversational Prompt Engineering
    • Complete RAG Pipeline with Performance Monitoring
  • Complete Web API with FastAPI
    • FastAPI Application Structure
    • Request Models with Validation
    • Search Endpoint Implementation
    • RAG Question-Answering Endpoint
    • System Statistics and Monitoring
    • Server Startup and Configuration
  • Demonstration and Sample Data
    • Realistic Sample Data Generation
    • Multitopic Sample Coverage
    • Sample Data Processing Pipeline
    • Comprehensive System Demonstration
    • Progressive Feature Demonstration
    • RAG Demonstration with Conditional Execution
    • Production Import Functionality
    • Application Entry Points
  • Conclusion: A Complete Personal Knowledge System
• 9. Vector Query Language
  • Core Concepts
    • Data Model
    • Basic Syntax Structure
  • Vector Operations
    • Similarity Search
      • Basic similarity search
      • Similarity search with metadata filtering
    • Hybrid Search
    • Range Search
    • Batch Operations
  • Vector Functions and Aggregations
    • Vector Functions
      • Basic vector operations
      • Vector arithmetic
    • Vector Aggregations
      • Compute centroids
      • Find representative vectors
        • Geometric median versus centroid
        • Robustness
• Index

pokaż cały spis treści

ukryj recenzje