Vector Databases Hayden Fisher Technologies

The book

Vector Databases

From First Principles to Production RAG Systems

609pages
37chapters
25labs
127909words
v1.0current release

Engines covered

One lens, five engines, one spatial database

Chroma Qdrant Weaviate Milvus pgvector PostGIS

Each engine gets the same treatment: architecture, storage, indexing, scaling, operations and honest weaknesses. Then a decision framework, and an abstraction layer that makes swapping one for another a configuration change rather than a rewrite.

Contents

Every chapter, in order

Chapter titles are public. The chapters themselves are the product.

— How to Use This Book {.unnumbered} 5 min
0 Setting Up the Laboratory 17 min
1 What Is Semantic Search? Foundations 17 min
2 What Is an Embedding? Foundations 35 min
3 Visualizing Embeddings Foundations 20 min
4 Distance Metrics Foundations 16 min
5 Chunking Strategies Foundations 16 min
6 Building an Embedding Pipeline Building a Vector Pipeline 14 min
7 Building a Local Search Engine Building a Vector Pipeline 14 min
8 Your First RAG Pipeline Building a Vector Pipeline 15 min
9 Metadata Filtering Building a Vector Pipeline 11 min
10 Hybrid Search Building a Vector Pipeline 12 min
11 Chroma Understanding Vector Databases 34 min
12 Qdrant Understanding Vector Databases 29 min
13 Weaviate Understanding Vector Databases 23 min
14 Milvus Understanding Vector Databases 19 min
15 PostgreSQL and pgvector Understanding Vector Databases 19 min
16 Choosing a Vector Database Understanding Vector Databases 18 min
17 ANN Algorithms Advanced Topics 20 min
18 Reranking and Multi-Stage Retrieval Advanced Topics 16 min
19 Evaluating RAG Systems Advanced Topics 16 min
20 Performance Engineering Advanced Topics 15 min
21 Vectors Meet Geometry Spatial AI 15 min
22 Geospatial RAG Spatial AI 37 min
23 Production Architecture Production 37 min
24 Capstone: The AI Knowledge Platform Production 36 min
F Glossary {#sec:appendix-f} Appendices 9 min
— Colophon {.unnumbered} 3 min

Try it

Ask the contents a question

The same list, searched the way the book teaches: your words become a vector, every chapter title becomes a vector, and the closest ones come back ranked by the angle between them.

A toy on purpose: hashed character trigrams and a cosine, a few dozen lines, no model. It matches letters, not meaning, so "maps and coordinates" misses Chapter 22 entirely. Chapter 2 explains what a real embedding adds.

    Type anything. It scores every chapter title against your query.

    Available now

    Get the Early Look

    The complete book, now.

    See pricing