Vector Databases
Learn to store, index, filter and operate vectors in real systems: pgvector with SQL, Qdrant and Chroma APIs, HNSW tuning, filtered and hybrid search, sharding, replication, capacity planning, security and choosing a database, with every example run for real.
Syllabus
What a Vector Database Is
- Why Vectors Need a Database
- The Landscape: Libraries, Extensions, Dedicated Engines, Managed Services
- Data Model: Collections, Records, Payloads and Namespaces
- What to Expect: Approximate, Eventually Consistent, Not a General Database
pgvector: Vectors Inside PostgreSQL
- Creating a Table With a vector Column
- Distance Operators and ORDER BY
- Filtering With WHERE: SQL and Vectors Together
- Transactions and Types: vector, halfvec
Indexing in Practice
- Without an Index: The Exact Sequential Scan
- Creating an HNSW Index and Reading the Plan
- Measuring and Tuning Recall (ef_search)
- Index Size, Build Time and IVFFlat
Filtered and Hybrid Search
- The Filtered-Search Problem
- Pre-Filtering, Partial Indexes and Partitions
- Hybrid Search: Full-Text Plus Vectors
- Filtering in a Dedicated Engine: Qdrant
Other Engines and How to Compare Them
- Chroma: A Developer-Friendly Embedded Store
- Comparing Engines: A Practical Checklist
- Benchmarking Honestly
Operating at Scale
- Bulk Ingestion and Batching
- Sharding: Splitting Data Across Machines
- Replication, Quorums and Read-Your-Writes
- Capacity Planning: Memory Is the Budget