Vector Databases and Retrieval Engineering — HNSW, IVF-PQ, and the Filter Collapse¶
A vector database is the piece of the RAG stack every job description now names and almost no interview tests properly: teams ship one, hit a recall cliff or an out-of-memory crash six weeks after launch, and discover that "just use pgvector" was a schema decision, not a tuning decision. This chapter treats it as the systems problem it is — an approximate index is a deliberate trade of correctness for speed, made once at build time and paid for continuously in memory, and every parameter you touch moves that trade in a specific, computable direction. …
🔒 La suite est en accès freemium — lecture complète 100 % gratuite
Tu lis ici l'aperçu libre. Le reste du chapitre (code, schémas, maths, exercices) fait partie du livre complet : crée un compte gratuit (30 secondes, aucun paiement) pour tout lire.