PostgreSQL in Operation — Ingestion, Indexing, and the Transactions That Keep Them Honest¶
Every retrieval system you put in front of an LLM is, underneath the vector search, an ordinary database being asked to do something unusual: absorb a constantly changing pile of heterogeneous documents — PDFs, wiki pages, spreadsheets, code — and keep a queryable index of them correct, current, and fast, while a service is reading from it thousands of times a second. Chapter 19 told you what a chunk, an embedding, and a vector index are; this chapter is about the machinery that gets a messy source document into that index exactly once, keeps it there correctly when the source changes or disappears, and survives the concurrency of a real service without corrupting a row or falling over under connection pressure. …
🔒 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.