Search: when the database stops being enough

Try simple first

For lists of thousands of records, a query with the right index is entirely enough. Add a dedicated search engine when there's a real need for relevance ranking, autocomplete, or filtering across dozens of fields at once.

What sets quality

Language processing: tokenization, handling full and defective spelling, abbreviations and synonyms. In Hebrew that's the main gap between search that feels smart and search that's frustrating.

Ranking: not just word matching but freshness, popularity and the user's context. Start simple and improve by what people actually search for.

Sync

The index is a copy. Define how it updates — by event or by periodic scan — and how to rebuild it from scratch. An index you can't rebuild is a time bomb.