Weaviate v1.38 adds Boost: re-rank search results without dropping them

victorialslocum · x · 2026-09-15

Weaviate's v1.38 introduces Boost, letting you promote or demote results without removing them from the result set.

Use cases include: preferring articles from one category without excluding others, favoring recent documents while keeping authoritative older ones, factoring in popularity without letting one viral result dominate, and pushing drafts down without hiding them.

How it works: Boost runs after the primary vector, hybrid, or BM25 search. The primary search retrieves a candidate pool controlled by depth; Boost evaluates its conditions, normalizes and blends the primary and boost scores, then re-sorts candidates in memory. Supported conditions include filter match (a soft WHERE clause) and numeric property weighting (likes, downloads, popularity, with optional limits).

Original post →

More from coding & agent

coding & agent channel →