- Optimizations for top-K queries with a WHERE filter (aka faceted search) yielding significant (up to 5X) speedups
- We now adjust the K value based on planner selectivity estimates
- This simple approach outperforms ParadeDB and implementation alternatives considered (e.g. using an index on facet fields) in benchmarks
- Chinese-language support via zkparser
- Robustness improvements for large corpuses
- Harden various tests that proved flaky on non-standard Postgres variants
- A host of stability improvements
Full Changelog: v1.3.1...v1.4.0