## Problem Vector search over-fetches, then filters by loading ALL project symbol IDs into HashSet. ## Location `src/index/brain.rs:198-208` ## Solution Store `project_id` in usearch index, filter during KNN search. ## Expected Impact Eliminate large DB query per search. Save 10-20ms for large projects.
Activity
ajianaz commentedon Jul 28, 2026
Meta-tracker. Sub-issue #415 is straightforward but low priority — project_id stored in vector cache already.