Upstream contribution · liliang-cn/cortexdb
cortexdb · Check vector dimensions before CortexDB index training
Problem
CortexDB already enforced dimensionality when adding or searching vectors. Its IVF training path could still panic on a short vector or produce centroids of the wrong width from a longer one.
Engineering response
Validate every training vector's dimension before running k-means or mutating index state. The error identifies the out-of-shape input.
Evidence and scope
The pull request records a baseline panic and passing Go package tests with short, long and nil-vector coverage. It does not report a full repository race run.