vector-index-bench/vibe

Supports quantized 8-bit and binary vector representations alongside full-precision indexing, HPC integration via Slurm and NUMA awareness, and GPU-accelerated algorithms. Built on containerized algorithm implementations (Apptainer/Singularity) with a modular architecture enabling straightforward addition of new vector search methods through standardized Python wrappers and hyperparameter configuration files. Evaluates both in-distribution and out-of-distribution embedding scenarios across diverse modalities (text, image, code) with automated result visualization and radar chart comparisons.