ailake-index 0.0.6

HNSW and IVF-PQ vector indexes with GPU acceleration for AI-Lake
Documentation

ailake-index — HNSW and IVF-PQ index lifecycle

Search backend priority (all detected at runtime, no build-time GPU SDK required):

  1. NVIDIA CUDA — cuBLAS SGEMM via libloading dlopen of libcudart + libcublas
  2. AMD ROCm — hipBLAS SGEMM via libloading dlopen of libamdhip64 + libhipblas
  3. CPU rayon — parallel brute-force, always available