HyperVectorDB
HyperVectorDB is a high-performance local vector database built in C designed to store vectors and associated documents for efficient similarity search. It natively supports a wide range of distance and similarity measures including Cosine Similarity, Jaccard Dissimilarity, Euclidean Distance, Manhattan Distance, Chebyshev Distance, and Canberra Distance. The system features query and response caching specifically optimized for Cosine Similarity to accelerate repeated identical queries, with automatic cache invalidation upon data changes. To maximize performance, HyperVectorDB offers automatic parallelization that distributes data across multiple files and memory regions to leverage async I/O and multithreading. Storage efficiency is handled through LZ4 compression when persisting data to disk. The library includes flexible indexing capabilities allowing users to index individual text strings, preprocess entire files line-by-line with custom filters, or apply custom pre and post-processing logic to refine con