VectorRAG.Net
An embedded, in-process vector database library for .NET that runs inside an application with no separate server, used for semantic search and retrieval-augmented generation.
VectorRAG.Net runs directly inside a .NET 8 application process rather than as an external service, avoiding HTTP calls, Docker containers and network latency. Its core search uses Random Hyperplane locality-sensitive hashing to generate candidates, with an exact rerank by dot product or cosine similarity.
It supports SIMD-optimized vector search, hybrid search combining vector similarity with BM25 keyword scoring, automatic document chunking, metadata filtering, and saving or loading a full database snapshot to a single file. It targets use cases such as RAG systems, internal knowledge-base search, offline desktop AI applications, and vector search on embedded or edge devices.
It installs as a NuGet package and requires a license key obtained from the vendor's site.