VectorRAG.Net

VectorRAG.Net is a .NET-native embedded vector database library engineered for high-speed semantic search and Retrieval-Augmented Generation (RAG). It runs in-process, eliminating network latency and external dependencies, making it ideal for real-time retrieval over millions of embeddings. The library uses Random Hyperplane LSH for fast candidate generation, followed by exact dot-product or cosine similarity reranking for accuracy. The library targets professional .NET developers—quantitative developers, AI specialists, and game programmers—who need low-latency, deterministic data processing. It leverages ArrayPool integration and SIMD acceleration where possible, minimizing garbage collection pressure and ensuring stable latency. VectorRAG.Net supports built-in document chunking with configurable strategies, metadata filtering, optional hybrid search combining vector and BM25 text retrieval, file-based persistence, runtime metrics for monitoring, and batch APIs. Unlike external vector databases (e.g., Pinecone, Qdrant), VectorRAG.Net operates embedded within your application. It integrates via NuGet, requires no separate service, and is optimized for high-throughput, in-process retrieval. The library is part of Principium's suite of performance-critical components, which also include QuantCore.Net for quantitative finance and GameAI.Net for game AI. It is designed for scenarios where every millisecond counts and architectural control is paramount. VectorRAG.Net comes with commercial licensing that allows free evaluation for development and prototyping. Pricing is available on request, making it suitable for enterprise teams that need predictable performance and direct vendor support.