Qdrant

Qdrant is a specialized vector similarity search engine and vector database written in Rust. Designed specifically for production retrieval-augmented generation (RAG) and recommendation systems, it provides sub-millisecond nearest neighbor search over millions of high-dimensional embeddings.

Unlike traditional databases with bolted-on vector indexes, Qdrant integrates payload filtering directly into the HNSW graph traversal, preventing search recall degradation when querying with boolean constraints. It supports scalar, product, and binary quantization to dramatically reduce RAM consumption, plus full support for sparse-dense hybrid retrieval (BM25 + embeddings). It runs self-hosted via Docker or on managed Qdrant Cloud with multi-region high availability.