Qdrant
An open-source vector database for AI retrieval. Run it locally in Docker or on your servers, with hybrid search, metadata filters and GPU indexing.
Qdrant is a self-hosted vector database for developers building semantic search, retrieval-augmented generation (RAG), recommendations and AI agent memory. It stores embeddings alongside JSON metadata, so applications can find similar content while restricting results by attributes such as location, text or numeric ranges. The Rust engine is open source under Apache 2.0 and runs locally in Docker or on your own servers.
Hybrid search combines semantic similarity with keyword matching in one query. Qdrant supports dense and sparse vectors, including BM25, SPLADE++ and miniCOIL, plus multiple embeddings per object for models such as ColBERT. Metadata filters apply during search. Relevance controls let developers boost scores with business rules or use Maximum Marginal Relevance to reduce repetitive results.
For larger datasets, quantization and on-disk storage reduce RAM use, while sharding and replication spread data across servers. New vectors become searchable without a full index rebuild. Qdrant uses CPU acceleration and supports NVIDIA and AMD GPUs for indexing. REST and gRPC APIs connect it to applications, and a built-in web interface lets developers inspect collections and test queries.
Self-hosted deployments keep stored vectors and metadata on your infrastructure. Qdrant Cloud is a separate managed service on AWS, GCP or Azure; Cloud Inference generates text and image embeddings there. Hybrid Cloud runs the data side on your Kubernetes infrastructure, while Private Cloud supports air-gapped deployments.
