Back to glossary

Term

Qdrant

Qdrant is an open-source vector database written in Rust — it stores embeddings for semantic search, RAG and recommendation systems and runs locally, on your own cluster or as a managed cloud.

Qdrant — explained in more detail

Qdrant launched as an open-source project in 2021 and is today among the most widely used self-hostable vector DBs. The engine is written in Rust and uses HNSW as its index structure — fast even at millions of vectors. Records (points) consist of a vector and arbitrary JSON payload; payload filters are a core feature and are evaluated directly inside the index rather than as a post-step. Deployment modes: single-node container, distributed cluster, embedded mode inside an application, or Qdrant Cloud (managed).

Example / Practical context

A RAG pipeline with Qdrant: use the Python client to write embeddings via client.upsert(collection_name="docs", points=[...]), then client.search(collection_name="docs", query_vector=q, query_filter=...) to query. Filters like {"must": [{"key": "tenant", "match": {"value": "acme"}}]} compose naturally — important for multi-tenant scenarios. Snapshot and backup APIs simplify migrations.

Pinecone offers comparable functionality but is cloud-only. Weaviate leans more on schema modelling with classes and properties; Milvus targets very large datasets with explicit GPU index support. Qdrant’s sweet spot: solid self-hosting with strong filter performance and manageable operational overhead. pgvector remains the right choice when vectors should live next to relational data in a Postgres database.

Entdecke mehr