Open Source
Other Vector DatabaseRustEmbeddingsRAGgRPCProduction
Qdrant
Production-grade vector database with filtering, scalar quantisation, and a REST + gRPC API. Available as a Docker image, managed cloud, or embedded.
Install
$docker pull qdrant/qdrant && docker run -p 6333:6333 qdrant/qdrant About
Qdrant is built in Rust for performance. Its HNSW index supports filtered search — you can combine approximate nearest neighbour search with exact metadata filters without pre-filtering or post-filtering penalty.
For production use, Qdrant supports on-disk payloads and indexes (datasets larger than RAM), scalar quantisation for memory efficiency, and horizontal sharding via its distributed mode.
Getting Started
docker run -p 6333:6333 qdrant/qdrant
pip install qdrant-client
from qdrant_client import QdrantClient
from qdrant_client.models import Distance, VectorParams, PointStruct
client = QdrantClient(url="http://localhost:6333")
client.create_collection("my_docs", vectors_config=VectorParams(size=1536, distance=Distance.COSINE))
client.upsert("my_docs", points=[PointStruct(id=1, vector=[0.1]*1536, payload={"text": "Hello"})])
results = client.search("my_docs", query_vector=[0.1]*1536, limit=5)
print(results)Features
Scalar and product quantisation — 4× memory reduction with minimal recall loss
Filtered HNSW — combine vector search with payload metadata filters
Named vectors — store multiple embeddings per point (multi-modal)
Sparse vectors — support for BM25/SPLADE for hybrid search
On-disk indexing — index datasets larger than RAM
Snapshots — full collection backup and restore
REST and gRPC APIs
Python, JavaScript, Go, Rust, and Java client libraries
Qdrant Cloud — managed hosted service
References & Links
Project Info
Language Rust
License Apache-2.0
Stars 22.0k ★
Forks 1.6k
Category Other
Topics
Vector DatabaseRustEmbeddingsRAGgRPCProduction
Ready to try Qdrant?
Free and open source. Install in one command, works with Claude Code, Cursor, Windsurf, and more.
$docker pull qdrant/qdrant && docker run -p 6333:6333 qdrant/qdrant