Qdrant

Qdrant is a high-performance, open-source vector search engine written in Rust, designed for production workloads that demand fast filtering, efficient memory usage, and reliable performance under load. It stands out from other vector databases with its advanced payload filtering that executes during the search itself (not as a post-filter), ensuring accurate results even with complex filter conditions.

• Advanced filtering — filter by metadata fields (geo, range, keyword, nested) during vector search, not after, ensuring you always get the requested number of results
• Quantization — reduce memory usage by 4-32x with scalar, product, and binary quantization while maintaining search quality
• Sparse vectors — store and search sparse vectors (BM25, SPLADE) alongside dense vectors for hybrid retrieval
• Multi-vector support — store multiple named vectors per point (e.g., title embedding + body embedding) with independent search
• Payload indexing — create indexes on payload fields for fast filtered search across millions of points
• Snapshot & replication — built-in snapshotting, replication, and sharding for production reliability

Backend engineers and ML teams building production RAG systems choose Qdrant when they need fast filtered search at scale — particularly in multi-tenant SaaS applications where every query must be scoped to a specific user, organization, or permission level. Qdrant's Rust-based engine delivers consistent low-latency performance.

Qdrant's combination of open-source self-hosting, a managed cloud, and advanced features like multi-vector search and built-in quantization make it a strong choice for teams that need production-grade vector search with full control over their infrastructure.

1. Run locally with Docker:

   docker pull qdrant/qdrant
   docker run -p 6333:6333 -p 6334:6334 qdrant/qdrant

2. Or use Qdrant Cloud — create a managed cluster at cloud.qdrant.io (free tier: 1GB cluster).
3. Install the Python client:

   pip install qdrant-client

4. Create a collection and upsert vectors:

   from qdrant_client import QdrantClient
   from qdrant_client.models import Distance, VectorParams, PointStruct
   
   client = QdrantClient('localhost', port=6333)
   
   client.create_collection(
       collection_name='documents',
       vectors_config=VectorParams(size=1536, distance=Distance.COSINE)
   )
   
   client.upsert(
       collection_name='documents',
       points=[
           PointStruct(id=1, vector=[0.1, 0.2, ...], payload={'source': 'docs', 'category': 'ml'}),
           PointStruct(id=2, vector=[0.3, 0.4, ...], payload={'source': 'faq', 'category': 'api'})
       ]
   )

5. Search with filtering:

   from qdrant_client.models import Filter, FieldCondition, MatchValue
   
   results = client.query_points(
       collection_name='documents',
       query=[0.1, 0.2, ...],
       query_filter=Filter(
           must=[FieldCondition(key='category', match=MatchValue(value='ml'))]
       ),
       limit=5
   )

Pricing: Open-source: free, self-hosted. Qdrant Cloud Free: 1GB cluster at no cost. Cloud: from ~$25/month for small clusters. Enterprise: dedicated clusters with SLAs and premium support. Full pricing details.