Hanzo Vector
High-performance vector database for semantic search and RAG
Drop-in replacement for Pinecone, Weaviate, and ChromaDB. Powered by Qdrant.
Built for production AI workloads
Semantic search, RAG, and recommendations -- all backed by a battle-tested vector engine.
High-Performance Search
HNSW-based approximate nearest neighbor search delivers sub-10ms queries across billions of vectors with tunable accuracy-speed tradeoffs.
Flexible Filtering
Combine vector similarity with payload-based metadata filters in a single query. Filter by any field without sacrificing search speed.
Multiple Distance Metrics
Choose cosine similarity, dot product, or Euclidean distance per collection. Match the metric to your embedding model for optimal results.
Payload Storage
Attach arbitrary JSON metadata to every vector. Store, filter, and retrieve rich context alongside your embeddings.
Quantization
Scalar, product, and binary quantization reduce memory usage by up to 32x while maintaining search quality. Fit more vectors per node.
Horizontal Scaling
Shard collections across nodes with automatic replication. Scale reads and writes independently as your data grows.
How it works
Three steps from raw embeddings to production-grade semantic search.
Store your embeddings
Create a collection with your chosen distance metric and dimensions. Push vector embeddings from any provider -- OpenAI, Cohere, HuggingFace, or your own models. Attach JSON metadata as payload.
Query with precision
Search by vector similarity, filter by metadata, or combine both in a single request. Quantization keeps queries fast even at massive scale. Results return in under 10ms.
Power your AI stack
Use Hanzo Vector as the retrieval layer for RAG pipelines, recommendation engines, or semantic search. Integrates directly with Hanzo Search and Hanzo Crawl for end-to-end workflows.
Works with any embedding provider
Native support for OpenAI, Cohere, HuggingFace, and Sentence Transformers. REST API and gRPC for everything else.
Pricing
Pay for hosted vector storage and search. Self-host Qdrant free forever.
Build
- 1M vectors
- Up to 3072 dimensions
- 10 collections
- Cosine, dot, Euclidean
- REST API + gRPC
Scale
- 10M vectors
- Up to 3072 dimensions
- 100 collections
- Quantization + sharding
- Priority support
Enterprise
- Unlimited vectors
- Unlimited dimensions
- Unlimited collections
- Dedicated clusters
- Dedicated support + SLA
Qdrant is open source (Apache 2.0). Self-host free forever. Pay only for our hosted API, managed clusters, and enterprise features.
25% of compute goes back to open source
Every deployment is SBOM-verified. Contributors to Qdrant earn a share of compute revenue — transparent, on-chain, and customizable by the community.
Give your AI applications long-term memory
Store once, search by meaning.
Semantic search, RAG, and recommendations -- all from one vector database.