Vector search, production-ready
Store and search high-dimensional embeddings for semantic search, RAG, and recommendations — a managed, horizontally-scalable vector database with a clean API and one-command deploy.
Open source (Apache-2.0), built on Qdrant. Self-host anywhere or run managed on Hanzo Cloud.
Embeddings in, relevant results out
One index powers the three workloads modern AI apps depend on — search, retrieval, and recommendations — behind a single API and one key.
Semantic search
Search by meaning, not keywords — approximate nearest-neighbor over your embeddings with rich payload filtering and hybrid scoring.
RAG memory
The retrieval layer for grounded LLM apps: attach real source passages to every answer and keep hallucination in check.
Recommendations
“More like this,” dedup, and candidate generation at scale — similarity and personalization served from the same collection.
Everything you need to ship retrieval
HNSW indexing
Fast approximate nearest-neighbor with tunable recall vs. latency, scaling to billions of vectors per collection.
Metadata filtering
Rich payload filters — geo, ranges, keywords — applied inside the ANN search, not bolted on afterward.
Hybrid search
Combine dense vectors with sparse and keyword signals for precision on the queries pure vectors miss.
Quantization
Scalar and product quantization cut memory and cost dramatically without collapsing recall.
One key for embed + store
Pair with Hanzo’s OpenAI-compatible embeddings API — one API key to embed, upsert, and search.
Horizontal scale
Sharding and replication for high availability; snapshots for backup and point-in-time restore.
Ship semantic search this week
Spin up a managed collection on Hanzo Cloud, or self-host the open-source engine anywhere.