Hanzo Vector

Hanzo Vector is a managed vector database for embeddings and semantic search. Provision a collection through the control plane, then upsert and query high-dimensional vectors with HNSW indexing, metadata filtering, and hybrid (dense + sparse) search.

Provision a Collection

A collection is a logical resource in the shared Hanzo Vector backend, scoped to your organization. Create one with a single call, specifying the embedding dimension and distance metric.

curl -X POST https://api.hanzo.ai/v1/vector \
  -H "Authorization: Bearer sk-..." \
  -H "X-Org-Id: org_a1b2c3" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "documents",
    "dimensions": 1536,
    "distance": "cosine"
  }'

dimensions must match your embedding model — 1536 for text-embedding-3-small, 1024 for many open models. Supported distances: cosine, euclidean, dot.

Upsert and Search

A point is a vector plus a JSON payload; a query returns nearest neighbours, optionally filtered on that payload.

Generate the vectors with Embeddings, then store them here. For full-text and hybrid keyword ranking, pair with Search.

Features

• HNSW approximate nearest-neighbor index with cosine, euclidean, and dot-product metrics
• Payload storage and filtering — combine vector similarity with structured metadata, ranges, and geo
• Sparse and multi-vector support for hybrid keyword + semantic ranking
• Scalar quantization to cut the memory footprint of large collections
• Snapshots for backup and migration

The backend is API-compatible with Qdrant, so any Qdrant client works against your collection's endpoint.

Related

• Hanzo Vector service reference — full collection, points, and snapshot API
• Embeddings — generate the vectors you store here
• Search — full-text and hybrid indexes
• API Keys — authenticate with sk- keys
• API Reference — every endpoint at api.hanzo.ai