Hybrid by default
Combine vectors, full-text ranking, filters, and boosts in one query.
36 REGIONS · 99.99% UPTIME
Vectors at the edge,
without the ops.
OrbitDB puts hybrid search next to your users. One API, automatic replication, predictable pricing.
Create a database See how it works
application ──write──▶ nearest region ──replicate──▶ global object store
│ │ │
└────────query──────────┴──── vector + text index ─────┘Built for production retrieval.
Scale to zero
Cold namespaces cost almost nothing and wake when the first query arrives.
Observable
Per-query traces explain latency, candidates, filters, and final ranking.
Production numbers
18ms
global p90 query latency
42B
documents indexed
12×
lower idle cost
Predictable pricing
Every search feature, on every plan. Compare the plans below.
sandbox
For prototypes and shared workspaces.
- Shared regional cluster
- Encrypted storage and scoped API credentials
- Developer forum and asynchronous email help
studio
For teams needing operational controls.
- Shared regional cluster with priority query capacity
- Workspace sign-in policies, change history, and network access rules
- Optional event export: $48/month
- Private engineering channel with weekday responses
dedicated
For reserved infrastructure and closer support.
- Isolated regional cluster with reserved query capacity
- Your cloud account, if preferred
- Event export, private service endpoints, and customer-controlled keys
- Named technical partner, always-on responses, and 99.95% availability target
FAQ
Can I import an existing index?
Yes. Stream JSONL, Parquet, or use the migration endpoint for live cutovers.
Do you support sparse vectors?
Dense, sparse, and multi-vector fields can be combined in the same ranking expression.
Where is data stored?
Choose a primary region and optional replicas. Data never leaves those regions.