Deployment models¶
AnvaiOps runs the stateless control plane at the edge and the stateful engine in your cloud. You choose how much of it you own.
flowchart LR
subgraph EDGE["Edge — global, stateless"]
UI[Console + docs]
ACCT[Account control plane]
end
subgraph CLOUD["Your cloud + region"]
ENG[ProximaDB engine]
OBJ[(Object store)]
ENG --> OBJ
end
User --> UI
User --> ACCT
UI -. data ops, direct to your cloud .-> ENG
Only UI bytes cross the edge; every data operation goes straight to the engine in your cloud. Your data never egresses through our control plane.
Residency tiers¶
| Tier | Where the engine runs | Where storage lives | For |
|---|---|---|---|
| Managed | Our managed cloud account, your region | Managed | Fastest onboarding |
| BYO-storage | Our managed compute | Your object store (S3 / Blob / GCS) | Data-residency needs |
| BYOC | Your cloud account | Your account | Regulated / on-prem |
What runs where¶
- Stateless (console UI, account/billing, auth) → the edge, globally.
- Stateful (the ProximaDB engine, your ingested data, query execution) → your cloud, pinned to a region.
This split is deliberate: compute goes to the data, never the reverse — so query traffic never pays cross-cloud egress.
Self-hosting the open engine
The core ProximaDB engine is open source and can be run entirely on your own — see the ProximaDB project for the OSS deployment path. AnvaiOps adds the managed control plane, governance, and metering on top.