connect("./data") or connect("s3://…") moves to the hosted service by changing only
the connect target. See the Quickstart for the one-line change.
What you get
- Fully managed. No servers to provision, patch, or scale. Storage grows with your
data, and table maintenance (
optimizeandgc) runs for you. - Object-storage-native. Your data is stored as standard Parquet on object storage, the same open format the engine uses locally. No proprietary lock-in.
- The same retrieval surface. BM25, vector kNN, hybrid, and SQL over one copy of your data, indexed once. See Search.
- Usage-based. You pay for storage and query usage rather than for provisioned, always-on capacity.
When to choose it vs. self-hosting
Infino is open source, and you can always run the engine yourself against local disk or your own object-storage bucket. See Connect & storage for the self-hosted backends. Choose Infino Cloud when you want retrieval as a service: you would rather not operate storage, tune caches, or run maintenance, and you want to go from an API key to a working table in minutes. It suits teams that want to focus on their application rather than on running infrastructure. Self-host the open-source engine when you need the data and compute to stay entirely inside your own environment, you already operate object storage, or you want to embed the engine directly in your process. The code you write is the same either way, so you can start self-hosted and move to the hosted service later without rewriting your application.Next steps
Quickstart
Connect, provision a database, and run your first search.
Authentication
API keys, the Authorization header, and key rotation.
Architecture
How the service separates control, compute, and storage.
REST API
Not using an SDK? Drive Infino Cloud over plain HTTP — the OpenAPI reference.
