Skip to content
Verified

How data is stored

Every project, and your own account workspace, keeps its records in a database that belongs to it alone.

Xuda does not pool every customer's records into one shared database. Every project you build, and the workspace your account lands on when you sign in, is its own database: created for that project alone, readable only through credentials scoped to it, and never merged with anyone else's.

One project, one database

The moment you create a project, whether from New project in the dashboard or by an AI-built app spinning one up for you, the platform provisions a database for it and nothing else. That database gets its own admin and standard access credentials, generated for this project alone, and the database is locked down so only those credentials can read or write it. No other project, and no other customer's project, holds a key that opens yours.

Every table you build in Xuda Studio lives inside that one database, as does the underlying record for every row you ever save. A project with fifty tables and a project with one table are both, at the storage level, a single database: what looks like several tables in Studio is several shapes of record inside that one place, which is also why a backup of the project (Backups) captures every table in it in one action.

Datacenters, if you use one

A project's database can also be replicated onto a datacenter you provision yourself, for a copy that survives a region going down. That is a hosting choice layered on top of the same one database per project model described here, see Datacenter.

Your own workspace works the same way

The workspace you land on at https://xuda.ai/dashboard, the six tabs covered in The workspace, is not a special case. It is backed by a project database exactly like one you build yourself. A team project you build in Studio can carry the same six tabs under its own URL, with its own chats, apps, contacts and profiles, kept in its own database, separate from your account's, prepared the first time you open it as a workspace.

How a call finds the right database

Every table read or write names an app_id, the id of the project the table belongs to. Two rules keep that id honest:

  • A deployed instance still points at the one owner. If a project has been deployed and the id you pass belongs to that running deployment rather than the project record itself, the platform resolves it back to the project that actually owns the data before touching a table. Reads and writes never fork into a separate copy per deployment, they land in the one database the project was created with.
  • Opening a workspace is checked against who owns it. Switching into a project's workspace view is refused outright for anyone who is not that project's owner, there is no fallback to a different account's data if the check fails. The same account id, workspace id and requester have to agree before a workspace's database is ever touched.
What this means day to day

You never choose a database, name a database, or configure where a table's rows are kept. You build a table in a project, and the project's own database is where it lives, for as long as the project exists.