Skip to main content
Available for self-hosted Business plan and above
Path: System Admin → Table operations → Schema integrity The field definitions Teable holds and the actual structure in the database are meant to match one another. When they drift apart, users see fields that will not open, link fields that return nothing, or a table whose reads and writes keep failing, with no visible cause. Schema integrity locates problems of this kind and repairs them. An instance admin can check any project in the instance without joining its space first, tenants on a customer-managed database (BYODB) included.

Run a Check

With no table selected, the page lists project checks. Search by project, space, table ID or name, then confirm the scope using its space, database and runtime. Projects in different spaces often share a name. Click Check project on that row, then Run Check in the Schema Integrity dialog. If the issue affects one table, select it using Choose a table at the top; a table check starts automatically. Choose Check whole project to expand the check to the project, or Review on an issue to inspect repair options.

Read the Results

Results are listed per field and rule, in four states: Start with Error: a fault a user reported almost always lands in this category. A Warning does not explain the current fault, but it can turn into an error as the field structure keeps changing, so it is worth clearing once the errors are handled.

Repair

You can Repair rule by rule, or work in bulk with Repair warnings only or Repair warnings and errors. While chasing a live fault, repair the errors one at a time and confirm the fault is gone before handling warnings, so that a new problem can be traced to a single change. A repair changes the table structure only; it does not change record content. Before running one, use the preview next to the repair button: Confirm repair details shows the reasoning and the SQL the dry run produced, and nothing is executed until you confirm. Some rules cannot be repaired automatically and show Manual instead; the dialog then explains why the problem needs a person. When the dry run returns no executable SQL, the dialog says so explicitly, and that case needs a person too. After repairing, click Re-check to confirm the problem is gone.
Last modified on October 27, 2026