All products

Check Metabase references with AI

Validate links between exported reports and the columns, collections or cards they depend on before importing a revised content tree.

You can check Metabase references with AI when an export looks well formed but a dashboard still points to a card that no longer exists. The metabase-semantic-checker skill runs the official reference checker against local content files and the database metadata they describe.

A team renaming a warehouse column can use the report to find saved questions that still refer to the old field. This check is relatively slow, so it is suited to an explicit investigation or a batch of edits rather than every small change.

What it does

  • Check dashboard and document references to saved cards.
  • Find missing collection parents and unresolved entity identifiers.
  • Inspect table and column references in native SQL.
  • Check query-builder joins and expressions against exported metadata.
  • Rerun the check once related corrections have been applied.

Install and use

  1. Install the skill and Docker. Obtain the official Metabase Enterprise image, the representation tree and its matching metadata export. The container can access the mounted project.
  2. Ask “Verify the column references in these saved questions after the payments table change.”
  3. Review the failing file and identifier. If the metadata is stale, refresh it before treating the reported reference as broken.

FAQ

Source

Made by Metabase. Original skill: metabase-semantic-checker, published on GitHub under the MIT licence.

View on GitHub