Data Assistant

By the end of this page you will know how to use the AI Data Assistant to create tables, modify schema, and generate sample data from plain-language requests — and how the approval step keeps you in control.

Prerequisites

• Admin or Developer role. The assistant is available to users who can manage data.
• A project open at Data Vault → Tables.

Open the assistant

In the Tables area, click AI Assistant at the bottom of the table list. A chat panel opens on the right; the button changes to Close Assistant.

What it can do

Describe what you want in the message box (for example, “Create a customers table with name, email, phone, address, and status”). Suggestion chips offer common starting points:

• Create a customers table
• Create related tables
• Add columns to a table
• Generate sample data
• Insert a record

Under the hood the assistant can read your schema, query data, create and modify tables, and generate realistic sample data. Sample data is produced with Faker — it doesn’t consume model credits. For the full set of underlying tools, see the database tools.

Approve before it runs

Any change to your schema or data pauses for your approval — nothing is executed until you say so.

Review the proposed operation

When the assistant is ready to create a table, change schema, or insert rows, it shows an approval card summarizing the operation and the SQL it will run.

Execute or reject

Click Execute to run the operation, or Reject to cancel it and refine your request. Once it runs, the table list refreshes to show the change.

Common issues

• The assistant panel isn’t available — you need a role that can manage data. Viewers don’t see it.
• Nothing changed after a request — check whether an approval card is waiting; operations don’t run until you approve them.
• A request was rejected or errored — rephrase with more detail (table name, column names and types) and try again.

Next

• Manage table data — work with the rows the assistant created.
• ER diagram — see how new tables relate.