Skip to main content
The schema editor lets you view your database structure and add descriptions that help AI understand what your data means. Better descriptions lead to more accurate dashboards.

Opening the schema editor

  1. Go to Data Sources
  2. Find your connection (must show AI Ready status)
  3. Click Edit Schema or open the menu and select Edit schema

Viewing your schema

The editor has two panes:
  • Left sidebar — List of all tables with column counts and estimated row counts. Use the search bar to filter by table name, description, or column name.
  • Right panel — Details for the selected table, showing every column with its type, constraints, and description.
Each column displays:
  • Data type (e.g., varchar, integer, timestamp)
  • Whether it’s nullable
  • Primary key or foreign key indicators
  • Unique constraints
  • Allowed values or check constraints (if any)

Adding descriptions

Descriptions tell the AI what your tables and columns represent. Without them, the AI can only guess from names and types.

Table descriptions

Click the description area under any table name and type a short explanation of what the table stores. Maximum 200 characters. Examples:
  • orders“Customer purchase orders with line items and totals”
  • events“User activity events tracked from the web app”

Column descriptions

Click the description area next to any column and explain what it contains. Examples:
  • mrr“Monthly recurring revenue in cents (USD)”
  • status“Order status: pending, confirmed, shipped, delivered, cancelled”
  • created_at“When the user first registered”
Columns with ambiguous names benefit most from descriptions. A column called type or status could mean anything — telling the AI what the values represent makes a big difference.

How descriptions improve dashboards

When you ask VexiData to generate a dashboard, the AI reads your schema including all descriptions. For example:
  • Without description: AI sees a column called amt of type integer and might misinterpret it
  • With description “Total order amount in cents (USD)”: AI knows to divide by 100 and format as currency