AI > Fine-tune AI

Learn how Fine-tune AI in Draxlr lets you describe tables and fields and choose which tables the AI can use, so it generates more accurate SQL for your data.

Fine-tune AI lets you add descriptions to your tables and fields, and choose which tables Draxlr AI is allowed to use. This gives the AI the business meaning behind your schema so it can generate more accurate, relevant SQL.

What is Fine-tune AI?

Fine-tune AI is schema-level context you provide to Draxlr AI. For each table you can:

  • Add a table description — explain what the table represents in plain language
  • Toggle Include in AI — control whether Draxlr AI can use the table when answering questions
  • Add field descriptions — explain what individual columns mean and how they are used

Without this, AI relies only on table and column names. With it, the AI understands intent and skips the tables you would rather keep out of its answers.

How to Use Fine-tune AI

The Fine-tune AI feature helps Draxlr's AI better understand your database by adding human readable descriptions to tables and fields, and by letting you control which tables the AI can use. This improves the accuracy of generated SQL queries, charts, and insights.

To use Fine-tune AI:

  1. Open Draxlr AI and go to Fine-tune AI.
  2. Use Search Table to find a table, then select it from the list (tables are grouped by schema).
  3. Click Add table description to explain what the table represents in simple terms, and press Enter to save.
  4. Use the Include in AI toggle to control whether Draxlr AI can use this table when answering questions. Disable it for tables the AI should ignore.
  5. Search or scroll the fields, click a field to add a short description of what it means and how it is used, then press Enter to save.
  6. Repeat this for the important tables and fields that you frequently query.

Once added, these descriptions are automatically used by the AI when generating SQL, choosing joins, applying filters, and naming charts. Tables with Include in AI disabled are excluded from AI answers.

Tips

  • Start with core business tables like orders, users, payments, or events.
  • Turn off Include in AI for system, auth, or migration tables the AI should never touch.
  • Keep descriptions short and clear, as if explaining to a teammate.
  • Explain business meaning, not technical details.
  • Clarify fields with ambiguous names (e.g., status, type, code).
  • Add context for IDs and foreign keys to help AI choose correct joins.
  • Update descriptions when schema or business logic changes.