Skip to main content

DB Copilot

The DB Copilot is the Autobot on the AutessaDB Tables page. Tell it what data you need to store — or what you want to know about data you already have — and it designs schemas, applies changes with your approval, and answers queries conversationally.

What it's for​

Designing a good schema means picking column types, deciding what's required, what should be vectorized for semantic search, and how tables link together. The DB Copilot does that from a plain-language description: "track customer support tickets with a status, priority, and a link to the customer" becomes a concrete table proposal you can inspect column by column. It also modifies existing schemas and queries your data, so it doubles as a quick way to explore what's in a table without building anything.

Open AutessaDB → Tables in the sidebar and click the Autobot button. The copilot opens as a resizable panel on the right. Conversations start fresh each time you open the panel.

Key concepts​

  • Reads run directly, writes need approval. Exploring — listing tables, fetching schemas, looking up select-column options, querying data — happens straight away, visible as expandable tool-call cards with arguments and results. Anything that changes a schema goes through a confirmation card first.
  • Schema change cards — proposals appear as an amber table change(s) proposed card. A Create proposal expands to the full column list (name, type, required, vectorized, foreign key), plus tags and select-column options. An Update proposal shows exactly which columns will be added or removed and any new select options. Apply Changes executes; Reject Changes discards. The chat input locks until you decide.
  • Design first, then build — for larger asks the copilot may present a design overview of several related tables before proposing the actual creations, so you can steer the model of your data early.
  • Credits — like the other copilots, it draws on your company's daily and weekly AI credit allowances.

When to use it​

  • New data models — describe the business objects and let it propose types, required flags, and foreign keys; adjust in the table editor afterwards if needed.
  • Schema evolution — "add a column for the contract renewal date" is faster in chat than in the editor, and the update card shows the exact diff before it lands.
  • Ad-hoc questions — ask about your data ("how many open tickets are assigned to nobody?") without writing a query.
  • Prefer the manual table editor when you're making a single precise change you already know exactly, or when editing triggers and computed columns, which have their own editors.

How to create a table with the copilot​

  1. On the Tables page, click Autobot.
  2. Describe the data: what each record represents, the fields you care about, which text should be semantically searchable, and any relationships to existing tables (name them).
  3. Review the Create card — check every column's type, the required flags, vectorized columns, and foreign keys.
  4. Click Apply Changes. The table appears in your list, ready for rows, triggers, and agent access like any hand-made table.