Letting AI design a table and suggest rules in Base by Zaweba

The blank table designer is the hardest screen in any database. You know you want to keep track of something, a set of tools on loan, say, but turning that into fields with the right types takes practice. Base by Zaweba has two AI helpers for this stage, both using your own AI account. Design a table turns a description into a table design. Suggest rules looks at a table you already have and proposes validation rules and input masks for it.

Neither does anything on its own. Each shows you what it proposes and waits for you to say yes.

Design a table

Describe what the table should hold, in ordinary words. "Tools we lend to members: the tool's name, its condition, who has it, when it went out and when it's due back." The AI turns that into a design with fields and types.

Before anything is made, Create table shows every field so you can check it, and Base runs its own design checks on the proposal. Only then is the table created. If a field has the wrong type, or the AI has missed something, change it in the table designer afterwards like any other table.

Look closely at links. "Who has it" should usually be a lookup to your members table rather than a typed name. Add or correct those lookups in the designer or from the relationships diagram, so the new table joins the rest of the database properly.

Suggest rules

Suggest rules works on the open table. It looks at the fields and proposes validation rules and input masks that would keep the data tidy: a quantity that can't be negative, a postcode mask, a percentage between 0 and 100, and so on, depending on what the table holds.

Apply rules lists each suggestion with a tick box and a short reason, so you can see why it was proposed and choose which to keep. Untick anything that doesn't suit your data. A rule that is usually true but not always is worth leaving out, because it will block a genuine record one day.

Base's usual safety applies. If a suggested rule is one that existing records already break, Base refuses it, just as it would if you typed the rule yourself. That's a useful prompt to find and tidy those records before trying again.

A sensible way to use them

Use Design a table for the first draft and your own judgement for the finish. Give each field a clear name, set required and unique where they matter, and add any value lists for fixed choices. Then run Suggest rules and pick the suggestions that match how your organisation really works.

For new users, the helpers double as a lesson. Seeing which types the AI chose for a description, and which rules it thought sensible, teaches the shape of a good table quickly.

Only the design leaves your PC

Only the database's design goes to your AI provider: tables, fields, types, rules, relationships and the names of queries, forms and scripts, plus the open table. Records stay on your PC unless you tick Include sample rows, which sends the first three of each table and is off again each time. For designing a table and suggesting rules, the design is all the AI needs.

Keeping control

Any change an AI helper makes is previewed first, and AI > Undo reverses AI changes for the session. Between that, Base's own checks and the backups it makes, there is little risk in trying a suggestion and stepping back if it isn't right.

The result is a table that is well formed from the start, with rules that protect it on every form, in every import and on the exported website.