Free

Free coded apps to suit all. No sign up, no email, usable offline, updates when I can, feedback welcome. Send feedback

Thinking

Can I make a database with sample data using AI?

Yes. Make sample data in Base by Zaweba asks your own AI account for ten made-up records, previews them and adds them all in one go or not at all.

Can I make a database with sample data using AI?

Ten made-up records at a time is what Base by Zaweba's AI helper will add, if you connect your own AI account. Make sample data creates ten invented records for the open table, Add records shows them to you first, and they go in together in a single transaction. If one of them is bad, none are added, so the table never ends up half filled.

Sample data is the fastest way to see whether a new database actually works.

Why test with made-up records

An empty table tells you very little. You can't see whether a form is laid out sensibly, whether a report's groups and totals come out right, or whether a query finds what you expect, until there are records in it. Typing test records by hand is tedious, and real records may be private.

Invented records fill the gap. With ten realistic looking customers or products in place, you can try every form, query and report you build.

Making the records

Open the table and choose Make sample data from the AI menu. The AI sees the table's design (field names, types and rules) and writes ten records to suit. Look through the preview, then press Add records.

Because the records go through the same table rules as anything else, a made-up price that breaks your "price must be 0 or more" rule is refused, and with it the whole batch. Run it again and you'll get a fresh set.

For linked tables, start with the ones that others depend on. Make customers before orders and products before order lines, so the lookups have something to point at.

Other ways to get test data

The five templates (Contacts, Inventory, Invoices and Customers, Book and Collection catalogue, and Simple CRM) come with optional example records, so if one of them fits, you get a working database with data from the start.

You can also import a CSV or Excel file, which is handy if you already have a spreadsheet of test data, or an old list you want to bring in. Columns are matched by name or caption, and bad rows are skipped and listed.

Clearing it out before going live

Made-up records must not end up mixed with real ones. Before you start using the database for real, delete the test records. In a small table you can do it in the datasheet. In a bigger one, a delete query does it in one go, telling you how many records it will remove and asking first. Afterwards, Database > Compact and repair tidies the file.

Another approach is to keep a copy of the database with test data for trying things out, and a clean copy for real use. Since each database is a single .obase file, that is just a matter of copying it.

What is sent to the AI

Only the design goes to your provider: table and field names, types, rules and relationships, plus the open table. Records are not sent unless you tick Include sample rows, which starts off every time. Changes made through the AI are previewed and can be reversed for the session with AI > Undo.

The AI menu stays greyed out until you connect an account, whether that is a provider such as Claude, OpenAI or Gemini, a local model such as Ollama, or copy and paste with a chat website. Everything else in Base by Zaweba works without one.