Draft a suite or test data with the assistant

Ask for a suite with checks, or rows of test data, and confirm the preview before anything is saved.

In short

The assistant can draft a suite with checks from your saved requests, or rows of test data for a suite. You read the preview and confirm; nothing is saved before that, and nothing is sent until you run it.

Before you start

Who: drafting needs Edit on the project. Asking uses your AI allowance or your own key.
Plan: AI is part of every plan. On Free the platform's key answers questions about how {{brand}} works, so drafting there uses your own AI key or a paid plan (see /help/ai/answers-left).
First: save the requests the suite should call.

Steps

Press "Ask AI" and say, for example: Create a smoke suite for the orders requests with checks.
Read the preview: the steps, and the checks on each.
Press "Confirm". The suite is saved.

Open a suite that has no data set, press "Ask AI" and say, for example: Generate ten rows of test data for this suite.
Read the preview: how many rows, and which columns.
Press "Confirm". The rows become the suite's data set.

Result: a suite, or a data set, you can run.

Good to know

The checks are written from the project's API document and the shape of earlier answers - field names and types, never a value. A step can only name a request already saved in the project. Read the checks before you rely on them.
The test values are made up. Each row is one run of every step, so check them before you run the suite.
Asking counts like any other question to the assistant. Confirming spends nothing more.
"Undo" deletes a drafted suite while nobody has run or changed it, and takes generated rows away while they are unchanged.

If something goes wrong

A suite that already has a data set is left alone: clear the data set first.
A step names a request that does not exist: save the request in the project first.