Skip to main content

Test Suites

A test suite is a named collection of test cases that can be executed together. Suites are the primary way to organize your tests by feature area, flow, or release scope — and to run a whole group of tests with a single action.

Purpose​

Individual test cases validate specific scenarios in isolation. Suites let you compose those cases into meaningful groups that reflect how you think about your bot:

  • A "Checkout Flow" suite covering all the scenarios in your ordering process
  • An "Onboarding" suite for new-user interactions
  • A "Regression" suite for the most critical flows you want to verify before every release

Every new test case is automatically added to the Default suite when created. You can move or copy cases to other suites at any time.

Execution Behavior​

When you run a suite, all test cases in it are executed. The system tracks each case's status in real time as execution proceeds.

If you trigger a suite while some of its test cases are already running, the platform only starts the ones that are not currently in progress. This makes it safe to re-trigger a suite if you are unsure whether a previous trigger registered.

Import & Export​

You can move test suites between bots and environments on its own, without exporting an entire bot version. This is useful for sharing a suite with another bot or promoting it from a testing environment to production.

Exporting​

Use the Export action to download your test suites as a .zip file. The export bundles all suites, its test cases, and the agents those cases rely on - input generators, custom conditions, and LLM-judge conditions.

Importing​

Use the Import button to upload one or more suite .zip files at once.

Importing is destructive: it replaces all existing test suites and test cases in the current bot with the contents of the uploaded files. Because this cannot be undone, you must confirm by typing the bot's name before the import runs.

References are re-resolved against the target bot automatically:

  • Agents (input generators, custom conditions, LLM-judge conditions) are matched by name. One that already exists in the target bot is reused; one that doesn't is created. Existing agents are never deleted.
  • Slot conditions are re-linked to the target bot's extraction fields by name, and tool-call conditions are matched to tools by name.

If a referenced extraction field or tool does not exist in the target bot, the import still completes and reports a warning listing what could not be resolved. Review those cases, as they may not run correctly until the missing field or tool is added.

An import is blocked while the bot is busy: if one of its test cases is currently being edited, if a test run is in progress, or if a bot-version restore is running. Wait for the other operation to finish and try again.

Tips​

Prefer smaller, focused suites. A suite that covers one feature area is easier to reason about and faster to run than one that covers everything. Run the relevant suite when you change something specific; run a broader suite before a release.

Rerun after changes. Any time you modify a flow, a prompt, an extraction field, or a script, rerun the suite that covers that area to catch regressions early.

Move test cases to match your structure. As your bot evolves, reorganize test cases into suites that reflect the current feature boundaries. A test case can be copied into multiple suites if it belongs to more than one area.