Your first test run
This walkthrough assumes ARTA is installed and running at http://localhost:38088. The flow below is the product’s core loop and runs entirely in the UI.
The fastest first win is the API-test path from an OpenAPI spec — it needs no browser automation against your app and grounds generation in a contract you already have.
1. Create a project
Section titled “1. Create a project”A project points ARTA at one system under test (SUT). Give it your SUT’s OpenAPI spec URL or base URL. From here ARTA can discover what actually exists: endpoints, response shapes, and (for UI testing) the DOM.
2. Import requirements
Section titled “2. Import requirements”Add the requirements you want verified — from Jira, an OpenAPI spec, or plain documents. Each requirement gets risk-scored (probability × impact, 1–9) so the highest-risk behavior is tested first, and decomposed into ATDD Gherkin acceptance criteria.
3. Generate
Section titled “3. Generate”Press Generate. For each requirement, ARTA produces test scripts for the
relevant runtimes — Playwright for UI flows, Newman collections for API
behavior — and runs every script through its grounding validators: any
selector, role, or endpoint that doesn’t exist in what discovery found is
rejected, retried with hints, and honestly marked BLOCKED if it can’t be
fixed. A blocked test is a visible outcome, not a silent gap.
4. Run
Section titled “4. Run”Press Run. ARTA dispatches the generated scripts across its execution runtimes and collects results, retrying with self-healing hints where a failure looks recoverable.
5. Read the report
Section titled “5. Read the report”The report is where ARTA differs from a raw test runner. Every failure is classified:
| Classification | Meaning |
|---|---|
sut_regression |
Your system broke — this is the signal |
test_gen_bug |
ARTA generated a bad test — our bug, not yours |
grounding_blocked |
The test couldn’t be grounded in discovered reality |
and every result is linked back through the traceability graph: requirement → acceptance criterion → test → execution → defect. A red report means something.
Where the generated tests live
Section titled “Where the generated tests live”Generated artifacts are standard formats — Playwright specs, Postman collections, k6 scripts — not a proprietary representation. They are yours to inspect, edit, and version.
- Core concepts — the ideas behind grounding and truthful reporting
- Configuration — providers, Jira defect filing, telemetry