Repeat counts run exactly as the test says
A repeat count is run as given, an until condition that already holds is refused before anything runs, and only evidenced iterations count toward a goal.
Wopee.io · Changelog
Product updates and improvements, as they ship to production.
Running in productionAPI v5.216.2CMD v3.424.0Agent v3.126.4since
A repeat count is run as given, an until condition that already holds is refused before anything runs, and only evidenced iterations count toward a goal.
Saved locators drop state such as item counts, keep every role option on replay, and a heal can no longer swap in a control with a different name.
Scenarios you create while a crawl runs are kept with their IDs; the crawl updates only its own entries, and crawled user stories keep their names.
Every reasoning-effort level from none to max is checked against the model it is set for, GPT-6 runs as a reasoning model, and AI chat honours it.
Each reported issue is checked against what the run did; a claim the record contradicts is withdrawn, and locator failures are listed as automation.
A run that executed every step and failed only on its defined assertions now gets a Playwright spec that asserts the expected values until the fix.
The agent now receives a screenshot after every interaction and can look at the page on demand; existing projects move to Observe unless they chose Off.
New Project and Advanced Setup are one dialog: demo apps pre-fill the URL, and framework, generation mode and name wait in an Advanced settings accordion.
Integrity gates catch runs that asserted nothing or skipped steps, mark them Incomplete in amber, and close stranded runs instead of leaving them hanging.
Expected text the crawl never saw is flagged, the agent's scenario picks run first, steps visit only crawled URLs, and broken images are reported.