Pay less for the agent's routine lookups
Run element lookups and key presses on a cheaper Azure deployment with its own reasoning effort, and pick only the effort levels each model accepts.
Wopee.io · Changelog
What's new in Wopee.io
Product updates and improvements, as they ship to production.
Running in productionAPI v5.240.5CMD v3.458.3Agent v3.143.16since
Game key presses count only if the game reacts
A key press into a canvas game is checked against the picture before and after, so a press the game ignored is reported as having no effect, not as a success.
A test that never ran opens on its steps
Select a test that has not run yet and its steps open instead of an empty run view; story names show in the scope picker and step lists indent cleanly.
"Repeat 3 times" means three real times
A repeat count runs exactly as written, a stop condition that already holds is refused before the run starts, and only repetitions with evidence count.
Self-healing that keeps the right button
Saved locators drop state such as item counts and keep every role option on replay, and a heal can no longer swap Finish for a differently named button.
Exploring again never wipes your edits
Scenarios you create while the agent explores keep their IDs; a new exploration updates only its own entries, and user stories keep their names.
Pay once, land straight back in your projects
After checkout you return to Wopee.io, your plan is active on arrival and you continue to Projects, even when the payment confirmation is slow to arrive.
AI settings that fit each model
Every reasoning level from none to max is checked against the model it is set for, GPT-6 runs as a reasoning model, and the AI chat respects the setting.
Bug reports show the actions behind them
Each reported issue is checked against what the run did: a claim the record contradicts is withdrawn, and locator failures are listed apart from bugs.
Found a bug? You still get the test code
A run that did every step and failed only on its checks now leaves a Playwright test asserting the expected values, so it turns green once the bug is fixed.