End-to-end test a desktop app with four HTTP calls

Web apps get Playwright. Desktop apps get a shrug: GUI test frameworks are heavy, flaky, and chained to a runner nobody wants to maintain.

Your app's CI does it differently. The build installs the binary in a container and launches it on a display. The test is plain HTTP: POST /api/v1/display/input/batch walks the critical path: open file, click the wizard, save: with waits between steps. Then GET /screenshot captures the result, and a diff against the baseline image decides pass or fail. On failure, POST /input/reset releases held keys, and the failing screenshot lands in the CI log.

No framework, no driver, no runner with X11 duct tape.

The test rig is four HTTP calls and an image diff.