// the find
chenglou/node-huxley
Codeless front-end testing.
Huxley is a visual regression tool for front-end code. You drive the UI once with `hux --record`, pressing enter to capture a screenshot at each step, and it saves the actions and images into a Huxleyfolder beside a Huxleyfile.json. Later runs replay the same steps through Selenium and diff the new screenshots against the recorded ones. It suits a team with stable pages or components that wants a safety net without writing selectors.
Recording means there are no selectors or test code to maintain. The clicks, keypresses and scrolls are stored as data in the Huxleyfolder, so the baseline and the test live together in version control. The core is small: record, replay, and a diff step, with the click, keypress and scroll simulators each in their own module and covered by a test. The default workflow includes git helpers and an unchanged-task filter, which suggests replays can skip tasks whose inputs did not change.
The last push was February 2017. The README still tells you to run a Selenium Server and a Python SimpleHTTPServer, which predates headless browsers and current driver setup, so expect to spend time on tooling before the first recording. The diff is pixel-level, so fonts, antialiasing and scrollbars differ across OS and browser versions, and baselines recorded on one machine will fail on CI unless you pin the environment. A failure only tells you a screenshot changed, not what changed or whether the change was intended, and every deliberate UI change means re-recording by hand, which gets tedious past a few dozen tasks.