Skip to content
Test Automation Recipes

Featured libraries

Several recipes on this site are powered by small, focused open-source libraries. Each one exists because a recipe needed it — they solve exactly one hard problem each.

playwright-audio-mocking

Section titled “playwright-audio-mocking”

Stream any audio file into a website’s microphone (getUserMedia()) from Playwright — on Chromium, Firefox and WebKit, with runtime play / pause / stop / loop control.

Chromium’s built-in fake-audio flag only supports WAV, only works on Chromium, and can’t switch files at runtime. This library replaces getUserMedia with a real Web Audio-backed MediaStream, so MediaRecorder, analysers, WebRTC and speech recognition all work unmodified.

agent-screen-observer

Section titled “agent-screen-observer”

Watches the screen while a test (or an AI agent) drives a UI, and reports visual events — toasts that fade away, error screens that flash by, late layout shifts, animated regions — as compact text with on-demand image evidence.

It uses classical computer vision (pixel diffing against a pre-action baseline) over a CDP screencast, so detection is deterministic: no LLM, no frame-rate guessing. A one-line summary costs tens of tokens instead of streaming full screenshots.

playwright-interactive-ui-test

Section titled “playwright-interactive-ui-test”

An experimental toolkit for testing highly interactive UIs — map panning with inertia, canvas rendering, physics animations — without touching the app’s DOM or JavaScript. All observation is pixel-based: gesture synthesis with velocity profiles, phase-correlation visual odometry to measure how far the view actually moved, and closed-loop control that pans by an exact pixel distance.