Skip to content
facetshot

Status and limits

facetshot is at version 0.1, and the config format may change before 1.0. This page is the one place that says what has and hasn’t been tried; other pages describe how things are meant to work.

Platform support

Section titled “Platform support”
macOS Linux Windows
Web pages in Chromium and WebKit tested expected, untested expected, untested
Web pages in Firefox not yet seen working (see below) expected, untested expected, untested
Storybook, Electron tested expected, untested expected, untested
Sessions and the background helper tested untested probably not working (see below)
Mac windows: screenshots, steps, sessions, logs tested not available not available
iOS simulator screenshots and label steps tested by hand; CI needs a macOS runner with Xcode not available not available
SwiftUI scenes tested; CI needs a macOS runner with Xcode not available not available
Flutter web tested; macOS not run end to end; iOS untested expected, untested (web) expected, untested (web)
Tauri the UI in WebKit tested; the real window run by hand, no automated test planned planned
Android planned planned planned

“Expected” means the feature uses only Playwright, which behaves the same on that system, but nobody has tried it there yet. Reports are welcome. No CI provider has been tried yet.

Known limits

Section titled “Known limits”

  • Installing: bun add -g (from npm and from a packed tarball), bunx facetshot from npm, a tagged version from GitHub, facetshot setup, init --agent all and uninstall --yes in a folder with no package.json were run on macOS. In 0.1.0, uninstall run through bunx wrongly says facetshot is installed in the project; the next version fixes it. uninstall --browsers hasn’t been run; it calls Playwright’s own uninstall.

  • Firefox: Playwright’s Firefox build hasn’t been confirmed working on macOS yet. moz-firefox (the installed Firefox app) is driven over WebDriver BiDi, which Playwright supports only experimentally.

  • New iOS steps, not yet run on a simulator: steps that name an element by role (role=button[name="Save"]) or by a point, long presses (and right-clicks, which become long presses), drags, and the app’s system log are written but haven’t been run on a simulator yet. There is no automated iOS test.

  • Sessions don’t open iOS simulators or Storybook stories yet. Use flows for those.

  • Flutter on macOS: Flutter 3.24’s own build tool fails flutter build macos with recent Xcode versions (it misreads lipo’s output), so a Flutter Mac app hasn’t been run end to end. Use a newer Flutter; the web build is not affected.

  • Tauri’s real window was run by hand on macOS: a flow pressed a link through Accessibility, and the Rust side’s println! came back in the logs. No automated test covers it.

  • Mac permissions: steps and sessions in Mac windows need Accessibility, and resizing a window from outside needs Automation. See Desktop apps.

  • iOS variants: on the iOS simulator, locale, right-to-left, personas, offline and reduced motion aren’t applied yet. See Variants on iOS.

  • Windows: the background helper listens on a Unix socket and stops work by process group. Neither is expected to work on Windows, so sessions there probably don’t, and runs may fall back to working in the calling process. The load brake is off, because Windows has no load average.

  • Tauri on Windows (WebView2) and Linux (WebKitGTK) isn’t supported.

  • FacetshotKit: the path for test targets hosted in an app (which can draw Liquid Glass and sheets) is written but not yet covered by tests. See the FacetshotKit README.

SourceIssuesReleasesLicense (MIT)