Skip to content

v0.6.0: Python browser-use action receipts

Choose a tag to compare

@LydiaTools LydiaTools released this 26 Oct 07:24

StepReceipt v0.6.0 adds an optional single-file Python bridge for browser-use ActionResult and your own independent outcome checks. It writes the existing portable JSON/HTML format and CI summaries. The Node recorder keeps zero runtime dependencies.

  • Four original synthetic local cases: disabled button, working save, false success toast, unavailable readback. Actual CLI exits: 1 / 0 / 1 / 1. Each case makes one native click; no automatic retry.
  • Six Python bridge tests and the existing thirteen Node unit tests passed. The exact source also passed all three GitHub CI jobs, including native browser-use on Linux/Chrome 155. Local native runs used macOS/Chrome 156.
  • Real report inputs, English/Chinese guides, a genuine screenshot and offline HTML reports are included. The optional Stagehand Page example already on main is preserved.

Try without installing anything: extract the ZIP and open docs/browser-use/disabled.html. For fresh native execution, follow the English guide or 中文指南. Native example dependencies are optional; browser-use is pinned to the installable PyPI 0.13.11.

Open the live disabled-click report · Bridge source

This measures native tool calls on original synthetic pages, without Agent.run, a model, an API key or a production account. The disabled-button issue was previously described by other contributors in browser-use #5964 and #6031; this project does not claim its discovery or republish their fixtures/model results. The bridge records supplied checks and does not fix upstream click behavior, enforce business idempotency or establish an overall agent success rate.

Source: b08c302bb397b5961443ed675fecce7615727377. ZIP SHA-256: fd981ff71c2401f45168300da07856554f8c316b630b16e50278e9ed717d6ec1. Both the source ZIP and SHA256SUMS.txt are attached. Older fixed release assets are unchanged.