Python + Playwright · three flows · USD 250
Catch the small break
before it ships.
Choose three important flows on one public website. I turn them into a small Chromium regression suite that runs at 1440 × 1000 and 390 × 844, saves JUnit results, and captures evidence when a check fails.
No credentials, source archive, or payment at the fit-check stage. We agree the exact URLs, flows, assertions, and delivery terms in writing before payment.
One public site · three flows / twelve checkpoints · two viewports · one correction pass.
A baseline you can rerun
Name. Test.
Hand off.
01Freeze up to three public user flows, twelve checkpoints, and the visible result that makes each one pass.
02Build a Python pytest/Playwright suite with a small page-object layer.
03Run the flows three consecutive times in one agreed environment at both Chromium viewports.
04Return the source, one-command runner, JUnit output, failure screenshots, issue ledger, and README.
What comes back
A test baseline your team can actually run.
The delivery is code and evidence, not a recording of someone clicking through the site once.
Agreed test map
Up to three named flows and twelve checkpoints, their public URLs, the two viewport expectations, and the visible pass condition for each.
Runnable suite
Python pytest/Playwright tests, a small page-object layer, pinned setup instructions, and one stable headless-Chrome command.
Reviewable evidence
Three consecutive runs in one agreed environment, JUnit XML, screenshots on assertion failure, a short issue and boundary ledger, plus a concise handoff README.
Executed first-party proof
The same pattern now guards a real trilingual site.
LazyTravel's Lanzhou site is generated from canonical multilingual content. Its public regression suite checks all eleven chapters, desktop and 390px mobile-width navigation, ruby controls, the interactive map, console errors, and failed requests.
Clean GitHub Actions run
Built from source, then passed three consecutive browser runs.
The workflow installs its own browser, builds the site locally, runs the same pytest suite three times, and uploads JUnit and failure artifacts. The implementation uses page objects and invocation-specific artifact folders, so reruns do not hide earlier failures.
This is LazyingArt's own site and workflow evidence, not a customer result or a promise that an unfamiliar site will have the same shape.
Companion recovery fixture
A reload, failed write, and account switch all get their own test.
A second project-owned specimen saves an unsent draft across reload, interrupts one write and retries it with the same id without a duplicate, then checks that changing accounts hides the previous account's cached state. The suite passed three consecutive Chromium runs and keeps JUnit plus a screenshot and Playwright trace when a check fails.
This is a synthetic fixture with no customer data or private AiMemo source. Authenticated staging, offline queues, and account-scoped storage are outside the fixed USD 250 public-site baseline and require a separately agreed scope.
A deliberately small first engagement
Public, stable flows first.
The fixed scope excludes login, private staging, credentials, CAPTCHA or MFA bypass, checkout or destructive actions, cross-browser coverage, native apps, site fixes or visual redesign, full manual-checklist conversion, test-data systems, performance, security or accessibility certification, production deployment, ongoing maintenance, and unlimited revisions.
You provide
The public site, three priority flows, expected visible results, supported language, and any known instability or third-party boundary.
We agree first
Exact URLs, actions, assertions, viewport sizes, repository or delivery format, timing, cancellation, and what counts as accepted.
Your site stays yours
You remain responsible for the site, its content, hosting, third-party services, and production decisions. No private customer data belongs in this baseline.
Working terms
Three flows are fixed before payment.
Timing
Delivery is within seven business days after payment and after the public URLs, pass conditions, target repository or delivery format, and required public assets are stable and available.
Correction pass
One consolidated list of factual corrections to the agreed checks is included within seven calendar days after delivery. New flows, pages, browsers, viewports, or site behavior are a new scope.
Cancellation and refund
Cancel before work begins for a full refund. Once work starts, the USD 250 price is allocated as USD 75 for the test map and setup, USD 125 for the runnable suite, and USD 50 for evidence and handoff; only undelivered items are refunded.
Support
Support is limited to written clarification during the correction window. Fixing the website itself, test-data setup, CI administration, and maintenance after site changes are separate work.
Browser Regression Baseline · USD 250
Start with the three flows you least want to break.
The free fit check asks for public URLs and expected behavior only. If the site fits, both sides accept the exact written scope before payment. Direct-site clients receive a Stripe request afterward; marketplace clients keep contract and payment on that marketplace.