QA Automation Tool

A centralized platform for automating web-application quality assurance. Runs a predefined suite of end-to-end checks, aggregates results in real time, and presents trends on a clean dashboard.

Live Dashboard →

---

Status & CI

• Scheduled tests: every 6 hours via GitHub Actions
• Cleanup tasks: every 3 days (Supabase & Vercel KV)
• Preview DB keepalive: every 3 days (test Supabase)
• Coverage:
  • Playwright tests for page structure, functionality and performance
  • Manual-mode support via Excel input

Preview DB Keepalive

The workflow Keep Test Supabase Alive inserts a heartbeat row every three days using scripts/keepalive-test-supabase.js. It expects SUPABASE_TEST_URL and SUPABASE_TEST_SERVICE_ROLE_KEY secrets. The script writes to a small keepalive table and prunes old rows to show ongoing activity. If that table doesn't exist, it falls back to a simple read query so the database still registers usage.

---

Overview

The QA Automation Tool:

• Executes a comprehensive set of UI and workflow checks across target URLs
• Stores results, progress and artifacts (screenshots, videos when needed) in Supabase and Vercel KV
• Visualizes outcomes and trends in an interactive Next.js dashboard
• Enables ad-hoc and scheduled runs, with AI-driven query support via Gemini

---

Technology Stack

Layer	Framework / Service
Application	Node.js, Next.js
Automation	Playwright
Data Storage	Supabase (Postgres) & Vercel KV
CI / Delivery	GitHub Actions, Vercel
Reporting	Chart.js, ExcelJS
AI Insights	Gemini with LangChain

---

Key Features

Dashboard

• Recent Runs: pass / fail / known-issue / N-A counts, trend charts, artifact links
• Ask Gemini: natural-language queries (passphrase-protected)
• Ad-hoc Execution: submit name, passphrase, optional Excel; trigger tests immediately

Scheduling & Cleanup

• Run QA: every 6 hours (cron)
• Cleanup: Supabase & KV data older than 60 days, every 3 days

Manual Validation

• Screaming Frog tab outlines steps to replicate tests manually (requires paid license)

---

Test Definitions & Protocol

ID	Name	Description
TC-01	Hero Overlay on Desktop	Hero text visible at 1920×1080 for homepage, category, product and custom hero banners
TC-02	Hero Static on Mobile	Hero text visible on 375×667 viewport for homepage, category, product and custom hero banners
TC-03	Header Presence	<header> or class containing “header”
TC-04	Navigation Presence	<nav> or class containing “nav”
TC-05	Main Content	<main> or class containing “main”
TC-06	Footer Presence	<footer> or class containing “footer”
TC-07	Main Video & Carousel	<video> or Vidyard iframe; play button opens modal; carousel videos show no "Video Not Found"
TC-08	Contact Form Opens	“Contact” button launches form overlay
TC-09	Rendering Error Check	No “A rendering error occurred” text
TC-10	Gatekeeper Redirect	Incognito + “Yes” click yields 200 status
TC-11	Insights Link Works	First insights/newsroom link returns HTTP 200
TC-12	DocCheck Login Route	Incognito → URL includes /account/doccheck-login
TC-13	DE Nav Redirect	“Produkte → Ultraschall → Mehr erfahren” → target site (see Known Issue below)
TC-14	HTTP Status Code Valid	Status 200 or valid redirect (301/302)
TC-15	DocCheck Login Present	HCP-gated pages (e.g. de-de DatScan): DocCheck login iframe populates
TC-16	Regional Microsite Loads	Dismiss language + geo-location modals on a product microsite; page renders

Result Statuses

Each test resolves to one of four states (shown by color on the dashboard):

• 🟢 Pass — the check succeeded.
• 🔴 Fail — a genuine regression; investigate.
• 🟡 Known Issue — a known, accepted limitation, not a regression. Today this is driven by GE's client-side GeoIP redirect: from the US-based CI runner, some region pages (/de-de, /en-gb, /fr-fr/…, /en-in/…) bounce to /en-us, so they can't be validated as their locale in CI. A real US browser (even fresh incognito) stays on the region page, so this affects automation only. The count is flaky run-to-run by nature, reflecting how often the runner is bounced. TC-13 is the canonical example.
• ⚪ N/A — the test isn't assigned to that URL.

See CHANGELOG.md for the full history.

Supported hero components for TC-01:

• div[id*="ge-homepage-hero"]
• .ge-homepage-hero-v2-component
• .ge-category-hero__container
• .hero-content-intro
• .product-heroV2-container

Method Summary

Both TC-01 and TC-02 use the heroTextVisible helper from utils/hero.js. The script iterates over the selectors above and returns true once any hero text element is visible. TC‑01 runs at a 1920×1080 viewport, while TC‑02 sets a 375×667 mobile viewport before checking visibility.

Screaming Frog Emulation

Screaming Frog cannot verify viewport‑specific behavior such as overlays, but it can confirm whether a hero element exists. Use Custom Extraction and create an extraction with a regular expression covering the selectors listed above, e.g. div\[id*="ge-homepage-hero"\] or section\.ge-homepage-hero-v2-component. After crawling, review the “Extraction” tab to see URLs containing these hero elements.

For details and any custom extraction, use the dashboard’s Test Definitions section.

---

Repository Architecture & Access

This is the private main repository containing QA automation source code, workflows, and configurations.

Public Artifacts Access

The project is frozen (scheduled runs paused 2026-10-25). Historical test results remain in the read-only, archived repository:

• qat-artifacts - archived; do not delete or make private, because the dashboard links to its run files
• No new results are published. The publish steps in run-qa.yml and publish-artifacts.yml target this repo and will fail harmlessly now that it is archived; point them elsewhere before resuming scheduled runs

How it Works

1. Code (this repo) ran tests via GitHub Actions on a schedule (now paused)
2. Dashboard serves the archived run snapshot in data/archive-runs.json
3. Result files for past runs are linked from the dashboard into the archived qat-artifacts repo

For assistance, reach out via company channels (Teams/Outlook).

---

License

All rights reserved © Ryan Smedstad.
Unauthorized copying, modification or distribution is prohibited. Last updated: 2025-09-01