Workflow
How the PW-Tests framework operates end-to-end: from CI triggers through test execution to failure resolution.
CI Pipeline Flow
Tests are triggered by CI schedules, deploy hooks, and manual runs. Each trigger type has its own suite configuration.
Schedule Triggers
| Schedule | Suite | Frequency |
|---|---|---|
| Smoke | smoke | Before deploys |
| PDT | pdt | After deploys |
| Shadow | shadow | Every 15 min |
| E2E | e2e | Daily |
| Content | content | Daily |
| Nightly | user-flows, mobile, visual | Nightly |
Pipeline Stages
- Install — npm ci, Playwright browser install
- Execute — run test suite with configured workers and retries
- Report — write JSONL logs, push to OpenSearch via reporter
- Alert — Slack notifications for failures (if SLACK_ALERTS=1)
- Heal — auto-healing attempts for known failure patterns
Failure Lifecycle
When a test fails, it enters a structured resolution flow.
Detection
Tests fail → reporter captures: screenshot, error category, stack trace, step data. Written to OpenSearch cncqa_tests-*.
Classification
Intelligence service assigns: category (SELECTOR_NOT_IN_DOM, CONSENT_BLOCKING, etc.), priority (P0-P3), owner (QA, Frontend, Backend, DevOps).
Resolution
- Auto-healed — unified healer detects pattern, proposes fix, creates MR
- Interactive heal —
npm run heal:claudefor complex cases - Manual fix — developer investigates and fixes
Verification
Fix applied → re-run test → verify passing → update fix database → close.
Slack Integration
Failure Alerts
Per-failure alerts with calm headline, verdict labels, clustered thread style. See docs/operations/SLACK-HUMANIZED-FAILURE-MESSAGES.md.
Recovery Replies
When a previously failing test starts passing, automated reply posted to the original failure thread with root cause and fix details.
Weekly Reports
Automated weekly summary posted to #qa-reports with pass rates, trends, and notable changes.