E2E test journeys & status.

How to read this page

The Playwright suite is partitioned by role project: each project loads a Keycloak storage state (e2e/.auth/<role>.json) and matches only the specs that role is responsible for. Multi-actor flows capture a Bearer token from a second role's storage state via the captureBearerForRole helper, so a single spec can drive cross-role interactions without context-switching the whole browser.

How to run

All commands are run from e2e/ against the default base URL https://dev.cplk.org (override with CPLK_E2E_BASE_URL).

# first time / refresh role storage states
pnpm test:auth-setup

# everything
pnpm test

# only journeys (~22 files)
pnpm test:journeys

# only one role's projects
npx playwright test --project=agent
npx playwright test --project=property-owner

# only the new journeys added in this round
npx playwright test --project=agent --grep "notification-preferences|system-announcements|support-ticket-lifecycle"

# point at a different env
CPLK_E2E_BASE_URL=https://localhost:3000 npx playwright test

Role projects

Journeys

A journey is a multi-step test.describe that walks an actor (or a small cast) through a complete lifecycle. Most journeys are self-contained: setup, action, assertion, cleanup. A few open side contexts to act as a second role within the same spec.

Property lifecycle

Agency, content & engagement

New journeys added in this round

Three additional journeys close real coverage gaps on dev. All three exercise public REST contracts that previously had unit-test coverage only.

Atomic spec areas

Outside the journeys, ~660 atomic specs cover the rest of the surface. The full inventory is in docs/E2E_TEST_COVERAGE.md; below is the summary by area.

Known gaps & backend issues

The Playwright suite is also a forcing function for backend issues. Every gap below has a corresponding active or test.skip test, so the suite flips green automatically when the backend ships the fix.

Deferred — infrastructure / shared-env constraints

• Email-channel notification delivery (requires SMTP / mailbox interception not wired against dev).
• R2 upload failure retry behaviour (requires S3 5xx mocking).
• Slow-network / offline behaviour (requires fault-injection harness).
• Cross-browser (Firefox / WebKit / Edge) — deliberately Chrome-only since the journey redesign.
• Two users editing the same property / shared ad-slot races (not deterministic on shared dev).

Backend gaps surfaced by the suite

Conventions in this suite

• One file per journey. Each journey lives in e2e/tests/journeys/*.journey.spec.ts and is named after the lifecycle it covers.
• Independent tests. Where possible, journey tests are not serial — each creates its own resource and cleans up in afterEach so a single flake doesn't cascade.
• Cross-role via Bearer tokens. Multi-actor journeys use captureBearerForRole to pull a JWT from another role's storageState rather than spinning up a second browser session.
• Unique title prefixes. Test artefacts use [E2E-XX] prefixes (e.g. [E2E-BE] for blog editor, [E2E-ST] for support tickets) so manual cleanup is straightforward if a spec crashes.
• The coverage-report.mjs script walks tests/**.spec.ts and rewrites the auto-generated section of docs/E2E_TEST_COVERAGE.md between markers. This HTML page is the human-friendly mirror — keep both in sync.

References

• Playwright config — e2e/playwright.config.ts
• Auth setup — e2e/setup/auth.setup.ts
• Credentials fixture — e2e/tests/fixtures/credentials.ts
• Live coverage runbook — docs/E2E_TEST_COVERAGE.md
• Backend state machines — backend/src/main/java/com/cplk/api/config/PropertyStateMachineConfig.java