Skip to content

testing.md

Latest commit

 

History

History
47 lines (29 loc) · 2.29 KB

testing.md

File metadata and controls

47 lines (29 loc) · 2.29 KB
Raw

Testing TLEF-CREATE

TLEF-CREATE uses three complementary test layers. Frontend Vitest tests cover components and state, backend Jest tests cover domain and API behavior, and Playwright checks a production frontend build against the real Express API.

Fast local checks

npm test -- --run
npm run test:backend:unit
npm run build

test:backend:unit uses routes/create/jest.unit.config.js and intentionally does not load the Mongo setup. The full test:backend command retains database-backed integration coverage.

Playwright E2E smoke

Install the Chromium binary once:

npm run test:e2e:install

Run the E2E suite:

npm run test:e2e

The Playwright configuration builds the Vite frontend, serves it on port 8092, starts Express on port 8051, authenticates an isolated e2e-admin session, creates a course through the browser, verifies the API and navigation result, and deletes that course afterward.

The backend launcher never intentionally uses the normal application database. It uses E2E_MONGODB_URI when provided; otherwise it derives a database name ending in -e2e from MONGODB_URI. It refuses to start if the resolved database name does not contain e2e.

Authentication uses /auth/auto-login only while AUTO_LOGIN_ENABLED=true. The E2E launcher enables it inside the isolated test process. Public deployments should leave the variable false. Playwright storage state is written under playwright/.auth/ and is ignored by Git.

On failure, Playwright retains the screenshot/video and records a trace on the first retry. Open the report with:

npx playwright show-report

What belongs in E2E

Use Playwright for browser/API contracts: authentication and roles, tab navigation, Blueprint save/restore, failure messages, SSE progress, preserving existing questions after a failed save, review, preview, and downloads.

Do not make pull-request E2E tests depend on a live commercial LLM. Model output is nondeterministic, rate-limited, and may incur cost. Test model streaming, retry, parsing, and budget handling with deterministic backend contract tests. Run a small fixed real-model workflow separately as a scheduled staging canary.

Start with Chromium for pull requests. Add Firefox/WebKit as a scheduled or release suite after the core smoke is stable.