Running the Tests

The project ships with tests on both sides: a Python suite for the backend (the data layer, pipeline, and API) and a JavaScript suite for the dashboard. They run independently, so you can run whichever side you're working on.

What's Covered

The backend tests exercise the parts where the real logic lives: the Postgres data layer (the datetime contract and the zip export), the ingestion poller, the per-student workers, the APTED edit-distance computation, the five trigger rules and their history, identity-switch detection, the episode segmenter, the block humanizer, the live-stream change signal, the failed-write outbox, the auth gate, the workspace/per-browser isolation and the daemon's fan-out, and every API endpoint through a test client.

The frontend tests cover the dashboard's formatting helpers and render the real component against a mocked API to check that students, alerts, the trigger-history grid, the toggle buttons, and the resilient-write path (retry, outbox park, red toast) behave.

You don't need to read every case to trust them. If they're green, the contract the daemon and dashboard depend on is intact.

Backend (pytest)

The test tools (pytest, pytest-cov, httpx) are in requirements.txt. The suite needs a Postgres it can reach — it connects using DATABASE_URL with the database name swapped to lm_dashboard_test (override with TEST_DATABASE_URL), and it truncates that database between tests, so your real data is never touched.

Point DATABASE_URL at a local Postgres (the prod compose DB is internal, so for running tests use a Postgres exposed on a host port — e.g. a small postgres container on localhost:5432), create the test database once, then run from the repo root:

createdb -O lmdash lm_dashboard_test    # once
pytest

For a coverage report:

pytest --cov=app --cov-report=term-missing

The suite is quick: there's no ML model to load and no heavy scientific dependencies to import, just FastAPI, psycopg, and APTED.

Frontend (vitest)

From the frontend/ directory:

cd frontend
npm install
npm test

That runs the suite once and exits. During development you can use npx vitest instead to keep it watching and re-running on save.