Testing
Each stack has its own test runner. There is no unified “pnpm test” — root pnpm check covers JS/TS lint+format+types, but not Python tests and not the Rust runner.
TS workspace (Vitest)#
Vitest runs per-app and per-package where tests exist.
# Per app / package:
pnpm --filter web test
pnpm --filter live test
pnpm --filter @pi-dash/ui test
# Across workspace via Turbo:
pnpm turbo run testTest files live next to source (*.test.ts / *.test.tsx) or in app-local tests/ dirs (e.g. apps/web/tests/, apps/live/tests/).
Storybook#
@pi-dash/ui ships Storybook for component-in-isolation development:
pnpm --filter @pi-dash/ui storybook # http://localhost:6006Build new components in Storybook first, wire them into an app second.
Django backend (pytest)#
apps/api/pytest.ini configures:
--reuse-db --nomigrations— DB persists between runs; migrations are skipped.- Markers:
unit | contract | smoke.
cd apps/api
./run_tests.sh # wraps tests/run_tests.sh
python run_tests.py -u # unit only
python run_tests.py -u -o -p # unit + coverage + parallel (-n auto)
python run_tests.py # full suite
pytest -m smoke # direct pytest, smoke marker only
pytest --create-db # rebuild DB — required for schema-altering changes
pytest apps/api/pi_dash/runner/tests/test_session.py::test_welcome_frame # one testMarkers#
| Marker | Meaning |
|---|---|
unit | Fast, isolated. Default for new tests. |
contract | Cross-module contract tests — wire-format, serializer parity. |
smoke | Coarse-grained end-to-end happy-path. |
Tests live under apps/api/pi_dash/tests/ and per-module tests/ (e.g. pi_dash/runner/tests/).
Test settings#
apple_pi_dash.settings.test is the test settings module — it tweaks middleware and disables some prod-only services so tests don’t need RabbitMQ etc.
Rust runner (cargo test)#
cd runner
cargo test # unit + integration
cargo test --release # release mode (slower compile, runtime closer to shipped binary)
cargo test test_name # filter
cargo clippy -- -D warnings # lint as a quality gateTest layout:
- Unit tests — inline
#[cfg(test)] mod tests { ... }in each module. Deterministic, table-driven where possible (protocol serde, approval policy, reconnect backoff, workspace resolve, config roundtrip). - Integration tests —
runner/tests/protocol_roundtrip.rsand friends. Every client/server protocol variant round-trips; router state machine invariants get exercised.
Manual QA matrix (per release)#
Per runner/README.md, every runner release is manually QA’d:
macOS arm64 / x64 + Linux x64 + Windows x64 → first-run
configure→install→start→ TUI shows connected → synthetic run via/api/runners/runs/→ approval prompt → decision.
Quality gates summary#
| Gate | Command | When it runs |
|---|---|---|
| TS format | oxfmt | Pre-commit (lint-staged) + pnpm check:format |
| TS lint | oxlint --fix --deny-warnings | Pre-commit + pnpm check:lint (per-app --max-warnings ceilings) |
| TS types | tsc --noEmit | pnpm check:types (per-package) |
| TS unit | vitest | pnpm test / per-package |
| Python format | ruff format | Manual (not wired to root) |
| Python lint | ruff check | Manual |
| Python tests | pytest | ./run_tests.sh |
| Rust lint | cargo clippy -- -D warnings | Manual |
| Rust tests | cargo test | Manual |
Everything is run in CI on PRs — the --deny-warnings and -D warnings settings make sure no warning slips through unreviewed.
Where to read next#
- 13 — Development workflow — the full command reference
apps/api/tests/andapps/api/run_tests.py— Python test harnessrunner/tests/— Rust integration tests