[CI and testing]

CI and testing

Abstract

AXIS quality gates live primarily under frontend/ (Bun tests) and frontend/worker/tests/, orchestrated by GitHub Actions (.github/workflows/ci.yml, axis-nightly.yml). Deep test conventions are documented in Testing reference and frontend/TESTING.md.

Conceptual model

Interface surface — local commands

cd frontend
bun run test                 # tests/ + worker/tests/
bun run test:unit
bun run test:coverage:gate   # lcov + scripts/check-coverage.mjs 95
bun run test:security
bun run test:e2e:smoke
bun run test:e2e:critical
bun run test:e2e
bun run test:all             # gate + security

Worker:

cd frontend/worker
npm run typecheck

Repo Makefile:

make test-frontend
make worker-typecheck

CI workflows

ci.yml (excerpted jobs)

axis-nightly.yml

Schedule: 0 4 * * * UTC + workflow_dispatch.

Internals

Coverage policy (summary)

• Gate: 95% lines on a scoped core (plugins, storage minus idb, store, results, sources, streams, data, selected chart helpers, worker auth/keys/runtime/scripts).
• Excluded from gate: heavy chart UI, pyodide boot, legacy JS, Solid .tsx surfaces (unit-tested elsewhere or deferred).

E2E design

• Builds production preview; baseURL http://127.0.0.1:4173.
• Mocks /run and Binance where possible so smoke does not need live Pro API.
• Selectors use data-testid (axis-btn-load, axis-manager, …).

Invariants & edge cases

1. No real network in unit tests — mock fetch / WebSocket.
2. Import ./setup first when touching store/plugins.
3. CI Bun version pinned in workflows (e.g. 1.3).
4. Playwright browsers installed with --with-deps on CI.

Failure modes

See also

• Testing reference
• Local dev