Testing and verification

The commands below cover the usual scopes. They are a menu rather than a requirement to run every suite.

Common checks

python -m pytest tests
python -m pytest tools\rhythm-lab\tests tools\audio-dedup\tests tools\audio-online\tests scripts\tests
npm --prefix .\frontend test
npm --prefix .\frontend run typecheck
npm --prefix .\frontend run build
npm --prefix .\docs\dj-track-similarity run check
git diff --check

The root Pytest configuration targets only tests/. Four other suites need naming on the command line:

Suite	Files	Covers
tools/rhythm-lab/tests	3	labelling, training, promotion, label transfer, Lab consumers
tools/audio-dedup/tests	4	CLI output, fingerprint candidates, the fingerprint benchmark, the spectral check
tools/audio-online/tests	11	config, auth, CLI, inputs, matching, sources, report, workbook format, HTTP
scripts/tests	7	Audio Doctor repair, database optimization, database QA, the LAN launcher, dedup, the CLAP checkpoint embedder, the search benchmark

Frontend tests run with node --test tests/*.test.mjs from frontend/. One of them, frontend/tests/testsExecuteCode.test.mjs, is a meta-guard rather than a feature test: it fails when a test asserts on the text of a source file instead of driving the module. Read it before adding a frontend test.

Use the ml, slow, and evaluation markers only when the changed behavior requires those optional dependencies or long-running paths.

npm run check runs strict Vale style checking for README.md plus the VitePress Markdown tree and the site build. Run npm run vale:sync once after a fresh checkout or when .vale.ini packages change. Use npm run lint:style when you want the same style report without failing the command.

Focused examples

• Audio Doctor: scripts\tests\test_repair_audio_metadata.py. Audio Doctor has no API route and no API test.
• Audio Dedup: scripts\tests\test_audio_dedup.py, tools\audio-dedup\tests, and tests\test_api_audio_dedup.py.
• Analysis staging: tests\test_sonara_staging.py and tests\test_ml_staging.py.
• Text search: tests\test_api_text_search.py, which also pins the warmup and feedback routes.
• Database validation: tests\test_api_database_validation.py and tests\test_database_validation_jobs.py.
• Supported formats: tests\test_supported_audio_formats.py, which pins all 14 extensions.
• Rhythm Lab: tools\rhythm-lab\tests\test_rhythm_lab.py.
• SONARA runtime and storage: tests\test_sonara_features.py, tests\test_sonara_native_batch.py, and tests\test_sonara_storage.py.
• Tempo, Camelot, and transitions: tests\test_tempo_resolution.py, tests\test_track_resolution.py, and tests\test_transition_diagnostics.py.
• Classifier compatibility: tests\test_classifier_productionization.py, tests\test_break_energy.py, and tools\rhythm-lab\tests\test_rhythm_lab.py.

Safety

Automated model, audio, and database tests use temporary SQLite and WAV fixtures plus stubs. They must not use project databases, source music files, downloaded model runs, or Audio Doctor and Audio Dedup apply modes.

Clean-machine backstop

.github/workflows/ci.yml runs the same suites on empty Windows runners. It is manual only, started from the Actions tab, and its scope is the base install plus dev and rhythm-lab. The ml extra stays out, so torch, CUDA wheels, and model checkpoints are absent and the tests that need torch skip themselves. Use it before a release, or when something works locally and you suspect the local environment is why.

Manual behavior check

After a source behavior change, exercise the matching surface. Open UI work in a browser and send a live HTTP request for API work. Invoke CLI commands directly, or use a minimal import driver for library code. Cover one happy path and one relevant failure path. Automated checks support this final exercise but do not replace it.