Validation and documentation¶

Find the implementation¶

PyDuck writes molecular and integral files. src/QuAcK/QuAcK.f90 reads the calculation inputs and dispatches to the R, U, G, or B driver. The drivers then call the method-specific directories in src/. Use the Fortran source index to locate individual files and procedure names.

Run the checks¶

The regular CI builds QuAcK and runs the light benchmark suite. Its workflow shows the database and fixture setup. Method-specific scripts in tests/ check GW–Parquet and Faddeev calculations. Read each script's arguments and fixtures before running it.

The four documentation examples are smoke tests. They check that the selected drivers produce finite results and that QuAcK finishes. Benchmark tests are the place for numerical reference comparisons.

Build the manual locally¶

From the repository root, with the PyDuck dependencies installed:

python -m pip install -r docs/requirements.txt
python utils/build_docs_reference.py
python -m mkdocs build --strict
python -m mkdocs serve

The reference generator reads the CLI help, shipped templates, curated method inventory, and tracked Fortran sources. It checks that every method flag has a row and that all inventory source links resolve in the checkout. Generated pages carry the Git revision. Keep the generated files out of commits.

When adding a flag, update the input reader and template, then record the active R/U/G paths and limitations in docs/methods.tsv. For a change to an existing flag, review its row and the relevant method guide. Where possible, record or extend a small test that exercises the normal entry point and checks an observable.

Publish¶

The documentation workflow builds on pull requests and pushes to master. To publish from the repository, choose GitHub Actions under Settings → Pages → Build and deployment. A build from master then deploys the manual from its own commit. The build also provides an HTML artifact for review.