Skip to content

openWFN 0.10.1

Release date: October 2, 2026. These historical notes describe the 0.10.1 packaging and reproducibility fixes. Formulas, numerical kernels, validation tolerances and data schemas were unchanged.

What changed

  • Replace development-era validation prose with a current-state evidence page.
  • Extend the canonical validation/manifest.json so current capability status, historical captures and resource-benchmark meaning are machine-readable in one place.
  • Mark the earlier three-molecule regression snapshot and grid-ESP convergence capture as historical evidence rather than current status authorities.
  • Package the complete eleven-molecule everyday-QC Molden corpus with the wheel and install it under everyday-qc/ through the existing openwfn examples install command.
  • Make the resource benchmark consume an explicit example directory and hash the actually imported openWFN package, allowing built-wheel and public-PyPI verification to test the installed artifact rather than repository source files.
  • Resolve explicit relative example-directory paths to absolute paths before per-workflow temporary working-directory changes, so installed-corpus verification uses the intended 11 inputs.
  • Require the complete eleven-input corpus and all 99 expected workflow records; an empty or incomplete corpus cannot pass.
  • Derive release version, tag and release-note path from package metadata rather than embedding a previous version in the publication workflow.
  • After publication, install openwfn[interop,resources] from public PyPI and rerun the same installed-corpus real-workflow benchmark used before publication.

Resource evidence wording

The release gate records that 99/99 prescribed CLI workflows completed successfully within resource limits; scientific result status is evaluated separately.

Those 99 runs are eleven real molecular inputs multiplied by nine prescribed command workflows. They test command completion and bounded resource behavior. They do not convert a partial or Experimental scientific result into a Validated one. Capability-specific reference tests and per-result diagnostics remain authoritative for scientific status.

Installed reproduction corpus

python -m pip install "openwfn[interop,resources]==0.10.1"
openwfn examples install installed-examples
python scripts/benchmark_resources.py \
  --examples-dir installed-examples/everyday-qc \
  --output resource-report.json

The packaged Molden files are byte-identical copies of the versioned examples/everyday-qc/ inputs used by repository validation.

Scientific scope

The scientific scope remains the documented 0.10 everyday-QC scope. Successful orbital-composition, Mayer, DOS, PDOS and native point-ESP results retain their documented validation status and reference boundaries. Conditional failures in normalization, conservation, conditioning or convergence remain partial or Experimental instead of being promoted by the package-level stable label.

Native Hirshfeld, spectroscopy/excited-state workflows, NTOs, analytic density derivatives and advanced topology/basin methods remain outside this release.