Skip to content

Tested documentation candidate

The learning paths were tested on macOS arm64 with Python 3.13.12 and uv 0.10.10. The direct wheel and the wheel rebuilt independently from its source archive have the same SHA-256:

812bd814329321ec3d2e6573e9a7191d59bd29f38a4f5ee4fea70f519893cd1c

The artifact is get_nick_done-0.1.0-py3-none-any.whl. Its uncompressed payload is 6,116,293 bytes. Each origin was installed in a fresh environment outside the repository and ran the documented dispatch, release, BookRec, expense and API examples: 122 offline blocks per origin, including external-runtime preparation and supplied checkpoints. The source archive's SHA-256 is:

0914a1a1741577524d97974ba21a5bbd4b99cf1748505c6feb7acc7bf9adc1a8

The native learning paths were also executed with local PyRel 1.30.1 on DuckDB 1.5.5 in a separate locked environment, Soufflé 2.5 (64-bit), and Lean 4.34.0. Native results and GND verdicts remain separate evidence; the generated Lean bundle retains its unproved metadata.

These identities bind the tested package, including the supplied synthetic resources. A CI preview also carries SOURCE_COMMIT, identifying its documentation source revision. Build your own candidate when either source changes; a matching version number alone does not identify tested bytes.

This is focused documentation and application validation. Final candidate sign-off and publication remain separate release steps. The support table distinguishes measured profiles from metadata compatibility and optional native/provider workflows.