Provenance and archive policy

Authoritative history

The consolidated repository continues the published asthanaa/autogen history. Work from the disconnected autogen2 checkout was imported as reviewed content; its unrelated root history was not merged.

The clean projected-AGP QPCCSD/PAV package originated in autogen2 commit 2b881e031a4b1fddcb3b36f9c42f463a4428195f. The scalable predecessor was 970dc4ac6291048c85bfd3a198196422ee928d42. These identifiers establish source provenance without making either snapshot an importable package inside the final tree.

Excluded generated plan

The unpublished legacy commit 0a851d3469a8476aa36d01b3437021c9c2815ace contained a 387,064,269-byte projected-codegen plan. That blob exceeds normal GitHub limits and is not needed by the canonical production QPCCSD/PAV runtime. It is excluded rather than rewritten into the consolidated branch.

Its path, byte size, SHA-256, legacy Git location, and regeneration command are recorded in the repository file provenance/excluded_large_objects.json. This compact record is authoritative for artifact identity; copying the plan back into the repository is prohibited.

Campaign and calculation archives

Raw molecular calculations remain on approved remote systems or in separately managed archives. Preserve the configuration, code commit, result/checkpoint hashes, host, dependency versions, scheduler receipt, residual certification, and projection-grid validation together. Do not commit whole run directories.

To promote a result into the test suite, create a compact fixture that records those identifiers and only the values required by the regression contract. Never replace an expected energy solely to make a test pass.