Metriplane v0.4.0.post2

Metriplane v0.4.0.post2

Metriplane v0.4.0.post2 is the corrected publication identity for the reduced Truth Recovery core release. It adds no product capability or assurance claim beyond that qualified scope.

Release identity:

• Release commit: 98ad98c57198aaa78bbf8e9d94c699cd02c72da7
• Release tree: 23178d0e9e9d571a198dc692b3d74fbbfe1eed2b
• Annotated tag object: b070c7563cae51d2889f2cea7fef3d6be79d5115
• Wheel SHA-256: 2e206db012b16d2021ae98fc52a03a831d99746c4e4ff9a4bc7eefa5102a7f38
• Source-distribution SHA-256: 579d2578126fbf3cad07e20e365549cda2d17f49b4bb7b57695fca778f14c0ba
• SHA256SUMS SHA-256: 487f0fa8061d3fbb08654ce84bbab5d7dea4415b292f7e2714d45db9872ef30f
• Tag/TestPyPI workflow: 34043220748 — successful
• Production workflow: 34045289769 — publication and production verification succeeded; its original final reconciliation job failed and was later reconciled through the controlled broker recovery described below
• DOI: none. Do not substitute the frozen v0.2.0 DOI.

Qualification and publication

The exact candidate passed the canonical locked toolchain, Ruff lint and format, strict mypy, tracked compilation, strict documentation, generated-inventory currentness, supported Linux and macOS gates, package metadata and distribution inspection, independent wheel and sdist installs, pip check, version/help/doctor/demo checks, and the complete installed-wheel evidence-to-regression proof with deterministic repeat and the declared tampered-bundle negative case.

• Canonical Linux complete suite: 2,636 passed / 16 skipped / 0 failed / 0 errors
• Supported macOS complete suites: 2,635 passed / 17 skipped / 0 failed / 0 errors
• Governed public census: 10,091 public rows / 2,293 public APIs / 3,580 manifest keys

The annotated-tag workflow built exactly one wheel and one source distribution, recorded the attached SHA256SUMS, published those retained bytes to TestPyPI, and verified a fresh staged install. The owner-only production workflow reused the same retained bytes. PyPI exposes exactly the wheel and sdist hashes listed above, and an independent fresh PyPI installation passed package checking, exact version/help, doctor, the camera-free demo, evidence-bundle verification, and generated regression execution.

After PyPI publication and production verification, GitHub exposed later generated cleanup steps as pending at the exact final reconciliation boundary. The broker failed closed and quarantined the existing publication lease; no second upload, rebuild, tag change, or release identity was created. A narrowly scoped, independently qualified broker repair re-proved the exact workflow, tag, main, state, rulesets, retained hashes, live PyPI bytes, and fresh install before deleting only the exact lease ref and updating the same App check to canonical released/success. The same repair then merged through protected main as PR #105. Workflow 34045289769 is therefore retained accurately as publication-successful but not globally green.

Install

python -m pip install "metriplane==0.4.0.post2"
metriplane demo

Scope and limitations

Metriplane accepts recorded, already-estimated state and supplied process rules. It does not establish physical source accuracy, certify safety, control a robot or machine, or establish generic compatibility with arbitrary robotics data. MP2-007 and MP2-014 through MP2-017 remain deferred to v0.4.1 Assurance Hardening. The frozen v0.2.0 research artifact and DOI remain separate and unchanged.

The immutable v0.4.0 tag records a qualification failure before registry publication. The immutable v0.4.0.post1 candidate reached verified TestPyPI staging but was retired unpublished after a required protected-main broker repair. Neither prior identity has a production package or GitHub Release, and none of their artifacts was reused for post2.