Add deterministic reproducibility test for layered admissibility artifact

[ProfRandom92]

Summary:
Add a focused deterministic reproducibility test that regenerates committed benchmark artifacts via the existing layered admissibility generator and fails on any JSON drift; the check is implemented as an explicit, extensible artifact spec list so future artifacts can be added deliberately.

Changed files:

• tests/test_artifact_reproducibility.py (new)

Testing:

• python scripts/generate_layered_admissibility_artifact.py — produced artifacts/layered_admissibility_results.json successfully.
• pytest tests/test_artifact_reproducibility.py -q — 1 passed.
• pytest tests/test_degradation_curve_generator.py -q — 13 passed.
• pytest tests/test_fixture_manifest.py -q — 8 passed.
• npm run check — completed; full Python test suite ran and reported 197 passed.

Risks:

• Low risk: this PR only adds a deterministic test and does not change generators, fixtures, or committed artifacts; the test uses logical JSON equality and will need explicit additions to the spec for any future deterministic artifacts.

Next:

• If new deterministic benchmark artifacts are introduced, add them to ARTIFACT_SPECS in tests/test_artifact_reproducibility.py to include them in the reproducibility gate.

---

Codex Task

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
comptextv7	Ready	Preview, Comment	May 19, 2026 3:07pm

[gemini-code-assist]

Code Review

This pull request introduces a new test suite to verify the reproducibility of committed artifacts, specifically for layered admissibility results. The feedback focuses on improving the robustness and maintainability of the tests by suggesting the use of a repository root constant for path resolution and refactoring the test logic to utilize pytest's parametrization for better isolation and reporting.

[gemini-code-assist]

Import pytest and define REPO_ROOT to allow for robust path resolution relative to the repository structure, ensuring tests are runnable from any directory.

import pytest

REPO_ROOT = Path(__file__).resolve().parents[1]

[gemini-code-assist]

Use REPO_ROOT to anchor the artifact path. This prevents the test from failing if executed from a directory other than the repository root.

Suggested change

	committed_path=Path("artifacts/layered_admissibility_results.json"),
	committed_path=REPO_ROOT / "artifacts/layered_admissibility_results.json",

[gemini-code-assist]

Refactor the test to use pytest.mark.parametrize. This improves test isolation and provides clearer reporting by treating each artifact as a separate test case, which aligns with the extensible design mentioned in the PR description.

Suggested change

	def test_committed_deterministic_artifacts_are_reproducible(tmp_path: Path) -> None:
	for spec in ARTIFACT_SPECS:
	regenerated_path = tmp_path / spec.committed_path.name
	spec.regenerate(regenerated_path)
	regenerated_payload = _load_json(regenerated_path)
	committed_payload = _load_json(spec.committed_path)
	assert regenerated_payload == committed_payload, (
	f"Deterministic artifact drift detected for '{spec.name}'. "
	f"Regenerate with: python scripts/generate_layered_admissibility_artifact.py"
	)
	@pytest.mark.parametrize("spec", ARTIFACT_SPECS, ids=lambda s: s.name)
	def test_committed_deterministic_artifacts_are_reproducible(spec: DeterministicArtifactSpec, tmp_path: Path) -> None:
	regenerated_path = tmp_path / spec.committed_path.name
	spec.regenerate(regenerated_path)
	regenerated_payload = _load_json(regenerated_path)
	committed_payload = _load_json(spec.committed_path)
	assert regenerated_payload == committed_payload, (
	f"Deterministic artifact drift detected for '{spec.name}'. "
	f"Regenerate with: python scripts/generate_layered_admissibility_artifact.py"
	)