Bootstrap the Graphcal QA corpus

[shunichironomura]

Summary

• establish corpus governance, privacy rules, provenance guidance, and the versioned manifest contract
• add strict structural validation with Pydantic, tests, read-only CI, and pre-commit.ci
• add Typer-based validate-corpus and test-graphcal commands with versioned JSON reports
• enforce typed health-only, stability-baseline, and reference-backed expectations
• add one minimal active example of each expectation class
• run validation for stacked pull requests

Provenance and review

• I classified each project as a source-backed reproduction or original synthetic complexity case.
• I performed a human privacy review and found no private project names, paths, source text, inputs, or results.
• I confirmed that all included material may be redistributed.
• I classified each expectation as reference-backed, stability-baseline, or health-only.
• Reference-backed values were independently calculated without relying on Graphcal output; human verification remains pending.
• I recorded producing revisions, exact semantic comparison rules, and five-run determinism evidence for the stability baseline.
• I recorded candidate-generation provenance; human review remains pending.
• I ran uv run validate-corpus and uv run -m pytest.

Evidence

• minimal-mass-budget is an original synthetic health-only infrastructure case.
• modular-power-budget is an original synthetic multi-file stability baseline. Its evidence records Graphcal and corpus revisions, checked-in inputs, exact comparison rules, and five identical canonical SHA-256 results.
• ideal-rocket-equation is a source-backed reproduction of the Tsiolkovsky equation. The independent Python calculation and assumptions cite NASA Glenn’s “Ideal Rocket Equation” and CGPM Declaration 2 (1901), DOI 10.59161/CGPM1901DECL2E.

Local verification:

• uv run -m pytest — 39 passed
• uvx pre-commit run --all-files
• uv run validate-corpus
• uv run test-graphcal "$(command -v graphcal)" --root . — all 3 cases passed