v0.1.1 — Cross-version reproduction check

This patch fixes the release verification check across Python versions.

Python 3.12 produced last-digit floating-point differences while all scientific tests passed. The checker now requires exact structure, counts, strings, and source hashes, with a 1e-12 relative/absolute tolerance for finite floating-point results. Two regression tests cover rounding differences and meaningful changes.

The 18-case dataset, 21,500 readings, 58 differing-rate timestamps, scientific calculations, and report findings are unchanged. All 19 tests pass locally.

Report · Project brief

Formal subgroup recognition and confirmed collaborators remain pending.

v0.1.0 — RadLab diagnostics and annotation checker

First release of Space Radiation Measurement Reliability.

• Reproducible audit of 18 DosTel1/DosTel2 instrument-day snapshots: 21,500 readings and 58 timestamps with differing rate values.
• Diagnostics remain available for all series; 16 are excluded from numerical integration.
• Reusable SAA interval checker with synthetic examples and input/output options.
• Exact source queries, preserved public CSV bytes, and SHA-256 hashes.
• 17 tests passed locally and from a fresh public clone; offline reproduction matches the published results.

Start with the report or project brief. Python 3.9+; no third-party dependencies.

These are exploratory diagnostics and numerical comparisons of reported rates, not validated physical-dose estimates. Formal AWG subgroup recognition and confirmed collaborators remain pending.