Design Decisions

Safety-First Release Strategy

• Public PyPI package is a non-functional educational stub.
• Functional source remains in the repo, with strict safety gates:
• Requires PE_PACKER_ALLOW_PACKING=1 and --force.
• Dry-run prints entropy/size and exits 0 without writing outputs.
• CI enforces these gates via python/tests/test_safety.py.

Tech Stack

• Rust core with PyO3 bindings (abi3) for stable wheels and performance.
• Python layer for CLI/UX, data orchestration, and training workflows.
• maturin for mixed Rust/Python builds; uv for dependency management.

CLI UX (Typer)

• Subcommands: pack, generate-training-data, analyze-dataset, validate.
• Clear error messages, explicit options, and verbose logging mode.
• Gating implemented at the CLI layer for least surprise.

Dataset Generation Philosophy

• Focus on educational, controlled variants.
• Record rich metadata for ML evaluation and reproducibility.
• Provide analyze-dataset to examine distributions and coverage.

Policy Alignment

• SECURITY.md defines purpose, ethics, and contact.
• LICENSE-APPENDIX.md clarifies educational use and liability.
• CI blocks unsafe publish flows; the stub workflow checks for native content.