Thanks for your interest! This project values measured claims above all:
every performance or savings number in the docs is generated by a script in
tests/ and lands in reports/ with raw data. Contributions
are held to the same bar.
- No unmeasured claims. If your change affects tokens, latency, cost or quality, include the harness (or extend an existing one) that measures it.
- Honest limits. If your feature has a failure mode, document it the way
benchmarks/ab-publico/results/RESULTS.mdreports refusals and quality losses — visible, not buried. - The kernel stays lean.
bioma_microis deliberately two primitives (hormonal bus + context apoptosis). New primitives need strong evidence.
git clone https://github.com/jonathascordeiro20/bioma-framework
cd bioma-framework
python -m venv .venv && . .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -e ".[all]" maturin pytest
cd bioma_micro && maturin develop --release && cd ..
pytest tests/test_kernel.py tests/test_gateway.py -q # fast, offline suiteOnline benchmarks (test_quality_preservation.py, e2e_claude_code.py) need
OPENROUTER_API_KEY in .env and spend real money —
they are opt-in, never run in CI.
- One logical change per PR; reference an issue when one exists.
- CI (
ci.yml) must be green: it runs the offline test suite on the three OSes. - Rust code:
cargo fmt+cargo clippy -- -D warningsinsidebioma_micro/. - Python: keep the existing style (type hints, docstrings with the why).
Bump versions (bioma_micro/Cargo.toml and/or root pyproject.toml), update
CHANGELOG.md, then git tag vX.Y.Z && git push origin vX.Y.Z — the
release.yml workflow builds abi3 wheels for the three OSes and publishes to
PyPI via Trusted Publishing (no tokens).