Omni is an open source runtime under active debugging. Please do not assume a path is healthy only because it returns a response. A useful contribution is one that improves clarity, reproducibility, observability, or reliability.
- Python 3.11+
- Node.js >=24.15 <25
- Rust toolchain
npm installOptional subproject installs:
pip install -r backend/python/requirements.txt
pip install -r omni-training/requirements.txtRun the Node-side tests:
npm run test:nodeRun the Python-side tests:
npm run test:pythonRun everything:
npm testRun the Python entrypoint:
python backend/python/main.pyRun the Rust API:
cargo run --manifest-path backend/rust/Cargo.tomlPlease keep PRs small and focused. Use the following branch naming convention:
feat/p5e-*— new features or phase itemsfix/p5e-*— bug fixesrefactor/p5e-*— refactoring without feature changeschore/p5e-*— CI, tooling, documentation
Never push directly to main. Always open a pull request and wait for CI to pass.
Recommended flow:
- Create a branch for one problem only.
- Add or update tests when behavior changes.
- Update documentation when the runtime behavior, contract, or repo structure changes.
- Explain what evidence you used to justify the change.
- Do not commit secrets, credentials, or local
.envvalues. - Do not remove failing tests just to make CI green.
- Do not hide degraded behavior behind vague success messaging.
- Do not rewrite large runtime subsystems without evidence and scoped tests.
- Prefer additive changes over breaking public contracts.
- If a change affects Rust/Python/Node boundaries, mention that explicitly in the PR.
- Rust/Python/Node runtime boundary debugging
- execution-lane reliability
- observability truthfulness
- test coverage around real runtime paths
- contributor onboarding and developer setup clarity
- documentation cleanup and public-facing explanations
- docs/public-debug/PROJECT_STATUS.md
- docs/public-debug/REPRODUCTION.md
- docs/public-debug/CONTRIBUTOR_TASKS.md
- docs/audits/brain-runtime-flow-map.md
- docs/architecture/runtime-flow.md
If you change behavior, run the most relevant tests you can locally and mention what you ran in the PR. If you cannot run a full validation, say what was skipped and why.