Fairbeam is developed with AI coding assistants, and AI-assisted contributions are welcome. This page says how the project uses AI and what we ask of contributors who do.
How Fairbeam is made
- The maintainer directs. İsmail Akdağ (opens in a new tab) maintains Fairbeam and decides what is built and what is merged.
- Much of the work is AI-assisted. A large part of the code, tests and documentation is written
with AI coding assistants. Commits written that way carry a
Co-Authored-By:trailer that names the model. - Every change is checked before it is merged. It passes the repository's automated checks
(
npm run check:*, the Python test suite) and is reviewed. User-facing changes are also tried in the app. - Numbers come from runs, not from models. Results, validation figures and timings in the documentation come from simulations that anyone can repeat. A project bundle records the generator, the versions and the run settings. A statement by an AI model is never evidence.
- The maintainer is responsible for everything in the repository and in the releases, whoever or whatever wrote it.
Contributing with AI assistance
You may use any suitable AI tool or model. If you do:
- Say so in the pull request. Name the tool and the model, and say what it did: wrote the code, the tests or the text, or only reviewed or explained. The PR template asks for this.
- Test it. Run the relevant checks and say what you ran and what you saw. For changes to physics or numerics, attach a reproducible comparison: a model, its settings, and the result before and after.
- Keep it focused. One concern per pull request. Large unreviewed rewrites, mass reformatting and drive-by "improvements" generated in bulk will be closed.
- Respect licences and data rights.
- Fairbeam is GPL-3.0-or-later. Don't submit code that reproduces third-party code under an incompatible licence.
- Don't submit data you have no right to share, such as results exported from licensed commercial software, private measurements or other people's designs.
- Don't paste other people's private information into AI tools.
- Check what you post. Read every issue, comment and review an AI wrote for you before you post it. Automated or bulk-generated issues, comments and pull requests will be closed.
Maintainers may close a contribution that doesn't follow these rules, and may ask for changes before reviewing. Coding agents working in this repository also read AGENTS.md (opens in a new tab).
Security
Report security problems privately to the maintainer, not in a public issue. Contact details are at https://akdag.dev (opens in a new tab).