Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
60 lines (44 loc) · 2.53 KB

CONTRIBUTING.md

File metadata and controls

60 lines (44 loc) · 2.53 KB
Raw

Contributing to Cecelia Feijoa

Thanks for your interest. Cecelia Feijoa is early software and is primarily maintained by its author, Dominik Schienstock. Contributions — bug reports, fixes, and ideas — are welcome. Please read this first.

Before you start

  • Questions / usage help → check the FAQ first.
  • Bugs and features → use the issue templates.
  • Larger changes → open an issue to discuss the design before writing code, so effort isn't wasted on something that doesn't fit the architecture.

Architecture is non-negotiable

The layer boundaries are deliberate and enforced: analysis logic lives only in the headless Cecelia.jl package — never in the API or the Vue frontend — and there is no fourth language. Read docs/ARCHITECTURE.md and CLAUDE.md before proposing structural changes. Departures from the ported R behaviour should be deliberate and documented, not incidental.

Development workflow

The full workflow — branching, commits, PRs, and releases — is documented in docs/DEV.md. In short:

  • Branch off the latest main with a conventional prefix (feat/, fix/, docs/, chore/, refactor/). Never commit or push to main directly — everything lands via a pull request.
  • Keep a branch scoped to one logical change.
  • Every change ships with tests and docs. The package must stay verifiable headlessly from the Julia REPL, without the GUI.

Running the tests

Tests run per layer via Pixi (all of these run in CI):

pixi run test-pkg       # Cecelia.jl package (headless)
pixi run test-api       # Julia API adapters
pixi run test-frontend  # frontend logic (Vitest)
pixi run test-py        # Python analysis env
pixi run test-mcp       # MCP observer client

See docs/INSTALL.md for the developer environment setup.

A note on AI-assisted development

This project was built almost entirely with an AI assistant under human direction — see How this software was built. That transparency expectation applies to contributions too: AI-assisted PRs are fine, but you are responsible for the correctness of what you submit, and scientific/visual claims must be validated by a human, not asserted by a model.

License

By contributing, you agree that your contributions are licensed under GPL-3.0-or-later, the license of this project.