Contribute to Matemium

Matemium welcomes code, documentation, testing, example projects, translations, bug reports, and carefully described visual failures.

Before contributing

Read:

• Architecture overview
• Source-available license
• Repository contribution guide

The license permits forks made to prepare contributions to the official project. It does not grant unrestricted redistribution of derivative applications.

Sustain the project

Matemium is a non-commercial project without product income. Financial contributions help cover distribution infrastructure, development tools, testing, and the work required to keep the engine and desktop application available.

The engine can render locally without paid AI. The most agentic authoring experiences and deeper automation, however, require model inference and supporting AI resources that the project’s independent maintainer cannot currently fund alone. Support helps Matemium develop and test those experiences without turning the project into a commercial product.

Support Matemium →

You can also contribute meaningfully without money: report defects, improve documentation, test releases, create cross-subject examples, or submit focused code changes.

Development setup

git clone https://github.com/fargonee/math.git
cd math
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Install FFmpeg and LaTeX for render tests.

Run tests

pytest

Optional render tests:

pytest -m slow

Contribution boundaries

1. Keep the engine generic.
2. Put subject-specific composition in project helpers.
3. Make new engine features reachable through CanvasBuilder.
4. Preserve the desktop → Rust → sidecar boundary.
5. Do not make the server import the rendering engine.
6. Include tests for new public behavior where practical.
7. Document public API changes.

Contribute a project recipe

A strong example project:

• Answers a clear conceptual question
• Demonstrates a capability intentionally
• Uses generic engine primitives
• Keeps subject-specific helpers beside the project
• Renders in a supported quality
• Includes accurate content
• Provides a useful poster frame and short public preview

Cross-subject projects are especially valuable because they test whether the visual reasoning model generalizes beyond its mathematical foundation.

Report a visual defect

Attach:

• Minimal project source
• Render timestamp
• Screenshot or short recording
• Expected result
• Platform and engine version

Visual evidence is more useful than “the animation looks wrong.”

Commit style

Use clear, imperative subjects:

• Add viewport-safe zoom cap
• Document multiple-tape authoring
• Fix labels during solid inspection

For substantial design changes, open an issue before building a large solution.