Contribute to Simjecture

Bring a scientific question, a reliable tool, or one improvement that makes the next investigation easier to run and assess. You do not need a large GPU, paid model access or a full research campaign to make a useful contribution.

Find a place to start

You bring…	A useful first contribution	Start here
Domain knowledge	Review a diagnostic, numerical control or claim in a recorded study	Recorded investigations
A scientific solver	Provide a working example, readable outputs and a bounded validation case	Add a capability
An experimental workflow	Commission it through the actual launcher, with inputs and diagnostics	Guided commissioning
A research question	Write the hypothesis, scope and a test that could falsify it	First investigation
Python or web experience	Reproduce a bug and add a focused fix with a regression check	Repository map
A fresh reader’s perspective	Fix an unclear setup step or improve an explanation	Documentation development

Open an issue to discuss a contribution. For a small, clear fix, a pull request with the problem, change and validation is enough. Discuss new scientific policies or substantial architecture changes before building them.

Make a scientific contribution inspectable

A useful tool contribution includes a runnable example, the model’s assumptions, how to read its outputs, and a numerical check whose failure is meaningful. Identify the upstream source, license and runtime requirements. Operator-supplied software can be integrated without redistributing its source or binaries.

A useful research record includes the human’s question, agent instructions, software/model identities, actual commands and results, numerical controls and review decisions. An unresolved result is welcome if the record makes clear what was learned and what still needs testing. Separate numerical reproduction from replaying recorded artifacts and from a fresh autonomous study.

The evidence guide explains the contract. The FLASH record shows why scientific review can remain useful after a campaign ends; the Kepler record provides a small example with no-key verification and numerical reproduction.

A concrete open example is the radiation tool-selection proposal: can an agent identify an inadequate material model and choose an appropriate source or generator without a prescribed tool chain? It is a proposed study, with operator-readiness requirements and no claimed result. Contributions to its material data, conversion checks or radiation commissioning would be useful.

Development and credit

CONTRIBUTING.md is the canonical guide for environment setup, checks, pull requests and licensing. Run the checks relevant to your change and report unavailable prerequisites separately from passes. Ordinary regression work should not require paid calls.

Keep contributor names and upstream attribution with the relevant tool or study. In a research record, distinguish human framing, implementation, numerical work, independent review and later editorial material. Identify the exact Simjecture version and commit when citing a result. Scientific authorship should reflect the actual contribution and be discussed with the study’s collaborators.