Contributor guide

Turn useful ideas into reviewable code and evidence.

Contributions should be small enough to inspect and useful enough to keep: a backend profile, compiler adapter, benchmark manifest, notebook, test, or documentation improvement.

Pick a concrete artifact and make the assumptions visible.

Profile

Add a hardware profile

Define coupling Hamiltonian family, topology, timing model, native operation assumptions, calibration notes, and validation status.

Adapter

Implement a circuit import path

Translate Qiskit, OpenQASM, TKET, QIR, or a minimal internal format into a project trace that can be tested.

Compiler

Build a pass

Prototype template synthesis, approximate compaction, SU(4)-aware routing, mapping bookkeeping, or pass-level diagnostics.

Benchmark

Create a manifest

Add circuits, parameters, expected outputs, environment notes, command lines, and result summaries in a reproducible format.

Validation

Add tests and checks

Protect behavior with unit tests, fixture comparisons, regression data, numerical tolerances, or smoke-test notebooks.

Docs

Improve the explanation layer

Clarify terminology, onboarding, architecture, setup, review rules, benchmark interpretation, or failure modes.

What a strong contribution should include.

01

Issue or intent

Connect the contribution to an issue, design note, benchmark need, or documented gap. Review starts faster when the purpose is explicit.

02

Runnable command

Include the smallest command, script, notebook cell, or test that lets a reviewer reproduce the behavior without guessing.

03

Assumption record

State backend assumptions, coupling model, topology, calibration limits, skipped cases, tolerances, and known failure modes.

04

Result artifact

Attach the manifest, output file, plot, table, profile diff, or documentation page that proves the contribution works.

05

Test coverage

Add the smallest check that prevents silent breakage. For numerical work, include tolerances and explain why they are reasonable.

06

Respectful review

Follow the Code of Conduct, explain trade-offs, respond to questions in good faith, and separate critique of ideas from critique of people.

Research software becomes credible when every claim is inspectable.

Reproducible

A reviewer can run it, see the same shape of output, and understand what changed.

commands · fixtures · notebooks

Scoped

The pull request solves one visible problem instead of bundling unrelated experiments together.

small PRs · clear commits

Explicit

The assumptions are written down: backend model, input circuits, thresholds, tolerances, and comparison baseline.

profiles · manifests · notes

Respectful

Scientific disagreement is welcome. Dismissive, harassing, or exclusionary conduct is not.

conduct · review · credit
GitHub logo

Open an issue, fork the repository, and submit a focused pull request.

The best first contribution is usually a backend profile, a benchmark manifest, a documentation fix, or a small test that makes the project easier to trust.