epiinfer-contact-tracing

Releasing

Publishing the Docker image (GitHub Container Registry)

The publish workflow builds a linux/amd64 + linux/arm64 image, smoke-tests it, and pushes it to ghcr.io/<owner>/epitrace. The base images are pinned by digest in docker/Dockerfile, and Dependabot proposes digest updates weekly.

Dry run first (pushes nothing). On GitHub: Actions → publish → Run workflow. It builds both architectures and runs the smoke test, so an arm64 build failure shows up before a release.
Release. Bump version in pyproject.toml, run uv lock, update CHANGELOG.md, commit, then:
```
git tag v0.1.0 && git push origin v0.1.0
```
The workflow refuses a tag that does not match pyproject.toml. It pushes 0.1.0, 0.1 and latest (a pre-release tag such as v0.2.0-rc1 pushes only its own tag).
Make the package public (first release only). New packages start private. In the repository’s Packages section open epitrace → Package settings → Change visibility.
Record the digest. The workflow summary prints ghcr.io/<owner>/epitrace@sha256:…. Quote it in the paper or release notes: tags can move, digests cannot.
Test the pull on a machine that is not logged in to GitHub: docker pull ghcr.io/<owner>/epitrace:0.1.0 && docker run --rm ghcr.io/<owner>/epitrace:0.1.0 --version

The Apptainer route needs no separate build: it converts the pushed image (apptainer pull docker://ghcr.io/<owner>/epitrace:0.1.0).

A citable archive

Connect the repository to Zenodo (Settings → GitHub), then publish a GitHub release. Zenodo archives the code and issues a DOI that can go in CITATION.cff.

This site is open source. Improve this page.