Build and publish the website¶

The website and its authoritative documentation live in dancher00.github.io. The simulation code lives in wasserman. The configured project URL is https://dancher00.github.io/wasserman/. Author information is left blank until the paper is published.

Build¶

python3 scripts/restore_reproduction_sources.py
python3 scripts/fetch_website_media.py --destination .asset-cache/website-media.tar.gz
python3 scripts/prepare_website.py restore-media --archive .asset-cache/website-media.tar.gz
python3 -m pip install -r website/docs-requirements.txt
npm ci --prefix website
python3 scripts/verify_presentation.py
npm run build:publication --prefix website

Scientific presentation records are bundled with the checkout; no research-archive download is required. The media fetcher verifies the 0.1.0 archive digest. During private preparation it uses Git credentials or GITHUB_TOKEN; after the repository is public, its published media release supports anonymous download. Restoration preserves existing files that differ.

The build combines the React project page and static, searchable documentation. It verifies media routes and refuses a Pages artifact of 950 MB or more. All films retain their full timelines; the editorial overview stays at 1080p. The media manifest records original and presentation hashes.

Preview and verify¶

npm run preview --prefix website -- --host 127.0.0.1 --port 5188
WM_SITE_TEST_PUBLICATION=1 npm test --prefix website
python3 scripts/prepare_website.py pages
node website/scripts/verify-pages.mjs

The final artifact is website/.publication/pages/: the account root redirects to /wasserman/, which contains the project, documentation and media. See asset sources and overview provenance.

Deployment¶

Run Build and optionally deploy website with deploy=false to build and check the artifact without publishing. Source checks and browser tests do not launch the simulator or training.

When publication is authorized, make the code and site repositories public, select GitHub Actions as the Pages source, and run the workflow with deploy=true. Confirm /wasserman/, documentation, paper and video links. The workflow does not change repository visibility. Precomputed research datasets and fitted models remain separate from the source release; see data and models.