Skip to content

Publishing

The public site is at https://www.christianmlange.com/rcwa-qnm-invdes/.

The project remains on `hoodlab01`. Build the static documentation and dashboard there, then publish to the existing Contabo documentation hub. Do not clone this repository onto Contabo or modify the edge proxy for ordinary documentation updates.

Publish the current documentation and artifacts

```bash ./scripts/publish_site.sh ```

This uses a separate ignored `.venv-docs/` environment, renders the dashboard from the current JSON results under `artifacts/`, performs a strict MkDocs build, and mirrors the generated `site/` directory to the public project path.

Update the dashboard after a new run

Write the run artifacts to a named directory under `artifacts/`, then publish:

```bash uv run rcwa-qnm initialize configs/workstation.yaml --output artifacts/planned-workstation-initial uv run rcwa-qnm solve configs/tiny.yaml --output artifacts/my-run-solve uv run rcwa-qnm pole configs/tiny.yaml --output artifacts/my-run-pole uv run rcwa-qnm optimize configs/tiny.yaml --steps 20 --output artifacts/my-run-opt ./scripts/publish_site.sh ```

The dashboard discovers every `geometry.json`, `response.json`, `pole.json`, and `history.json` it finds. The newest planned geometry is shown as a zero-step run, separately from prior integration records.

Preview without publishing

```bash UV_PROJECT_ENVIRONMENT=.venv-docs ~/.local/bin/uv sync --group docs python3 scripts/build_dashboard.py UV_PROJECT_ENVIRONMENT=.venv-docs ~/.local/bin/uv run --group docs mkdocs serve ```

Agent checklist

  1. Keep source pages in `docs/`, dashboard generation in `scripts/build_dashboard.py`, and raw numerical artifacts in `artifacts/`.
  2. Never present a tiny/integration run as a converged device result.
  3. Run `./scripts/publish_site.sh` only after results and prose have been reviewed.
  4. Confirm the public URL loads and that the dashboard lists the intended latest artifacts.