Migrating from manubot

If you have an existing manubot manuscript built on rootstock and you want to move it to quartobot, here’s the translation. The pattern is preserved — prose source in git, auto-resolved citations, per-commit permalinks, PR-reviewed prose. What you change is the build pipeline and a few filename conventions.

Before you start

Don’t delete anything. Migrate on a branch (or a separate worktree); keep the manubot setup intact until the new one renders correctly.
Decide the project shape. Most manubot manuscripts map cleanly onto Quarto’s manuscript template (single paper.qmd plus references). Longer works — theses, reviews, books in progress — map better onto the book template (chapter-per-file). Pick one; you can switch later, the file shape is the only thing that changes.
What carries over unchanged: your prose, your cite keys, your hand-curated references, your figures. What changes: the build pipeline, the metadata file’s shape, the directory layout.

Step 1 — Install the quartobot CLI

uv tool install git+https://github.com/quartobot/quartobot
quartobot --version

Puts quartobot on your user PATH (~/.local/bin/ typically). Quarto’s pre-render subprocess needs to find it there. See Install for uvx, editable, and post-v0.1-tag pip install paths.

Step 2 — Scaffold the Quarto project

In the same repo, alongside your existing content/, build/, and ci/ directories:

quartobot init

init writes:

_quarto.yml with the pre-render hook wired to quartobot resolve
references.bib (seed for hand-curated entries — empty)
_version-banner.html.template + _version-banner.html (dev placeholder)
.github/workflows/render.yml (ten-line caller of the upstream reusable workflow)
.github/workflows/pr-closed.yml (preview teardown)
.gitignore augments

For a book project, pass --project-type book — same files, plus chapters/ is where your content lives.

quartobot init --project-type book

Step 3 — Move your content

Manuscript shape (single document)

Manubot’s content/01-abstract.md, 02-introduction.md, … get concatenated at build into one manuscript. Quarto’s manuscript template uses one paper.qmd (or whatever you name it) for the same thing. The mechanical translation:

# Order matters — match manubot's numeric prefix order.
cat content/*.md > paper.qmd

Then move the YAML front-matter from content/metadata.yaml into the top of paper.qmd between --- fences (see Step 4).

The .md/.qmd syntax overlap is high enough that prose usually needs no edits. Things to spot-check after the concat:

Heading levels. Manubot expects # for the document title and ## for top-level sections; Quarto’s manuscript template treats the YAML title: field as the title and uses # for top-level sections. Drop the first # Title line and let Quarto render the title from the YAML.
Cite keys. No change — @doi:, @pmid:, @arxiv:, @isbn: etc. all work identically. Bare DOIs (@10.1371/…) also work.
Figure references. Manubot uses ![](images/foo.png); Quarto adds named cross-references via {#fig-foo}. If you want auto-numbered figure cross-references, add the ID; if your prose says “Figure 3” by hand, no change needed.
Math. Both use $…$ and $$…$$. No change.

Book shape (chapter-per-file)

Rename your manubot content files into chapters/:

mkdir -p chapters
for f in content/[0-9]*.md; do
  base=$(basename "$f" .md)
  # Drop the leading "01-" / "02-" prefix if you want clean URLs;
  # keep it if you want to preserve manubot's filenames.
  cp "$f" "chapters/${base#[0-9]*-}.qmd"
done

Edit _quarto.yml to list the chapters in order under book.chapters:. Move metadata.yaml content into the book: block in _quarto.yml (see Step 4).

Step 4 — Move your metadata

Manubot’s content/metadata.yaml looks like:

title: My manuscript title
keywords:
  - reproducibility
authors:
  - github: alice
    name: Alice Author
    initials: AA
    orcid: 0000-0000-0000-0000
    email: alice@example.org
    corresponding: true
    affiliations:
      - Department of Examples, University of Demonstration

Quarto’s equivalent goes either in the file’s YAML front-matter (for the manuscript shape) or under book: in _quarto.yml (for the book shape). The field names mostly map:

title: "My manuscript title"
keywords:
  - reproducibility
author:
  - name: Alice Author
    affiliation: Department of Examples, University of Demonstration
    orcid: 0000-0000-0000-0000
    email: alice@example.org
    corresponding: true

Differences worth flagging:

authors: → author: (singular key, still takes a list — Quarto’s convention).
github: and initials: don’t have direct equivalents. Quarto doesn’t render them; if you want them on the manuscript, add them in an author block extension or drop them. Manubot uses these for contribution attribution in the rendered author block — see Quarto’s attribution: field for a similar surface.
affiliations: (list per author in manubot) → affiliation: (string, singular) for the simple case, or affiliations: (list of refs into a top-level affiliations: block) for multiple affiliations per author. Quarto’s affiliation reference shape is documented at quarto.org/docs/journals/authors.html.
funders: / acknowledgments: go in the front-matter the same way manubot puts them in metadata; Quarto reads them per the journal template you’re targeting.

Step 5 — Move your hand-curated references

Manubot’s content/manual-references.json (CSL JSON) or content/manual-references.yaml (CSL YAML) or content/manual-references.bib all map onto references.bib (BibTeX) or stay as their original format — pandoc citeproc reads .bib, .json (CSL JSON), and .yaml (CSL YAML) interchangeably.

The simplest move:

# If you had a .bib already, just copy it.
cp content/manual-references.bib references.bib

Or convert CSL JSON to BibTeX with pandoc-citeproc:

pandoc -f csljson -t biblatex content/manual-references.json -o references.bib

Then make sure _quarto.yml lists it under bibliography:. The template quartobot init writes already does:

bibliography:
  - references.bib    # hand-curated
  - references.json   # auto-resolved by `quartobot resolve`

If you had manual-references.json populated by manubot’s resolver runs (not hand-curated), discard it. The pre-render hook regenerates the auto-resolved references.json on next render from the cite keys in your prose.

Step 6 — Swap the CI pipeline

Delete build/ and the manubot CI workflow:

git rm -r build/
git rm ci/build.yaml 2>/dev/null  # or wherever your manubot workflow lives
git rm .github/workflows/manubot.yaml 2>/dev/null

quartobot init already wrote .github/workflows/render.yml — a ten-line caller of the upstream reusable workflow. That workflow handles render → permalink staging → gh-pages deploy → sticky PR comment, the same surface manubot’s build/build.sh + build/webpage.sh covered.

The deployed URL layout is preserved: https://<owner>.github.io/<repo>/v/<sha>/ for per-commit snapshots, https://<owner>.github.io/<repo>/ for latest main, https://<owner>.github.io/<repo>/pr/<n>/ for open PR previews. Existing links into your manuscript’s snapshot URLs keep working.

Step 7 — Local test

quarto render

The pre-render hook runs quartobot resolve first, writes references.json. Then pandoc renders the document; citeproc reads both bibliographies. Open index.html (manuscript) or _book/index.html (book) and check that:

Every @doi: / @pmid: / @arxiv: cite resolved (no [Unresolved citation: …] markers).
Hand-curated cites from references.bib rendered.
Figures show up.
The version banner is the dev placeholder (gets replaced by CI on push to main).

quartobot validate . runs the same checks the CI lint step does:

quartobot validate .

Expect 5 checks; failures usually mean the pre-render line in _quarto.yml is malformed or references.json isn’t in the bibliography: list.

Step 8 — Push

git add .
git commit -m "Migrate to quartobot"
git push

CI takes it from there. On first push to main, enable GitHub Pages in Settings → Pages → Source: GitHub Actions (or “Deploy from a branch → gh-pages” depending on which reusable workflow path you’re on). If the manubot setup was already deploying to gh-pages with the same /v/<sha>/ layout, the deploy slots in without breaking existing snapshot URLs.

What to delete after the dust settles

Once the new pipeline has rendered cleanly on main:

content/ directory
build/ directory
ci/ directory (if manubot-specific)
Old manubot workflow files
Any webpage/ source-controlled checkout (the gh-pages branch is managed by CI now)

Keep the gh-pages branch intact — it has your historical snapshots at /v/<old-sha>/, and the new pipeline writes new ones alongside.