Octavo

The quantitative social science starter pack

日本語の README

Write your papers, slides and lecture notes in Markdown, keep the analysis in Quarto, and let Octavo place the analysis's numbers, figures and tables in every document — then typeset them to PDF (with Typst) and Word, or LaTeX if you have TeX. Citations come from one .bib file, in any CSL style.

What you get

A page of a paper and a slide, both built from the example project (octavo init demo --all --example). The coefficient, the table and the figure come from the analysis; the equation, table and figure numbers and the references are filled in when it is typeset.

• No number is typed by hand. The analysis registers it (ov_value("n_obs", nrow(d))) and the manuscript calls it by name ({{n_obs}}). Re-run the analysis and every document follows. octavo check points out numbers typed into the prose, and values that are only placeholders.
• R or Python. The analysis is a Quarto .qmd in R or in Python; both come with the same helpers (ov_value, ov_figure, ov_table), and the project's .venv and renv are set up for you.
• One source, several documents. A paper in your journal's layout, a talk, and lecture notes that become an A4 handout plus one slide deck per session — with speaker scripts, and Word for coauthors.
• References by label: @fig-trend becomes "Figure 2.1", and stays right when you reorder sections.
• Diagrams without TikZ: octavo new figure gives you a Typst file to draw boxes and arrows in; it becomes a figure like any other.
• Tables by hand, edited as tables: octavo new table gives you a CSV that the VS Code extension edits as a grid (Excel works too); it becomes a table like the analysis's.
• Submission and beyond: word limits, blind review, a submission zip, reading a coauthor's tracked changes in Word, "what moved since I submitted", and a replication package.
• A VS Code extension with a live PDF preview that follows the cursor, a sidebar for everything above, and one-click setup.
• Ready for your AI assistant: every project keeps its working rules in AGENTS.md — never type a result into the manuscript, data/raw is read-only, and so on — and Claude Code, GitHub Copilot, Codex and Antigravity all read it, so the assistant knows them from the start.
• Linux, macOS and Windows; English and Japanese.

Install

With VS Code, install the Octavo extension and press Set up when it asks: it installs pandoc, Typst, Quarto, R, the fonts and the octavo command.

From a terminal:

curl -LsSf https://astral.sh/uv/install.sh | sh   # uv, if you don't have it yet
uv tool install octavo-kit                        # the octavo command
octavo setup                                      # pandoc / Typst / Quarto / fonts / R
octavo doctor                                     # reports anything still missing

Getting started

With VS Code:

1. Open a folder and the Octavo icon in the activity bar. In a folder that is not a project yet it offers New project: choose where, the language, the parts you need (an analysis in R or Python, a paper, slides, lecture notes) and whether to fill them with examples. An analysis gets its environment set up right after.
2. Open a manuscript and press the PDF button at the top right of the editor: the PDF appears beside it, is rebuilt on every save, and follows your cursor.
3. Add more from the + at the top of the Octavo sidebar.

From a terminal, start with the example project — an analysis, a paper, slides and lecture notes:

octavo init demo --all --example
cd demo
octavo env                    # the project's analysis environment (.venv and renv)
octavo build --compile        # runs the analysis, then typesets everything to PDF

Then your own project, starting with the parts you need:

octavo init study --with analysis,paper
cd study
octavo env                    # the project's analysis environment (.venv and renv)
octavo build paper --compile  # after writing analysis/analysis.qmd and papers/paper/paper.md
octavo check                  # before you submit

In the analysis, register what the paper shows; in the manuscript, refer to it:

ov_value("n_obs", nrow(d))
ov_figure(p, "trend")

ov_value("n_obs", len(d))
ov_figure(fig, "trend")

The sample has {{n_obs}} cases (@fig-trend).

![Trend](../../assets/figures/trend.png){#fig-trend}

Documentation

The guide covers everything:

1. Install — Linux, macOS and Windows, step by step
2. Your first project — octavo init and octavo new, the analysis environment
3. Writing manuscripts — references, numbered blocks, math, figures, tables
4. The analysis — numbers, figures and tables from Quarto
5. Citations
6. Papers — checking, blind review, submitting, revisions
7. Slides and lecture notes — sessions, per-session handouts
8. Making it yours — templates, Word styles, fonts
9. Command reference
10. VS Code
11. What to check on your own machine

Contributing and license

Issues and pull requests are welcome — see CONTRIBUTING.md. MIT License.