codif-ape-docs

Technical note on INSEE's MLOps pipeline for the automatic coding of economic activity (APE / NAF / NACE), as a bilingual Quarto website.

Live site: https://inseefrlab.github.io/codif-ape-docs/

Structure

• index.qmd: root page, redirects to en/, or to fr/ if the reader last chose French
• en/*.qmd: English version
• fr/*.qmd: French version, with the same file names and the same anchors (#sec-…, #tbl-…, fig-…)
• references.bib: bibliography shared by both versions
• style/styles.scss: theme (Bootswatch litera with a few rules)
• style/nav.html: the single, collapsible table of contents in the left sidebar, and the EN | FR switch in the navbar, which swaps the page to the other language in place (same page, same section, no reload)
• scripts/outline.py: post-render script that writes _site/outline.json, the plan of the note (chapters, sections, sub-sections) read by style/nav.html
• codif-ape-prez/: submodule holding the presentations. It is not rendered here: the site links to https://inseefrlab.github.io/codif-ape-prez/.

Every change must be made in both languages.

Authors

Each chapter lists its own authors in its front matter, shown under the title. Set them in both languages (en/<chapter>.qmd and fr/<chapter>.qmd):

author:
  - name: Jane Doe
    affiliations:
      - Insee
  - name: John Doe
    affiliations:
      - Insee
      - CREST

Quarto also accepts email, orcid, url and corresponding: true for each author.

Build locally

quarto preview

Deployment

On each push to main, .github/workflows/publish.yaml renders the site and publishes it to the gh-pages branch. Under Settings → Pages, GitHub Pages must serve from the gh-pages branch.