robotframework-browser.org

The website for the Robot Framework Browser library. Nuxt 4 + Nuxt Content 3, prerendered to static files, served from GitHub Pages.

Maintaining it? Read CONTRIBUTING.md. It covers editing content, what is generated and must not be hand-edited, and how a release reaches the site.

Commands

pnpm install
pnpm dev            # http://localhost:3000
pnpm test           # vitest
pnpm libdoc         # transform content/libdoc/*.json into app/generated/
pnpm generate       # runs libdoc, then static output into .output/public
pnpm check:bundle   # fails if server-only weight reached the client
pnpm verify         # test + generate + check:bundle

• /styleguide renders every design token and every component in every state.
• /keywords is the keyword reference: 151 keyword pages and 81 type pages.

Documents

Design spec	What we are building and every settled decision (D1–D23)
P1 plan	Foundation phase task ledger

How it fits together

• Content is Markdown and JSON in content/, edited by pull request. No CMS.
• Design tokens live in app/assets/css/tokens.css and nowhere else. A test fails the build if a colour literal appears anywhere else under app/.
• Themes. Light is the default; dark is a designed token swap. The --term-* plate tokens are identical in both, so code blocks look the same either way and the syntax theme only has to be defined once.
• Syntax highlighting runs at build time via Shiki, using Robot Framework grammars vendored from robotcode (Apache-2.0, see NOTICE). No highlighter reaches the browser.
• Components use radio inputs and CSS for state, so Editor tabs and the ComparisonSplit expansion work with JavaScript disabled.
• The keyword reference is generated from Libdoc JSON at build time. Drop a new content/libdoc/Browser-<version>.json in, point LATEST at it, and pnpm libdoc emits a 36 KB index plus one payload per keyword and per type. Groups come from each keyword's defining module via content/keyword-groups.json, because Libdoc carries no group field.

Deployment

Pushing to main runs the tests, generates the site, and deploys to GitHub Pages. Every pull request gets its own surge.sh preview, commented on the PR. When the library releases, a scheduled workflow opens a pull request here with the new keyword reference, release note and version numbers.

public/CNAME carries the custom domain, which is what preserves it across a deployment from an artifact.

Deployment is gated on the repository variable PAGES_LIVE. Until the cutover it builds and tests on every push and publishes nothing, because Pages has one source per repository and this one still serves the previous site from legacy-site.