website/

Ridu website

The public, consumer-facing home, documentation, guides, and API reference for Ridu. Repository architecture and maintainer documentation remain as plain Markdown under docs/.

Develop

This package has its own Bun lockfile and sits outside the framework's root Bun workspace.

bun run dev:website
bun run --cwd website verify

Run these commands from the repository root. dev:website installs both locked dependency trees and compiles @riducms/build before starting Astro. The website consumes that package’s canonical UnoCSS utilities through its published export; its output must exist before Astro loads the config. Only the build package is compiled; the admin and other runtime packages are not required.

verify checks formatting and types, runs focused unit tests, builds all static routes, validates links and anchors, checks the lazy search catalog, and guards the total generated HTML size.

Production and Cloudflare Pages

From a fresh checkout, run:

bun run build:website

This uses the same preparation as local development and writes the static site to website/dist. It needs no existing node_modules, package dist, or NODE_PATH override.

Configure Cloudflare Pages with:

Use these settings for production and previews. The build command owns dependency installation and compilation in order; Cloudflare’s automatic dependency install stays disabled.

Cloudflare supports these explicit version overrides and the dependency-install switch in its build image configuration. If a build logs a Bun list-all HTTP 403, check whether the pinned Bun version subsequently installs. The upstream asdf-bun plugin queries GitHub’s releases API to discover versions, then downloads an explicit version from a separate release-asset URL. A discovery failure can therefore precede a successful installation. The status alone does not establish rate limiting. Keep the pin when installation succeeds; diagnose a later missing @riducms/build/dist/uno/index.js through the preparation command above.

bun run check:website:clean copies current source files into a temporary checkout with an empty Bun cache and no installed dependencies or generated package output, runs the production command without NODE_PATH, and validates the emitted routes, assets, and search catalog. It runs in make documentation-check, the local release gate, and the manual release workflow before publishing.

Source map

• src/content/docs/ and src/content/guides/ contain ordinary consumer documentation as schema-validated Markdown.
• src/config/documentation.ts derives routes, navigation, pagination, outlines, and search records from those collections. The interactive Project Structure guide is merged into the same model.
• src/reference/authoring/ contains the reviewed module registry, stable-ID editorial overlays, CLI records, and route lock. src/reference/generated/catalog.json is deterministically extracted from the public Go, TypeScript, and Svelte sources; bun run reference:check rejects drift.
• src/features/project-structure/ owns the interactive guide's data, components, controllers, validation, and bespoke diagram CSS.
• src/components/ is grouped by owning surface: chrome, code, docs, and reference.
• src/styles/ is limited to tokens, base/accessibility rules, generated code chrome, and Markdown prose. Ordinary page and component layout uses website-local UnoCSS utilities.
• src/pages/search-index.json.ts emits the search catalog once. The site header fetches it lazily on first use instead of embedding hundreds of links in every page.

Author documentation

Each Markdown document declares its product, title, description, ordering, and navigation group in frontmatter. Stable section anchors use an explicit suffix:

## Access decisions {#access-decisions}

Labelled code fences use ordinary metadata:

```go title="cms/config.go"
var Config = ridu.Config{}
```

Plain Markdown is the default. Keep API signatures and other structured contracts out of prose, and introduce MDX only when a document genuinely needs an embedded interactive component.