# CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

## Commands

```bash
pnpm dev            # Tailwind watch + Hugo dev server in parallel
pnpm build          # build:css then hugo --minify → public/
pnpm build:css      # Tailwind only → static/css/main.css
pnpm watch:css      # Tailwind watch (CSS only)
hugo server -D      # Hugo dev server (drafts included)
hugo --minify       # Production build
python3 scripts/import-posts.py  # Import posts from ~/projects/ginger/posts.csv
```

CSS must be compiled before Hugo runs. `pnpm build` does both in order. In dev, `pnpm dev` runs both concurrently.

## Architecture

**Hugo Extended** static site with **Tailwind CSS v4** (no config file — design tokens live in `@theme {}` in `assets/css/main.css`).

### CSS pipeline

Tailwind reads `assets/css/main.css` and writes to `static/css/main.css` (gitignored — build artefact). The `@theme {}` block defines all design tokens (`--color-paper`, `--color-roux`, `--font-display`, etc.). Global resets are in `@layer base` so Tailwind utility classes correctly override them.

### Content model

- `content/posts/<slug>/index.md` — page bundles; each has a `plate` number, `issues`, `categories` (single item), `tags`, `featured` flag, and a `.png` image resource.
- `content/issues/<id>/_index.md` — branch bundle for an issue; metadata (`title`, `description`, `issueNumber`, `season`, `status`) drives the homepage hero and issue list page.
- `content/imprint.md` — standalone page using `layout: imprint`.

Taxonomies: `categories`, `tags`, `issues` (defined in `hugo.toml`).

### Layout hierarchy

```
layouts/baseof.html          ← shell: head, header, main, footer, lightbox, roux-data JSON
layouts/index.html           ← homepage: current-issue hero + 5 featured plates
layouts/_default/list.html   ← two branches: kind="taxonomy" (term cards) | kind="term"/section (paginated post grid)
layouts/_default/single.html ← post plate page (triggers lightbox via window.__ROUX_OPEN_SLUG)
layouts/_default/imprint.html← prose page layout
layouts/issues/list.html     ← paginated post grid scoped to an issue
layouts/issues/terms.html    ← issue index, reads issueIds from hugo.toml params
layouts/404.html             ← editorial 404 page
```

Key partials: `header.html`, `footer.html`, `card.html` (reusable post card), `lightbox.html`, `pagination.html`, `logo.html`.

### JavaScript (`static/js/app.js`)

Single vanilla JS file, no build step. Loaded from `baseof.html`. Reads `#roux-data` (JSON injected by Hugo containing all posts) at runtime.

Responsibilities:
- **Inverted search index** — weighted token scoring across title, tags, categories, description; prefix matching; renders into `#searchpop`.
- **Lightbox** — full-screen plate viewer with keyboard nav, thumbnail strip, share/copy URL, meta panel. Opened automatically when `window.__ROUX_OPEN_SLUG` is set (direct URL to a post).
- **Ribbon** dismiss, tab sync, view transitions (`document.startViewTransition`), masthead date.

### Data flow: homepage featured plates

The homepage shows exactly the posts with `featured: true` in frontmatter. Mark posts via `featured: true`; the template filters with `where .Site.RegularPages ".Params.featured" true`.

### Content import

`scripts/import-posts.py` reads `~/projects/ginger/posts.csv` and `~/projects/ginger/images/final/selected/`. It creates/updates page bundles deterministically (sorted by filename → stable plate numbers). Run it from the repo root after updating the upstream CSV or image set.

### Hugo params (hugo.toml)

`issueNumber`, `issueName`, `issueSeason`, `issueBlurb` are fallback values used when `$.Site.GetPage "/issues/01"` returns nil. `issueIds` drives the issues terms page. `umamiSrc`/`umamiId` are injected only in production (`hugo.IsProduction`).

### nginx

`nginx.conf` at repo root is the production server config (used in Docker). `error_page 404 /404.html` is already wired.