A complete Eleventy blogging solution. SEO / GEO / AI ready.

Zero client-side JavaScript by default. Full dark mode. SEO-ready with JSON-LD, Open Graph, sitemaps, and RSS. GEO-optimized with structured data for local and global discoverability. AI-ready with llms.txt, content negotiation, and machine-readable markdown views. Full-text search powered by SQLite and PHP. An optional PHP contact form with honeypot anti-spam. Everything you need to start writing — nothing you need to strip out.

• Eleventy 3 with ESM, Liquid templates, and Markdown blog posts
• CSS-first dark mode — three-layer priority system using CSS :has() with radio buttons, no JavaScript required for theme switching
• Full-text search — SQLite FTS5 index built at compile time, PHP search handler with fuzzy matching cascade and BM25 ranking
• JSON-LD structured data — Person, WebSite, BlogPosting, BreadcrumbList, FAQPage schemas
• Open Graph and Twitter Cards — full meta tags for social sharing
• Dynamic social card images — SVG-based OG images generated per blog post
• RSS 2.0 feed with XSL stylesheet for human-readable browser rendering
• XML sitemap with automatic lastmod from git history
• llms.txt — LLM agent discovery with per-page markdown views
• Content negotiation — serve markdown to clients that send Accept: text/markdown (Apache .htaccess)
• Syntax highlighting — Prism via @11ty/eleventy-plugin-syntaxhighlight with theme colors that follow the dark mode system
• HTML and CSS minification in production (html-minifier-next + Terser for inline JS, clean-css for external stylesheets)
• Computed modified dates from git history with a global site-wide floor
• PHP contact form — honeypot anti-spam, input sanitization, open redirect prevention, mock email fallback for local development
• Docker setup for local development with PHP and Apache
• Clean URLs — directory-style (/blog/my-post/ not /blog/my-post.html)
• No client-side JavaScript required — JS only adds localStorage theme persistence and mobile navigation; everything works without it
• ESLint for Markdown files
• GitHub Actions CI workflow

Requirements: Node.js 18+ and npm.

git clone https://github.com/jcubic/complite.git
cd complite
npm install
npm run dev

The dev server starts at http://localhost:8080/. Edit files in src/ and the site rebuilds automatically.

npm run build

The output goes to _site/. It is a fully self-contained static site that can be deployed to any static hosting provider.

All site metadata lives in src/_data/:

• site.json — site title, description, URL, language, default author, email, founding date
• users.json — author profiles keyed by nickname (name, bio, photo, social links)
• person.json — JSON-LD Person schema for the site owner

Update site.json with your domain before deploying. The url field is used for canonical URLs, Open Graph tags, sitemaps, and RSS.

The modified field in site.json is a site-wide floor date. Update it whenever a non-content change (CSS, layout, config) affects all pages — this ensures the sitemap reflects the change.

Create Markdown files in src/blog/posts/. Each post needs front matter:

• title — post title
• description — used in meta tags and post cards
• date — publication date (YYYY-MM-DD)
• author — nickname matching a key in users.json
• tags — array of tags (these generate tag archive pages automatically)
• keywords — comma-separated, used in meta keywords tag

Optional front matter:

• modified — explicit last-modified date (overrides git-inferred date)
• featuredImage — path to cover image (used in Open Graph meta tags only, not displayed in the article body)
• unlisted: true — adds noindex, nofollow and excludes from collections
• faq — array of {question, answer} objects, generates FAQPage JSON-LD schema

complite includes a full-text search system powered by SQLite FTS5 and PHP.

1. At build time, Eleventy runs scripts/build-search-index.mjs after generating the site. The script walks all HTML files in _site/, extracts titles and content (stripping navigation, headers, footers), and inserts them into a SQLite database with an FTS5 virtual table using the unicode61 tokenizer.

2. At request time, the PHP search handler (src/static/search/index.php) reads the built HTML search page and replaces a <!-- search-results-placeholder --> comment with search results. It uses a three-tier fuzzy search cascade:

   • Tier 1: FTS5 MATCH with all terms ANDed using prefix expansion ("word"*)
   • Tier 2: FTS5 MATCH with terms ORed using prefix expansion (if AND returned nothing)
   • Tier 3: SQL LIKE fallback for partial substring matching

   Results are ranked using FTS5's bm25() function, and matching context is extracted via snippet() with <mark> highlighting.

Search indexing is controlled by the SEARCH_INDEX environment variable:

The search page at /search/ works without the index — it displays a message asking the user to rebuild with indexing enabled. This means development mode works normally without waiting for index rebuilds.

The search page requires:

• PHP 8.0+ with the SQLite3 extension (included by default in most PHP builds)
• Apache with mod_rewrite for the /search/ → index.php rewrite

The .htaccess file blocks direct access to the search.db file (returns 403 Forbidden).

If you don't need search, set SEARCH_INDEX=0 in your build environment and remove the search icon from the nav partial. The rest of the site is unaffected.

The theme system has three layers of priority, each overriding the one below:

1. System preference (CSS only) — @media (prefers-color-scheme: dark) applies dark variables automatically. No JavaScript needed.
2. JavaScript data-theme attribute — on page load, JavaScript reads the user's saved preference from localStorage and sets data-theme="dark" or data-theme="light" on <html>. This overrides the system preference.
3. CSS :has() radio buttons (CSS only, highest priority) — the theme toggle is a pair of hidden radio inputs. When a user clicks the sun/moon label, CSS :has(#mode_dark:checked) or :has(#mode_light:checked) applies the correct variables instantly, with no JavaScript involved.

The result: the site works fully without JavaScript (system preference), clicking the toggle works without JavaScript (CSS :has()), and JavaScript only adds persistence across page loads via localStorage.

Structured data is injected as application/ld+json script blocks:

• Homepage — @graph with Person, WebPage, WebSite, BreadcrumbList
• Blog posts — @graph with Person, BlogPosting (dates, author, headline, image, speakable), WebPage, Blog, WebSite, BreadcrumbList
• Author pages — Person with sameAs links, ProfilePage

Add faq to a post's front matter to generate a FAQPage schema automatically.

Every page outputs og:title, og:description, og:image, og:url, and og:type. Blog posts additionally include article:published_time, article:modified_time, and article:author. Twitter Card tags use summary_large_image.

/sitemap.xml is generated from all pages. The lastmod date is computed from the newer of the page's git commit date and the global site.modified floor. Pages with eleventyExcludeFromCollections or unlisted: true are excluded.

/feed.xml is an RSS 2.0 feed with an Atom self-link. It includes a pretty-feed XSL stylesheet so the feed is human-readable in browsers. Post links include UTM parameters for tracking.

The default robots.txt allows all crawlers including AI bots (GPTBot, ClaudeBot, Google-Extended, Applebot-Extended). Edit src/static/robots.txt to restrict access.

The site generates llms.txt for LLM agent discovery:

• /llms.txt — index with site description and links to all pages in markdown format
• /blog/{slug}.md — raw markdown for each blog post
• /about.md, /blog.md, /contact.md, /privacy.md — markdown views of static pages
• /llms-full.txt — full text of the entire site in one file

Content negotiation via .htaccess also serves the markdown version when a client sends Accept: text/markdown.

Blog posts get auto-generated Open Graph images via eleventy-plugin-svg-social-card. The plugin renders the SVG template in src/card/social-card.svg with per-post data (title, author, date) in a headless Chromium and screenshots it to a 1200×630 PNG. The {% card "emit" %} shortcode in head.liquid returns the URL for og:image and twitter:image meta tags.

Customize the card design by editing the SVG template. A static fallback image (src/static/img/social-card.png) is used for non-article pages.

The contact form is the only feature that requires a server-side runtime. Everything else is purely static.

Requirements for the contact form: Apache with mod_rewrite and mod_headers, PHP 8.0+, and Composer.

Install PHP dependencies:

composer install

Copy the example config and edit it with your email:

cp api/config.example.json api/config.json

Edit api/config.json:

{
    "contact_email": "you@example.com",
    "site_url": "https://yourdomain.com/"
}

If api/config.json is missing, the contact form will display an error telling the user to set up the configuration.

• The contact page is output as .php (not .html) so it can inject form feedback messages
• Eleventy generates the HTML, and a thin PHP wrapper reads it and replaces a <!-- form-message-placeholder --> comment with success/error messages
• The form handler (api/contact/index.php) validates input, checks the honeypot field, sanitizes headers against injection, and sends the email via PHP mail()
• On local development without a mail server, emails are saved to api/mail/ as HTML files (mock mail fallback)
• Bot attempts are logged to api/logs/bot.log

• Honeypot hidden field to catch bots
• Header injection prevention (sanitizeHeader() strips CR/LF)
• Open redirect prevention (sanitizeRedirectUrl() validates path prefix)
• config.json, api/lib/, api/logs/, and api/mail/ are blocked by .htaccess

If you don't need the contact form or search, and deploy to a platform without PHP (Netlify, Vercel, Cloudflare Pages), simply remove the src/contact/ and src/static/search/ directories and the api/ directory. The rest of the site works as a purely static site with no server-side dependencies.

Docker provides a local Apache + PHP environment for testing the contact form, search, .htaccess rules, and content negotiation.

Requirements: Docker and Docker Compose.

npm run build
docker compose up -d

The site is available at http://localhost:8080/. The _site/ directory is mounted as a volume, so rebuilding with npm run build updates the site without restarting the container.

To stop:

docker compose down

The included .htaccess file provides:

• Markdown content negotiation — Accept: text/markdown serves .md files
• HTTP Link headers pointing to the sitemap and llms.txt
• Search page rewrite — /search/ routes to index.php
• Custom 404 page
• Security rules blocking access to config files, PHP libraries, logs, vendor directory, and .db files

This requires Apache with mod_rewrite and mod_headers. If you deploy to Nginx, Netlify, Vercel, or another platform, you will need equivalent configuration for your server.

complite/
├── .eleventy.js                 # Eleventy configuration (collections, filters, transforms)
├── package.json
├── composer.json                # PHP dependencies (only needed for contact form)
├── Dockerfile                   # Apache + PHP for local dev
├── docker-compose.yaml
├── api/                         # PHP backend (contact form)
│   ├── contact/index.php        # Form handler
│   ├── lib/common.php           # Shared utilities (config, email, security)
│   ├── config.example.json      # Copy to config.json with your settings
│   └── logs/                    # Bot attempt logs (gitignored)
├── scripts/                     # Build and validation scripts
│   ├── build-search-index.mjs   # SQLite FTS5 indexer (runs after build)
│   ├── validate-jsonld.mjs
│   └── validate-social-card.mjs
├── src/
│   ├── _data/                   # Site metadata
│   │   ├── site.json            # Title, URL, description, author
│   │   ├── users.json           # Author profiles
│   │   └── person.json          # JSON-LD Person schema
│   ├── _includes/
│   │   ├── head.liquid          # <head> with meta, OG, fonts, CSS
│   │   ├── layouts/             # Page layouts (base, home, article, blog, author, ...)
│   │   └── partials/            # Reusable components (nav, footer, post-card, schemas, ...)
│   ├── blog/posts/              # Markdown blog articles
│   ├── authors/                 # Author profile pages
│   ├── static/                  # Copied to output root
│   │   ├── css/style.css        # Single stylesheet with dark mode variables
│   │   ├── css/prism-tomorrow.css # Syntax highlighting (uses CSS variables)
│   │   ├── favicon/             # Favicon set (SVG, PNG, ICO, webmanifest)
│   │   ├── img/                 # Images and avatars
│   │   ├── search/index.php     # PHP search handler (fuzzy FTS5 + LIKE fallback)
│   │   ├── .htaccess            # Content negotiation + security + search rewrite
│   │   ├── robots.txt
│   │   ├── pretty-feed.xsl     # RSS stylesheet
│   │   └── contact/index.php   # PHP wrapper for form messages
│   ├── card/social-card.svg     # OG image SVG template
│   ├── index.liquid             # Homepage
│   ├── search.liquid            # Search page
│   ├── about.md                 # About page
│   ├── privacy.md               # Privacy policy
│   ├── 404.md                   # Not found page
│   ├── sitemap.liquid           # XML sitemap
│   ├── feed.liquid              # RSS feed
│   ├── llms.txt.liquid          # LLM discovery index
│   ├── llms-posts.liquid        # Per-post markdown views
│   ├── llms-full.txt.liquid     # Full site text
│   └── src.11tydata.js          # Computed modified dates from git
└── _site/                       # Build output (gitignored)

Copyright (c) 2026 Jakub T. Jankiewicz

Released under the MIT License. See LICENSE for details.