Lanternlog
Lanternlog is one small server that hosts many blogs, each on its own domain, and makes every page easy for search engines and AI assistants to find, read and cite. It is a single Rust binary with an SQLite file, and you publish to it through a small HTTP API, which suits scripts and AI agents as well as people.
What every blog gets
- Pages that render in the HTML. No JavaScript is needed to read a post, so crawlers that don't run scripts, including most AI crawlers, see everything.
- Structured data on every page:
BlogPosting or WebPage with author, dates, word count and image, plus BreadcrumbList on posts and Blog (or WebSite, with a custom front page) on the home page.
- A share image for every post, a 1200×630 card drawn from the post's title, with Open Graph and Twitter tags pointing at it.
- Canonical addresses. Extra domains and
www. redirect to the main one, and paths without a trailing slash redirect to the one with it.
- Files for crawlers and assistants:
sitemap.xml, an Atom feed at feed.xml with full posts, robots.txt, llms.txt and llms-full.txt, and a Markdown copy of every page at /slug.md.
- IndexNow. Each blog has its own key, and publishing a post tells Bing, Yandex and the other IndexNow engines about it within seconds. Google reads the sitemap.
- An outline for long posts. Headings get stable ids, and posts with three or more get a table of contents.
- A way back to your product. Each blog names the project it writes about, and every post ends with a call to action that links to it.
- Speed. Pages are rendered once and served from memory, gzipped and with ETags, until the next write. Lighthouse scores 100 for performance, accessibility, best practices and SEO.
- Made for reading. A book-style serif for posts, light and dark themes with a toggle that remembers the reader's choice, and link colors adjusted to stay readable whatever accent a blog picks.
- Your own brand. Each blog can set its logo (drawn into its share cards too), favicon, share image, header links, head tags, footer text and font, and drop the Lanternlog credit.
Run it
With Docker:
docker build -t lanternlog .
docker run -d --name lanternlog -p 8080:8080 -v lanternlog:/data -e LANTERNLOG_TOKEN="$(openssl rand -hex 32)" lanternlog
From source, with Rust installed:
cargo build --release
LANTERNLOG_TOKEN=change-me ./target/release/lanternlog
Point each blog's domain at the server, and put a TLS proxy in front of it (any proxy that passes the Host header works). On Tugport, deploy the repository as one service and add every blog's domain to it; Tugport issues the certificates.
| Variable | Default | What it does |
|---|
PORT | 8080 | Port to listen on |
LANTERNLOG_DATA | ./data | Folder for lanternlog.db and uploaded images |
LANTERNLOG_TOKEN | none | Bearer token for the admin API; the API is off without it |
LANTERNLOG_SCHEME | https | Scheme used in addresses in pages and feeds |
LANTERNLOG_INDEXNOW | on | off stops IndexNow pings |
LANTERNLOG_FONTS | none | Extra folder of fonts for share images (Inter is in the Docker image) |
Publish
Every call needs Authorization: Bearer $LANTERNLOG_TOKEN, and can go to any of the server's domains.
Create or change a blog:
curl -X PUT https://blog.example.com/api/sites/blog.example.com \
-H "Authorization: Bearer $LANTERNLOG_TOKEN" -H "Content-Type: application/json" -d '{
"title": "Example Notes",
"description": "What we learn running Example in production.",
"aliases": ["www.blog.example.com"],
"author": "Jane Doe", "author_url": "https://jane.example",
"project_name": "Example", "project_url": "https://example.com",
"cta": "Example deploys an app with one command.",
"accent": "#e4572e", "lang": "en"
}'
Publish a post (the slug comes from the address; status is draft until you say published):
curl -X PUT https://blog.example.com/api/sites/blog.example.com/posts/one-server \
-H "Authorization: Bearer $LANTERNLOG_TOKEN" -H "Content-Type: application/json" -d '{
"title": "Moving to one server",
"description": "Why we put every service on one machine, and what it cost.",
"body": "We moved **every service** to one box.\n\n## Why one box\n\n...",
"tags": ["Self-hosting"],
"status": "published"
}'
The answer holds the post's address. Set "kind": "page" for pages such as About: they appear in the header instead of the post list and the feed. To give a blog a front page of its own, publish a page and set the blog's "home" to its slug; the page then shows at /, with the latest posts below it. The docs list every setting, including the branding ones (logo, favicon, og_image, links, head, footer, font and credit).
| Call | What it does |
|---|
GET /api/sites | All blogs |
GET, PUT, DELETE /api/sites/{host} | One blog; PUT creates or replaces its settings |
GET /api/sites/{host}/posts | Its posts and pages, drafts included |
GET, PUT, DELETE /api/sites/{host}/posts/{slug} | One post; PUT creates or replaces it |
POST /api/sites/{host}/media | Uploads an image (the raw bytes, with its Content-Type) and answers with its address |
GET /api/health | {"ok": true, "version"} |
Posts are Markdown with tables, footnotes, task lists and heading ids (## Setup {#setup}). Raw HTML passes through, so only give the token to writers you trust.
What it leaves out
Lanternlog has no web editor, comments, newsletters or themes beyond an accent color. Posts come in through the API, and anything that needs a browser is meant to live on your product's own site.
This project's website
lanternlog.com runs on Lanternlog. Its pages are in site/, and site/publish.py publishes them through the API.
License
MIT
