Feeds and Sitemaps

Astral ships built-in plugins for feed and sitemap generation in static sites.

Feed

plugin Astral.Plugin.Feed,
  site_url: "https://example.com",
  title: "My Blog",
  author: "Astral",
  collection: :posts

The feed plugin renders a generated route, usually /feed.xml, from collection entries. It reads entry titles, dates, descriptions, authors, and draft status from schema-normalized entry.data, so declare those fields in the collection schema when you want them in feeds.

Sitemap

plugin Astral.Plugin.Sitemap,
  site_url: "https://example.com",
  changefreq: :weekly,
  priority: fn page -> if page.route_path == "/", do: 1.0, else: 0.7 end

The sitemap plugin renders /sitemap.xml from discovered and generated routes. For collection entry pages, default <lastmod> uses schema-normalized entry.data[:updated] or entry.data[:date]; standalone pages use their page frontmatter metadata.

llms.txt

Generate a metadata-only index following the llms.txt proposal:

plugin Astral.Plugin.LLMs,
  site_url: "https://example.com",
  title: "My Site",
  description: "Documentation and news about the project."

By default, /llms.txt contains a single “Pages” section with links in URL order. Titles and optional descriptions come from page metadata or normalized collection entry data. Without a title, the first level-one Markdown heading is used, then the route path (Home for /). Site identity is explicit, not guessed from HTML or the Mix application name. Without site_url, URLs are root-relative; the starter uses this mode until you configure your deployment URL.

Drafts, error pages, non-HTML outputs, and pages declaring noindex: true, robots: "noindex", or llms: false in metadata are omitted. Use exclude: ["/legal/"] for additional route exclusions. Generated routes override source pages at the same output destination, matching normal build behavior. Generated HTML routes can provide titles and descriptions through their metadata map.

For larger sites, curate sections without duplicating link metadata:

plugin Astral.Plugin.LLMs,
  title: "My Site",
  sections: [
    {"Start here", ["/", "/about/"]},
    {"Writing", [{:collection, :posts}]},
    {"Optional", [
      %{url: "/legal/", title: "Legal information"},
      %{url: "https://github.com/example/project", title: "Source code"}
    ]}
  ]

Sections replace automatic selection. Their order and explicit link order are preserved; collection selections are sorted by URL. Map selectors can override :title and :description; external links require an explicit title. Unknown local routes and collections raise errors. Exclusions still apply to curated sections. Empty sections are omitted, and sections: [] produces only site identity.

The same generated route works in static builds and development. Only llms.txt is produced: no page-body rendering or conversion, Markdown alternates, or llms-full.txt dumps. This index is discovery guidance, not crawler permissions or access control; robots.txt remains separate.

Custom generated XML routes

For site-specific generated files, implement Astral.Plugin and return routes:

defmodule MySite.FeedPlugin do
  @behaviour Astral.Plugin

  @impl true
  def name, do: "feed"

  @impl true
  def routes(site) do
    [Astral.Route.new("/feed.xml", site.config, content_type: "application/atom+xml")]
  end

  @impl true
  def render_route(%Astral.Route{path: "/feed.xml"}, site) do
    {:ok, MySite.Feed.render(site.entries.posts)}
  end

  def render_route(_route, _site), do: nil
end

Use this pattern for feeds, JSON indexes, search documents, or other static generated assets. Prefer entry.data for schema-normalized collection values and entry.metadata only when you intentionally need raw frontmatter.