mdbook-content-loader

Zero-runtime-fetch content collections for mdBook

Injects your content-collections.json (generated by mdbook-content-collections) directly into your HTML as window.CONTENT_COLLECTIONS — no network request, no path hacks, works offline.

By default, the data is injected on your landing page (index.md) only. You can optionally enable the old behavior and inject it on every page via a config flag.

This gives you true Astro/Zola-style content collections inside mdBook: instant, reliable access to posts, blog entries, notes, changelogs — anywhere in your theme.

Why this exists

The original mdbook-content-collections example used fetch('content-collections.json') in the browser.

That works… until it doesn’t:

This crate fixes all of that permanently.

What you get

A single, ready-to-query global on every page (or just the landing page, by default)

<script>
  window.CONTENT_COLLECTIONS = { ...ready to use... }
</script>

• window.CONTENT_COLLECTIONS.entries: all published items, sorted newest → oldest

• window.CONTENT_COLLECTIONS.collections: per‑collection buckets (blog, notes, changelog, etc.)

• window.CONTENT_COLLECTIONS.generated_at: build timestamp for cache busting or debugging

Each entry has the full shape from mdbook-content-collections (path, title, date, description, tags, optional collection, preview_html, and draft stripped out), so your theme code can immediately render lists, sidebars, tag clouds, and “related posts” without any extra parsing or network calls.

Installation

cargo install mdbook-content-loader
# If not installed, install mdbook-content-collections
# cargo install mdbook-content-collections

Setup

Add the following to your book.toml:

[preprocessor.content-collections]
# Generates content-collections.json

[preprocessor.content-loader]
command = "mdbook-content-loader"
after = ["content-collections"]   # make sure JSON is generated first

# Behavior:
# - default: inject window.CONTENT_COLLECTIONS only into index.md
# - set inject_all = true to inject into every chapter
# inject_all = true

The after key ensures the JSON exists before this loader runs, and inject_all is an optional boolean that restores the old “every page” behavior. mdBook exposes this table under [preprocessor.content-loader] to the preprocessor via PreprocessorContext, which is where this flag is read.

Works best with an empty index.md, create one by adding this as the first line of your SUMMARY.md:

()[index.md]

Then run mdbook build to create it.

Frontmatter Example

mdbook-content-loader injects the collection blog into index.md by default, so for a chapter to be listed on the landing, the frontmatter would look like:

---
title: "My first note"
date: 2025-12-07
collection: "blog"
description: "Short note about using mdbook-content-loader."
draft: false
---

Chapters with the latest date will be listed first. Chapters with draft: true will be skipped.

Usage in your theme

Drop this anywhere in theme/index.hbs (or any .hbs file).

For example, placing it directly after this block works, and you'll have a landing page of your "Latest Edited Posts" placed right above the prev/next buttons:

<nav class="nav-wrapper" aria-label="Page navigation">
<!-- Mobile navigation buttons -->
{{#if previous}}
    <a rel="prev" href="{{ path_to_root }}{{previous.link}}" class="mobile-nav-chapters previous" title="Previous chapter" aria-label="Previous chapter" aria-keyshortcuts="Left">
        {{#if (eq ../text_direction "rtl")}}
        {{fa "solid" "angle-right"}}
        {{else}}
        {{fa "solid" "angle-left"}}
        {{/if}}
    </a>
{{/if}}
// <div id="latest-posts"></div>

Add the following block:

<div id="latest-posts"></div>

<script>
  if (!window.CONTENT_COLLECTIONS) {
    console.warn("mdbook-content-loader: data not loaded");
  } else {
    const posts =
      window.CONTENT_COLLECTIONS.collections.blog ||
      window.CONTENT_COLLECTIONS.collections.posts ||
      [];

    const list = document.getElementById("latest-posts");
    const ul = document.createElement("ul");

    posts.slice(0, 5).forEach((post) => {
      const li = document.createElement("li");
      li.innerHTML = `
        <h3><a href="$$ {post.path.replace(/\.md $$/, '.html')}">${post.title}</a></h3>
        ${post.date ? `<time>${new Date(post.date).toISOString().slice(0, 10)}</time>` : ""}
        <div class="preview">${post.preview_html}</div>
      `;
      ul.appendChild(li);
    });

    list.appendChild(ul);
  }
</script>

Entries with collection: "blog" end up in window.CONTENT_COLLECTIONS.collections.blog.

Entries with collection: "posts" or with no collection at all end up in window.CONTENT_COLLECTIONS.collections.posts as a general “all posts” bucket.

The “Latest posts” example code first prefers collections.blog and only falls back to collections.posts if there are no blog items.

Want a sidebar? Tag cloud? Search index? Related posts? Just read from window.CONTENT_COLLECTIONS.

Common patterns

// Latest 5 blog posts
window.CONTENT_COLLECTIONS.collections.blog?.slice(0, 5);

// All posts (any collection)
window.CONTENT_COLLECTIONS.entries;

// Posts with tag "rust"
window.CONTENT_COLLECTIONS.entries.filter((p) => p.tags.includes("rust"));

Data Shape of content-collections.json

Same as mdbook-content-collections, plus:

• entries: all entries from content-collections.json, sorted newest → oldest

• collections: grouped views of the same entries, for example:

• collections.blog: all entries where collection === "blog"

• collections.posts: all entries (fallback/all posts)

• generated_at: ISO 8601 timestamp string of when the data was injected into window.CONTENT_COLLECTIONS

• All draft: true entries removed

See: mdbook-content-collections

Example Project built with content-collections / content-loader

• mdbook-kanagawa-theme

License

Apache License 2.0