Making a blog website with SvelteKit

Created: | Updated:

List of projects used.

gray-matter - parse front-matter
markdown-it - parse markdown
markdown-it-prism > Prism - syntax highlighting

Changes to directory structure after initializing SvelteKit project.


└─ routes
	 └─ +layout.svelte
	 └─ +page.server.js
	 └─ index.svelte
	 └─ blog
     ├─ [slug].svelte
     └─ [slug].json.js

Create articles folder in your root directory. It is home for all future blog posts. Then create the first article, “SvelteKit is amazing”.

mkdir articles
cd articles

Add front-matter and content to the article.

title: SvelteKit is amazing
description: SvelteKit is absolutely amazing.
created: 2021-06-10
  - 'SvelteKit'
  - 'Markdown'

Why SvelteKit is absolutely amazing?

File name should match article title ( transformed to lowercase, and empty strings replaced with _ ).


Two server endpoints are to be made.
One for a blog page ~/routes/blog/[slug]/+page.server.js.
And one for an index page ~/routes/+page.server.js (getting blogs metadata).

So, inside the load function for fetching a blog, following steps are performed.

  1. Load corresponding .md file as string.
  2. Process separately front-matter and content.
  3. Parse front-matter with 'gray-matter’ as JSON object.
  4. Parse file content with ‘markdown-it’ returning HTML as string.
  5. Return processed data.
/* ~/src/routes/blog/[slug]/+page.server.js */
import fs from 'fs';
import mi from 'markdown-it';
import prism from 'markdown-it-prism';
import 'prism-svelte';
import matter from 'gray-matter';

const md = mi({
	html: true,
	linkify: true,
	typographer: true,

// Remember old renderer, if overridden, or proxy to default renderer
const defaultRender =
	md.renderer.rules.link_open ||
	function (tokens, idx, options, env, self) {
		return self.renderToken(tokens, idx, options);

md.use(prism, {});
/** @typedef {import("../../../typings/types").BlogMetadata} DM*/

/** @type {import('./$types').PageServerLoad} */
export async function load({ params }) {
	const { slug } = params;

	const doc = await fs.promises.readFile(`articles/${slug}.md`, 'utf8');
	// console.log(doc)

	const { data, content } = matter(doc);
	/** @type {DM} */
	const metadata = data;
	metadata.tags = => t.toLowerCase());

	/* 3. Process content {String}*/
	const html = md.render(content);

	return {
		blog: { html, metadata },

As all blog posts are placed in articles folder, and only there, and only in .md format, the load function for fetching blogs metadata is much simpler. Steps:

  1. Get file names of all blog posts.
  2. Load each file and parse front-matter.
  3. Sort by creation date.
  4. Return blogs as array.
/* ~/src/routes/+page.server.js */
import fs from 'fs';
import matter from 'gray-matter';

/** @type {import('./$types').PageServerLoad} */
export async function load() {
	const fileNames = await fs.promises.readdir('articles');

	const blogs = await Promise.all( (fileName) => {
			const doc = await fs.promises.readFile(`articles/${fileName}`, 'utf8');
			const { data } = matter(doc);
			/** @type {import('../../typings/types').BlogMetadata } */
			const md = data;
			return md;
	blogs.sort((a, b) => b.created.getTime() - a.created.getTime());

	return { blogs };


<!-- +layout.svelte -->

<a href="/">Index</a>
	<slot />

As the load function fetches corresponding endpoints, the index page receives all posts from it.

Then all tags are listed and blogs are filtered by route param (?tag=) if there is one.

<!-- index.svelte -->

	import { browser } from '$app/env';
	import { page } from '$app/stores';

	/** @type {import('./$types').PageServerData} */
	export let data;
	const blogs = data.blogs;

	const tagSet = new Set();
	blogs.forEach((blog) => {
		blog.tags.forEach((tag) => tagSet.add(tag));
	const tags = [...tagSet].sort();

	let tag;

	$: blogsFilteredByTag = tagSet.has(tag)
		? blogs.filter((p) => p.tags.includes(tag))
		: blogs;

	let unsub;

	if (browser) {
		unsub = page.subscribe(({ url }) => {
			tag = url.searchParams.get('tag');
			// console.log({ tag })

	onDestroy(() => {
		unsub && unsub();


	{#each tags as tag}
	<li><a href="/?tag={tag}"> #{tag} </a></li>

	{#each blogsFilteredByTag as blog}
		<a href="/blog/{blog.title.replaceAll(' ', '_').toLowerCase()}">


		<div class="tags">
			{#each blog.tags as tag}
			<a href="/?tag={tag}"> #{tag} </a>

Just fetching and displaying a blog.

	export let blog;
	const blog =;

	<meta name="description" content="{blog.metadata.description}" />

		<time>{new Date(blog.metadata.created).toLocaleDateString()}</time>
	{@html blog.html}

	{#each blog.metadata.tags as tag}
	<li><a href="/?tag={tag}"> #{tag} </a></li>

Types for vscode.

/* ~/typings/types.d.ts' */

export interface BlogMetadata {
	title: string;
	description: string;
	created: Date;
	tags: string[];


Install @sveltejs/adapter-static and update svelte.config.js as instructed. Then npm run build.

All pages will be prerendered.
Only the index page will load JS, make hydration, turning the website into SPA.
Every blog page will go with 0 JS, so traditional navigation and no hydration.