Searching Posts and Documentation with Minisearch

Makerkit's blog and documentation pages come with built-in search functionality that does not reply on complex and expensive cloud services such as Algolia and ElasticSearch.

Instead, we used Minisearch, a lightweight Javascript Search Engine.

How does Minisearch work?

Instead of indexing our posts at runtime, we index all our documentation and blog articles at build time, right after the Next.js application has finished building.

Minisearch creates a simple index file in the public folder. The API loads the index and responds to the user's query.

By default, we index two directories: _docs and _posts. Of course, it's easy to extend this to other folders if you create more types of content.

Below you can find our indexer script that takes care of indexing recursively our .mdx files in the folders provided:

const FIELDS = ['content', 'path', 'tag', 'title', 'collection', 'slug'];
export async function searchIndexer() {
 const miniSearch = new MiniSearch({
   fields: FIELDS,
   storeFields: FIELDS,
 });
 const engine = new SearchEngine(miniSearch);
 await engine.indexDirectory(`_posts`, (doc) => {
   return { ...doc, tag: `blog` };
 });
 await engine.indexDirectory(`_docs`, (doc) => {
   return { ...doc, tag: `docs` };
 });
 await engine.export();
 process.exit();
}

Assuming you have another folder named _guides, you could add:

await engine.indexDirectory(`guides`, (doc) => {
 return { ...doc, tag: `guide` };
});

Does it scale?

It's important to keep the index not larger than few MBs, otherwise it will exceed the maximum memory available to the API function.

That could be more or less a few hundreds pieces of content (eg. articles), so you may have some room before having to migrate.