Meilisearch

Search starts as WHERE title LIKE '%term%'. It works until someone misspells a word, searches two words in the wrong order, or expects results as they type — and then the options look like a week with Elasticsearch or a bill from Algolia.

Meilisearch sits between them: a single Rust binary, typo-tolerant and fast by default, with an API you can learn in an afternoon. Open source, self-hostable, with a paid cloud if you’d rather not run it.

Run and index link

  docker run -p 7700:7700 -e MEILI_MASTER_KEY=aSampleMasterKey \
  getmeili/meilisearch:latest
  

  npm install meilisearch
  

  import { MeiliSearch } from 'meilisearch';
const client = new MeiliSearch({ host: 'http://localhost:7700', apiKey: process.env.MEILI_KEY });

const index = client.index('posts');

await index.addDocuments([
  { id: 1, title: 'Deploying with Railway', tags: ['deploy'], published_at: 1735689600 },
  { id: 2, title: 'Postgres branching with Neon', tags: ['database'], published_at: 1738368000 },
]);

const results = await index.search('postgrs branchng');   // typos are fine
  

That last line is the demo that sells it: misspellings still match, and the response comes back in a handful of milliseconds.

Configure the three things that matter link

  await index.updateSettings({
  searchableAttributes: ['title', 'summary', 'body'],   // order = importance
  filterableAttributes: ['tags', 'author_id', 'published'],
  sortableAttributes: ['published_at'],
});
  

• searchableAttributes in priority order — a match in the title should outrank one buried in the body.
• filterableAttributes for anything you’ll narrow by. Filters are exact; search is fuzzy.
• sortableAttributes for anything you’ll order by.

  await index.search('postgres', {
  filter: 'tags = database AND published = true',
  sort: ['published_at:desc'],
  limit: 20,
  attributesToHighlight: ['title'],
});
  

Keys, and the one that must not leak link

Generate a search-only key for the browser and keep the master key on the server. It’s the same mistake as a database password in client code, and it’s easy to make because the quickstart hands you the master key first.

  const searchKey = await client.createKey({
  actions: ['search'],
  indexes: ['posts'],
  expiresAt: null,
});
  

Keeping the index in step with the database link

Postgres stays the source of truth; the index is derived and disposable. Two patterns:

• Reindex on write. After a create or update, push the document — ideally from a background job like Inngest so a slow index doesn’t slow the request.
• Rebuild periodically. A nightly full reindex into a new index, then swap it, keeps drift from accumulating. Cheap, and it makes the “did the index miss something?” worry go away.

Choosing among the options link

Genuinely start with Postgres. A tsvector column and a GIN index covers more products than people expect, and one fewer service to run is worth a lot. Move here when typo-tolerance, ranking, or search-as-you-type becomes a real requirement.

Next link

Search tells you what people looked for; analytics tells you what they did → PostHog