Meilisearch — Lightning-Fast Open Source Search Engine

What Meilisearch Does

• Instant Search: Sub-50ms search responses even on large datasets
• Typo Tolerance: Automatic handling of typos and misspellings
• Faceted Search: Filter by categories, prices, dates, and custom attributes
• Hybrid Search: Combine keyword search with AI embeddings for semantic understanding
• Synonyms: Define synonyms for better search results
• Ranking Rules: Customizable ranking with business logic
• Multi-Language: Built-in support for 20+ languages including Chinese, Japanese, Arabic
• Geosearch: Search by geographic distance and bounding box
• Index Management: Create, update, and delete indexes with REST API
• API Keys: Granular access control with scoped API keys

Architecture

┌──────────────┐     ┌──────────────┐     ┌──────────────┐
│  Your App    │────▶│  Meilisearch │────▶│  Local Disk  │
│  (Search UI) │     │  Engine      │     │  (LMDB)      │
└──────────────┘     │  (Rust)      │     └──────────────┘
                     └──────────────┘

Meilisearch is a single binary written in Rust. No external dependencies — just a single process that stores data in an embedded LMDB database.

Self-Hosting

Docker Compose

services:
  meilisearch:
    image: getmeili/meilisearch:latest
    ports:
      - "7700:7700"
    environment:
      MEILI_MASTER_KEY: your-master-key-min-16-chars
      MEILI_ENV: production
      MEILI_NO_ANALYTICS: "true"
    volumes:
      - meili-data:/meili_data
    restart: unless-stopped

volumes:
  meili-data:

Basic Usage

1. Create Index and Add Documents

# Add documents to an index
curl -X POST 'http://localhost:7700/indexes/products/documents' 
  -H "Authorization: Bearer your-master-key" 
  -H "Content-Type: application/json" 
  --data-binary '[
    {
      "id": 1,
      "name": "iPhone 15 Pro",
      "price": 999,
      "brand": "Apple",
      "tags": ["phone", "premium"]
    },
    {
      "id": 2,
      "name": "Samsung Galaxy S24",
      "price": 899,
      "brand": "Samsung",
      "tags": ["phone", "android"]
    }
  ]'

2. Search

# Simple search
curl 'http://localhost:7700/indexes/products/search?q=iphone' 
  -H "Authorization: Bearer your-master-key"

# Advanced search with filters
curl -X POST 'http://localhost:7700/indexes/products/search' 
  -H "Authorization: Bearer your-master-key" 
  -H "Content-Type: application/json" 
  -d '{
    "q": "phone",
    "filter": "price < 1000 AND brand = "Apple"",
    "sort": ["price:asc"],
    "limit": 10,
    "attributesToHighlight": ["name"]
  }'

3. Typo Tolerance in Action

Query: "iphne" → Finds "iPhone" (automatic typo correction)
Query: "samsong" → Finds "Samsung"
Query: "prroducts" → Finds "products"

SDK Examples

JavaScript

import { MeiliSearch } from 'meilisearch';

const client = new MeiliSearch({
  host: 'http://localhost:7700',
  apiKey: 'your-master-key',
});

// Add documents
await client.index('products').addDocuments([
  { id: 1, name: 'iPhone 15', price: 999 },
]);

// Search
const results = await client.index('products').search('iphone', {
  limit: 10,
  filter: 'price < 1500',
});

Python

import meilisearch

client = meilisearch.Client('http://localhost:7700', 'your-master-key')

# Add documents
client.index('products').add_documents([
    {'id': 1, 'name': 'iPhone 15', 'price': 999}
])

# Search
results = client.index('products').search('iphone', {
    'limit': 10,
    'filter': 'price < 1500'
})

Key Features

Hybrid Search (AI + Keyword)

{
  "q": "comfortable running shoes",
  "hybrid": {
    "embedder": "openai",
    "semanticRatio": 0.5
  }
}

Combines traditional keyword matching with AI semantic understanding for best results.

Faceted Search

{
  "q": "phone",
  "facets": ["brand", "price_range", "tags"]
}

Response includes facet counts for building filter UIs.

Ranking Rules

{
  "rankingRules": [
    "words",
    "typo",
    "proximity",
    "attribute",
    "sort",
    "exactness",
    "popularity:desc"
  ]
}

Instant Search UI

Use InstantMeilisearch for React/Vue/vanilla JS:

import { InstantSearch, SearchBox, Hits } from 'react-instantsearch-dom';
import { instantMeiliSearch } from '@meilisearch/instant-meilisearch';

const searchClient = instantMeiliSearch('http://localhost:7700', 'your-key');

function App() {
  return (
    <InstantSearch indexName="products" searchClient={searchClient}>
      <SearchBox />
      <Hits hitComponent={ProductHit} />
    </InstantSearch>
  );
}

Meilisearch vs Alternatives

FAQ

Q: What dataset size is Meilisearch suitable for? A: It's suited to small-to-medium datasets (hundreds of thousands to tens of millions of documents). The official recommendation is under 10M documents per instance. For hundred-million-scale data, ElasticSearch or Qdrant is a better fit.

Q: How well does it handle Chinese search? A: Excellent. Meilisearch has built-in Chinese tokenization (jieba) with solid support for Simplified, Traditional, and mixed Chinese-English queries — no extra configuration required.

Q: What's the main difference compared to Algolia? A: Algolia is a paid hosted service; Meilisearch is free and self-hostable. Features are similar and search speed is on par. For cost-sensitive projects, Meilisearch is a perfect alternative.

Sources & Credits

• GitHub: meilisearch/meilisearch — 57K+ ⭐ | MIT
• Website: meilisearch.com