Install Meilisearch on Linux

sudo systemctl daemon-reload
sudo systemctl enable --now meilisearch

Verify it is running

systemctl status meilisearch

You should see active (running). Then confirm the API responds:

curl -s http://127.0.0.1:7700/health

Expected output: {"status":"available"}

Create an Index and Add Documents

Meilisearch creates an index automatically when you first push documents to it. Replace MASTER_KEY with your actual key in all subsequent requests, or create a scoped API key (recommended for application use).

curl -s -X POST 'http://127.0.0.1:7700/indexes/products/documents' \
  -H 'Authorization: Bearer MASTER_KEY' \
  -H 'Content-Type: application/json' \
  --data-binary '[
    {"id": 1, "name": "Mechanical Keyboard", "category": "peripherals", "price": 129.99},
    {"id": 2, "name": "USB-C Hub", "category": "peripherals", "price": 39.99},
    {"id": 3, "name": "27-inch Monitor", "category": "displays", "price": 349.00}
  ]'

Meilisearch processes document additions asynchronously. The response contains a taskUid. Check task status:

curl -s 'http://127.0.0.1:7700/tasks/0' \
  -H 'Authorization: Bearer MASTER_KEY'

When "status":"succeeded" appears, the documents are indexed and searchable.

Run a search

curl -s -X POST 'http://127.0.0.1:7700/indexes/products/search' \
  -H 'Authorization: Bearer MASTER_KEY' \
  -H 'Content-Type: application/json' \
  --data-binary '{"q": "keybord"}'

Notice the deliberate typo — Meilisearch still returns the keyboard result. That is typo tolerance working by default.

Configure Typo Tolerance

Meilisearch enables typo tolerance automatically, but you can tune it per index. The key settings are the minimum word length for allowing one or two typos, and which fields to exclude.

curl -s -X PATCH 'http://127.0.0.1:7700/indexes/products/settings/typo-tolerance' \
  -H 'Authorization: Bearer MASTER_KEY' \
  -H 'Content-Type: application/json' \
  --data-binary '{
    "enabled": true,
    "minWordSizeForTypos": {
      "oneTypo": 4,
      "twoTypos": 8
    },
    "disableOnAttributes": ["price"]
  }'

This means words with 4+ characters can have one typo, words with 8+ characters can have two typos, and the price field is never matched by typo. After sending this PATCH, poll the returned task until it succeeds before testing.

To disable typo tolerance entirely on an index (useful for SKU or ID fields):

curl -s -X PATCH 'http://127.0.0.1:7700/indexes/skus/settings/typo-tolerance' \
  -H 'Authorization: Bearer MASTER_KEY' \
  -H 'Content-Type: application/json' \
  --data-binary '{"enabled": false}'

Sync Data from PostgreSQL

For production workloads, you need a reliable way to keep Meilisearch in sync with your source-of-truth database. The two practical approaches are a polling script or a trigger-based approach using pg_notify. The polling approach is simpler and shown here.

Install dependencies

On Debian/Ubuntu:

sudo apt install python3-pip python3-venv
python3 -m venv /opt/meili-sync
/opt/meili-sync/bin/pip install psycopg2-binary meilisearch requests

On Fedora/RHEL:

sudo dnf install python3-pip
python3 -m venv /opt/meili-sync
/opt/meili-sync/bin/pip install psycopg2-binary meilisearch requests

Write the sync script

sudo tee /opt/meili-sync/sync.py > /dev/null <<'PYEOF'
import psycopg2
import meilisearch
import os

PG_DSN = os.environ["PG_DSN"]  # e.g. "host=localhost dbname=myapp user=app password=secret"
MEILI_URL = os.environ.get("MEILI_URL", "http://127.0.0.1:7700")
MEILI_KEY = os.environ["MEILI_MASTER_KEY"]
INDEX_NAME = "products"

client = meilisearch.Client(MEILI_URL, MEILI_KEY)
index = client.index(INDEX_NAME)

with psycopg2.connect(PG_DSN) as conn:
    with conn.cursor() as cur:
        cur.execute("SELECT id, name, category, price FROM products")
        rows = cur.fetchall()
        columns = [desc[0] for desc in cur.description]
        docs = [dict(zip(columns, row)) for row in rows]

task = index.add_documents(docs, primary_key="id")
print(f"Enqueued task {task.task_uid} with {len(docs)} documents")
PYEOF

This performs a full re-index on each run. For large tables, add a WHERE updated_at > last_sync_ts clause and track the high-water mark in a small state file or a dedicated Postgres table.

Schedule the sync with a systemd timer

sudo tee /etc/systemd/system/meili-sync.service > /dev/null < /dev/null <

sudo systemctl daemon-reload
sudo systemctl enable --now meili-sync.timer

systemctl list-timers meili-sync.timer

sudo ufw allow 443/tcp

sudo firewall-cmd --permanent --add-service=https
sudo firewall-cmd --reload