Skip to main content
Ayase Quart includes native database search, but you can add a dedicated search engine for faster and more advanced search capabilities.

​
Search options

Ayase Quart provides two search methods:
  • Vanilla search - Native SQL-based search (always available)
  • Index search - Full-text search using dedicated search engines (optional)
Index search is optional but recommended for better performance on large archives.

​
Supported search engines

EngineStatusNotes
LNXFully supportedRecommended, extremely fast (<2ms for 100 results)
MeilisearchPartial supportLow RAM usage, good disk efficiency
TypeSensePartial supportC++ based, caching support
Only LNX 0.9.0 is currently supported. LNX 0.10.0 is incomplete and not recommended.
LNX offers the best performance with sub-millisecond search times.
1

Review Docker configuration

Navigate to the LNX directory and review the configuration:
cd ~/ayase-quart/index_search/lnx/
cat docker-compose.yml
The default configuration:
services:
  lnx:
    image: chillfish8/lnx:0.9.0
    network_mode: 'host'
    stdin_open: true
    environment:
      - AUTHORIZATION_KEY=hello
      - LOG_LEVEL=info
    volumes:
      - "./data:/index"
Change AUTHORIZATION_KEY from the default value for security.
2

Configure Ayase Quart

Edit config.toml to enable index search:
[index_search]
enabled = true
provider = 'lnx'
host = 'http://localhost:8000'
headers = { content-type = 'application/json', Authorization = 'hello' }
multi_board_search = true

[index_search.lnx]
max_concurrency = 4
reader_threads = 8
writer_threads = 8
writer_buffer = 2_147_483_648
  • max_concurrency: Range 4-16. Keep low for large datasets to prevent I/O contention.
  • reader_threads / writer_threads: Set to your vCPU count.
  • writer_buffer: Bytes before disk flush. Default 256MB, shown is 2GB.
3

Start LNX container

In Terminal A, start the LNX container:
cd ~/ayase-quart/index_search/lnx/
sudo docker-compose up
Don’t use -d flag initially so you can verify data is being populated.
4

Create search index

In Terminal B, create the search index:
cd ~/ayase-quart
source venv/bin/activate
ayaseq search index create
5

Load board data

Load your boards into the search index:
ayaseq search load full a b c g gif
Replace board names with the ones you’re archiving. You should see progress bars, and Terminal A should show LNX processing data.
6

Run in background

After loading completes, stop the container (Ctrl+C in Terminal A) and restart in detached mode:
sudo docker-compose up -d

​
Automation with systemd

Automate LNX startup on server reboot with a systemd service.
1

Create startup script

Create /usr/local/bin/start_lnx.sh:
#!/usr/bin/env bash
set -e

# Clear caches for better performance
free
sync
echo 1 > /proc/sys/vm/drop_caches
free

cd /mnt/aq/index_search/lnx/
/usr/bin/docker compose up -d
Make it executable:
sudo chmod +x /usr/local/bin/start_lnx.sh
Update the path to match your LNX installation directory.
2

Create systemd service

Create /etc/systemd/system/start_lnx.service:
[Unit]
Description=Drop caches and start lnx docker container
Requires=mnt.mount docker.service
After=mnt.mount docker.service

[Service]
Type=oneshot
WorkingDirectory=/mnt/aq/index_search/lnx/
ExecStart=/usr/local/bin/start_lnx.sh
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target
Adjust Requires and After based on your mount points. The example waits for /mnt to be mounted.
3

Enable service

Enable and start the service:
sudo systemctl daemon-reload
sudo systemctl enable start_lnx.service
sudo systemctl start start_lnx.service
sudo systemctl status start_lnx.service

​
Search index management

Manage your search index with the ayaseq search commands.

​
Create index

Initialize a new search index: 
ayaseq search index create

​
Delete index

Wipe all index data: 
ayaseq search index delete

​
Reset index

Delete and recreate the index: 
ayaseq search index reset

​
Load boards

Load specific boards into the index: 
ayaseq search load full a b c g gif pol

​
Incremental updates

Update the index with new posts: 
ayaseq search load incr a b c
Run continuously with cron interval (in seconds): 
ayaseq search load incr a b c --cron 300
Configure the native SQL-based search in config.toml: 
[vanilla_search]
enabled = true
highlight = false
hits_per_page = 50
max_hits = 1_000
multi_board_search = false

​
Search engine comparison

​
LNX

Pros:
  • Extremely fast (<2ms for 100 results)
  • Uses Tantivy for indexing
  • Consistent performance under load
Cons:
  • No boolean or null support
  • Single writer lock required
  • Documentation lacking
  • Version 0.10.0 incomplete

​
Meilisearch

Pros:
  • Very low RAM usage
  • Leverages disk efficiently
  • Built-in debug UI
Cons:
  • No infix search
  • Slow ingestion (bottlenecked on single core)
  • Single node only
  • Only considers first 10 words of search queries

​
TypeSense

Pros:
  • Caching support for reduced load
  • C++ based for performance
Cons:
  • No null support
  • No primary key concept (duplicates possible)
  • Limited result sets (100 works, 10,000 fails)
  • Requires authentication even in dev mode

​
Next steps

Moderation guide

Set up content moderation and reporting

Production deployment

Deploy Ayase Quart to production

Build docs developers (and LLMs) love

Get started for freeTalk to us