Apr 30, 2023 23:00:00

Open source that operates at super high speed and can be self-hosted Japanese is also OK search engine `` Meilisearch '' finally released the first stable version after 3 years from development, so I tried using it

Meilisearch , a high-performance search engine with functions such as sequential search, typo resistance, faceted search, geosearch, and multi-tenancy, released its first stable version, version 1.0, in February 2023. I immediately ran it locally and checked the Japanese search performance.

Meilisearch 1.0: the next stage in search

https://blog.meilisearch.com/v1-enterprise-ready-stable/

An English demo is available on the official website , so you can easily check the usability. When I started typing 'mario' as a test, the sequential search was enabled, so results were returned even in the middle.

Entering 'mario' hits 36 results. The time it took to search was 4 milliseconds, an amazing speed.

According to

a survey by Deloitte and Google , a mere 0.1 second improvement in page display speed increased the purchase rate of retail sites by 8.4% ...

In the case of travel sites, the application rate increased by 10.1%. As you can see, speed is very important in the world of the web.

Meilisearch is designed from scratch with speed pursuit in mind, so it can return search results at explosive speed. Searches are not limited to English, Meilisearch supports all languages, and has special optimizations for Japanese, Chinese, Hebrew, Thai, and Korean. Explosive search is possible in various languages.

Other features of Meilisearch include:

・ Typo resistance
Even if the input is slightly wrong, the result is what the user expects.

・Faceted search
The search can be narrowed down by the user's selection of predefined search conditions, such as searching only for items with a specific tag.

・Geosearch
Search results can be filtered and sorted according to the user's current location information.

・Multi-tenancy
Even if the information that can be viewed by each user is different, it is possible to set the search target to only the information that the user who is searching can access.

◆ I actually used it
I was able to fully experience its speed and responsiveness with the demo version, but I will try whether Meilisearch can demonstrate its ability even when searching Japanese data with self-hosting. This time we plan to test the operation using WSL2. The installation of WSL2 is summarized by Microsoft below.

Install WSL | Microsoft Learn
https://learn.microsoft.com/ja-jp/windows/wsl/install

If you check the quick start guide of Meilisearch , it seems that you can install it with the following command.

[code]curl -L https://install.meilisearch.com | sh[/code]

It took 4 seconds from command input to completion of installation. It's already blazing fast.

If you enter the command './meilisearch' as it is, the server will start. Since the address of the dashboard is displayed, copy it ......

Paste it into your browser. I clicked 'Select an index', but it says 'no index found' because I haven't added any data yet.

This time, we have extracted the title of each animation from

the list of new animations starting in the spring of 2023 as Japanese data. Since it is necessary to set the primary key when saving the data, add 'id' and save it collectively in JSON format .

Send the data to meilisearch with the curl command as per

the documentation instructions .
[code] curl \
-X POST 'http://localhost:7700/indexes/anime/documents?primaryKey=id' \
-H 'Content-Type: application/json' \
--data-binary @anime.json[/code]

Then, the place where 'Select an index' display was automatically switched to 'anime', and the data saved earlier appeared at the bottom.

The performance is good, such as 0 millisecond search time.

Although it is strong against typos, it seems to be weak against misconversions, and there are no hits due to differences in hiragana and katakana and mistakes in kanji conversion.