v0.30 of Meilisearch was released earlier this month. With it, our primary goal was to make the Meilisearch workflow as smooth as possible. After all, we are on the road to v1, and our destination is closer than ever. For that, having index deployment with zero downtime was a crucial step. We’re happy to say that v0.30 solves this problem, making it easier than ever to update an index in production. Read on to learn more about how the feature works!

  
The challenge
----------------

Suppose you have changes in your database that you need to synchronize with your Meilisearch index in production. What do you do? Well, if you were working on Meilisearch v0.29, you might:

### Update your index

🙅‍♀️  Simple, right? But if you’ve ever tried it, then you should already know…

![It's a trap!](https://media.tenor.com/FtNhFA1nSo0AAAAC/star-wars-its-a-trap.gif)

Updating your production index while it receives search queries can result in inconsistencies with results or even missing information. You don’t want to offer that search experience, do you? Otherwise, you wouldn't be using Meilisearch 😁

### Delete the index, create a new index with the same name, and re-index the data

⌛ Each step will take some time. Even if it’s just a few seconds, it’s a few seconds of downtime… And that’s at least three requests! A few seconds here plus a few seconds there can result in minutes of downtime! No, thank you very much.

### Create a new index with another name, stage the changes there, then modify all the clients to point to the new index

🤔 Well, it’s not the worst option. But again, updating the clients leads to downtime 😞 In some cases, it isn’t even up to the developer when the client update will be pushed to production. Take an iOS app for example: Apple has to review the new version before you can release it, which can be a major drag on your release plan.

### Create an intermediate redirection layer to avoid downtime

🧞 Very clever! While this is a good approach, it's not the most straightforward. It requires architectural knowledge, time, and extra tools.

Aren't you glad you never have to do any of these things again?

The silver bullet
-----------------

Meilisearch has always been committed to offering the best developer experience.

Our team has been working hard for the past months, and after several iterations, they’ve come up with a solution that integrates seamlessly into the developer’s stack: **index swapping**.

Let me show you how it works.

Let’s say you have an index in production—`indexA`—in which your clients search. You want to sync the changes in your primary database with Meilisearch. To do so, you would follow the steps below.

### Step 1: Create a new index with up-to-date data

First, you need to create an index—let’s call it `indexA_new`—representing the new version of `indexA` that you want to deploy to your search clients. Add up-to-date documents from your database, and update the index settings if necessary.

Don’t forget to check that all the tasks related to the index creation have successfully completed, including `indexCreation`, `settingsUpdate`, and `documentAdditionOrUpdate` task types. You can use the `/tasks` [route](https://docs.meilisearch.com/reference/api/tasks.html#tasks) to get information about their progress.

### Step 2: Test the new index

Before sending your index to production, you want to be sure everything is working as expected.

![Meme: image of a city on fire in the middle of a thunderstorm, the text reading: "text in production" they said, "what could go wrong" they said ](https://unable-actionable-car.media.strapiapp.com/test_in_production_meme_1_63b857925c.jpg)

Make sure to update the settings for any new fields you may have introduced with the updated data. For example, you may want to add new fields to `searchableAttributes`, `filterableAttributes`, and/or `sortableAttributes`. Don't forget to double-check the relevancy of search results any time you add new data!

👉 Keep in mind that the `searchableAttributes` list not only designates the fields that are searchable, but also dictates the [attribute ranking order](https://docs.meilisearch.com/learn/core_concepts/relevancy.html#attribute-ranking-order). Make sure any new fields you may have introduced are added to the list in the right order.

### Step 3: Swap indexes

Once your `indexA_new` has been successfully created, filled with data, and tested, it’s ready to be deployed with an **index swap**. To do so, send a `POST` request to the `/swap-indexes` endpoint. Specify the indexes you want to swap in the payload. Since it’s a swap, the order doesn’t matter.

```bash
curl \
-X POST 'http://localhost:7700/swap-indexes' \
-H 'Content-Type: application/json' \
--data-binary '[
	{ "indexes": ["indexA", "indexA_new"] }
]'
```

👉 In a [protected](https://docs.meilisearch.com/learn/security/master_api_keys.html#protecting-a-meilisearch-instance) Meilisearch instance, the API key used to swap indexes must have access to the `indexes.swap` action as well as the indexes you want to swap. If not, Meilisearch will throw an `invalid_api_key` [error](https://docs.meilisearch.com/reference/errors/error_codes.html#invalid-api-key). For more information about creating API keys with specific permissions, see the [documentation](https://docs.meilisearch.com/learn/security/master_api_keys.html#creating-an-api-key).

You can use the response's `taskUid` to track the status of your request with the `GET /tasks/{task_uid}` [endpoint](https://docs.meilisearch.com/reference/api/tasks.html#get-tasks). A successful index swap should look like this:

```JSON
{
  "uid": 23,
  "indexUid": null,
  "status":"succeeded",
  "type":"indexSwap",
  "details":{
	"swaps": [
  	     {"indexes": ["indexA", "indexA_new"]},
	]
  },
  "duration": "PT1S",
  "enqueuedAt": "2021-08-10T14:29:17.000000Z",
  "startedAt": "2021-08-10T14:29:18.000000Z",
  "finishedAt": "2021-08-10T14:29:19.000000Z"
}

```

Your indexes have been swapped without any downtime! The documents, settings, and task history of `indexA`—except for any enqueued tasks—have been swapped with those of `indexA_new`. Every mention of `indexA` in the task history has been replaced by `indexA_new` and vice-versa (`enqueued` tasks are left unmodified).

![Animation illustrating index swapping: an index receives search queries while the another is being created in the background. When the second index is ready, they swap, then the unnecessary index is deleted.](https://unable-actionable-car.media.strapiapp.com/ezgif_4_520c610091_8f72080b3b.gif)

After the swap, `indexA_new` holds the outdated content. You can delete it or keep it as a backup, should something go wrong and you need to swap back. Better safe than sorry!

And that’s it! Only three steps, two if you like to live dangerously. Could this get any better?

The cherries—yes, plural—on the cake 🍰
---------------------------------------

What if I tell you that you can swap several indexes with just one request?

I am not kidding. A single request can swap as many index pairs as you wish. Meilisearch can deploy all changes at the same time. Clients will access the new version of all indexes at once without any downtime.

```bash
curl \
  -X POST 'http://localhost:7700/swap-indexes' \
  -H 'Content-Type: application/json' \
  --data-binary '[
    {
        "indexes": ["indexA", "indexA_new"]
    },
    {
        "indexes": ["indexB", "indexB_new"]
    },
    {
        "indexes": ["indexC", "indexC_new"]
    }
]'

```

In the example above, three swap operations will occur simultaneously and atomically.

Wait, what?

Yes, you can read that again. It’s atomic! Either all indexes are successfully swapped, or none are. Either all the content is swapped, or none is.

Why is that important? It prevents partial changes in the database, ensuring consistency and, thus, a top-notch search experience.

Conclusion
----------

As I mentioned earlier, this feature has been in the works for several months, and it all started with user feedback. The [“swap indexes” card](https://roadmap.meilisearch.com/c/65-swap-indexes) on our roadmap got 38 votes and almost as many comments explaining use cases where this feature is a must-have. Here are some of our favorites:

*   “It would be very useful when trying to change rules on production database to tweak for the best results”
*   “Would help if index was accidentally created with wrong name or need to rename/change it for various reasons.”
*   “We need it to clean out all deleted items.”

These are just a few examples. This kind of insight is extremely helpful for our team, as it allows us to shape the product to fit the users' needs.

While very convenient—you can have an overall view of all the feature ideas submitted, in the making, and released—the [roadmap](https://roadmap.meilisearch.com/) is just one of the options to provide feedback.

The [product discussions](https://github.com/meilisearch/product/discussions) on GitHub are, in my opinion, one of the best places to explain your needs. You can have direct contact with the Meilisearch product team, and, being public, it allows anyone to participate and enrich the process.

The last option is our brand-new [Discord server](https://discord.gg/meilisearch). Don’t hesitate to join us and talk about what you’ve built with Meilisearch, your use case, and your specific needs.

No matter which option you choose, we look forward to hearing from you!

Zero downtime index deployment

Meilisearch v0.30 is out and brings with it the ability to create numbered page selectors. You can read more about the new features and updates from v0.30 in our [release blog post](/blog/whats-new-in-v0-30/)his article will go over the different ways of navigating search results and the pros and cons of each method.

### What is pagination, and what are some common alternatives?

Let’s say you go to an ecommerce website to buy a new pair of shoes. It would be quite overwhelming if you got 700 pairs of shoes to choose from as soon as the page loads—not to mention that fetching every pair of shoes would certainly slow down the site.  

To avoid this, most websites will break search results up into chunks, such as pages, and load a certain number at a time based on input from the user.  

There are three main ways to display large datasets: pagination, infinite scroll, and load more. Each method has benefits and downsides that make it suited for certain use cases and content types. For example, Google’s use of numbered pagination makes it easier to find a specific result again, while Twitter, Facebook, and Instagram use infinite scroll to keep users glued to their screens. Google uses load more on mobile devices to display search results.  

Changing from one type of pagination to another can completely alter user experience.  Imagine waking up to a paginated version of Twitter, where you have to continuously click a “Next” button to view more posts in your feed (thanks, Elon). This would completely change the way you interact with the site!  

In making the choice of how to display search results, you decide how much content to show users at one time, what action they have to take to view more, and how convenient it is to find a specific result again.

### Pagination

Pagination works by breaking search results down into separate pages. There are two main types of pagination interfaces: numbered ones, where you find a list of numbered pages at the bottom of each page, and unnumbered ones, where you navigate from page to page using the “Previous” and “Next” buttons.  

Pagination is one of the most popular ways to display search results and is particularly common on ecommerce websites. Pagination breaks down content, making it easier to process. Since each page is distinct, users can easily return to items they liked or share them with others. This helps users to search for something specific instead of browsing through a never-ending list of items.  

The biggest downside to pagination is that most users don’t look beyond the first few pages and hence never get to look at more results. When was the last time you went to the second page of Google’s search results?

![](https://media.tenor.com/zmDrt3416pcAAAAC/think-winniethepooh.gif)

Since the results are replaced with every new page, users need to switch between pages to compare results. In addition, pagination isn’t very mobile-friendly. Mobile users prefer load more or infinite scroll as it’s more intuitive. It also requires additional actions from the user: clicking the “Next” button and waiting for the page to load. This may also give the impression that pagination is slow.

#### Numbered page selection

Page numbers add clarity to the search process and guide users to what they’re searching for. Numbered page selection offers numerous advantages, such as:  

*   Users can be shown the total number of pages of results, allowing them to estimate how long they will need to find a particular item. (“I have 25 pages worth of shoes to sift through!”)
*   With numbered pages, it’s clear that some pages are more relevant than others. (“I didn’t find those shoes on the first five pages, so this website probably doesn’t sell them.”)
*   Users can remember the positioning of an item, and numbered pages allow them to return to it faster. (“I liked the pink shoes on page 3.”)
*   Users can generally stop in the middle of a session and continue from where they left off. (“I stopped at page 5 yesterday.”)

![](https://unable-actionable-car.media.strapiapp.com/Screen_Shot_2022_11_29_at_16_12_39_22ff2f4296.png)

#### “Previous” and “Next” buttons

In numberless pagination, users navigate via “Next” and “Previous” buttons instead of a list of numbered pages. This alternative to numbered pagination is often used for mobile displays where there isn’t enough space to display page numbers. Since there are only two buttons, it offers less precision, as users can’t jump to specific pages. However, it offers better performance as it isn’t necessary to calculate the total number of search results.  

![](https://lh4.googleusercontent.com/7g66Jnt6E-GR1esTMZrmCv6MDvHMOn-FprG5cHxYQhmHGQLS8vYFvlmOlpaBdjmf_pc285dIACe3SeUNSj2UJ9C2JMfk4cxoxbqFuS_eD4q9qZp1h2yhIaz-9ng7xB3ly39_PvvtUNqm6L0a5euKFYvZSn9ttg5NzKPQMmz4X0k37XiGqD7RnkuVwT9XKg)

### Infinite scroll

Infinite scroll has become increasingly popular over the past decade due to the rise in smartphones. It loads more content seamlessly as the user scrolls, reducing friction. It’s also more intuitive than pagination; you just keep scrolling to view more content! Additionally, since all the content is loaded to a single page, it may seem faster than traditional pagination.

![](/blog/content/images/2022/11/IMG_3391.GIF

The biggest downside to infinite scroll is poor usability.  The page length increases infinitely, making it difficult to find previously viewed results again, nor can users bookmark or share a particular set of items on the list.  Page performance slows down as more content is loaded, making the page “heavier”. Meanwhile, the site footer may be inaccessible since infinite scroll pushes it endlessly down as more content is loaded. Since Googlebot cannot emulate scrolling, infinite scroll also has a negative impact on SEO.  

However, the greatest impact by far is on user behavior. At some point, users of infinite scroll will stop paying attention to individual results and begin to focus more on scrolling, leading to decreased engagement. You can read more about this in [Baymard Institute's study](https://www.smashingmagazine.com/2016/03/pagination-infinite-scrolling-load-more-buttons/) on design patterns for loading products on desktops and mobile devices.  

Based on the above, infinite scroll may not be the best solution for ecommerce websites. How do you find the pink shoes you liked 15-ish scrolls ago?  

There is one use case where infinite scroll cannot be beat. For entertainment websites whose main purpose is to keep the users with no particular aim on the page for as long as possible, infinite scroll all but guarantees a longer time on site. This is why infinite scroll is closely linked with social media and other business models that rely on advertising or data harvesting.

### Load more

Like infinite scroll, load more is a mobile-friendly option that works by extending one page infinitely. However, unlike infinite scroll, load more allows users to choose whether they want to view more results by clicking a “Load more” button at the bottom of the page. This critical difference gives users more control, making it easier, for example, to access the footer or keep track of a specific result. Studies show that it also increases the amount of attention users give to each individual item.  

Load more shares many downsides with infinite scroll, such as poor page performance. Like infinite scroll, it’s not ideal for SEO as Googlebot cannot click the “Load more” button. This means it cannot crawl or index all of your content.  

![](https://lh5.googleusercontent.com/-xmPwy0C6ZdgZoHNMV4fmRdxR_g9scVY5J0ktGY_k0oh24Un80WooYE9zSRo_TxNhPJE5--hAQJ-UJIRPejnFZt4pcNca6l4z6JedvwrjdKucMBAurVeBgMgIGYWS9unr2cOEA-1CZYBwlQ0ySexv2HqYSWqCSemBCVK5_jb-VweLwczjNcXQjl3ghxdmA)

### So, what do I use?

The short answer: it depends on the kind of product experience you want.  

**Pagination**

*   Your users are looking for highly specific results
*   SEO is very important to you

**Infinite scroll**

*   You’re building a social media site
*   Your users want to explore a large amount of content

**Load more**

*   You want to hide your content behind a paywall
*   You want to build an ethical mobile experience
*   You want to gain some of the benefits of both pagination and infinite scroll  
    

Or you could go hybrid! Consider Ikea: they use a load more button but also display the total number of search results and a progress bar, enabling users to see how many results they’ve viewed so far.  

![](https://unable-actionable-car.media.strapiapp.com/Screen_Shot_2022_11_29_at_18_45_56_68c9b2fc30.png)

While we have covered the three main ways to display search results today, they are far from set in stone. At the end of the day, your goal is to help your users find the content they need. If you understand which key goals you want to accomplish with your search navigation interface and which pitfalls you want to avoid, you may be able to design something even better than what’s been done before. Good luck, and let us know what you create with Meilisearch!

Navigating search results: pagination vs infinite scroll vs load more

These past few years, our team has observed the growth of our community across multiple online spaces such as Twitter, GitHub, LinkedIn, and especially our public Slack workspace. We were and are thrilled to see more and more people interested in our work and our company, and applied ourselves to making everyone feel welcome. Eventually, however, we began to consider a new option: the possibility of migrating our community from Slack to [our very own brand new Discord server.](https://discord.meilisearch.com) It was only a matter of time, since Discord has established itself as _the_ reference in terms of community platforms.

Such a change had to be carefully considered since it meant migrating our quite large Slack community onto a brand new platform. We didn't want to go through with it without a good reason and even better plan of action.

**First of all, why did we decide to move to Discord?**

Discord is difficult to ignore. It has become an extremely popular and highly democratized platform, with an average of 150 million active monthly users\*. Discord’s large user base was a very good argument in its favor, but it wasn't the only reason for our decision. Discord is already the happy home of various online communities from within the tech space, including our friends at [AppWrite](https://appwrite.io/) and [Orbit](https://orbit.love/). As such, the platform is very well designed for maintainers. Some of its benefits include:

*   A large variety of moderation tools, many of which can be automated thanks to bots
*   A variety of roles that can be assigned with different permissions, to both valorize community members and identify moderators
*   The option to make the Code of Conduct mandatory before joining the servers to help protect the space from bad actors
*   Unlimited and expiration-free invites
*   Many inclusive options, such as allowing people to assign themselves their preferred pronouns

When considering all of this, the general consensus was that a move to Discord was the best option for both our community members and the moderators.

Luckily, we aren't the first ones to consider this move and go through with it! After taking the decision, our next step was to diligently do our homework: research how other companies had handled the change and apply their best practices.

![](https://unable-actionable-car.media.strapiapp.com/Screenshot_2022_11_28_at_16_26_14_a0a03dd987.png)

**Secondly, how to manage the migration?**

As we embarked on the actual migration, the very first step taken well in advance was to prepare our future Discord server. We asked around internally to figure out which features would be expected and whether each team wanted its own space. Based on the initial feedback, we created the first proposal and invited the team to have a look around and share their impressions. At this time some team members raised issues around inclusivity, especially concerning the text-to-speech feature, and we did our best to address these in order to ensure a good experience for as many people as possible. Based on this second wave of feedback, the server was adjusted accordingly and the design was locked in.

There was also the question of timing. Our original plan was to have a post-new year 2023 migration. However, we had to consider the impact of other upcoming announcements. We recently announced and are soon to release our Cloud offering and v.0.30 of our search engine, which we expect to translate into an increase in awareness and an influx of newcomers to our community. We were concerned about having users join us only to be told a month later they would have to migrate to a completely different platform. Alongside this, we were preparing for our first sponsorship of the [Advent of Code](https://adventofcode.com/), a month long event offering challenging programming puzzles, for which we also wanted to create an [original coding challenge](https://aoc.meilisearch.com/). It is common for tech-oriented servers to host their very own Advent of Code channel; without a Discord, it wasn't something we could easily offer.

A December migration seemed very interesting from a strategic standpoint, the only issue being the short turnaround to communicate the upcoming changes to our Slack users.

**Finally, what happens now with our Slack?**

As we mentioned earlier, this change was greatly motivated by our desire to offer a more friendly space to our community members. Our aim isn't to take people by surprise and rush them around. Our Slack workspace will remain open for the next few months so that members who missed the initial announcement won’t simply be cut out and left to wonder what exactly happened. We will keep publishing regular announcements throughout these next few months to explain the situation and instruct people on how they can join us on this new platform. These efforts will also be supported on our various social media to get the word around as much as possible. We strongly encourage our community members to ask any questions they might have so that we can answer these queries transparently and reassure anyone feeling uncertain about this decision.

This migration was very carefully considered by Meilisearch, and we look forward to hearing your opinion about our new server. We intend to take full advantage of Discord's new options, and are excited to have you [joining us](https://discord.meilisearch.com) in this new adventure. Thank you for your continued support!

\*[https://www.cloudwards.net/discord-statistics/](https://www.cloudwards.net/discord-statistics/)

From Slack to Discord: our migration

In this article, let's take a look at some of the most significant changes in Meilisearch's latest update. You can read [the full changelog on GitHub](https://github.com/meilisearch/meilisearch/releases/tag/v0.30.0), but we’ll go over the main ones in this article.

Zero downtime index deployment
------------------------------

Syncing your database changes with your Meilisearch index results in downtime. You cannot afford a “broken” search: user experience is paramount. You’ve come up with some workaround, but it takes time and resources. There must be a simpler way to handle this, right?

v0.30 introduces index swapping. You can now deploy a new index version with zero downtime!

To synchronize the changes between your primary database and your Meilisearch index  —`indexA`— in production, you just need to take the following steps:

*   Step 1: Create a new version of the index —`indexA_new`— containing all those changes
*   Step 2: Swap indexes, which means you replace the documents, settings, and task history of `indexA` with those of `indexA_new`. This is the request you need to send for the magic to happen:

```bash
curl \
  -X POST 'http://localhost:7700/swap-indexes'\
  -H 'Content-Type: application/json' \
  --data-binary '[
      { "indexes": ["indexA", "indexA_new"] }
  ]'
```

⏲️ Wait for the task’ status to update to `succeeded`...

🎉 Tada! Your changes are now in production! Your users are still searching in `indexA`, but it holds the data of `indexA_new` now, and `indexA_new` holds the data of `indexA`.

New features: task cancelation and deletion
-------------------------------------------

v0.30 gives you more control over asynchronous operations: you can now cancel tasks and delete your task history.

### Task cancelation

The inability to cancel tasks has been one of the recurring problems users have raised over time. Not being able to interrupt tasks enqueued by mistake or taking too long to process can be very frustrating.

With v0.30, this problem is now a thing of the past. With the new API route `/tasks/cancel`, you can cancel enqueued or processing tasks based on their:

*   `uid`
*   `status`
*   `type`
*   `indexUid`
*   the date they were enqueued or started (`beforeEnqueuedAt`, `afterEnqueuedAt`, `beforeStartedAt`, `afterStartedAt`)

You can learn more about tasks in our [documentation](https://docs.meilisearch.com/learn/advanced/asynchronous_operations.html#understanding-tasks).

To cancel all enqueued tasks that add or update documents to an index, you should send the following request:

```bash
curl \
  -X POST 'http://localhost:7700/tasks/cancel?statuses=enqueued&types=documentAdditionOrUpdate’

```

### Task deletion

Succeeded or failed —and now canceled— task information can take up much-needed disk space in instances with a large number of asynchronous operations. Clean up finished tasks with the new `DELETE /tasks` endpoint.

Just as for task cancelation, task deletion can be done based on the tasks’:

*   `uid`
*   `status`
*   `type`
*   `indexUid`
*   the date they were enqueued or started (`beforeEnqueuedAt`, `afterEnqueuedAt`, `beforeStartedAt`, `afterStartedAt`)

But you can also delete them based on the date they finished processing and what task canceled them:

*   `beforeFinishedAt`
*   `afterFinishedAt`
*   `canceledBy`

The following request deletes all tasks processed before October 11, 2020, at 11:49:53 am.

```bash
curl \
  -X DELETE 'http://localhost:7700/tasks?beforeFinishedAt=2020-10-11T11:49:53.000Z'

```

New feature: configuration file support
---------------------------------------

Thanks to our dear contributor [mlemesle](https://github.com/mlemesle/), a new way of configuring your Meilisearch instance has entered the chat. Along with environment variables and CLI options, you can use a configuration file 🥳

Configuration files can be easily shared and versioned. They allow you to manage all configuration options in one place and give an overall view at a glance.

To do so, set the desired instance options in a single  `.toml`  file:

```toml
env = "production"
master_key = "YOUR_MASTER_KEY_VALUE"

```

You can use the `--config-file-path` CLI option or the `MEILI_CONFIG_FILE_PATH` environment variable to specify a configuration file. Otherwise, Meilisearch will use the file `./config.toml` in the working directory as the default configuration file. You can download a default configuration file with the following command:

```json
curl https://raw.githubusercontent.com/meilisearch/meilisearch/main/config.toml > config.toml
```
<Callout>
  Please note that specifying the `config-file-path` key in the configuration file will throw an error.
</Callout>

If no configuration file is provided, the `Config file path` is set to `none`.

Improvement: search result pagination
-------------------------------------

v0.30 introduces a numbered pagination mode for users who need pagination interfaces.

Up to v0.29, when using the search endpoint, Meilisearch only returned `estimatedTotalHits`, an estimate of how many documents matched the query. For this reason, creating interfaces with a page selector was not recommended.

You can now use the new search parameters `page` or `hitsPerPage` to get the exhaustive number of hits and the total number of pages for a query.

```bash
curl \
  -X POST 'http://localhost:7700/indexes/movies/search' \
  -H 'Content-Type: application/json' \
  --data-binary '{ "q": "american ninja", "page": 1 }'

```

The above request will fetch the first page of results for the given query and will return the following response:

```JSON
{
  "hits": [
        // … 20 hits
    ],
  "query": "american ninja",
  "processingTimeMs": 2,
  "hitsPerPage": 20,
  "page": 1,
  "totalPages": 49,
  "totalHits": 966
}

```

As you can see, it returns the exhaustive number of search results (`totalHits`) and the number of search results pages (`totalPages`), which allows you to create numbered pagination interfaces easily.

  
👉 Note that `totalHits` and `totalPages` are calculated based on the [pagination.maxTotalHits index setting](https://www.meilisearch.com/docs/reference/api/settings#pagination).

Breaking changes
----------------

### Field rename

When [deleting documents by batch](https://docs.meilisearch.com/reference/api/documents.html#delete-documents-by-batch), the corresponding task object contains a field with the number of document ids received. This field, previously called `receivedDocumentIds`, has been renamed to `providedIds`.

### New error

In cases where users try accessing the `/keys` route without setting a master key, Meilisearch now throws a more helpful error  `missing_master_key`:

```JSON
{
    "message": "Meilisearch is running without a master key. To access this API endpoint, you must have set a master key at launch.",
    "code": "missing_master_key",
    "type": "auth",
    "link": "https://docs.meilisearch.com/errors#missing_master_key"
}
```

### Task filter and error naming

To keep naming consistent with the value they can have, we have pluralized the query parameters used to filter tasks:

*   `indexUid` becomes `indexUids`
*   `type` becomes  `types`
*   `status` becomes  `statuses`

We have also renamed the following error codes:

*   `invalid_task_type` to `invalid_task_types_filter`
*   `invalid_task_status` to `invalid_task_statuses_filter`

Hacktoberfest contributions
---------------------------

We are infinitely grateful for all the contributions we received during Hacktoberfest. Most of the changes mentioned above were from these contributions, and we would like to highlight two more. We would not have released them in this version if it were not for our contributors.

*   Support for compressed API requests (Gzip, Brotli, Zlib) by [@mou](https://github.com/mou)
*   Apple Silicon binaries are available thanks to [@jeertmans](https://github.com/jeertmans)

Other changes
-------------

*   Meilisearch has been binding by default to `127.0.0.1:7700` , which is an IPv4 address. In order to make Meilisearch reachable through both [IPv4](https://en.wikipedia.org/wiki/IPv4) (`127.0.0.1:7700`) and [IPv6](https://en.wikipedia.org/wiki/IPv6) (`[::1]:7700`) addresses, we have changed the default address to `localhost:7700`.
*   All task objects now contain the `canceledBy` and `error` fields
*   New [task filters](https://docs.meilisearch.com/learn/advanced/asynchronous_operations.html#filtering-tasks) are available for the `GET /tasks` endpoint:
    *   `uids`
    *   `beforeEnqueuedAt` / `afterEnqueuedAt`
    *   `beforeStartedAt` / `afterStartedAt`
    *   `beforeFinishedAt` / `afterFinishedAt`
    *   `canceledBy`
*   We have introduced a new task type: `snapshotCreation`. Snapshots tasks are now visible in [the task list](https://docs.meilisearch.com/reference/api/tasks.html#get-tasks)
*   The indexing speed has been significantly increased
*   We have reduced both the size of the Meilisearch binary and the size the Meilisearch instance takes up in disk space
*   All instance options now have their corresponding environment variable. Some related to the snapshot and dump features were missing

A big thank you to all our contributors! This project wouldn't have gotten very far without your help 🙏

And that’s it for v0.30! Remember to check the [changelog](https://github.com/meilisearch/meilisearch/releases/tag/v0.30.0) for the full release notes.

What's new in v0.30

Time to close the door behind Hacktoberfest 2022! What an event it was, the time has come to sum up our learnings for this year to better prepare for the future.

For the third time in a row, everyone at Meilisearch was very excited to be participating in Hacktoberferst, probably one of the most popular celebrations of the open-source community online.

We were very grateful for everyone support and contributions throughout the month, please find below the recap we’ve shared internally.

Our teams really enjoyed taking part into the event, leaving us no choice but to go for one more round next year: Hacktoberfest 2023, here we come!

Here, you can find the slides of the presentation with a few comments for context:

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_636027ab45.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_3_082f4c1b7d.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_4_54734eb823.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_5_f0c74dca0b.jpg)

It’s important to keep in mind the way we count our PRs might be slightly different from the official counting. In the following graph, we counted any PR merged during October, and we included the PRs that were validated but not merged in time on October 31st. Some PRs were open before October 1st but still benefited from the “Hacktoberfest approved” tag since they were completed during October.

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_6_cbd60a838b.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_7_7b0c2029be.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_8_d35b08263e.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_9_4c4e155893.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_10_1_78e13c206c.jpg)

We were truly blown away by the number of contributions we received this year! We didn’t quite reach our fourth goal, but we set it as such, knowing it was definitely going to be a challenge. At the same time, we were thrilled to see we weren’t that far away from it!  
  
These objectives being hit isn’t just an ego boost for us. It also translates directly into monetary donations for various organizations we’ve selected before the start of the event. The following organizations were selected on the basis of causes that are important to us. We also tried to find some groups that were geographically close to some of the employees of Meilisearch. This second criterion is because we wish to establish a stronger link with the organizations we donate to. We’d love to remain in touch with them and maybe see how our donation- by extension yours!- has potentially improved things for them.  
  
As we mentioned briefly, these donations wouldn’t have been achievable without your support and contributions throughout the whole month of October! We wish to extend a big thank you to everyone for that.

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_11_1_8804545d22.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_12_a474cd83ac.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_13_625ef09d3b.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_14_00a1b66b56.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_15_1_ba4bae4f61.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_18_2_c65de291a9.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_19_70512c60c4.jpg)

![](https://unable-actionable-car.media.strapiapp.com/External_Hacktoberfest_2022_A_recap_20_ad1a2df74e.jpg)

Hacktoberfest 2022 recap

<Callout>
  This post was originally published in May 2020 by guest author [Riccardo Giorato](/blog/author/riccardo/)t the time, Meilisearch was on v0.09. It has been updated by [Carolina Ferreira](/blog/author/carolina/)o work with Meilisearch v1. You can find the first version of the post on [GitHub](https://github.com/meilisearch/tutorials/tree/main/src/react-decathlon/original-tutorial/original-tutorial.md).
</Callout>

Introduction
------------

In this tutorial, you'll learn how to easily create a search-based web app with instant and reliable results thanks to the power of Meilisearch.

We will cover the basic steps to add our data to Meilisearch, create a custom front-end search, and move on to customization at the end.

For this tutorial, we are going to create an instant search experience for a sports brand. Here is a preview of what you will be building:

![Searching for multiple items on the decathlon demo using Meilisearch](https://unable-actionable-car.media.strapiapp.com/decathlon_d14c32764d.gif)

Prerequisites
-------------

Before getting started, ensure that you have [Node.js](https://nodejs.org/en/) >= 18 installed on your machine.

You can follow this tutorial and write the code as you go through the steps, using this [GitHub project](https://github.com/meilisearch/tutorials/tree/main/src/react-decathlon).

Finally, this tutorial assumes that you are already familiar with React. If that is not the case, you can check the [React Documentation](https://reactjs.org/docs/getting-started.html) to learn more.

Getting Started
---------------

### Clone the repository

Clone the GitHub repository using the following command:

```json
git clone https://github.com/meilisearch/tutorials.git
cd src/react-decathlon

```

### Run a new Docker image

If you cloned the repository to set up the Meilisearch instance, just execute the following commands inside the main folder:

```json
npm install
npm run setup_meili
```

If you didn't clone the repository and want to launch Meilisearch using Docker, execute this command:

```json
docker run -it --rm \
    -p 7700:7700 \
    -e MEILI_ENV='development' \
    -v $(pwd)/meili_data:/meili_data \
    getmeili/meilisearch:v1.0
```
<Callout>
  By default, Meilisearch's API is unprotected. You will need a master key in production. You can learn more about it in our [documentation](https://docs.meilisearch.com/learn/security/master_api_keys.html).
</Callout>

You can check if Meilisearch is running by visiting: [http://localhost:7700/](http://localhost:7700/)

Want to avoid a local installation? To quickly create a best-in-class search experience, we offer the convenience of [Meilisearch Cloud](https://cloud.meilisearch.com/?ref=blog.meilisearch.com), a hosted and fully‑managed version of Meilisearch. There's a free 14-day trial, no credit card required 😉

### Create an index in Meilisearch

An index is an entity where documents are stored, like an array of objects with some specific settings attached to it, and a unique [primary key](https://docs.meilisearch.com/learn/core_concepts/primary_key.html#primary-key).

Each document indexed must have a primary field, a special field that must be present in all documents. This field holds the unique value of the document: its id.

Meilisearch can infer the primary key from your dataset, provided that it contains the `id` substring. You can also set it explicitly.

Below is a sample document to add to Meilisearch.

```javascript
{
  "id": 100013768717,
  "name": "Fitness Foldable Shoe Bag",
  "url": "https://www.decathlon.com/products/gym-foldable-shoe-bag",
  "vendor": "Domyos",
  "category": "Sport bag",
  "tags": [
    "Artistic Gymnastics",
    "Boy's",
    "CARDIO_FITNESS_ACCESSORIES",
    "CARDIO_FITNESS_BAGS",
    "CODE_R3: 11782"
  ],
  "images": "https://cdn.shopify.com/s/files/1/1330/6287/products/sac_20a_20chaussure_20kaki_20_7C_20001_20_7C_20PSHOT_20_490180e6-44e4-4340-8e3d-c29eb70c6ac8.jpg?v=1584683232",
  "creation_date": "2020-04-03T15:58:48-07:00",
  "price": "2.49"
}

```

You can easily create this index with a REST client like [Postman](https://www.postman.com/), but in this tutorial, we will use the [Meilisearch Javascript SDK](https://github.com/meilisearch/meilisearch-js/) to do it directly from Node.js.

```javascript
const { MeiliSearch } = require('meilisearch')

;(async () => {
  try {
    const config = {
      host: 'http://localhost:7700'
    };

    const meili = new MeiliSearch(config);
    
    await meili.createIndex('decathlon'); 
    
    // or you can set the primary key explicitly: 
    // await meili.createIndex({ uid: "decathlon", primaryKey: "id" });
        
  } catch (e) {
    console.error(e);
    console.log("Meili error: ", e.message);
  }
})();

```

You can read more about the properties of indexes in the [Meilisearch documentation](https://docs.meilisearch.com/learn/core_concepts/indexes.html).

### Index documents

Meilisearch receives documents in JSON format and stores them for searching purposes. These documents are composed of fields that can hold any type of data. Meilisearch also accepts datasets in the following formats: NDJSON and CSV. You can read more about the format in the [documentation](https://docs.meilisearch.com/learn/core_concepts/documents.html#upload).

For this tutorial, you can download this dataset full of sportswear items: [**decathlon.json**](https://github.com/meilisearch/tutorials/tree/main/src/react-decathlon/backend/decathlon.json)

Use the following script to upload all the objects from this JSON file to Meilisearch. Remember to change the path to your JSON file before running it!

```javascript
const { MeiliSearch } = require('meilisearch')

;(async () => {
  try {
    const config = {
      host: 'http://localhost:7700'
    };

    const meili = new MeiliSearch(config);

    const decathlon = require("../decathlon.json"); // path to json file

    const index = meili.index("decathlon");
    
    await index.addDocuments(decathlon);
        
  } catch (e) {
    console.error(e);
    console.log("Meili error: ", e.message);
  }
})();

```

Prepare the React app
---------------------

We'll need a standard React app. You can use the project you cloned before in the [Getting started](/blog/instant-search-react-app/#getting-started)ection.

If you prefer to start from an empty app, you can create your own using [Create React App](https://create-react-app.dev/) with the command below. You can name the application however you desire.

```json
npx create-react-app meili_react_demo
cd meili_react_demo

```

### Include Tailwind CSS

To speed up the styling process, add [Tailwind CSS](https://tailwindcss.com/) style directly into the `<head>` element of the index.html file:

```html
  <script src="https://cdn.tailwindcss.com"></script>

```

### Configure App.js state

Then, modify the `App.js` file using this code to set up a simple search form and a few state variables to handle every aspect of the search.

```javascript
import React, { useState, useEffect } from 'react'
import { MeiliSearch } from 'meilisearch'
import Item from './components/Item'

// TODO configure the MeiliSearch Client

const index = client.index('decathlon')

function App () {
  const [searchedWord, setSearch] = useState('')
  const [resultSearch, setResults] = useState([])

  // TODO add function to send searchedWord to Meilisearch

  return (
    <div className='mx-auto'>
      <div className='header font-sans text-white items-center justify-center'>
        <header className='py-12'>
          <img
            className='h-20 w-auto items-center justify-center p-2 mx-auto'
            src='/wide_logo.png'
            style={{ filter: 'invert(0%)' }}
            alt='Decathlon logo'
          />
          <h1 className='flex flex-wrap flex-grow text-3xl w-full justify-center p-4'>
            Stop looking for an item — find it and work hard!
          </h1>
          <div className='border rounded overflow-hidden w-full flex justify-center mx-auto searchBox mt-6'>
            <button className='flex items-center justify-center px-4 shadow-md bg-white text-black'>
              <svg
                className='h-4 w-4 text-grey-dark'
                fill='currentColor'
                xmlns='http://www.w3.org/2000/svg'
                viewBox='0 0 24 24'
              >
                <path d='M16.32 14.9l5.39 5.4a1 1 0 0 1-1.42 1.4l-5.38-5.38a8 8 0 1 1 1.41-1.41zM10 16a6 6 0 1 0 0-12 6 6 0 0 0 0 12z' />
              </svg>
            </button>
            <input
              type='text'
              value={searchedWord}
              onChange={(event) => setSearch(event.target.value)}
              className='px-6 py-4 w-full text-black'
              placeholder='Product, sport, color, …'
            />
          </div>
        </header>
      </div>
      <div>
        <div className='flex flex-wrap searchResults'>
          // TODO iterate over the search results to display them with the Item component
        </div>
      </div>
    </div>
  )
}

export default App


```

This code should output this beautiful header with a search form.

![Decathlon header with a search bar](https://unable-actionable-car.media.strapiapp.com/download_60983dd13f.png)

Search results in React
-----------------------

Connecting React with Meilisearch using the [Javascript SDK](https://github.com/meilisearch/meilisearch-js/) is a simple operation that can be done in just a few steps.

### Meilisearch client

Install the Meilisearch SDK using the following command:

```json
// if you use npm
npm install meilisearch
// if you use yarn
yarn add meilisearch

```

Set up the Meilisearch client with the server URL. In our case, it was the localhost Docker machine. Finally, load the right index from the backend.

Replace this comment in `App.js` with the code snippet below:

"// TODO configure the Meilisearch Client"

```javascript
import { MeiliSearch } from "meilisearch";

const client = new MeiliSearch({
  host: "http://localhost:7700/",
});

const index = client.index("decathlon");

```

### Send the search query

Add a `useEffect` hook to execute the search of the typed words into Meilisearch. All the results will be set to a simple state variable called `resultsSearch`.

Replace this comment in `App.js` with the code snippet below:

"// TODO add function to send searchedWord to Meilisearch"

```javascript
  useEffect(() => {
    // Create a scoped async function in the hook
    async function searchWithMeili() {
      const search = await index.search(searchedWord);
      setResults(search.hits);
    }
    // Execute the created function directly
    searchWithMeili();
  }, [searchedWord]);

```

### Showcase the results

You will iterate over the JSON objects returned by Meilisearch —they'll have the same structure as the uploaded JSON objects— and you'll display them in an `Item` component, linking to the product pages.

Let's create the `Item` component that will help us display our products. Create a components folder with an `Item.js` file inside, and copy-paste the following code snippet:

```javascript
function Item ({ url, image, name, category, vendor, price, id }) {
  return (
    <div className='flex w-full sm:w-1/2 md:w-1/3 lg:w-1/4 xl:w-1/6 p-3' key={id}>
      <a className='flex-1 rounded overflow-hidden shadow-lg' href={url}>
        <img
          className='w-full h-48 object-cover'
          src={image}
          alt={name}
          onError={(e) => {
            e.target.onerror = null
            e.target.src = '/wide_logo.png'
          }}
        />
        <div className='px-6 py-3'>
          <div className='font-bold text-sm mb-1 text-gray-600 capitalize'>
            {category}
          </div>
          {name}
          <div className='font-bold text-xl mb-2 text-gray-800'>
            {vendor} -
          </div>
          <p className='text-black text-xl font-bold text-base py-2'>
            $ {price}
          </p>
        </div>
      </a>
    </div>
  )
}

export default Item

```

Then, replace this comment in `App.js` with the code snippet below:

```javascript
{resultSearch?.map((result) => (
    <Item
      url={result.url}
      image={result.images}
      name={result.name}
      category={result.category}
      vendor={result.vendor}
      price={result.price}
      key={result.id}
      />
))}

```

You can have a look at the fullcode on [GitHub](https://github.com/meilisearch/tutorials/tree/main/src/react-decathlon/src/App.js).

![](https://unable-actionable-car.media.strapiapp.com/showcase_1578bbacab.png)

Configure the search
--------------------

With Meilisearch, you get a ton of customization options for fine-tuning your search experience. We will take a look at a few features here. You can read more about them in the [documentation](https://docs.meilisearch.com/learn/getting_started/customizing_relevancy.html).

### Search Ranking

We will start with changing the ranking rules, the criteria Meilisearch uses to sort the documents you uploaded whenever a search query is made. The order of the ranking rules influences the relevancy of your search results. You can learn more about it in the [documentation](https://docs.meilisearch.com/learn/core_concepts/relevancy.html#ranking-rules).

Let's use the following order:

```json
[
  "words",
  "typo",
  "proximity",
  "attribute",
  "sort",
  "exactness",
  “creation_date:desc”
]

```

This uses the default order along with a custom rule: `creation_date`. This rule ranks items by their creation date if all previous values are identical.

### S**earchable attributes**

You can also configure [searchable attributes](https://docs.meilisearch.com/learn/configuration/displayed_searchable_attributes.html#searchable-fields). These are attributes whose values Meilisearch searches for matching query words. By default, all attributes are searchable, but you can configure it so that it only searches the `name`, `vendor`, `category`, and `tags` fields, leaving out `images` and `URL`:

```json
searchableAttributes: ["name", "vendor", "category", "tags"]
```

### Displayed attributes

Displayed attributes are attributes that Meilisearch can return to the user in the front-end application with the [displayedAttributes](https://docs.meilisearch.com/learn/configuration/displayed_searchable_attributes.html#displayed-fields) array. Like searchable attributes, all attributes are displayed by default.

```json
    "displayedAttributes": [
      "name",
      "vendor",
      "category",
      "tags",
      "images",
      "url"
    ]
```

### Upload the new settings to Meilisearch

It’s time to customize our Meilisearch index with the search settings explained above.

```javascript
const { MeiliSearch } = require('meilisearch')

;(async () => {
  try {
    const config = {
      host: 'http://localhost:7700'
    };

    const meili = new MeiliSearch(config);
    
    const index = meili.index("decathlon");

    const newSettings = {
      rankingRules: [
        "words",
        "typo",
        "proximity",
        "attribute",
        "sort",
        "exactness",
        "creation_date:desc"
      ],
      searchableAttributes: ["name", "vendor", "category", "tags"],
      displayedAttributes: [
        "name",
        "vendor",
        "category",
        "tags",
        "images",
        "url"
      ]
    };

    await index.updateSettings(newSettings);
        
  } catch (e) {
    console.error(e);
    console.log("Meili error: ", e.message);
  }
})();


```

Conclusion
----------

This quick search wouldn’t be possible without an incredible team working on this great project night and day! If you enjoy contributing to the Meilisearch family, check out the following repositories:

*   [Meilisearch](https://github.com/meilisearch/MeiliSearch)
*   [meilisearch-js](https://github.com/meilisearch/meilisearch-js)

* * *

Get these monthly updates straight to your inbox by subscribing to our [newsletter](https://meilisearch.us2.list-manage.com/subscribe?u=27870f7b71c908a8b359599fb&id=79582d828e&ref=blog.meilisearch.com).

For more things Meilisearch, join our developers community on [Discord](https://discord.meilisearch.com/?ref=blog.meilisearch.com). You can learn more about the product by checking out the [roadmap](https://roadmap.meilisearch.com/?ref=blog.meilisearch.com) and participating in [product discussions](https://github.com/meilisearch/product/discussions?ref=blog.meilisearch.com).

How to implement instant search in your React app

We are happy to announce that Meilisearch has raised a 15M$ Series A. This round was led by [Felicis](https://www.felicis.com/news/meili) with the participation of [Guillermo Rauch](https://twitter.com/rauchg), CEO of Vercel, and our existing investors [CRV](https://www.crv.com/), [Mango Capital](https://www.mangocapitalinc.com/), [LocalGlobe](https://localglobe.vc/), and [Seedcamp](https://seedcamp.com/). This new financing will help us to fulfill our mission—making search more accessible and more efficient for everyone—while allowing us to bring Meilisearch to an even larger audience worldwide.

A brief history of Meilisearch
------------------------------

We founded Meilisearch in 2019 because we were dissatisfied with the search solutions available on the market. We noticed how slow and irrelevant search results could degrade user experience in apps and websites. There had to be a better way.

While working on search technology at large e-commerce companies, we gained experience in building efficient search engines. So we decided to create our own search engine from scratch in Rust, and we open-sourced it to help every other developer struggling like us.

![A4 - 2.png](https://unable-actionable-car.media.strapiapp.com/A4_2_f84c47f309.png)

Meilisearch timeline

In 2020, we introduced Meilisearch to the world. In no time at all, hundreds of developers and organizations worldwide had embraced it. After witnessing such quick user adoption, we realized that there was a significant product-market fit.

![star-history-20221013.png](https://unable-actionable-car.media.strapiapp.com/star_history_20221013_2787264894.png)

GitHub star growth

Since then, we have experienced incredible growth in open-source adoption—one of the fastest growth rates of any COSS project. Our GitHub star traction is an indicator of how trendy Meilisearch has become, as well as how eagerly awaited our product has been by a massive community. Every month there are more than 100K users downloading Meilisearch and more than 10K applications using it.

This will remain unchanged
--------------------------

Meilisearch is and will always be open source. It’s part of our DNA and will stay in our company culture and how we think about products for as long as we exist.

Working in open source is incredibly rewarding, particularly with a community like ours. Our community members actively contribute to our codebase and our roadmap. Through their feedback, they help us consider use cases we wouldn’t have thought of. They’re our best ambassadors and introduce Meilisearch usage within their companies. I don’t know how Meilisearch could exist without the support of our community.

![Meilisearch {"<3"} Open Source](https://unable-actionable-car.media.strapiapp.com/A4_4_56bbfc7ad3.png)

Meilisearch {"<3"} Open Source

Creating a company has one huge advantage: you can shape it as you want. Meaning that you have full power over not only the direction the product is growing in, but also the culture the company will have. I’m so proud of the fantastic team we have created at Meilisearch. Through our joint effort, we have succeeded in creating a truly remarkable place to work.

![A4 - 1.png](https://unable-actionable-car.media.strapiapp.com/A4_1_be49eed673.png)

Meilisearch perks and benefits

Growing a company is hard. Keeping our culture is and will be hard. That’s why we spend so much time on HR-related decisions; we know the only way to keep our culture is by ensuring that employees grow with the company and carefully hiring those who fit with what we already have. We have big objectives, but we do not want to lose anyone on the road to obtaining them.

Looking forward
---------------

When we started building Meilisearch, it was with the idea to create a search solution that could fit both individual developers’ and enterprises’ needs. While we spent the last three years focusing on creating the best tool for developers and growing our open-source community, we are now entering the next phase: scaling Meilisearch’s team, product, and business for enterprises.

It’s super simple to use and put Meilisearch into production. However, most enterprises don’t want to manage software they didn’t create. Enterprises are also asking for more than just Meilisearch, the open-source search engine: they need a search solution with scalability, geo-replication, and analytics. To answer this demand, we have developed Meilisearch Cloud, a hosted version of Meilisearch that allows users to instantaneously have access to our technology in a few clicks. We would like to thank our beta testers, without whom Meilisearch Cloud would not exist as it does today!

Let me now give you a brief overview of Meilisearch’s aspirations for the future.

### Fast everywhere

We are, without a doubt, in a global world where the users and the data can be anywhere. One core principle of Meilisearch is to return an answer to any search query in a few milliseconds. But what if the service is hosted in the US and the search request originates from the EU? The experience will be degraded due to latency. That is why, especially with Meilisearch Cloud, we would love to break this barrier by creating a globally distributed search service. It’ll be a huge and very exciting challenge for our teams to make this possible.

We don't have small dreams. In the future, we would like to see Meilisearch running on the edge! We have already worked on a bunch of prototypes, and we would love to see how this technology could improve the search experience for our clients’ apps.

### Easier than ever

In the early days of Meilisearch, we understood that developer experience is key to success. Today, we are super proud to have a product so simple that it’s possible for any developer to try it in just a few minutes. Developer experience will remain a primary focus of our product and engineering team.

There are two parts to integrating Meilisearch. The first one is connecting Meilisearch to your data and making sure it’s synchronized. The second is building an interface that allows your users access to this data through our search API.

Concerning the first one, we will make sure that it’s easy for any developer to connect Meilisearch to their stacks. We already have SDKs for most languages, integrations for the most known web frameworks, plugins, and some awesome tools. But we will go even further. Being able to connect Meilisearch to your data in a few clicks will become a reality.

On the frontend side, it’s already possible to create a custom search bar thanks to our SDKs and libraries, particularly [instant-meilisearch](https://github.com/meilisearch/meilisearch-js-plugins/tree/main/packages/instant-meilisearch). Again, we would go further. We want every developer to be able to install a high-quality search bar into their website or application with no effort. That is why we would like to simplify the front-end installation of Meilisearch to require only a single line of JavaScript.

### Smarter than before

Everyone is different, which means that for the same request, different people should receive different answers based on their preferences. Meilisearch already parses most languages perfectly—we are always improving this part thanks to the community. The next big milestone for us would be to bring even more logic into the comprehension of user intent and the personalization of users’ results based on their previous searches or external data.

Not only is each person different, but each company is too. Most companies have totally different use cases that require different search behavior. A Notion search and a Booking search will not have the same requirement or the same business logic. To be honest, that’s one of the biggest challenges regarding the search topic. Today, even if we accommodate most of the regular use cases, we have a clear view of how to make Meilisearch customizable depending on the business logic. This will allow enterprises to put their ML models right into the Meilisearch engine.

What’s next
-----------

These projects will take time. So I'd like to tease you a little bit about some of the exciting things coming up in the next few months.

The very first project is one that has been talked about for months. One of the most important projects started since the creation of the search engine. I'm talking about [Meilisearch Cloud](https://cloud.meilisearch.com/register). For the many people waiting in the waiting list and the companies that can't wait for the GA of Meilisearch Cloud, your wish will soon be granted! Our team is working hard to get it out as soon as possible.

We have a second big announcement, and it will make every developer super happy. We are super proud to announce that Meilisearch V1 is planned!

If you look hard enough, you will find traces of the release day. Even if Meilisearch has been technically stable for years, we made a bet not to go beyond V1 in order to be able to iterate as much as possible on our developer experience. This was a great success. Now that we have stabilized our API, it is time to make V1 a reality.

As we are constantly releasing new features and projects, I recommend you to subscribe to our newsletter or follow the specific repos that interest you on Github. If you have specific needs or ideas for us, don't hesitate to leave feedback on our [roadmap](https://roadmap.meilisearch.com/tabs/5-ideas).

**Join us**
-----------

As you may have already understood, we are ambitious at Meilisearch. Thanks to this last fundraising, we have everything we need to go further. The only thing we are missing now is you.

If you want to know more about how it is to work at Meilisearch, I invite you to read our [blog](/blog/tag/company-news/)r discuss it with the team on our [community Slack](https://slack.meilisearch.com). If you’re already convinced about joining Meilisearch, we have not released every open position, so don’t hesitate to make a spontaneous application. We will hire a lot of engineers, PM, PMM, and other roles in the months and years to come. [https://jobs.lever.co/meili](https://jobs.lever.co/meili)

Press kit
---------

**Read our Press Release below**

[Meilisearch series A PR October 2022](/blog/content/files/2022/10/Meilisearch-series-A-PR----October-2022-1.pdf)("Download")

**Media Coverage:**

*   Techcrunch - [Meilisearch lands $15M investment to grow its ‘search-as-a-service’ business](https://techcrunch.com/2022/10/10/meilisearch-lands-15m-investment-to-grow-its-search-as-a-service-business/)
*   Felicis Ventures - [Searching for the Edge: How Meilisearch Is Building the Search Engine of the Future](https://www.felicis.com/news/meili)

Meilisearch raises a $15M Series A to offer enterprise-focused products and services

Today, we sit down with one of Meilisearch’s Rust developers, [Many](https://twitter.com/ManyTheFish), to talk about  the work he’s been doing with [Charabia](https://github.com/meilisearch/charabia) to improve Meilisearch’s language support. 

A bit of historical context
===========================

Though Many’s first contact with Meilisearch’s language parsing started two years ago, he only began focussing on tokenization around the end of 2021. You might be asking yourself: what is tokenization? Well, Many tells us it is one of the most important steps when indexing documents and, in short, means breaking down search terms into units our engine can process more efficiently. The part of Meilisearch’s engine that takes care of tokenization is called a tokenizer.

Many felt the way we handled content in different languages wasn’t efficient and that, even if Meilisearch already worked very well with languages like English and French, the same could not be said about other language groups. Many explains that a lot of it comes down to the different ways we type in different languages, as well as the various ways in which we make mistakes when writing in them. A typo in Japanese or Chinese, for example, might follow a different logic than a typo in Italian or Portuguese.

A different perspective
=======================

One main issue was immediately clear to Many: it would be impossible for us at Meilisearch to master all the languages in the world and improve our tokenizer alone. Luckily,  the open-source community is a diverse and generous bunch, with people from literally all over the world fluent in more languages we can possibly name. So Many’s focus shifted from directly improving language support to making contributing as simple and painless as possible.

Two main aspects go into improving a language with our tokenizer. First, we can work on **segmentation**, which means understanding where a word starts and where it ends.  
  
_“It appears pretty obvious for English or European speakers: “split on white spaces and you get your words!"; but it becomes harder when you face other types of Languages like Chinese, where there is no explicit separator between characters”_  
  
Second, we can work on **normalization**, which encompasses canonical modifications (upper or lower case), compatible equivalence (recognising the same character under different forms; e.g: ツ and ッ), and transliteration (transposing from one alphabet to the other, e.g.: Cyrillic into Latin)—we avoid this last one  as much as possible because it often results in information loss.

_“Here it’s the same, each Language has its specificities and we have to adapt the normalization for each Language. Take the case of capitalization: we have two versions of each Latin character, a capitalized version, and a lowercased version, but this specificity doesn’t exist for all Language in the world!”_

Many’s job, then, is less about writing a lot of code, and more about researching and understanding the ins and outs of a particular language so he can judge whether a contribution is going to work well with Meilisearch.  
He explains: 

_“This all boils down to the contrast between traditional translation and a field called information research (IR). Whereas translation focuses on preserving meaning, IR can afford to be less strict and account for mistakes such as misspelled words. Focussing on IR over translation allows us to offer search options that might be irrelevant in translation work.”_ 

![](https://unable-actionable-car.media.strapiapp.com/Word_explanation_0d5b8b3fac.png)

If a user typed “mais,” they would see results for both  “but” and “corn” because Meilisearch  takes into account that a user might have forgotten or simply did not feel like using the appropriate accent when typing.

So, whenever we wish to improve support for a language, it’s not only about having the right language experts contributing to our project. It also involves an enormous investment of research time on Many’s side, so he can determine whether a change to the tokenization process of a specific language is going to return more relevant search results.

What languages have we worked on so far?
========================================

As you might have read on our [blog post about v0.29 release](/blog/whats-new-in-v0-29/)n the past two months we’ve done extensive work on improving our support for Thai. This is a feature we’re very proud of and which was only possible thanks to the open-source community.  There’s a lot of room for improvement, though: Thai segmentation in Meilisearch is great, but normalization still requires work.

We also recently improved Meilisearch’s Hebrew normalisation. Before that, we made good advances on  Japanese segmentation, but want to polish Japanese normalization to better account for things like the differences between  ツ and ッ,or ダメ and だめ.

Right now, Many is very interested in Chinese tokenization, a language that comes with  its own set of challenges. For example, Chinese comes with a number of variants depending on the location where it’s spoken: Mandarin, mainland China’s various regional uses, and Cantonese, to name just a few. Though all these variations share mostly the same characters and meanings, they are not always used in the same way, can correspond to different sounds, and thus aren’t typed by everyone in the same fashion—all of which makes it really tricky when it comes to IR. At this point, we’re unable to have competing normalizations, which led us to base normalization on the dialect with the biggest number of users.

![](https://unable-actionable-car.media.strapiapp.com/rs7352_thinkstockphotos_493800479_low_1_45c0b2cea2.jpg)

Future ambitions
================

Accommodating competing normalizations for the same language (the tolerance of multiple variants of a language) is one of Many’s top priorities right now. 

He also feels the number of languages we support is too low, so an important step is devising strategies to get more contributors involved. You might ask: what languages Many wants to focus next? He tells us it’s a complex choice: he needs to balance his own personal preferences, the languages our contributors bring forward, as well as a strategic choice based on the number of speakers of a language.

Not all languages are equal, he explains. Some are easier to handle: agglutinative languages like Turkish, for example, could be next! _“However, if a contributor would like to push forward their language, we are happy to consider their suggestion and reprioritize,”_ assures Many.

Next on the list: parsing accents, diacritics, and other non-spacing marks. Remember our “maïs” and “mais” examples earlier? At this point, the tokenizer simply remove accents, but some users may want to take these diacritics into account. Accomplishing this isn’t technically very complex, according to Many: in short, it requires establishing two distinct normalization processes—one loose (which overlooks accents) and one strict (which doesn’t ignore anything). The challenge is integrating them without slowing down the search terribly or doubling the size of the index.

How can I contribute?
=====================

Want to help? You can contribute in various ways, many of which have little to do with coding:

*   Upvote a [discussion](https://github.com/meilisearch/product/discussions): this helps Many prioritize a specific topic or language
*   Create an [issue](https://github.com/meilisearch/charabia/issues) on GitHub explaining language-related problems you experience:  for example, if you notice Meilisearch is returning erroneous results when you search for a specific word or expression
*   Know a language like the back of your hand? We’d [love to hear](https://github.com/meilisearch/product/discussions) your thoughts on challenges and possible solutions when creating segmenters and normalizers
*   [Propose tokenization libraries](https://github.com/meilisearch/charabia/issues): it’s very helpful to know about existing tools you like and think works very well for a given language
*   [Submit a PR](https://github.com/meilisearch/meilisearch/pulls) solving a problem you experienced when using Meilisearch with a specific language. Tip: clearly explaining the problem and the solution in the PR itself or in a [GitHub issue](https://github.com/meilisearch/meilisearch/issues) definitely increases the chances of getting your contribution accepted 

If you’re participating in `Hacktoberfest` right now, most of the [issues listed on Charabia’s repository](https://github.com/meilisearch/charabia/issues) are eligible for the event! PRs without a corresponding issue are also welcomed—we’ll be happy to tag it as “hacktoberfest approved” if your contribution is accepted.

Last but not least: our work wouldn’t be possible without the fantastic input we receive from the community. Thank you so much for all your effort and generosity!

Improving Meilisearch’s language support

Hi 🙂 I’m Morgane, the HR Manager at Meilisearch, a search engine startup based in France. I joined Meilisearch because I was inspired by the founders’ vision of how they wanted to treat the people working with them. Before that, I worked for several medium to large-sized startups, but struggled with company cultures that glorified their founders and profits at the cost of fostering burnout. When I started working at Meilisearch, I was actually relieved: here were the creators of a tech startup using words like “people” and “fairness” before they ever mentioned “profit” or “IPO.”

As the HR manager at a young startup, one of the first projects on my to-do list was to solve the salary problem. We had heard about salary grids being used to solve similar problems, as well as something called a career framework that could help us clarify expectations for evolution within the company. We wanted to give all that a try. We did some research and were inspired by [Buffer’s](https://buffer.com/resources/career-framework/) solution to implement our own salary grid and career framework.

![https://media.giphy.com/media/9thduMKTL4Dcc/giphy.gif](https://media.giphy.com/media/9thduMKTL4Dcc/giphy.gif)

Two years down the road, I can confidently say this was a wise choice. Using a career framework and salary grid helped us solve specific problems that are faced by most distributed tech teams while also helping us express our company values of fairness and transparency.

Publishing this article is a way for us both to thank Buffer and similar companies for being transparent about their internal structure and to give back in the same way to new companies looking to solve similar problems. Scrolling down, you will read about our philosophy around salaries, current implementation, and potential improvements we’ve identified.

The challenges of compensation for a distributed tech company
=============================================================

Today we are a tech company distributed between Europe, South America, Africa, and the Middle East. As such, here are some of the challenges we are facing:

*   **Overbid:** Paying too much for software engineers. We mostly hire software engineers or people from the tech space. This means profiles can be rare, expensive, or both.
*   **Gender gap:** In 2021, a survey from [Hired](https://hired.com/h/wage-inequality-report/2021/) showed that in the tech space specifically, men were offered higher salaries than women for the same job title at the same company 59% of the time. One way to fight this kind of culture is by being transparent about what we pay people, de-emphasizing the role of negotiation in salary determination.
*   **Competitive hires:** Remote work has become ubiquitous in the tech industry during the past few years. Today, engineers can work from anywhere in the world. This puts us in direct competition with ********everyone********, including Silicon Valley-based tech companies that offer very high salaries.
*   **Diversity:** similar to the gender gap, salary systems that are opaque or negotiation-based tend to favor groups that are already highly privileged.
*   **Cost of living:** As a remote company, we can hire people anywhere worldwide, provided the time difference is manageable. This poses ethical issues regarding salaries: how do we make sure we pay our employees fairly given differences in cost of living, taxes, and market value?

Developing a salary grid is part of our solution to address issues such as internal retention, external attractiveness, and diversity & inclusion.

The principles behind our salary grid
=====================================

In January 2021, we implemented a career evolution framework and a salary grid. According to the grid, different teams and jobs are rewarded differently, based on data from [Figures](https://figures.hr/). We do regular benchmarking to make sure we stay relevant to the market and show team members we have fairness at heart.

The grid is [public](https://docs.google.com/spreadsheets/d/1yds9Wg4eAHQ0hmD4bvQspwqxp6UArXikXnxZEiAjxwg/edit?usp=sharing), so everyone individually knows where they stand. It’s a good way of avoiding arbitrary negotiations so that the best salary doesn’t always go to the pushiest negotiator. This initiative fits with important parts of our values and culture, such as openness, equality, and transparency.

![](https://unable-actionable-car.media.strapiapp.com/roman_synkevych_a_Sj_J_a_Cz_R7_Q_unsplash_f7c9afe080.jpg)

Regarding location, we chose to offer equal pay for equal work by basing gross salaries on the total cost for Meilisearch. So whether you’re a Product Manager working with us remotely in Cairo or Montreal, the base for calculating your salary will be the same: the only possible difference in your net salary will be the local taxes.

Moreover, with transparency in mind, we are able to show our grid to candidates in order to be completely transparent when we make an offer. We don’t want offers to be based on a candidate’s minimum salary expectation because this can perpetuate existing inequality.

It’s also important for us to specify that our [4-day work-week](/blog/how-we-implemented-the-4-day-work-week/ doesn’t impact salaries. As a Meilisearch employee, you’ll be paid as if you were working five days per week.

Our journey into compensation models
====================================

Our first salaries were set on French standards, with what we knew of the industry. A few months later, in an initiative to have a more data-oriented approach, we subscribed to [Figures](https://figures.hr/), a salary benchmarking tool that helped us reevaluate all our salaries at the end of 2021 to the top 25%, according to the data at our disposal.

After raising salaries to the top 25%, we thought our troubles were over. However, our salaries were based on European data; this meant that our salaries were insufficient to attract candidates from other parts of the world, and to retain our employees who had gained more experience and become desirable hires. This pushed us last September to reevaluate compensation for Software Engineering and Technical Writing based on an aggregation of data from Figures and a benchmark of salaries announced in job descriptions from tech companies in the US (for instance, the job board from [The Pragmatic Engineer](https://blog.pragmaticengineer.com/)).

Most people we have hired and the candidates we interview make it clear they’re interested in the potential of Meilisearch and our company culture above anything else. Having said that, let’s be honest: a large difference in salary can shift the balance of interest pretty quickly. **It’s important for us to have a competitive salary grid for our fully remote & distributed team and future hires.**

This is what our grid for Software Engineering looks like today:

|     |                     |              | Total cost for company | TCC       | TCC       | TCC       |
|-----|---------------------|--------------|-------------------------|-----------|-----------|-----------|
|     | IC                  | Manager      | 1                       | 2         | 3         | 4         |
|     | Software Engineering|              |                         |           |           |           |
| 1   | Entry               |              | €67,410                 | €73,149   | €78,888   | €84,628   |
| 2   | Intermediate        |              | €90,367                 | €96,106   | €101,845  | €106,933  |
| 3   | Advanced            | Lead         | €112,022               | €117,110  | €122,198  | €127,287  |
| 4   | Highly skilled      | Manager      | €132,375               | €138,133  | €143,891  | €149,649  |
| 5   | Specialist          | Senior Manager | €155,407             | €161,165  | €166,923  | €172,681  |
| 6   | Principal           | Director     | €178,439               | €184,197  | €189,955  | €195,713  |


How is it applied internally?
=============================

👉 **Our base: the total cost for the company** 🌎

Each team member’s salary is calculated according to their total cost for Meilisearch. The `total cost for company` value is used to calculate a gross salary or contractor fee. We calculate employer costs in each country based on the data provided by [Oyster HR](https://www.oysterhr.com/), which acts as an intermediary for international contracts.

The gross salary for an employee is then defined by considering employer costs and/or country-specific taxes. Here is a visual example of what this might look like:

|  | Total annual cost for company | Gross annual salary - Pre-tax |
| --- | --- | --- |
| France | €122,000.00 | €83,500.00 |
| Nigeria | €122,000 (₦52M) | €92,109 (₦40M) |
| USA | €122,000 ($120,183) | €96,790 ($95,634) |

Our formula for gross salaries is the following:

```markdown
total cost for company, minus: 

 - employer costs in specific country 
 - lunch stipend (same for everyone)
 - working from home stipend (same for everyone)
 - health insurance (if not included in employer costs)
 - Oyster HR fee

= gross salary for employees

```

👉 **What about contractors?** 🤨

Unfortunately, we lack data to automatize calculations for contractor fees. We start from the `total cost for the company` value, subtract the fees from Oyster HR, and offer a lump sum to the contractor who then handles the rest themselves.

This solution isn’t perfect
===========================

And it never will be. While we do have a solution that we find works well to compensate employees fairly and equally, we’ve identified areas of improvement that we will need to address going forward.

From an ideological standpoint, the evolution of our salary grids has marked a further distance between the salaries of developers and non-developers. Although these roles are not always cut and dry, we can see that software engineering as a skill is being valued very highly in the market. In time, this may create a sense of unwanted inequality. It also reminds us once again how dependent we are on the laws of the current market trends and obliged to react—dare I say _obey_?—to a senseless overbid on salaries in the tech space.

This market trend causes another problem for us: offering more money to stay relevant to candidates internationally. We know we will need to but are unfortunately limited in our current stage of development.

![](https://unable-actionable-car.media.strapiapp.com/brian_wangenheim_TEN_18_DN_8_KP_8_unsplash_e6b81cc129.jpg)

Finally, we can still improve our data-based approach to compensation. We will need, for instance, clear and relevant data regarding developer advocacy and technical writing jobs. These two fields may be well-known in Anglophone countries but aren’t very developed in Europe, thus making salary information scarce. For this, we are adapting this grid as we go along: meeting candidates and discussing their expectations help us gather insights.

All in all, we know there is more room to grow within this approach. However, we and all companies have leeway to create the future of work as we want to see it. One brick at a time, we’re getting there.

Thanks for reading
==================

Defining our policy and implementing this grid resulted from hours spent on the internet searching what other companies were doing, many internal discussions, and a few video calls with peers from tech startups.

With this article, I hope to help other businesses reflect on their compensation system. I’m also hoping to spark conversations and share ideas. Feel free to [reach out to me](https://www.linkedin.com/in/morgane-neff-8b1334102/) if this is something you’re interested in discussing!

Also… we’re hiring 🤩 If you want to join us, check are our [open roles](https://jobs.lever.co/meili/).

Meilisearch latest news and company updates