Share the article
Welcome to the October 2026 edition of Meilisearch updates! This month we've got faster facet distribution with sampledCount, indexing that holds up when your embedder struggles, an IP allowlist for your Cloud projects, our first official Drupal module, and a demo and updated guides for agentic search.
Product updates: what's new at Meilisearch
Faster facet distribution on large catalogs with sampledCount (v1.56)
Filters like brand, color, or size help your end users narrow down results. But if your catalog has hundreds of brands and thousands of colors and sizes, counting every facet value to find the top ones can slow searches down. Meilisearch v1.56 adds sampledCount, a new way to sort facet values by count. Instead of counting everything, it samples your data to find the top values, so your facet distribution loads up to 10x faster.
- What you get: your top facet values, faster, with an exact count for each value it returns.
- What to know: the list itself is an estimate, so some values a full count would return can be skipped. If you need the complete list,
countis still there. - How to use it: available via the API. Use
sampledCountinstead ofcountwhen sorting facet values.

See the benchmarks in the pull request and the full details in the v1.56.0 release notes.
Indexing that copes better with a struggling embedder (v1.56)
Your indexing shouldn't be stuck waiting on an embedder that's having a bad day. v1.56 handles embedder outages better:
- Small updates still go through when your embedder is slow.
- Big jobs stop early instead of hanging when the embedder is overloaded or down, so you find out fast and can fix it.
- When your embedder says how long to wait by returning the
Retry-Afterheader (in seconds), Meilisearch waits instead of retrying over and over.
Meilisearch now also adjusts how fast it sends requests to your embedder and gives up earlier during long outages. If you use embedders, v1.56 is worth upgrading to. It's available now on Meilisearch Cloud.
v1.56 also brings a general search performance improvement from community contributor @itsamine27 (thank you!) and a fix for the binary quantization state after deleting an embedder.

Restrict access to your Cloud project with an IP allowlist
You can now limit which IPv4 addresses and CIDR blocks can reach a Meilisearch Cloud project. Requests from other addresses get a 403 Forbidden. Each project can have up to 100 CIDR entries.
To turn it on, open your project in the Cloud dashboard, go to Project Settings → Security, and click Enable add-on. The IP allowlist is a paid add-on at $10/month per project. Enterprise customers can contact the Meilisearch team to enable it.

If you disable the add-on, your entries keep restricting access until the end of the billing cycle. To stop filtering right away, delete all entries. Disable the IP allowlist
Clearer facet timings and a faster rankingScoreThreshold (v1.55)
- More detail in
showPerformanceDetails: responses now break out facet distribution timing into "compute facet distribution" and "compute facet stats", so you can see what facets cost in a slow search. - Faster
rankingScoreThreshold: search now stops at the first document that scores below your threshold, instead of evaluating the whole database. See the upgrade notes below. - More robust highlighting: documents with raw Unicode codepoints no longer cause a panic during highlighting when a segmenter fails.
More Cloud updates
- Delete organizations: organization owners can now delete an organization, as long as it has no active projects.
- v1.55 and v1.56 are live on Meilisearch Cloud.
Upgrade notes
- Upgrade recommended (v1.56): if you use embedders, we recommend upgrading to v1.56 for more reliable indexing when your embedder is slow, overloaded, or down.
- Breaking change (v1.55): setting
binaryQuantization: nullon an embedder now returns an error. Disabling binary quantization through a settings update is no longer allowed, because the old behavior could corrupt stored embeddings. - Relevancy change (v1.55): the
rankingScoreThresholdperformance fix modifies relevancy in some circumstances. If you rely on this parameter, check your results after upgrading.
Community & ecosystem
An official Meilisearch module for Drupal (alpha)
Meilisearch now has an official Search API backend for Drupal, maintained by Meilisearch. It works with Drupal 10.3+ and 11. The current release is 1.0.0-alpha2. Because it's still in alpha, it isn't covered by Drupal's security advisory policy yet. Give it a try and let us know what you think.
See it in action in Meili Kitchen, a recipe search demo built with Drupal 11, Search API, and Meilisearch.
SDK and integration releases
- meilisearch-java v0.22.1: adds support for dynamic search rules and
showPerformanceDetails. Release - meilisearch-php v2.0.0-beta.8: the next beta of the v2 PHP SDK, with a new
MultiSearchQueryclass and typed index stats. It includes breaking changes, so check the migration guide. Thanks to @murilolivorato and @norkunas for their contributions. Release - meilisearch-rails v0.17.0: settings are now only called at initialization, thanks to a fix from @StephaneRob. Release
New in the docs
- Laravel Scout: the guide covers semantic and hybrid search.
- Agentic search: updated guides on agentic search and MCP, including when to use the Vercel AI SDK. To see it in action, try the what2watch demo, a chatbot built with the AI SDK, and read Laurent's post on building search tools for AI agents.
That's all for October. See you next month!




