5418: Cache embeddings in search r=Kerollmops a=dureuill
# Pull Request
## Related issue
TBD
## What does this PR do?
- Adds a cache for embeddings produced in search
- The cache is disabled by default, and can be enabled following the instructions [here](https://github.com/orgs/meilisearch/discussions/818).
- Had to accommodate the `timeout` test for openai that uses a mock that simulates a timeout on subsequent responses: since the test was reusing the same query, the cache would kick-in and no request would be made to the mock, meaning no timeout any longer and so a failing test 😅
- `Embedder::embed_search` now accepts a reference instead of an owned `String`.
## Manual testing
- I created 4 indexes on a fresh DB with the same settings (one embedder from openai)
- I sent 1/4 of movies.json to each index
- I sent a federated search request against all 4 indexes, with the same query for each index, using the embedder of each index.
Results:
- The first call took 400ms to 1s. Before this change, it took in the 3s range.
- Any repeated call with the same query took in the range of 25ms.
- Looking at the details at trace log level, I can see that the first index that needs the embedding is taking most of the 400ms in `embed_one`. The other indexes report that the query text is found in the cache and they each take a few µs.
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
5369: exhaustive facet search r=ManyTheFish a=ManyTheFish
Fixes#5403
This PR adds an `exhaustiveFacetCount` field to the `/facet-search` API allowing the end-user to have a better facet count when having a distinct attribute set in the index settings.
# Usage
`POST /index/:index_uid/facet-search`
**Body:**
```json
{
"facetQuery": "blob",
"facetName": "genres",
"q": "",
"exhaustiveFacetCount": true
}
```
# Prototype Docker images
```sh
$ docker pull getmeili/meilisearch:prototype-exhaustive-facet-search-00
```
Co-authored-by: ManyTheFish <many@meilisearch.com>
5398: Bump ring from 0.17.8 to 0.17.13 r=Kerollmops a=dependabot[bot]
Bumps [ring](https://github.com/briansmith/ring) from 0.17.8 to 0.17.13.
<details>
<summary>Changelog</summary>
<p><em>Sourced from <a href="https://github.com/briansmith/ring/blob/main/RELEASES.md">ring's changelog</a>.</em></p>
<blockquote>
<h1>Version 0.17.13 (2025-03-06)</h1>
<p>Increased MSRV to 1.66.0 to avoid bugs in earlier versions so that we can
safely use <code>core::arch::x86_64::__cpuid</code> and <code>core::arch::x86::__cpuid</code> from
Rust in future releases.</p>
<p>AVX2-based VAES-CLMUL implementation. This will be a notable performance
improvement for most newish x86-64 systems. This will likely raise the minimum
binutils version supported for very old Linux distros.</p>
<h1>Version 0.17.12 (2025-03-05)</h1>
<p>Bug fix: <a href="https://redirect.github.com/briansmith/ring/pull/2447">briansmith/ring#2447</a> for denial of service (DoS).</p>
<ul>
<li>
<p>Fixes a panic in <code>ring::aead::quic::HeaderProtectionKey::new_mask()</code> when
integer overflow checking is enabled. In the QUIC protocol, an attacker can
induce this panic by sending a specially-crafted packet. Even unintentionally
it is likely to occur in 1 out of every 2**32 packets sent and/or received.</p>
</li>
<li>
<p>Fixes a panic on 64-bit targets in <code>ring::aead::{AES_128_GCM, AES_256_GCM}</code>
when overflow checking is enabled, when encrypting/decrypting approximately
68,719,476,700 bytes (about 64 gigabytes) of data in a single chunk. Protocols
like TLS and SSH are not affected by this because those protocols break large
amounts of data into small chunks. Similarly, most applications will not
attempt to encrypt/decrypt 64GB of data in one chunk.</p>
</li>
</ul>
<p>Overflow checking is not enabled in release mode by default, but
<code>RUSTFLAGS="-C overflow-checks"</code> or <code>overflow-checks = true</code> in the Cargo.toml
profile can override this. Overflow checking is usually enabled by default in
debug mode.</p>
</blockquote>
</details>
<details>
<summary>Commits</summary>
<ul>
<li>See full diff in <a href="https://github.com/briansmith/ring/commits">compare view</a></li>
</ul>
</details>
<br />
[](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores)
Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting ``@dependabot` rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- ``@dependabot` rebase` will rebase this PR
- ``@dependabot` recreate` will recreate this PR, overwriting any edits that have been made to it
- ``@dependabot` merge` will merge this PR after your CI passes on it
- ``@dependabot` squash and merge` will squash and merge this PR after your CI passes on it
- ``@dependabot` cancel merge` will cancel a previously requested merge and block automerging
- ``@dependabot` reopen` will reopen this PR if it is closed
- ``@dependabot` close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually
- ``@dependabot` show <dependency name> ignore conditions` will show all of the ignore conditions of the specified dependency
- ``@dependabot` ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
- ``@dependabot` ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
- ``@dependabot` ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)
You can disable automated security fix PRs for this repo from the [Security Alerts page](https://github.com/meilisearch/meilisearch/network/alerts).
</details>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
5413: Make sure to delete useless prefixes r=ManyTheFish a=Kerollmops
We discovered a bug where the new indexer was still writing empty roaring bitmaps instead of deleting the prefix entry from the prefix database.
Co-authored-by: Kerollmops <clement@meilisearch.com>
5254: Granular Filterable attribute settings r=ManyTheFish a=ManyTheFish
# Related
**Issue:** https://github.com/meilisearch/meilisearch/issues/5163
**PRD:** https://meilisearch.notion.site/API-usage-Settings-to-opt-out-indexing-features-filterableAttributes-1764b06b651f80aba8bdf359b2df3ca8
# Summary
Change the `filterableAttributes` settings to let the user choose which facet feature he wants to activate or not.
Deactivating a feature will avoid some database computation in the indexing process and save time and disk size.
# Example
`PATCH /indexes/:index_uid/settings`
```json
{
"filterableAttributes": [
{
"patterns": [
"cattos",
"doggos.age"
],
"features": {
"facetSearch": false,
"filter": {
"equality": true,
"comparison": false
}
}
}
]
}
```
# Impact on the codebase
- Settings API:
- `/settings`
- `/settings/filterable-attributes`
- OpenAPI
- may impact the LocalizedAttributesRules due to the AttributePatterns factorization
- Database:
- Filterable attributes format changed
- Faceted field_ids are no more stored in the database
- FieldIdsMap has no more unexisting fields
- Search:
- Search using filters
- Facet search
- `Attributes` ranking rule
- Distinct attribute
- Facet distribution
- Settings reindexing:
- searchable
- facet
- vector
- geo
- Document indexing:
- searchable
- facet
- vector
- geo
- Dump import
# Note for the reviewers
The changes are huge and have been split in different commits with a dedicated explanation, I suggest reviewing the commit 1by1
Co-authored-by: ManyTheFish <many@meilisearch.com>
5401: Make composite embedders an experimental feature r=irevoire a=dureuill
# Pull Request
## Related issue
Fixes#5343
## What does this PR do?
- Introduce new `compositeEmbedders` experimental feature
- Guard `source = "composite"` and `searchEmbedder`, `indexingEmbedder` behind enabling the feature.
- Update tests accordingly
## Dumpless upgrade
- Adding an experimental feature is never a breaking change
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
