Revert the removal of the container

Improve S3 upload by showing errors in the task queue

.github/workflows/publish-release-assets.yml

@@ -65,9 +65,9 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [macos-13, windows-2022]
+        os: [macos-14, windows-2022]
         include:
-          - os: macos-13
+          - os: macos-14
             artifact_name: meilisearch
             asset_name: meilisearch-macos-amd64
           - os: windows-2022
@@ -90,7 +90,7 @@ jobs:
 
   publish-macos-apple-silicon:
     name: Publish binary for macOS silicon
-    runs-on: macos-13
+    runs-on: macos-14
     needs: check-version
     strategy:
       matrix:

.github/workflows/test-suite.yml

@@ -47,7 +47,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [macos-13, windows-2022]
+        os: [macos-14, windows-2022]
     steps:
       - uses: actions/checkout@v5
       - name: Cache dependencies
@@ -143,7 +143,7 @@ jobs:
   # We run tests in debug also, to make sure that the debug_assertions are hit
   test-debug:
     name: Run tests in debug
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-latest-large
     container:
       # Use ubuntu-22.04 to compile with glibc 2.35
       image: ubuntu:22.04
@@ -154,8 +154,8 @@ jobs:
           apt-get update && apt-get install -y curl
           apt-get install build-essential -y
       - uses: dtolnay/rust-toolchain@1.89
-      - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.8.0
+      - name: Run df -h
+        run: df -h
       - name: Run tests in debug
         uses: actions-rs/cargo@v1
         with:

Cargo.toml

@@ -23,7 +23,7 @@ members = [
 ]
 
 [workspace.package]
-version = "1.24.0"
+version = "1.26.0"
 authors = [
     "Quentin de Quelen <quentin@dequelen.me>",
     "Clément Renault <clement@meilisearch.com>",

crates/benchmarks/Cargo.toml

@@ -11,27 +11,27 @@ edition.workspace = true
 license.workspace = true
 
 [dependencies]
-anyhow = "1.0.98"
-bumpalo = "3.18.1"
-csv = "1.3.1"
-memmap2 = "0.9.7"
+anyhow = "1.0.100"
+bumpalo = "3.19.0"
+csv = "1.4.0"
+memmap2 = "0.9.9"
 milli = { path = "../milli" }
-mimalloc = { version = "0.1.47", default-features = false }
-serde_json = { version = "1.0.140", features = ["preserve_order"] }
-tempfile = "3.20.0"
+mimalloc = { version = "0.1.48", default-features = false }
+serde_json = { version = "1.0.145", features = ["preserve_order"] }
+tempfile = "3.23.0"
 
 [dev-dependencies]
-criterion = { version = "0.6.0", features = ["html_reports"] }
+criterion = { version = "0.7.0", features = ["html_reports"] }
 rand = "0.8.5"
 rand_chacha = "0.3.1"
 roaring = "0.10.12"
 
 [build-dependencies]
-anyhow = "1.0.98"
-bytes = "1.10.1"
-convert_case = "0.8.0"
-flate2 = "1.1.2"
-reqwest = { version = "0.12.20", features = ["blocking", "rustls-tls"], default-features = false }
+anyhow = "1.0.100"
+bytes = "1.11.0"
+convert_case = "0.9.0"
+flate2 = "1.1.5"
+reqwest = { version = "0.12.24", features = ["blocking", "rustls-tls"], default-features = false }
 
 [features]
 default = ["milli/all-tokenizations"]

crates/build-info/Cargo.toml

@@ -11,8 +11,8 @@ license.workspace = true
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-time = { version = "0.3.41", features = ["parsing"] }
+time = { version = "0.3.44", features = ["parsing"] }
 
 [build-dependencies]
-anyhow = "1.0.98"
+anyhow = "1.0.100"
 vergen-git2 = "1.0.7"

crates/dump/Cargo.toml

@@ -11,21 +11,21 @@ readme.workspace = true
 license.workspace = true
 
 [dependencies]
-anyhow = "1.0.98"
-flate2 = "1.1.2"
+anyhow = "1.0.100"
+flate2 = "1.1.5"
 http = "1.3.1"
 meilisearch-types = { path = "../meilisearch-types" }
 once_cell = "1.21.3"
-regex = "1.11.1"
+regex = "1.12.2"
 roaring = { version = "0.10.12", features = ["serde"] }
-serde = { version = "1.0.219", features = ["derive"] }
-serde_json = { version = "1.0.140", features = ["preserve_order"] }
+serde = { version = "1.0.228", features = ["derive"] }
+serde_json = { version = "1.0.145", features = ["preserve_order"] }
 tar = "0.4.44"
-tempfile = "3.20.0"
-thiserror = "2.0.12"
-time = { version = "0.3.41", features = ["serde-well-known", "formatting", "parsing", "macros"] }
+tempfile = "3.23.0"
+thiserror = "2.0.17"
+time = { version = "0.3.44", features = ["serde-well-known", "formatting", "parsing", "macros"] }
 tracing = "0.1.41"
-uuid = { version = "1.17.0", features = ["serde", "v4"] }
+uuid = { version = "1.18.1", features = ["serde", "v4"] }
 
 [dev-dependencies]
 big_s = "1.0.2"

crates/dump/src/lib.rs

@@ -96,6 +96,8 @@ pub struct TaskDump {
     pub finished_at: Option<OffsetDateTime>,
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub network: Option<TaskNetwork>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub custom_metadata: Option<String>,
 }
 
 // A `Kind` specific version made for the dump. If modified you may break the dump.
@@ -178,6 +180,7 @@ impl From<Task> for TaskDump {
             started_at: task.started_at,
             finished_at: task.finished_at,
             network: task.network,
+            custom_metadata: task.custom_metadata,
         }
     }
 }
@@ -396,6 +399,7 @@ pub(crate) mod test {
                     started_at: Some(datetime!(2022-11-20 0:00 UTC)),
                     finished_at: Some(datetime!(2022-11-21 0:00 UTC)),
                     network: None,
+                    custom_metadata: None,
                 },
                 None,
             ),
@@ -421,6 +425,7 @@ pub(crate) mod test {
                     started_at: None,
                     finished_at: None,
                     network: None,
+                    custom_metadata: None,
                 },
                 Some(vec![
                     json!({ "id": 4, "race": "leonberg" }).as_object().unwrap().clone(),
@@ -441,6 +446,7 @@ pub(crate) mod test {
                     started_at: None,
                     finished_at: None,
                     network: None,
+                    custom_metadata: None,
                 },
                 None,
             ),

crates/dump/src/reader/compat/v5_to_v6.rs

@@ -164,6 +164,7 @@ impl CompatV5ToV6 {
                     started_at: task_view.started_at,
                     finished_at: task_view.finished_at,
                     network: None,
+                    custom_metadata: None,
                 };
 
                 (task, content_file)

crates/file-store/Cargo.toml

@@ -11,7 +11,7 @@ edition.workspace = true
 license.workspace = true
 
 [dependencies]
-tempfile = "3.20.0"
-thiserror = "2.0.12"
+tempfile = "3.23.0"
+thiserror = "2.0.17"
 tracing = "0.1.41"
-uuid = { version = "1.17.0", features = ["serde", "v4"] }
+uuid = { version = "1.18.1", features = ["serde", "v4"] }

crates/flatten-serde-json/Cargo.toml

@@ -16,7 +16,7 @@ license.workspace = true
 serde_json = "1.0"
 
 [dev-dependencies]
-criterion = { version = "0.6.0", features = ["html_reports"] }
+criterion = { version = "0.7.0", features = ["html_reports"] }
 
 [[bench]]
 name = "benchmarks"

crates/fuzzers/Cargo.toml

@@ -11,12 +11,12 @@ edition.workspace = true
 license.workspace = true
 
 [dependencies]
-arbitrary = { version = "1.4.1", features = ["derive"] }
-bumpalo = "3.18.1"
-clap = { version = "4.5.40", features = ["derive"] }
+arbitrary = { version = "1.4.2", features = ["derive"] }
+bumpalo = "3.19.0"
+clap = { version = "4.5.52", features = ["derive"] }
 either = "1.15.0"
 fastrand = "2.3.0"
 milli = { path = "../milli" }
-serde = { version = "1.0.219", features = ["derive"] }
-serde_json = { version = "1.0.140", features = ["preserve_order"] }
-tempfile = "3.20.0"
+serde = { version = "1.0.228", features = ["derive"] }
+serde_json = { version = "1.0.145", features = ["preserve_order"] }
+tempfile = "3.23.0"

crates/index-scheduler/Cargo.toml

@@ -11,33 +11,33 @@ edition.workspace = true
 license.workspace = true
 
 [dependencies]
-anyhow = "1.0.98"
+anyhow = "1.0.100"
 bincode = "1.3.3"
 byte-unit = "5.1.6"
-bytes = "1.10.1"
-bumpalo = "3.18.1"
+bytes = "1.11.0"
+bumpalo = "3.19.0"
 bumparaw-collections = "0.1.4"
-convert_case = "0.8.0"
-csv = "1.3.1"
+convert_case = "0.9.0"
+csv = "1.4.0"
 derive_builder = "0.20.2"
 dump = { path = "../dump" }
-enum-iterator = "2.1.0"
+enum-iterator = "2.3.0"
 file-store = { path = "../file-store" }
-flate2 = "1.1.2"
-indexmap = "2.9.0"
+flate2 = "1.1.5"
+indexmap = "2.12.0"
 meilisearch-auth = { path = "../meilisearch-auth" }
 meilisearch-types = { path = "../meilisearch-types" }
-memmap2 = "0.9.7"
+memmap2 = "0.9.9"
 page_size = "0.6.0"
-rayon = "1.10.0"
+rayon = "1.11.0"
 roaring = { version = "0.10.12", features = ["serde"] }
-serde = { version = "1.0.219", features = ["derive"] }
-serde_json = { version = "1.0.140", features = ["preserve_order"] }
+serde = { version = "1.0.228", features = ["derive"] }
+serde_json = { version = "1.0.145", features = ["preserve_order"] }
 tar = "0.4.44"
 synchronoise = "1.0.1"
-tempfile = "3.20.0"
-thiserror = "2.0.12"
-time = { version = "0.3.41", features = [
+tempfile = "3.23.0"
+thiserror = "2.0.17"
+time = { version = "0.3.44", features = [
     "serde-well-known",
     "formatting",
     "parsing",
@@ -45,11 +45,11 @@ time = { version = "0.3.41", features = [
 ] }
 tracing = "0.1.41"
 ureq = "2.12.1"
-uuid = { version = "1.17.0", features = ["serde", "v4"] }
+uuid = { version = "1.18.1", features = ["serde", "v4"] }
 backoff = "0.4.0"
-reqwest = { version = "0.12.23", features = ["rustls-tls", "http2"], default-features = false }
+reqwest = { version = "0.12.24", features = ["rustls-tls", "http2"], default-features = false }
 rusty-s3 = "0.8.1"
-tokio = { version = "1.47.1", features = ["full"] }
+tokio = { version = "1.48.0", features = ["full"] }
 
 [dev-dependencies]
 big_s = "1.0.2"

crates/index-scheduler/src/dump.rs

@@ -150,6 +150,7 @@ impl<'a> Dump<'a> {
             details: task.details,
             status: task.status,
             network: task.network,
+            custom_metadata: task.custom_metadata,
             kind: match task.kind {
                 KindDump::DocumentImport {
                     primary_key,

crates/index-scheduler/src/insta_snapshot.rs

@@ -232,6 +232,7 @@ pub fn snapshot_task(task: &Task) -> String {
         status,
         kind,
         network,
+        custom_metadata,
     } = task;
     snap.push('{');
     snap.push_str(&format!("uid: {uid}, "));
@@ -252,6 +253,9 @@ pub fn snapshot_task(task: &Task) -> String {
     if let Some(network) = network {
         snap.push_str(&format!("network: {network:?}, "))
     }
+    if let Some(custom_metadata) = custom_metadata {
+        snap.push_str(&format!("custom_metadata: {custom_metadata:?}"))
+    }
 
     snap.push('}');
     snap

crates/index-scheduler/src/lib.rs

@@ -756,6 +756,19 @@ impl IndexScheduler {
         kind: KindWithContent,
         task_id: Option<TaskId>,
         dry_run: bool,
+    ) -> Result<Task> {
+        self.register_with_custom_metadata(kind, task_id, None, dry_run)
+    }
+
+    /// Register a new task in the scheduler, with metadata.
+    ///
+    /// If it fails and data was associated with the task, it tries to delete the associated data.
+    pub fn register_with_custom_metadata(
+        &self,
+        kind: KindWithContent,
+        task_id: Option<TaskId>,
+        custom_metadata: Option<String>,
+        dry_run: bool,
     ) -> Result<Task> {
         // if the task doesn't delete or cancel anything and 40% of the task queue is full, we must refuse to enqueue the incoming task
         if !matches!(&kind, KindWithContent::TaskDeletion { tasks, .. } | KindWithContent::TaskCancelation { tasks, .. } if !tasks.is_empty())
@@ -766,7 +779,7 @@ impl IndexScheduler {
         }
 
         let mut wtxn = self.env.write_txn()?;
-        let task = self.queue.register(&mut wtxn, &kind, task_id, dry_run)?;
+        let task = self.queue.register(&mut wtxn, &kind, task_id, custom_metadata, dry_run)?;
 
         // If the registered task is a task cancelation
         // we inform the processing tasks to stop (if necessary).

crates/index-scheduler/src/queue/mod.rs

@@ -257,6 +257,7 @@ impl Queue {
         wtxn: &mut RwTxn,
         kind: &KindWithContent,
         task_id: Option<TaskId>,
+        custom_metadata: Option<String>,
         dry_run: bool,
     ) -> Result<Task> {
         let next_task_id = self.tasks.next_task_id(wtxn)?;
@@ -280,6 +281,7 @@ impl Queue {
             status: Status::Enqueued,
             kind: kind.clone(),
             network: None,
+            custom_metadata,
         };
         // For deletion and cancelation tasks, we want to make extra sure that they
         // don't attempt to delete/cancel tasks that are newer than themselves.
@@ -344,6 +346,7 @@ impl Queue {
                 tasks: to_delete,
             },
             None,
+            None,
             false,
         )?;
 

crates/index-scheduler/src/scheduler/process_snapshot_creation.rs

@@ -438,12 +438,15 @@ async fn multipart_stream_to_s3(
     db_name: String,
     reader: std::io::PipeReader,
 ) -> Result<(), Error> {
-    use std::{collections::VecDeque, os::fd::OwnedFd, path::PathBuf};
+    use std::collections::VecDeque;
+    use std::io;
+    use std::os::fd::OwnedFd;
+    use std::path::PathBuf;
 
     use bytes::{Bytes, BytesMut};
     use reqwest::{Client, Response};
-    use rusty_s3::S3Action as _;
-    use rusty_s3::{actions::CreateMultipartUpload, Bucket, BucketError, Credentials, UrlStyle};
+    use rusty_s3::actions::CreateMultipartUpload;
+    use rusty_s3::{Bucket, BucketError, Credentials, S3Action as _, UrlStyle};
     use tokio::task::JoinHandle;
 
     let reader = OwnedFd::from(reader);
@@ -517,7 +520,6 @@ async fn multipart_stream_to_s3(
         while buffer.len() < (s3_multipart_part_size as usize / 2) {
             // Wait for the pipe to be readable
 
-            use std::io;
             reader.readable().await?;
 
             match reader.try_read_buf(&mut buffer) {
@@ -581,15 +583,17 @@ async fn multipart_stream_to_s3(
         async move {
             match client.post(url).body(body).send().await {
                 Ok(resp) if resp.status().is_client_error() => {
-                    resp.error_for_status().map_err(backoff::Error::Permanent)
+                    Err(backoff::Error::Permanent(Error::S3Error {
+                        status: resp.status(),
+                        body: resp.text().await.unwrap_or_default(),
+                    }))
                 }
                 Ok(resp) => Ok(resp),
-                Err(e) => Err(backoff::Error::transient(e)),
+                Err(e) => Err(backoff::Error::transient(Error::S3HttpError(e))),
             }
         }
     })
-    .await
-    .map_err(Error::S3HttpError)?;
+    .await?;
 
     let status = resp.status();
     let body = resp.text().await.map_err(|e| Error::S3Error { status, body: e.to_string() })?;

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/after_processing_everything.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, batch_uid: 0, status: succeeded, details: { from: (1, 12, 0), to: (1, 24, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, batch_uid: 0, status: succeeded, details: { from: (1, 12, 0), to: (1, 26, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, batch_uid: 1, status: succeeded, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 2 {uid: 2, batch_uid: 2, status: succeeded, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
 3 {uid: 3, batch_uid: 3, status: failed, error: ResponseError { code: 200, message: "Index `doggo` already exists.", error_code: "index_already_exists", error_type: "invalid_request", error_link: "https://docs.meilisearch.com/errors#index_already_exists" }, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
@@ -57,7 +57,7 @@ girafo: { number_of_documents: 0, field_distribution: {} }
 [timestamp] [4,]
 ----------------------------------------------------------------------
 ### All Batches:
-0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.24.0"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
+0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.26.0"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
 1 {uid: 1, details: {"primaryKey":"mouse"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"indexCreation":1},"indexUids":{"catto":1}}, stop reason: "created batch containing only task with id 1 of type `indexCreation` that cannot be batched with any other task.", }
 2 {uid: 2, details: {"primaryKey":"bone"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"indexCreation":1},"indexUids":{"doggo":1}}, stop reason: "created batch containing only task with id 2 of type `indexCreation` that cannot be batched with any other task.", }
 3 {uid: 3, details: {"primaryKey":"bone"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"indexCreation":1},"indexUids":{"doggo":1}}, stop reason: "created batch containing only task with id 3 of type `indexCreation` that cannot be batched with any other task.", }

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/register_automatic_upgrade_task.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, status: enqueued, details: { from: (1, 12, 0), to: (1, 24, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, status: enqueued, details: { from: (1, 12, 0), to: (1, 26, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 ----------------------------------------------------------------------
 ### Status:
 enqueued [0,]

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/registered_a_task_while_the_upgrade_task_is_enqueued.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, status: enqueued, details: { from: (1, 12, 0), to: (1, 24, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, status: enqueued, details: { from: (1, 12, 0), to: (1, 26, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, status: enqueued, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 ----------------------------------------------------------------------
 ### Status:

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/upgrade_task_failed.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, batch_uid: 0, status: failed, error: ResponseError { code: 200, message: "Planned failure for tests.", error_code: "internal", error_type: "internal", error_link: "https://docs.meilisearch.com/errors#internal" }, details: { from: (1, 12, 0), to: (1, 24, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, batch_uid: 0, status: failed, error: ResponseError { code: 200, message: "Planned failure for tests.", error_code: "internal", error_type: "internal", error_link: "https://docs.meilisearch.com/errors#internal" }, details: { from: (1, 12, 0), to: (1, 26, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, status: enqueued, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 ----------------------------------------------------------------------
 ### Status:
@@ -37,7 +37,7 @@ catto [1,]
 [timestamp] [0,]
 ----------------------------------------------------------------------
 ### All Batches:
-0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.24.0"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
+0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.26.0"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
 ----------------------------------------------------------------------
 ### Batch to tasks mapping:
 0 [0,]

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/upgrade_task_failed_again.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, batch_uid: 0, status: failed, error: ResponseError { code: 200, message: "Planned failure for tests.", error_code: "internal", error_type: "internal", error_link: "https://docs.meilisearch.com/errors#internal" }, details: { from: (1, 12, 0), to: (1, 24, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, batch_uid: 0, status: failed, error: ResponseError { code: 200, message: "Planned failure for tests.", error_code: "internal", error_type: "internal", error_link: "https://docs.meilisearch.com/errors#internal" }, details: { from: (1, 12, 0), to: (1, 26, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, status: enqueued, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 2 {uid: 2, status: enqueued, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
 ----------------------------------------------------------------------
@@ -40,7 +40,7 @@ doggo [2,]
 [timestamp] [0,]
 ----------------------------------------------------------------------
 ### All Batches:
-0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.24.0"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
+0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.26.0"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
 ----------------------------------------------------------------------
 ### Batch to tasks mapping:
 0 [0,]

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/upgrade_task_succeeded.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, batch_uid: 0, status: succeeded, details: { from: (1, 12, 0), to: (1, 24, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, batch_uid: 0, status: succeeded, details: { from: (1, 12, 0), to: (1, 26, 0) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, status: enqueued, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 2 {uid: 2, status: enqueued, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
 3 {uid: 3, status: enqueued, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
@@ -43,7 +43,7 @@ doggo [2,3,]
 [timestamp] [0,]
 ----------------------------------------------------------------------
 ### All Batches:
-0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.24.0"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
+0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.26.0"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
 ----------------------------------------------------------------------
 ### Batch to tasks mapping:
 0 [0,]

crates/index-scheduler/src/upgrade/mod.rs

@@ -48,6 +48,8 @@ pub fn upgrade_index_scheduler(
         (1, 22, _) => 0,
         (1, 23, _) => 0,
         (1, 24, _) => 0,
+        (1, 25, _) => 0,
+        (1, 26, _) => 0,
         (major, minor, patch) => {
             if major > current_major
                 || (major == current_major && minor > current_minor)
@@ -98,6 +100,7 @@ pub fn upgrade_index_scheduler(
             status: Status::Enqueued,
             kind: KindWithContent::UpgradeDatabase { from },
             network: None,
+            custom_metadata: None,
         },
     )?;
     wtxn.commit()?;

crates/index-scheduler/src/utils.rs

@@ -379,6 +379,7 @@ impl crate::IndexScheduler {
                 status,
                 kind,
                 network: _,
+                custom_metadata: _,
             } = task;
             assert_eq!(uid, task.uid);
             if task.status != Status::Enqueued {

crates/json-depth-checker/Cargo.toml

@@ -15,7 +15,7 @@ license.workspace = true
 serde_json = "1.0"
 
 [dev-dependencies]
-criterion = "0.6.0"
+criterion = "0.7.0"
 
 [[bench]]
 name = "depth"

crates/meili-snap/Cargo.toml

@@ -13,7 +13,7 @@ license.workspace = true
 [dependencies]
 # fixed version due to format breakages in v1.40
 insta = { version = "=1.39.0", features = ["json", "redactions"] }
-md5 = "0.7.0"
+md5 = "0.8.0"
 once_cell = "1.21"
-regex-lite = "0.1.6"
-uuid = { version = "1.17.0", features = ["v4"] }
+regex-lite = "0.1.8"
+uuid = { version = "1.18.1", features = ["v4"] }

crates/meilisearch-auth/Cargo.toml

@@ -12,15 +12,15 @@ license.workspace = true
 
 [dependencies]
 base64 = "0.22.1"
-enum-iterator = "2.1.0"
+enum-iterator = "2.3.0"
 hmac = "0.12.1"
 maplit = "1.0.2"
 meilisearch-types = { path = "../meilisearch-types" }
 rand = "0.8.5"
 roaring = { version = "0.10.12", features = ["serde"] }
-serde = { version = "1.0.219", features = ["derive"] }
-serde_json = { version = "1.0.140", features = ["preserve_order"] }
+serde = { version = "1.0.228", features = ["derive"] }
+serde_json = { version = "1.0.145", features = ["preserve_order"] }
 sha2 = "0.10.9"
-thiserror = "2.0.12"
-time = { version = "0.3.41", features = ["serde-well-known", "formatting", "parsing", "macros"] }
-uuid = { version = "1.17.0", features = ["serde", "v4"] }
+thiserror = "2.0.17"
+time = { version = "0.3.44", features = ["serde-well-known", "formatting", "parsing", "macros"] }
+uuid = { version = "1.18.1", features = ["serde", "v4"] }

crates/meilisearch-types/Cargo.toml

@@ -11,38 +11,38 @@ edition.workspace = true
 license.workspace = true
 
 [dependencies]
-actix-web = { version = "4.11.0", default-features = false }
-anyhow = "1.0.98"
-bumpalo = "3.18.1"
+actix-web = { version = "4.12.0", default-features = false }
+anyhow = "1.0.100"
+bumpalo = "3.19.0"
 bumparaw-collections = "0.1.4"
 byte-unit = { version = "5.1.6", features = ["serde"] }
-convert_case = "0.8.0"
-csv = "1.3.1"
-deserr = { version = "0.6.3", features = ["actix-web"] }
+convert_case = "0.9.0"
+csv = "1.4.0"
+deserr = { version = "0.6.4", features = ["actix-web"] }
 either = { version = "1.15.0", features = ["serde"] }
-enum-iterator = "2.1.0"
+enum-iterator = "2.3.0"
 file-store = { path = "../file-store" }
-flate2 = "1.1.2"
+flate2 = "1.1.5"
 fst = "0.4.7"
-memmap2 = "0.9.7"
+memmap2 = "0.9.9"
 milli = { path = "../milli" }
 roaring = { version = "0.10.12", features = ["serde"] }
 rustc-hash = "2.1.1"
-serde = { version = "1.0.219", features = ["derive"] }
+serde = { version = "1.0.228", features = ["derive"] }
 serde-cs = "0.2.4"
-serde_json = { version = "1.0.140", features = ["preserve_order"] }
+serde_json = { version = "1.0.145", features = ["preserve_order"] }
 tar = "0.4.44"
-tempfile = "3.20.0"
-thiserror = "2.0.12"
-time = { version = "0.3.41", features = [
+tempfile = "3.23.0"
+thiserror = "2.0.17"
+time = { version = "0.3.44", features = [
     "serde-well-known",
     "formatting",
     "parsing",
     "macros",
 ] }
-tokio = "1.45"
+tokio = "1.48"
 utoipa = { version = "5.4.0", features = ["macros"] }
-uuid = { version = "1.17.0", features = ["serde", "v4"] }
+uuid = { version = "1.18.1", features = ["serde", "v4"] }
 
 [dev-dependencies]
 # fixed version due to format breakages in v1.40

crates/meilisearch-types/src/error.rs

@@ -254,6 +254,7 @@ InvalidSearchHybridQuery                       , InvalidRequest       , BAD_REQU
 InvalidIndexLimit                              , InvalidRequest       , BAD_REQUEST ;
 InvalidIndexOffset                             , InvalidRequest       , BAD_REQUEST ;
 InvalidIndexPrimaryKey                         , InvalidRequest       , BAD_REQUEST ;
+InvalidIndexCustomMetadata                     , InvalidRequest       , BAD_REQUEST ;
 InvalidIndexUid                                , InvalidRequest       , BAD_REQUEST ;
 InvalidMultiSearchFacets                       , InvalidRequest       , BAD_REQUEST ;
 InvalidMultiSearchFacetsByIndex                , InvalidRequest       , BAD_REQUEST ;

crates/meilisearch-types/src/task_view.rs

@@ -55,6 +55,9 @@ pub struct TaskView {
 
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub network: Option<TaskNetwork>,
+
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub custom_metadata: Option<String>,
 }
 
 impl TaskView {
@@ -73,6 +76,7 @@ impl TaskView {
             started_at: task.started_at,
             finished_at: task.finished_at,
             network: task.network.clone(),
+            custom_metadata: task.custom_metadata.clone(),
         }
     }
 }

crates/meilisearch-types/src/tasks.rs

@@ -45,6 +45,9 @@ pub struct Task {
 
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub network: Option<TaskNetwork>,
+
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub custom_metadata: Option<String>,
 }
 
 impl Task {

crates/meilisearch/Cargo.toml

@@ -14,91 +14,91 @@ default-run = "meilisearch"
 
 [dependencies]
 actix-cors = "0.7.1"
-actix-http = { version = "3.11.0", default-features = false, features = [
+actix-http = { version = "3.11.2", default-features = false, features = [
     "compress-brotli",
     "compress-gzip",
     "rustls-0_23",
 ] }
 actix-utils = "3.0.1"
-actix-web = { version = "4.11.0", default-features = false, features = [
+actix-web = { version = "4.12.0", default-features = false, features = [
     "macros",
     "compress-brotli",
     "compress-gzip",
     "cookies",
     "rustls-0_23",
 ] }
-anyhow = { version = "1.0.98", features = ["backtrace"] }
-bstr = "1.12.0"
+anyhow = { version = "1.0.100", features = ["backtrace"] }
+bstr = "1.12.1"
 byte-unit = { version = "5.1.6", features = ["serde"] }
-bytes = "1.10.1"
-bumpalo = "3.18.1"
-clap = { version = "4.5.40", features = ["derive", "env"] }
+bytes = "1.11.0"
+bumpalo = "3.19.0"
+clap = { version = "4.5.52", features = ["derive", "env"] }
 crossbeam-channel = "0.5.15"
-deserr = { version = "0.6.3", features = ["actix-web"] }
+deserr = { version = "0.6.4", features = ["actix-web"] }
 dump = { path = "../dump" }
 either = "1.15.0"
 file-store = { path = "../file-store" }
-flate2 = "1.1.2"
+flate2 = "1.1.5"
 fst = "0.4.7"
 futures = "0.3.31"
 futures-util = "0.3.31"
 index-scheduler = { path = "../index-scheduler" }
-indexmap = { version = "2.9.0", features = ["serde"] }
-is-terminal = "0.4.16"
+indexmap = { version = "2.12.0", features = ["serde"] }
+is-terminal = "0.4.17"
 itertools = "0.14.0"
 jsonwebtoken = "9.3.1"
 lazy_static = "1.5.0"
 meilisearch-auth = { path = "../meilisearch-auth" }
 meilisearch-types = { path = "../meilisearch-types" }
-memmap2 = "0.9.7"
-mimalloc = { version = "0.1.47", default-features = false }
+memmap2 = "0.9.9"
+mimalloc = { version = "0.1.48", default-features = false }
 mime = "0.3.17"
 num_cpus = "1.17.0"
 obkv = "0.3.0"
 once_cell = "1.21.3"
-ordered-float = "5.0.0"
-parking_lot = "0.12.4"
+ordered-float = "5.1.0"
+parking_lot = "0.12.5"
 permissive-json-pointer = { path = "../permissive-json-pointer" }
 pin-project-lite = "0.2.16"
 platform-dirs = "0.3.0"
 prometheus = { version = "0.14.0", features = ["process"] }
 rand = "0.8.5"
-rayon = "1.10.0"
-regex = "1.11.1"
-reqwest = { version = "0.12.20", features = [
+rayon = "1.11.0"
+regex = "1.12.2"
+reqwest = { version = "0.12.24", features = [
     "rustls-tls",
     "json",
 ], default-features = false }
-rustls = { version = "0.23.28", features = ["ring"], default-features = false }
-rustls-pki-types = { version = "1.12.0", features = ["alloc"] }
+rustls = { version = "0.23.35", features = ["ring"], default-features = false }
+rustls-pki-types = { version = "1.13.0", features = ["alloc"] }
 rustls-pemfile = "2.2.0"
 segment = { version = "0.2.6" }
-serde = { version = "1.0.219", features = ["derive"] }
-serde_json = { version = "1.0.140", features = ["preserve_order"] }
+serde = { version = "1.0.228", features = ["derive"] }
+serde_json = { version = "1.0.145", features = ["preserve_order"] }
 sha2 = "0.10.9"
 siphasher = "1.0.1"
 slice-group-by = "0.3.1"
-static-files = { version = "0.2.5", optional = true }
-sysinfo = "0.35.2"
+static-files = { version = "0.3.1", optional = true }
+sysinfo = "0.37.2"
 tar = "0.4.44"
-tempfile = "3.20.0"
-thiserror = "2.0.12"
-time = { version = "0.3.41", features = [
+tempfile = "3.23.0"
+thiserror = "2.0.17"
+time = { version = "0.3.44", features = [
     "serde-well-known",
     "formatting",
     "parsing",
     "macros",
 ] }
-tokio = { version = "1.45.1", features = ["full"] }
-toml = "0.8.23"
-uuid = { version = "1.18.0", features = ["serde", "v4", "v7"] }
+tokio = { version = "1.48.0", features = ["full"] }
+toml = "0.9.8"
+uuid = { version = "1.18.1", features = ["serde", "v4", "v7"] }
 serde_urlencoded = "0.7.1"
 termcolor = "1.4.1"
-url = { version = "2.5.4", features = ["serde"] }
+url = { version = "2.5.7", features = ["serde"] }
 tracing = "0.1.41"
 tracing-subscriber = { version = "0.3.20", features = ["json"] }
 tracing-trace = { version = "0.1.0", path = "../tracing-trace" }
-tracing-actix-web = "0.7.18"
+tracing-actix-web = "0.7.19"
 build-info = { version = "1.7.0", path = "../build-info" }
 roaring = "0.10.12"
 mopa-maintained = "0.2.3"
@@ -114,35 +114,35 @@ utoipa = { version = "5.4.0", features = [
 utoipa-scalar = { version = "0.3.0", optional = true, features = ["actix-web"] }
 async-openai = { git = "https://github.com/meilisearch/async-openai", branch = "better-error-handling" }
 secrecy = "0.10.3"
-actix-web-lab = { version = "0.24.1", default-features = false }
+actix-web-lab = { version = "0.24.3", default-features = false }
 urlencoding = "2.1.3"
 backoff = { version = "0.4.0", features = ["tokio"] }
 
 
 [dev-dependencies]
-actix-rt = "2.10.0"
-brotli = "8.0.1"
+actix-rt = "2.11.0"
+brotli = "8.0.2"
 # fixed version due to format breakages in v1.40
 insta = { version = "=1.39.0", features = ["redactions"] }
 manifest-dir-macros = "0.1.18"
 maplit = "1.0.2"
 meili-snap = { path = "../meili-snap" }
 temp-env = "0.3.6"
-wiremock = "0.6.3"
+wiremock = "0.6.5"
 yaup = "0.3.1"
 
 [build-dependencies]
-anyhow = { version = "1.0.98", optional = true }
-cargo_toml = { version = "0.22.1", optional = true }
+anyhow = { version = "1.0.100", optional = true }
+cargo_toml = { version = "0.22.3", optional = true }
 hex = { version = "0.4.3", optional = true }
-reqwest = { version = "0.12.20", features = [
+reqwest = { version = "0.12.24", features = [
     "blocking",
     "rustls-tls",
 ], default-features = false, optional = true }
 sha-1 = { version = "0.10.1", optional = true }
-static-files = { version = "0.2.5", optional = true }
-tempfile = { version = "3.20.0", optional = true }
-zip = { version = "4.1.0", optional = true }
+static-files = { version = "0.3.1", optional = true }
+tempfile = { version = "3.23.0", optional = true }
+zip = { version = "6.0.0", optional = true }
 
 [features]
 default = ["meilisearch-types/all-tokenizations", "mini-dashboard"]

crates/meilisearch/src/analytics/segment_analytics.rs

@@ -195,7 +195,7 @@ struct Infos {
     experimental_enable_logs_route: bool,
     experimental_reduce_indexing_memory_usage: bool,
     experimental_max_number_of_batched_tasks: usize,
-    experimental_limit_batched_tasks_total_size: u64,
+    experimental_limit_batched_tasks_total_size: Option<u64>,
     experimental_network: bool,
     experimental_multimodal: bool,
     experimental_chat_completions: bool,
@@ -359,7 +359,7 @@ impl Infos {
             http_payload_size_limit,
             experimental_max_number_of_batched_tasks,
             experimental_limit_batched_tasks_total_size:
-                experimental_limit_batched_tasks_total_size.into(),
+                experimental_limit_batched_tasks_total_size.map(|size| size.as_u64()),
             task_queue_webhook: task_webhook_url.is_some(),
             task_webhook_authorization_header: task_webhook_authorization_header.is_some(),
             log_level: log_level.to_string(),

crates/meilisearch/src/lib.rs

@@ -230,7 +230,11 @@ pub fn setup_meilisearch(
         cleanup_enabled: !opt.experimental_replication_parameters,
         max_number_of_tasks: 1_000_000,
         max_number_of_batched_tasks: opt.experimental_max_number_of_batched_tasks,
-        batched_tasks_size_limit: opt.experimental_limit_batched_tasks_total_size.into(),
+        batched_tasks_size_limit: opt.experimental_limit_batched_tasks_total_size.map_or_else(
+            // By default, we use half of the available memory to determine the size of batched tasks
+            || opt.indexer_options.max_indexing_memory.map_or(u64::MAX, |mem| mem.as_u64() / 2),
+            |size| size.as_u64(),
+        ),
         index_growth_amount: byte_unit::Byte::from_str("10GiB").unwrap().as_u64() as usize,
         index_count: DEFAULT_INDEX_COUNT,
         instance_features: opt.to_instance_features(),

crates/meilisearch/src/option.rs

@@ -473,11 +473,13 @@ pub struct Opt {
     #[serde(default = "default_limit_batched_tasks")]
     pub experimental_max_number_of_batched_tasks: usize,
 
-    /// Experimentally reduces the maximum total size, in bytes, of tasks that will be processed at once,
-    /// see: <https://github.com/orgs/meilisearch/discussions/801>
-    #[clap(long, env = MEILI_EXPERIMENTAL_LIMIT_BATCHED_TASKS_TOTAL_SIZE, default_value_t = default_limit_batched_tasks_total_size())]
-    #[serde(default = "default_limit_batched_tasks_total_size")]
-    pub experimental_limit_batched_tasks_total_size: Byte,
+    /// Experimentally controls the maximum total size, in bytes, of tasks that will be processed
+    /// simultaneously. When unspecified, defaults to half of the maximum indexing memory.
+    ///
+    /// See: <https://github.com/orgs/meilisearch/discussions/801>
+    #[clap(long, env = MEILI_EXPERIMENTAL_LIMIT_BATCHED_TASKS_TOTAL_SIZE)]
+    #[serde(default)]
+    pub experimental_limit_batched_tasks_total_size: Option<Byte>,
 
     /// Enables experimental caching of search query embeddings. The value represents the maximal number of entries in the cache of each
     /// distinct embedder.
@@ -701,10 +703,12 @@ impl Opt {
             MEILI_EXPERIMENTAL_MAX_NUMBER_OF_BATCHED_TASKS,
             experimental_max_number_of_batched_tasks.to_string(),
         );
-        export_to_env_if_not_present(
-            MEILI_EXPERIMENTAL_LIMIT_BATCHED_TASKS_TOTAL_SIZE,
-            experimental_limit_batched_tasks_total_size.to_string(),
-        );
+        if let Some(limit) = experimental_limit_batched_tasks_total_size {
+            export_to_env_if_not_present(
+                MEILI_EXPERIMENTAL_LIMIT_BATCHED_TASKS_TOTAL_SIZE,
+                limit.to_string(),
+            );
+        }
         export_to_env_if_not_present(
             MEILI_EXPERIMENTAL_EMBEDDING_CACHE_ENTRIES,
             experimental_embedding_cache_entries.to_string(),
@@ -1273,10 +1277,6 @@ fn default_limit_batched_tasks() -> usize {
     usize::MAX
 }
 
-fn default_limit_batched_tasks_total_size() -> Byte {
-    Byte::from_u64(u64::MAX)
-}
-
 fn default_embedding_cache_entries() -> usize {
     0
 }

crates/meilisearch/src/personalization/mod.rs

@@ -1,14 +1,14 @@
-use crate::search::{Personalize, SearchResult};
-use meilisearch_types::{
-    error::{Code, ErrorCode, ResponseError},
-    milli::TimeBudget,
-};
+use std::time::Duration;
+
+use meilisearch_types::error::{Code, ErrorCode, ResponseError};
+use meilisearch_types::milli::TimeBudget;
 use rand::Rng;
 use reqwest::Client;
 use serde::{Deserialize, Serialize};
-use std::time::Duration;
 use tracing::{debug, info, warn};
 
+use crate::search::{Personalize, SearchResult};
+
 const COHERE_API_URL: &str = "https://api.cohere.ai/v1/rerank";
 const MAX_RETRIES: u32 = 10;
 

crates/meilisearch/src/routes/indexes/documents.rs

@@ -333,10 +333,12 @@ impl Aggregate for DocumentsDeletionAggregator {
 pub async fn delete_document(
     index_scheduler: GuardedData<ActionPolicy<{ actions::DOCUMENTS_DELETE }>, Data<IndexScheduler>>,
     path: web::Path<DocumentParam>,
+    params: AwebQueryParameter<CustomMetadataQuery, DeserrQueryParamError>,
     req: HttpRequest,
     opt: web::Data<Opt>,
     analytics: web::Data<Analytics>,
 ) -> Result<HttpResponse, ResponseError> {
+    let CustomMetadataQuery { custom_metadata } = params.into_inner();
     let DocumentParam { index_uid, document_id } = path.into_inner();
     let index_uid = IndexUid::try_from(index_uid)?;
     let network = index_scheduler.network();
@@ -359,7 +361,10 @@ pub async fn delete_document(
     let dry_run = is_dry_run(&req, &opt)?;
     let task = {
         let index_scheduler = index_scheduler.clone();
-        tokio::task::spawn_blocking(move || index_scheduler.register(task, uid, dry_run)).await??
+        tokio::task::spawn_blocking(move || {
+            index_scheduler.register_with_custom_metadata(task, uid, custom_metadata, dry_run)
+        })
+        .await??
     };
 
     if network.sharding && !dry_run {
@@ -678,6 +683,19 @@ pub struct UpdateDocumentsQuery {
     #[param(value_type = char, default = ",", example = ";")]
     #[deserr(default, try_from(char) = from_char_csv_delimiter -> DeserrQueryParamError<InvalidDocumentCsvDelimiter>, error = DeserrQueryParamError<InvalidDocumentCsvDelimiter>)]
     pub csv_delimiter: Option<u8>,
+
+    #[param(example = "custom")]
+    #[deserr(default, error = DeserrQueryParamError<InvalidIndexCustomMetadata>)]
+    pub custom_metadata: Option<String>,
+}
+
+#[derive(Deserialize, Debug, Deserr, IntoParams)]
+#[deserr(error = DeserrQueryParamError, rename_all = camelCase, deny_unknown_fields)]
+#[into_params(parameter_in = Query, rename_all = "camelCase")]
+pub struct CustomMetadataQuery {
+    #[param(example = "custom")]
+    #[deserr(default, error = DeserrQueryParamError<InvalidIndexCustomMetadata>)]
+    pub custom_metadata: Option<String>,
 }
 
 fn from_char_csv_delimiter(
@@ -819,6 +837,7 @@ pub async fn replace_documents(
         body,
         IndexDocumentsMethod::ReplaceDocuments,
         uid,
+        params.custom_metadata,
         dry_run,
         allow_index_creation,
         &req,
@@ -921,6 +940,7 @@ pub async fn update_documents(
         body,
         IndexDocumentsMethod::UpdateDocuments,
         uid,
+        params.custom_metadata,
         dry_run,
         allow_index_creation,
         &req,
@@ -940,6 +960,7 @@ async fn document_addition(
     body: Payload,
     method: IndexDocumentsMethod,
     task_id: Option<TaskId>,
+    custom_metadata: Option<String>,
     dry_run: bool,
     allow_index_creation: bool,
     req: &HttpRequest,
@@ -1065,8 +1086,10 @@ async fn document_addition(
     };
 
     let scheduler = index_scheduler.clone();
-    let task = match tokio::task::spawn_blocking(move || scheduler.register(task, task_id, dry_run))
-        .await?
+    let task = match tokio::task::spawn_blocking(move || {
+        scheduler.register_with_custom_metadata(task, task_id, custom_metadata, dry_run)
+    })
+    .await?
     {
         Ok(task) => task,
         Err(e) => {
@@ -1130,7 +1153,7 @@ async fn copy_body_to_file(
 /// Delete a set of documents based on an array of document ids.
 #[utoipa::path(
     post,
-    path = "{indexUid}/delete-batch",
+    path = "{indexUid}/documents/delete-batch",
     tag = "Documents",
     security(("Bearer" = ["documents.delete", "documents.*", "*"])),
     params(
@@ -1161,11 +1184,14 @@ pub async fn delete_documents_batch(
     index_scheduler: GuardedData<ActionPolicy<{ actions::DOCUMENTS_DELETE }>, Data<IndexScheduler>>,
     index_uid: web::Path<String>,
     body: web::Json<Vec<Value>>,
+    params: AwebQueryParameter<CustomMetadataQuery, DeserrQueryParamError>,
     req: HttpRequest,
     opt: web::Data<Opt>,
     analytics: web::Data<Analytics>,
 ) -> Result<HttpResponse, ResponseError> {
     debug!(parameters = ?body, "Delete documents by batch");
+    let CustomMetadataQuery { custom_metadata } = params.into_inner();
+
     let index_uid = IndexUid::try_from(index_uid.into_inner())?;
     let network = index_scheduler.network();
 
@@ -1190,7 +1216,10 @@ pub async fn delete_documents_batch(
     let dry_run = is_dry_run(&req, &opt)?;
     let task = {
         let index_scheduler = index_scheduler.clone();
-        tokio::task::spawn_blocking(move || index_scheduler.register(task, uid, dry_run)).await??
+        tokio::task::spawn_blocking(move || {
+            index_scheduler.register_with_custom_metadata(task, uid, custom_metadata, dry_run)
+        })
+        .await??
     };
 
     if network.sharding && !dry_run {
@@ -1244,12 +1273,15 @@ pub struct DocumentDeletionByFilter {
 pub async fn delete_documents_by_filter(
     index_scheduler: GuardedData<ActionPolicy<{ actions::DOCUMENTS_DELETE }>, Data<IndexScheduler>>,
     index_uid: web::Path<String>,
+    params: AwebQueryParameter<CustomMetadataQuery, DeserrQueryParamError>,
     body: AwebJson<DocumentDeletionByFilter, DeserrJsonError>,
     req: HttpRequest,
     opt: web::Data<Opt>,
     analytics: web::Data<Analytics>,
 ) -> Result<HttpResponse, ResponseError> {
     debug!(parameters = ?body, "Delete documents by filter");
+    let CustomMetadataQuery { custom_metadata } = params.into_inner();
+
     let index_uid = IndexUid::try_from(index_uid.into_inner())?;
     let index_uid = index_uid.into_inner();
     let filter = body.into_inner();
@@ -1282,7 +1314,10 @@ pub async fn delete_documents_by_filter(
     let dry_run = is_dry_run(&req, &opt)?;
     let task = {
         let index_scheduler = index_scheduler.clone();
-        tokio::task::spawn_blocking(move || index_scheduler.register(task, uid, dry_run)).await??
+        tokio::task::spawn_blocking(move || {
+            index_scheduler.register_with_custom_metadata(task, uid, custom_metadata, dry_run)
+        })
+        .await??
     };
 
     if network.sharding && !dry_run {
@@ -1372,12 +1407,14 @@ impl Aggregate for EditDocumentsByFunctionAggregator {
 pub async fn edit_documents_by_function(
     index_scheduler: GuardedData<ActionPolicy<{ actions::DOCUMENTS_ALL }>, Data<IndexScheduler>>,
     index_uid: web::Path<String>,
-    params: AwebJson<DocumentEditionByFunction, DeserrJsonError>,
+    params: AwebQueryParameter<CustomMetadataQuery, DeserrQueryParamError>,
+    body: AwebJson<DocumentEditionByFunction, DeserrJsonError>,
     req: HttpRequest,
     opt: web::Data<Opt>,
     analytics: web::Data<Analytics>,
 ) -> Result<HttpResponse, ResponseError> {
-    debug!(parameters = ?params, "Edit documents by function");
+    debug!(parameters = ?body, "Edit documents by function");
+    let CustomMetadataQuery { custom_metadata } = params.into_inner();
 
     index_scheduler
         .features()
@@ -1387,23 +1424,23 @@ pub async fn edit_documents_by_function(
 
     let index_uid = IndexUid::try_from(index_uid.into_inner())?;
     let index_uid = index_uid.into_inner();
-    let params = params.into_inner();
+    let body = body.into_inner();
 
     analytics.publish(
         EditDocumentsByFunctionAggregator {
-            filtered: params.filter.is_some(),
-            with_context: params.context.is_some(),
+            filtered: body.filter.is_some(),
+            with_context: body.context.is_some(),
             index_creation: index_scheduler.index(&index_uid).is_err(),
         },
         &req,
     );
 
     let engine = milli::rhai::Engine::new();
-    if let Err(e) = engine.compile(&params.function) {
+    if let Err(e) = engine.compile(&body.function) {
         return Err(ResponseError::from_msg(e.to_string(), Code::BadRequest));
     }
 
-    if let Some(ref filter) = params.filter {
+    if let Some(ref filter) = body.filter {
         // we ensure the filter is well formed before enqueuing it
         crate::search::parse_filter(
             filter,
@@ -1414,8 +1451,8 @@ pub async fn edit_documents_by_function(
     }
     let task = KindWithContent::DocumentEdition {
         index_uid: index_uid.clone(),
-        filter_expr: params.filter.clone(),
-        context: match params.context.clone() {
+        filter_expr: body.filter.clone(),
+        context: match body.context.clone() {
             Some(Value::Object(m)) => Some(m),
             None => None,
             _ => {
@@ -1425,18 +1462,21 @@ pub async fn edit_documents_by_function(
                 ))
             }
         },
-        function: params.function.clone(),
+        function: body.function.clone(),
     };
 
     let uid = get_task_id(&req, &opt)?;
     let dry_run = is_dry_run(&req, &opt)?;
     let task = {
         let index_scheduler = index_scheduler.clone();
-        tokio::task::spawn_blocking(move || index_scheduler.register(task, uid, dry_run)).await??
+        tokio::task::spawn_blocking(move || {
+            index_scheduler.register_with_custom_metadata(task, uid, custom_metadata, dry_run)
+        })
+        .await??
     };
 
     if network.sharding && !dry_run {
-        proxy(&index_scheduler, &index_uid, &req, network, Body::Inline(params), &task).await?;
+        proxy(&index_scheduler, &index_uid, &req, network, Body::Inline(body), &task).await?;
     }
 
     let task: SummarizedTaskView = task.into();
@@ -1477,12 +1517,14 @@ pub async fn edit_documents_by_function(
 pub async fn clear_all_documents(
     index_scheduler: GuardedData<ActionPolicy<{ actions::DOCUMENTS_DELETE }>, Data<IndexScheduler>>,
     index_uid: web::Path<String>,
+    params: AwebQueryParameter<CustomMetadataQuery, DeserrQueryParamError>,
     req: HttpRequest,
     opt: web::Data<Opt>,
     analytics: web::Data<Analytics>,
 ) -> Result<HttpResponse, ResponseError> {
     let index_uid = IndexUid::try_from(index_uid.into_inner())?;
     let network = index_scheduler.network();
+    let CustomMetadataQuery { custom_metadata } = params.into_inner();
 
     analytics.publish(
         DocumentsDeletionAggregator {
@@ -1501,7 +1543,10 @@ pub async fn clear_all_documents(
     let task = {
         let index_scheduler = index_scheduler.clone();
 
-        tokio::task::spawn_blocking(move || index_scheduler.register(task, uid, dry_run)).await??
+        tokio::task::spawn_blocking(move || {
+            index_scheduler.register_with_custom_metadata(task, uid, custom_metadata, dry_run)
+        })
+        .await??
     };
 
     if network.sharding && !dry_run {

crates/meilisearch/src/routes/mod.rs

@@ -218,6 +218,8 @@ pub struct SummarizedTaskView {
         deserialize_with = "time::serde::rfc3339::deserialize"
     )]
     enqueued_at: OffsetDateTime,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    custom_metadata: Option<String>,
 }
 
 impl From<Task> for SummarizedTaskView {
@@ -228,6 +230,7 @@ impl From<Task> for SummarizedTaskView {
             status: task.status,
             kind: task.kind.as_kind(),
             enqueued_at: task.enqueued_at,
+            custom_metadata: task.custom_metadata,
         }
     }
 }

crates/meilisearch/src/search/federated/types.rs

@@ -18,10 +18,9 @@ use serde::{Deserialize, Serialize};
 use utoipa::ToSchema;
 use uuid::Uuid;
 
-use crate::search::SearchMetadata;
-
 use super::super::{ComputedFacets, FacetStats, HitsInfo, SearchHit, SearchQueryWithIndex};
 use crate::milli::vector::Embedding;
+use crate::search::SearchMetadata;
 
 pub const DEFAULT_FEDERATED_WEIGHT: f64 = 1.0;
 

crates/meilisearch/tests/common/index.rs

@@ -91,7 +91,16 @@ impl<'a> Index<'a, Owned> {
         documents: Value,
         primary_key: Option<&str>,
     ) -> (Value, StatusCode) {
-        self._add_documents(documents, primary_key).await
+        self._add_documents(documents, primary_key, None).await
+    }
+
+    pub async fn add_documents_with_custom_metadata(
+        &self,
+        documents: Value,
+        primary_key: Option<&str>,
+        custom_metadata: Option<&str>,
+    ) -> (Value, StatusCode) {
+        self._add_documents(documents, primary_key, custom_metadata).await
     }
 
     pub async fn raw_add_documents(
@@ -352,12 +361,25 @@ impl<State> Index<'_, State> {
         &self,
         documents: Value,
         primary_key: Option<&str>,
+        custom_metadata: Option<&str>,
     ) -> (Value, StatusCode) {
-        let url = match primary_key {
-            Some(key) => {
-                format!("/indexes/{}/documents?primaryKey={}", urlencode(self.uid.as_ref()), key)
+        let url = match (primary_key, custom_metadata) {
+            (Some(key), Some(meta)) => {
+                format!(
+                    "/indexes/{}/documents?primaryKey={key}&customMetadata={meta}",
+                    urlencode(self.uid.as_ref()),
+                )
             }
-            None => format!("/indexes/{}/documents", urlencode(self.uid.as_ref())),
+            (None, Some(meta)) => {
+                format!(
+                    "/indexes/{}/documents?&customMetadata={meta}",
+                    urlencode(self.uid.as_ref()),
+                )
+            }
+            (Some(key), None) => {
+                format!("/indexes/{}/documents?&primaryKey={key}", urlencode(self.uid.as_ref()),)
+            }
+            (None, None) => format!("/indexes/{}/documents", urlencode(self.uid.as_ref())),
         };
         self.service.post_encoded(url, documents, self.encoder).await
     }

crates/meilisearch/tests/common/mod.rs

@@ -241,7 +241,7 @@ pub async fn shared_index_with_documents() -> &'static Index<'static, Shared> {
         let server = Server::new_shared();
         let index = server._index("SHARED_DOCUMENTS").to_shared();
         let documents = DOCUMENTS.clone();
-        let (response, _code) = index._add_documents(documents, None).await;
+        let (response, _code) = index._add_documents(documents, None, None).await;
         server.wait_task(response.uid()).await.succeeded();
         let (response, _code) = index
             ._update_settings(
@@ -284,7 +284,7 @@ pub async fn shared_index_with_score_documents() -> &'static Index<'static, Shar
         let server = Server::new_shared();
         let index = server._index("SHARED_SCORE_DOCUMENTS").to_shared();
         let documents = SCORE_DOCUMENTS.clone();
-        let (response, _code) = index._add_documents(documents, None).await;
+        let (response, _code) = index._add_documents(documents, None, None).await;
         server.wait_task(response.uid()).await.succeeded();
         let (response, _code) = index
             ._update_settings(
@@ -361,7 +361,7 @@ pub async fn shared_index_with_nested_documents() -> &'static Index<'static, Sha
         let server = Server::new_shared();
         let index = server._index("SHARED_NESTED_DOCUMENTS").to_shared();
         let documents = NESTED_DOCUMENTS.clone();
-        let (response, _code) = index._add_documents(documents, None).await;
+        let (response, _code) = index._add_documents(documents, None, None).await;
         server.wait_task(response.uid()).await.succeeded();
         let (response, _code) = index
             ._update_settings(
@@ -508,7 +508,7 @@ pub async fn shared_index_with_geo_documents() -> &'static Index<'static, Shared
         .get_or_init(|| async {
             let server = Server::new_shared();
             let index = server._index("SHARED_GEO_DOCUMENTS").to_shared();
-            let (response, _code) = index._add_documents(GEO_DOCUMENTS.clone(), None).await;
+            let (response, _code) = index._add_documents(GEO_DOCUMENTS.clone(), None, None).await;
             server.wait_task(response.uid()).await.succeeded();
 
             let (response, _code) = index
@@ -531,7 +531,7 @@ pub async fn shared_index_geojson_documents() -> &'static Index<'static, Shared>
             let index = server._index("SHARED_GEOJSON_DOCUMENTS").to_shared();
             let countries = include_str!("../documents/geojson/assets/countries.json");
             let lille = serde_json::from_str::<serde_json::Value>(countries).unwrap();
-            let (response, _code) = index._add_documents(Value(lille), Some("name")).await;
+            let (response, _code) = index._add_documents(Value(lille), Some("name"), None).await;
             server.wait_task(response.uid()).await.succeeded();
 
             let (response, _code) =

crates/meilisearch/tests/search/hybrid.rs

@@ -137,6 +137,60 @@ static SIMPLE_SEARCH_DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     }])
 });
 
+static MANY_DOCS: Lazy<Value> = Lazy::new(|| {
+    json!([
+    {
+        "title": "Shazam!",
+        "desc": "a Captain Marvel ersatz",
+        "id": "1",
+    },
+    {
+        "title": "Captain Planet",
+        "desc": "He's not part of the Marvel Cinematic Universe",
+        "id": "2",
+    },
+    {
+        "title": "Captain Marvel",
+        "desc": "a Shazam ersatz",
+        "id": "3",
+    },
+        {
+        "title": "Captain Marvel",
+        "desc": "a Shazam ersatz",
+        "id": "4",
+    },
+        {
+        "title": "Captain Marvel",
+        "desc": "a Shazam ersatz",
+        "id": "5",
+    },
+        {
+        "title": "Captain Marvel",
+        "desc": "a Shazam ersatz",
+        "id": "6",
+    },
+        {
+        "title": "Captain Marvel",
+        "desc": "a Shazam ersatz",
+        "id": "7",
+    },
+        {
+        "title": "Captain Marvel",
+        "desc": "a Shazam ersatz",
+        "id": "8",
+    },
+        {
+        "title": "Captain Marvel",
+        "desc": "a Shazam ersatz",
+        "id": "9",
+    },
+        {
+        "title": "Captain Marvel",
+        "desc": "a Shazam ersatz",
+        "id": "10",
+    }])
+});
+
 #[actix_rt::test]
 async fn simple_search() {
     let server = Server::new_shared();
@@ -449,6 +503,38 @@ async fn simple_search_hf() {
     snapshot!(response["semanticHitCount"], @"3");
 }
 
+#[actix_rt::test]
+async fn issue_5976_missing_docs_hf() {
+    let server = Server::new_shared();
+    let index = index_with_documents_hf(server, &MANY_DOCS).await;
+    let (response, code) = index
+        .search_post(
+            json!({"q": "Wonder replacement", "hybrid": {"embedder": "default", "semanticRatio": 1.0}, "retrieveVectors": true}),
+        )
+        .await;
+    snapshot!(code, @"200 OK");
+    let are_empty: Vec<_> = response["hits"]
+        .as_array()
+        .unwrap()
+        .iter()
+        .map(|hit| hit["_vectors"]["default"]["embeddings"].as_array().unwrap().is_empty())
+        .collect();
+    snapshot!(json!(are_empty), @r###"
+    [
+      false,
+      false,
+      false,
+      false,
+      false,
+      false,
+      false,
+      false,
+      false,
+      false
+    ]
+    "###);
+}
+
 #[actix_rt::test]
 async fn distribution_shift() {
     let server = Server::new_shared();

crates/meilisearch/tests/search/multi/proxy.rs

@@ -3141,3 +3141,513 @@ fn fail(override_response_body: Option<&str>) -> ResponseTemplate {
         response.set_body_json(json!({"error": "provoked error", "code": "test_error", "link": "https://docs.meilisearch.com/errors#test_error"}))
     }
 }
+
+#[actix_rt::test]
+async fn remote_auto_sharding() {
+    let ms0 = Server::new().await;
+    let ms1 = Server::new().await;
+    let ms2 = Server::new().await;
+
+    // enable feature
+
+    let (response, code) = ms0.set_features(json!({"network": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response["network"]), @"true");
+    let (response, code) = ms1.set_features(json!({"network": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response["network"]), @"true");
+    let (response, code) = ms2.set_features(json!({"network": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response["network"]), @"true");
+
+    // set self & sharding
+
+    let (response, code) = ms0.set_network(json!({"self": "ms0", "sharding": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response), @r###"
+    {
+      "self": "ms0",
+      "remotes": {},
+      "sharding": true
+    }
+    "###);
+    let (response, code) = ms1.set_network(json!({"self": "ms1", "sharding": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response), @r###"
+    {
+      "self": "ms1",
+      "remotes": {},
+      "sharding": true
+    }
+    "###);
+    let (response, code) = ms2.set_network(json!({"self": "ms2", "sharding": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response), @r###"
+    {
+      "self": "ms2",
+      "remotes": {},
+      "sharding": true
+    }
+    "###);
+
+    // wrap servers
+    let ms0 = Arc::new(ms0);
+    let ms1 = Arc::new(ms1);
+    let ms2 = Arc::new(ms2);
+
+    let rms0 = LocalMeili::new(ms0.clone()).await;
+    let rms1 = LocalMeili::new(ms1.clone()).await;
+    let rms2 = LocalMeili::new(ms2.clone()).await;
+
+    // set network
+    let network = json!({"remotes": {
+        "ms0": {
+            "url": rms0.url()
+        },
+        "ms1": {
+            "url": rms1.url()
+        },
+        "ms2": {
+            "url": rms2.url()
+        }
+    }});
+
+    println!("{}", serde_json::to_string_pretty(&network).unwrap());
+
+    let (_response, status_code) = ms0.set_network(network.clone()).await;
+    snapshot!(status_code, @"200 OK");
+    let (_response, status_code) = ms1.set_network(network.clone()).await;
+    snapshot!(status_code, @"200 OK");
+    let (_response, status_code) = ms2.set_network(network.clone()).await;
+    snapshot!(status_code, @"200 OK");
+
+    // add documents
+    let documents = SCORE_DOCUMENTS.clone();
+    let documents = documents.as_array().unwrap();
+    let index0 = ms0.index("test");
+    let _index1 = ms1.index("test");
+    let _index2 = ms2.index("test");
+
+    let (task, _status_code) = index0.add_documents(json!(documents), None).await;
+
+    let t0 = task.uid();
+    let (t, _) = ms0.get_task(task.uid()).await;
+    let t1 = t["network"]["remote_tasks"]["ms1"]["taskUid"].as_u64().unwrap();
+    let t2 = t["network"]["remote_tasks"]["ms2"]["taskUid"].as_u64().unwrap();
+
+    ms0.wait_task(t0).await.succeeded();
+    ms1.wait_task(t1).await.succeeded();
+    ms2.wait_task(t2).await.succeeded();
+
+    // perform multi-search
+    let query = "badman returns";
+    let request = json!({
+        "federation": {},
+        "queries": [
+            {
+                "q": query,
+                "indexUid": "test",
+                "federationOptions": {
+                    "remote": "ms0"
+                }
+            },
+            {
+                "q": query,
+                "indexUid": "test",
+                "federationOptions": {
+                    "remote": "ms1"
+                }
+            },
+            {
+                "q": query,
+                "indexUid": "test",
+                "federationOptions": {
+                    "remote": "ms2"
+                }
+            },
+        ]
+    });
+
+    let (response, _status_code) = ms0.multi_search(request.clone()).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response, { ".processingTimeMs" => "[time]", ".requestUid" => "[uuid]" }), @r###"
+    {
+      "hits": [
+        {
+          "title": "Batman Returns",
+          "id": "C",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 2,
+            "weightedRankingScore": 0.8317901234567902,
+            "remote": "ms2"
+          }
+        },
+        {
+          "title": "Batman the dark knight returns: Part 1",
+          "id": "A",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 1,
+            "weightedRankingScore": 0.7028218694885362,
+            "remote": "ms1"
+          }
+        },
+        {
+          "title": "Batman the dark knight returns: Part 2",
+          "id": "B",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 1,
+            "weightedRankingScore": 0.7028218694885362,
+            "remote": "ms1"
+          }
+        },
+        {
+          "title": "Badman",
+          "id": "E",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 2,
+            "weightedRankingScore": 0.5,
+            "remote": "ms2"
+          }
+        },
+        {
+          "title": "Batman",
+          "id": "D",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 0,
+            "weightedRankingScore": 0.23106060606060605,
+            "remote": "ms0"
+          }
+        }
+      ],
+      "processingTimeMs": "[time]",
+      "limit": 20,
+      "offset": 0,
+      "estimatedTotalHits": 5,
+      "requestUid": "[uuid]",
+      "remoteErrors": {}
+    }
+    "###);
+    let (response, _status_code) = ms1.multi_search(request.clone()).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response, { ".processingTimeMs" => "[time]", ".requestUid" => "[uuid]" }), @r###"
+    {
+      "hits": [
+        {
+          "title": "Batman Returns",
+          "id": "C",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 2,
+            "weightedRankingScore": 0.8317901234567902,
+            "remote": "ms2"
+          }
+        },
+        {
+          "title": "Batman the dark knight returns: Part 1",
+          "id": "A",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 1,
+            "weightedRankingScore": 0.7028218694885362,
+            "remote": "ms1"
+          }
+        },
+        {
+          "title": "Batman the dark knight returns: Part 2",
+          "id": "B",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 1,
+            "weightedRankingScore": 0.7028218694885362,
+            "remote": "ms1"
+          }
+        },
+        {
+          "title": "Badman",
+          "id": "E",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 2,
+            "weightedRankingScore": 0.5,
+            "remote": "ms2"
+          }
+        },
+        {
+          "title": "Batman",
+          "id": "D",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 0,
+            "weightedRankingScore": 0.23106060606060605,
+            "remote": "ms0"
+          }
+        }
+      ],
+      "processingTimeMs": "[time]",
+      "limit": 20,
+      "offset": 0,
+      "estimatedTotalHits": 5,
+      "requestUid": "[uuid]",
+      "remoteErrors": {}
+    }
+    "###);
+    let (response, _status_code) = ms2.multi_search(request.clone()).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response, { ".processingTimeMs" => "[time]", ".requestUid" => "[uuid]" }), @r###"
+    {
+      "hits": [
+        {
+          "title": "Batman Returns",
+          "id": "C",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 2,
+            "weightedRankingScore": 0.8317901234567902,
+            "remote": "ms2"
+          }
+        },
+        {
+          "title": "Batman the dark knight returns: Part 1",
+          "id": "A",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 1,
+            "weightedRankingScore": 0.7028218694885362,
+            "remote": "ms1"
+          }
+        },
+        {
+          "title": "Batman the dark knight returns: Part 2",
+          "id": "B",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 1,
+            "weightedRankingScore": 0.7028218694885362,
+            "remote": "ms1"
+          }
+        },
+        {
+          "title": "Badman",
+          "id": "E",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 2,
+            "weightedRankingScore": 0.5,
+            "remote": "ms2"
+          }
+        },
+        {
+          "title": "Batman",
+          "id": "D",
+          "_federation": {
+            "indexUid": "test",
+            "queriesPosition": 0,
+            "weightedRankingScore": 0.23106060606060605,
+            "remote": "ms0"
+          }
+        }
+      ],
+      "processingTimeMs": "[time]",
+      "limit": 20,
+      "offset": 0,
+      "estimatedTotalHits": 5,
+      "requestUid": "[uuid]",
+      "remoteErrors": {}
+    }
+    "###);
+}
+
+#[actix_rt::test]
+async fn remote_auto_sharding_with_custom_metadata() {
+    let ms0 = Server::new().await;
+    let ms1 = Server::new().await;
+    let ms2 = Server::new().await;
+
+    // enable feature
+
+    let (response, code) = ms0.set_features(json!({"network": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response["network"]), @"true");
+    let (response, code) = ms1.set_features(json!({"network": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response["network"]), @"true");
+    let (response, code) = ms2.set_features(json!({"network": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response["network"]), @"true");
+
+    // set self & sharding
+
+    let (response, code) = ms0.set_network(json!({"self": "ms0", "sharding": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response), @r###"
+    {
+      "self": "ms0",
+      "remotes": {},
+      "sharding": true
+    }
+    "###);
+    let (response, code) = ms1.set_network(json!({"self": "ms1", "sharding": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response), @r###"
+    {
+      "self": "ms1",
+      "remotes": {},
+      "sharding": true
+    }
+    "###);
+    let (response, code) = ms2.set_network(json!({"self": "ms2", "sharding": true})).await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response), @r###"
+    {
+      "self": "ms2",
+      "remotes": {},
+      "sharding": true
+    }
+    "###);
+
+    // wrap servers
+    let ms0 = Arc::new(ms0);
+    let ms1 = Arc::new(ms1);
+    let ms2 = Arc::new(ms2);
+
+    let rms0 = LocalMeili::new(ms0.clone()).await;
+    let rms1 = LocalMeili::new(ms1.clone()).await;
+    let rms2 = LocalMeili::new(ms2.clone()).await;
+
+    // set network
+    let network = json!({"remotes": {
+        "ms0": {
+            "url": rms0.url()
+        },
+        "ms1": {
+            "url": rms1.url()
+        },
+        "ms2": {
+            "url": rms2.url()
+        }
+    }});
+
+    println!("{}", serde_json::to_string_pretty(&network).unwrap());
+
+    let (_response, status_code) = ms0.set_network(network.clone()).await;
+    snapshot!(status_code, @"200 OK");
+    let (_response, status_code) = ms1.set_network(network.clone()).await;
+    snapshot!(status_code, @"200 OK");
+    let (_response, status_code) = ms2.set_network(network.clone()).await;
+    snapshot!(status_code, @"200 OK");
+
+    // add documents
+    let documents = SCORE_DOCUMENTS.clone();
+    let documents = documents.as_array().unwrap();
+    let index0 = ms0.index("test");
+    let _index1 = ms1.index("test");
+    let _index2 = ms2.index("test");
+
+    let (task, _status_code) = index0
+        .add_documents_with_custom_metadata(
+            json!(documents),
+            None,
+            Some("remote_auto_sharding_with_custom_metadata"),
+        )
+        .await;
+
+    let t0 = task.uid();
+    let (t, _) = ms0.get_task(task.uid()).await;
+    let t1 = t["network"]["remote_tasks"]["ms1"]["taskUid"].as_u64().unwrap();
+    let t2 = t["network"]["remote_tasks"]["ms2"]["taskUid"].as_u64().unwrap();
+
+    let t = ms0.wait_task(t0).await.succeeded();
+    snapshot!(t, @r###"
+    {
+      "uid": "[uid]",
+      "batchUid": "[batch_uid]",
+      "indexUid": "test",
+      "status": "succeeded",
+      "type": "documentAdditionOrUpdate",
+      "canceledBy": null,
+      "details": {
+        "receivedDocuments": 5,
+        "indexedDocuments": 1
+      },
+      "error": null,
+      "duration": "[duration]",
+      "enqueuedAt": "[date]",
+      "startedAt": "[date]",
+      "finishedAt": "[date]",
+      "network": {
+        "remote_tasks": {
+          "ms1": {
+            "taskUid": 0,
+            "error": null
+          },
+          "ms2": {
+            "taskUid": 0,
+            "error": null
+          }
+        }
+      },
+      "customMetadata": "remote_auto_sharding_with_custom_metadata"
+    }
+    "###);
+
+    let t = ms1.wait_task(t1).await.succeeded();
+    snapshot!(t, @r###"
+    {
+      "uid": "[uid]",
+      "batchUid": "[batch_uid]",
+      "indexUid": "test",
+      "status": "succeeded",
+      "type": "documentAdditionOrUpdate",
+      "canceledBy": null,
+      "details": {
+        "receivedDocuments": 5,
+        "indexedDocuments": 2
+      },
+      "error": null,
+      "duration": "[duration]",
+      "enqueuedAt": "[date]",
+      "startedAt": "[date]",
+      "finishedAt": "[date]",
+      "network": {
+        "origin": {
+          "remoteName": "ms0",
+          "taskUid": 0
+        }
+      },
+      "customMetadata": "remote_auto_sharding_with_custom_metadata"
+    }
+    "###);
+
+    let t = ms2.wait_task(t2).await.succeeded();
+    snapshot!(t, @r###"
+    {
+      "uid": "[uid]",
+      "batchUid": "[batch_uid]",
+      "indexUid": "test",
+      "status": "succeeded",
+      "type": "documentAdditionOrUpdate",
+      "canceledBy": null,
+      "details": {
+        "receivedDocuments": 5,
+        "indexedDocuments": 2
+      },
+      "error": null,
+      "duration": "[duration]",
+      "enqueuedAt": "[date]",
+      "startedAt": "[date]",
+      "finishedAt": "[date]",
+      "network": {
+        "origin": {
+          "remoteName": "ms0",
+          "taskUid": 0
+        }
+      },
+      "customMetadata": "remote_auto_sharding_with_custom_metadata"
+    }
+    "###);
+}

crates/meilisearch/tests/tasks/webhook.rs

@@ -656,3 +656,119 @@ async fn forbidden_fields() {
     }
     "#);
 }
+
+#[actix_web::test]
+async fn receive_custom_metadata() {
+    let WebhookHandle { server_handle: handle1, url: url1, receiver: mut receiver1 } =
+        create_webhook_server().await;
+    let WebhookHandle { server_handle: handle2, url: url2, receiver: mut receiver2 } =
+        create_webhook_server().await;
+    let WebhookHandle { server_handle: handle3, url: url3, receiver: mut receiver3 } =
+        create_webhook_server().await;
+
+    let db_path = tempfile::tempdir().unwrap();
+    let server = Server::new_with_options(Opt {
+        task_webhook_url: Some(Url::parse(&url3).unwrap()),
+        ..default_settings(db_path.path())
+    })
+    .await
+    .unwrap();
+
+    for url in [url1, url2] {
+        let (value, code) = server.create_webhook(json!({ "url": url })).await;
+        snapshot!(code, @"201 Created");
+        snapshot!(json_string!(value, { ".uuid" => "[uuid]", ".url" => "[ignored]" }), @r#"
+        {
+          "uuid": "[uuid]",
+          "isEditable": true,
+          "url": "[ignored]",
+          "headers": {}
+        }
+        "#);
+    }
+    let index = server.index("tamo");
+    let (response, code) = index
+        .add_documents_with_custom_metadata(
+            json!({ "id": 1, "doggo": "bone" }),
+            None,
+            Some("test_meta"),
+        )
+        .await;
+
+    snapshot!(response, @r###"
+    {
+      "taskUid": 0,
+      "indexUid": "tamo",
+      "status": "enqueued",
+      "type": "documentAdditionOrUpdate",
+      "enqueuedAt": "[date]",
+      "customMetadata": "test_meta"
+    }
+    "###);
+    snapshot!(code, @"202 Accepted");
+
+    let mut count1 = 0;
+    let mut count2 = 0;
+    let mut count3 = 0;
+    while count1 == 0 || count2 == 0 || count3 == 0 {
+        tokio::select! {
+            msg = receiver1.recv() => {
+              if let Some(msg) = msg {
+                count1 += 1;
+                check_metadata(msg);
+              }
+           },
+            msg = receiver2.recv() => {
+              if let Some(msg) = msg {
+                count2 += 1;
+                check_metadata(msg);
+              }
+             },
+            msg = receiver3.recv() => {
+              if let Some(msg) = msg {
+                count3 += 1;
+                check_metadata(msg);
+              }
+            },
+        }
+    }
+
+    assert_eq!(count1, 1);
+    assert_eq!(count2, 1);
+    assert_eq!(count3, 1);
+
+    handle1.abort();
+    handle2.abort();
+    handle3.abort();
+}
+
+fn check_metadata(msg: Vec<u8>) {
+    let msg = String::from_utf8(msg).unwrap();
+    let tasks = msg.split('\n');
+    for task in tasks {
+        if task.is_empty() {
+            continue;
+        }
+        let task: serde_json::Value = serde_json::from_str(task).unwrap();
+        snapshot!(common::Value(task), @r###"
+        {
+          "uid": "[uid]",
+          "batchUid": "[batch_uid]",
+          "indexUid": "tamo",
+          "status": "succeeded",
+          "type": "documentAdditionOrUpdate",
+          "canceledBy": null,
+          "details": {
+            "receivedDocuments": 1,
+            "indexedDocuments": 1
+          },
+          "error": null,
+          "duration": "[duration]",
+          "enqueuedAt": "[date]",
+          "startedAt": "[date]",
+          "finishedAt": "[date]",
+          "customMetadata": "test_meta"
+        }
+        "###);
+    }
+}

crates/meilisearch/tests/upgrade/mod.rs

@@ -43,7 +43,7 @@ async fn version_too_old() {
     std::fs::write(db_path.join("VERSION"), "1.11.9999").unwrap();
     let options = Opt { experimental_dumpless_upgrade: true, ..default_settings };
     let err = Server::new_with_options(options).await.map(|_| ()).unwrap_err();
-    snapshot!(err, @"Database version 1.11.9999 is too old for the experimental dumpless upgrade feature. Please generate a dump using the v1.11.9999 and import it in the v1.24.0");
+    snapshot!(err, @"Database version 1.11.9999 is too old for the experimental dumpless upgrade feature. Please generate a dump using the v1.11.9999 and import it in the v1.26.0");
 }
 
 #[actix_rt::test]
@@ -58,7 +58,7 @@ async fn version_requires_downgrade() {
     std::fs::write(db_path.join("VERSION"), format!("{major}.{minor}.{patch}")).unwrap();
     let options = Opt { experimental_dumpless_upgrade: true, ..default_settings };
     let err = Server::new_with_options(options).await.map(|_| ()).unwrap_err();
-    snapshot!(err, @"Database version 1.24.1 is higher than the Meilisearch version 1.24.0. Downgrade is not supported");
+    snapshot!(err, @"Database version 1.26.1 is higher than the Meilisearch version 1.26.0. Downgrade is not supported");
 }
 
 #[actix_rt::test]

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/batches_filter_afterEnqueuedAt_equal_2025-01-16T16_47_41.snap

@@ -8,7 +8,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "progress": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.24.0"
+        "upgradeTo": "v1.26.0"
       },
       "stats": {
         "totalNbTasks": 1,

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/batches_filter_afterFinishedAt_equal_2025-01-16T16_47_41.snap

@@ -8,7 +8,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "progress": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.24.0"
+        "upgradeTo": "v1.26.0"
       },
       "stats": {
         "totalNbTasks": 1,

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/batches_filter_afterStartedAt_equal_2025-01-16T16_47_41.snap

@@ -8,7 +8,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "progress": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.24.0"
+        "upgradeTo": "v1.26.0"
       },
       "stats": {
         "totalNbTasks": 1,

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/tasks_filter_afterEnqueuedAt_equal_2025-01-16T16_47_41.snap

@@ -12,7 +12,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "canceledBy": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.24.0"
+        "upgradeTo": "v1.26.0"
       },
       "error": null,
       "duration": "[duration]",

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/tasks_filter_afterFinishedAt_equal_2025-01-16T16_47_41.snap

@@ -12,7 +12,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "canceledBy": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.24.0"
+        "upgradeTo": "v1.26.0"
       },
       "error": null,
       "duration": "[duration]",

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/tasks_filter_afterStartedAt_equal_2025-01-16T16_47_41.snap

@@ -12,7 +12,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "canceledBy": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.24.0"
+        "upgradeTo": "v1.26.0"
       },
       "error": null,
       "duration": "[duration]",

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/the_whole_batch_queue_once_everything_has_been_processed.snap

@@ -8,7 +8,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "progress": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.24.0"
+        "upgradeTo": "v1.26.0"
       },
       "stats": {
         "totalNbTasks": 1,

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/the_whole_task_queue_once_everything_has_been_processed.snap

@@ -12,7 +12,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "canceledBy": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.24.0"
+        "upgradeTo": "v1.26.0"
       },
       "error": null,
       "duration": "[duration]",

crates/meilitool/Cargo.toml

@@ -9,15 +9,15 @@ edition.workspace = true
 license.workspace = true
 
 [dependencies]
-anyhow = "1.0.98"
-clap = { version = "4.5.40", features = ["derive"] }
+anyhow = "1.0.100"
+clap = { version = "4.5.52", features = ["derive"] }
 dump = { path = "../dump" }
 file-store = { path = "../file-store" }
-indexmap = { version = "2.9.0", features = ["serde"] }
+indexmap = { version = "2.12.0", features = ["serde"] }
 meilisearch-auth = { path = "../meilisearch-auth" }
 meilisearch-types = { path = "../meilisearch-types" }
-serde = { version = "1.0.219", features = ["derive"] }
-serde_json = { version = "1.0.140", features = ["preserve_order"] }
-tempfile = "3.20.0"
-time = { version = "0.3.41", features = ["formatting", "parsing", "alloc"] }
-uuid = { version = "1.17.0", features = ["v4"], default-features = false }
+serde = { version = "1.0.228", features = ["derive"] }
+serde_json = { version = "1.0.145", features = ["preserve_order"] }
+tempfile = "3.23.0"
+time = { version = "0.3.44", features = ["formatting", "parsing", "alloc"] }
+uuid = { version = "1.18.1", features = ["v4"], default-features = false }

crates/milli/Cargo.toml

@@ -15,15 +15,15 @@ license.workspace = true
 big_s = "1.0.2"
 bimap = { version = "0.6.3", features = ["serde"] }
 bincode = "1.3.3"
-bstr = "1.12.0"
-bytemuck = { version = "1.23.1", features = ["extern_crate_alloc"] }
+bstr = "1.12.1"
+bytemuck = { version = "1.24.0", features = ["extern_crate_alloc"] }
 byteorder = "1.5.0"
 charabia = { version = "0.9.8", default-features = false }
 cellulite = "0.3.1-nested-rtxns-2"
 concat-arrays = "0.1.2"
-convert_case = "0.8.0"
+convert_case = "0.9.0"
 crossbeam-channel = "0.5.15"
-deserr = "0.6.3"
+deserr = "0.6.4"
 either = { version = "1.15.0", features = ["serde"] }
 flatten-serde-json = { path = "../flatten-serde-json" }
 fst = "0.4.7"
@@ -38,51 +38,52 @@ heed = { version = "0.22.1-nested-rtxns-6", default-features = false, features =
     "serde-json",
     "serde-bincode",
 ] }
-indexmap = { version = "2.9.0", features = ["serde"] }
+indexmap = { version = "2.12.0", features = ["serde"] }
 json-depth-checker = { path = "../json-depth-checker" }
 levenshtein_automata = { version = "0.2.1", features = ["fst_automaton"] }
-memchr = "2.7.5"
-memmap2 = "0.9.7"
+memchr = "2.7.6"
+memmap2 = "0.9.9"
 obkv = "0.3.0"
 once_cell = "1.21.3"
-ordered-float = "5.0.0"
-rayon = "1.10.0"
+ordered-float = "5.1.0"
+rayon = "1.11.0"
 roaring = { version = "0.10.12", features = ["serde"] }
 rstar = { version = "0.12.2", features = ["serde"] }
-serde = { version = "1.0.219", features = ["derive"] }
-serde_json = { version = "1.0.140", features = ["preserve_order", "raw_value"] }
+serde = { version = "1.0.228", features = ["derive"] }
+serde_json = { version = "1.0.145", features = ["preserve_order", "raw_value"] }
 slice-group-by = "0.3.1"
-smallstr = { version = "0.3.0", features = ["serde"] }
+smallstr = { version = "0.3.1", features = ["serde"] }
 smallvec = "1.15.1"
 smartstring = "1.0.1"
-tempfile = "3.20.0"
-thiserror = "2.0.12"
-time = { version = "0.3.41", features = [
+tempfile = "3.23.0"
+thiserror = "2.0.17"
+time = { version = "0.3.44", features = [
     "serde-well-known",
     "formatting",
     "parsing",
     "macros",
 ] }
-uuid = { version = "1.17.0", features = ["v4"] }
+uuid = { version = "1.18.1", features = ["v4"] }
 
 filter-parser = { path = "../filter-parser" }
 
 # documents words self-join
 itertools = "0.14.0"
 
-csv = "1.3.1"
+csv = "1.4.0"
 candle-core = { version = "0.9.1" }
 candle-transformers = { version = "0.9.1" }
 candle-nn = { version = "0.9.1" }
-tokenizers = { git = "https://github.com/huggingface/tokenizers.git", tag = "v0.15.2", version = "0.15.2", default-features = false, features = [
+tokenizers = { version = "0.22.1", default-features = false, features = [
     "onig",
 ] }
 hf-hub = { git = "https://github.com/dureuill/hf-hub.git", branch = "rust_tls", default-features = false, features = [
     "online",
 ] }
-tiktoken-rs = "0.7.0"
+safetensors = "0.6.2"
+tiktoken-rs = "0.9.1"
 liquid = "0.26.11"
-rhai = { version = "1.22.2", features = [
+rhai = { version = "1.23.6", features = [
     "serde",
     "no_module",
     "no_custom_syntax",
@@ -94,15 +95,14 @@ hannoy = { version = "0.0.9-nested-rtxns-2", features = ["arroy"] }
 rand = "0.8.5"
 tracing = "0.1.41"
 ureq = { version = "2.12.1", features = ["json"] }
-url = "2.5.4"
-hashbrown = "0.15.4"
-bumpalo = "3.18.1"
+url = "2.5.7"
+hashbrown = "0.15.5"
+bumpalo = "3.19.0"
 bumparaw-collections = "0.1.4"
 steppe = { version = "0.4", default-features = false }
 thread_local = "1.1.9"
-allocator-api2 = "0.3.0"
 rustc-hash = "2.1.1"
-enum-iterator = "2.1.0"
+enum-iterator = "2.3.0"
 bbqueue = { git = "https://github.com/meilisearch/bbqueue" }
 flume = { version = "0.11.1", default-features = false }
 utoipa = { version = "5.4.0", features = [
@@ -112,21 +112,21 @@ utoipa = { version = "5.4.0", features = [
     "time",
     "openapi_extensions",
 ] }
-lru = "0.14.0"
-twox-hash = { version = "2.1.1", default-features = false, features = [
+lru = "0.16.2"
+twox-hash = { version = "2.1.2", default-features = false, features = [
     "std",
     "xxhash3_64",
     "xxhash64",
 ] }
-geo-types = "0.7.16"
+geo-types = "0.7.17"
 zerometry = "0.3.0"
 
 [dev-dependencies]
-mimalloc = { version = "0.1.47", default-features = false }
+mimalloc = { version = "0.1.48", default-features = false }
 # fixed version due to format breakages in v1.40
 insta = "=1.39.0"
 maplit = "1.0.2"
-md5 = "0.7.0"
+md5 = "0.8.0"
 meili-snap = { path = "../meili-snap" }
 rand = { version = "0.8.5", features = ["small_rng"] }
 

crates/milli/src/update/index_documents/extract/extract_vector_points.rs

@@ -1173,6 +1173,7 @@ pub fn extract_embeddings_from_fragments<R: io::Read + io::Seek>(
         request_threads,
         &doc_alloc,
         embedder_stats,
+        false,
         on_embed,
     );
 

crates/milli/src/update/new/extract/vectors/mod.rs

@@ -35,6 +35,7 @@ pub struct EmbeddingExtractor<'a, 'b> {
     possible_embedding_mistakes: PossibleEmbeddingMistakes,
     embedder_stats: &'a EmbedderStats,
     threads: &'a ThreadPoolNoAbort,
+    failure_modes: EmbedderFailureModes,
 }
 
 impl<'a, 'b> EmbeddingExtractor<'a, 'b> {
@@ -46,7 +47,15 @@ impl<'a, 'b> EmbeddingExtractor<'a, 'b> {
         threads: &'a ThreadPoolNoAbort,
     ) -> Self {
         let possible_embedding_mistakes = PossibleEmbeddingMistakes::new(field_distribution);
-        Self { embedders, sender, threads, possible_embedding_mistakes, embedder_stats }
+        let failure_modes = EmbedderFailureModes::from_env();
+        Self {
+            embedders,
+            sender,
+            threads,
+            possible_embedding_mistakes,
+            embedder_stats,
+            failure_modes,
+        }
     }
 }
 
@@ -91,6 +100,7 @@ impl<'extractor> Extractor<'extractor> for EmbeddingExtractor<'_, '_> {
                 self.threads,
                 self.sender,
                 &context.doc_alloc,
+                self.failure_modes,
             ))
         }
 
@@ -267,6 +277,7 @@ pub struct SettingsChangeEmbeddingExtractor<'a, 'b, SD> {
     sender: EmbeddingSender<'a, 'b>,
     possible_embedding_mistakes: PossibleEmbeddingMistakes,
     threads: &'a ThreadPoolNoAbort,
+    failure_modes: EmbedderFailureModes,
 }
 
 impl<'a, 'b, SD: SettingsDelta> SettingsChangeEmbeddingExtractor<'a, 'b, SD> {
@@ -279,7 +290,16 @@ impl<'a, 'b, SD: SettingsDelta> SettingsChangeEmbeddingExtractor<'a, 'b, SD> {
         threads: &'a ThreadPoolNoAbort,
     ) -> Self {
         let possible_embedding_mistakes = PossibleEmbeddingMistakes::new(field_distribution);
-        Self { settings_delta, embedder_stats, sender, threads, possible_embedding_mistakes }
+        let failure_modes = EmbedderFailureModes::from_env();
+
+        Self {
+            settings_delta,
+            embedder_stats,
+            sender,
+            threads,
+            possible_embedding_mistakes,
+            failure_modes,
+        }
     }
 }
 
@@ -336,6 +356,7 @@ impl<'extractor, SD: SettingsDelta + Sync> SettingsChangeExtractor<'extractor>
                     self.threads,
                     self.sender,
                     &context.doc_alloc,
+                    self.failure_modes,
                 ),
                 reindex_action,
             ));
@@ -539,6 +560,7 @@ struct Chunks<'a, 'b, 'extractor> {
 enum ChunkType<'a, 'b> {
     DocumentTemplate {
         document_template: &'a Prompt,
+        ignore_document_template_failures: bool,
         session: EmbedSession<'a, OnEmbeddingDocumentUpdates<'a, 'b>, &'a str>,
     },
     Fragments {
@@ -559,6 +581,7 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
         threads: &'a ThreadPoolNoAbort,
         sender: EmbeddingSender<'a, 'b>,
         doc_alloc: &'a Bump,
+        failure_modes: EmbedderFailureModes,
     ) -> Self {
         let embedder = &runtime.embedder;
         let dimensions = embedder.dimensions();
@@ -567,12 +590,14 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
         let kind = if fragments.is_empty() {
             ChunkType::DocumentTemplate {
                 document_template: &runtime.document_template,
+                ignore_document_template_failures: failure_modes.ignore_document_template_failures,
                 session: EmbedSession::new(
                     &runtime.embedder,
                     embedder_name,
                     threads,
                     doc_alloc,
                     embedder_stats,
+                    failure_modes.ignore_embedder_failures,
                     OnEmbeddingDocumentUpdates {
                         embedder_id: embedder_info.embedder_id,
                         sender,
@@ -589,6 +614,7 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
                     threads,
                     doc_alloc,
                     embedder_stats,
+                    failure_modes.ignore_embedder_failures,
                     OnEmbeddingDocumentUpdates {
                         embedder_id: embedder_info.embedder_id,
                         sender,
@@ -693,7 +719,11 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
                     },
                 )?;
             }
-            ChunkType::DocumentTemplate { document_template, session } => {
+            ChunkType::DocumentTemplate {
+                document_template,
+                ignore_document_template_failures,
+                session,
+            } => {
                 let doc_alloc = session.doc_alloc();
 
                 let old_embedder = settings_delta.old_embedders().get(session.embedder_name());
@@ -702,6 +732,7 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
                 } else {
                     old_embedder.as_ref().map(|old_embedder| &old_embedder.document_template)
                 };
+
                 let extractor =
                     DocumentTemplateExtractor::new(document_template, doc_alloc, fields_ids_map);
                 let old_extractor = old_document_template.map(|old_document_template| {
@@ -710,7 +741,15 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
                 let metadata =
                     Metadata { docid, external_docid, extractor_id: extractor.extractor_id() };
 
-                match extractor.diff_settings(document, &external_docid, old_extractor.as_ref())? {
+                let extractor_diff = if *ignore_document_template_failures {
+                    let extractor = extractor.ignore_errors();
+                    let old_extractor = old_extractor.map(DocumentTemplateExtractor::ignore_errors);
+                    extractor.diff_settings(document, &external_docid, old_extractor.as_ref())?
+                } else {
+                    extractor.diff_settings(document, &external_docid, old_extractor.as_ref())?
+                };
+
+                match extractor_diff {
                     ExtractorDiff::Removed => {
                         if old_is_user_provided || full_reindex {
                             session.on_embed_mut().clear_vectors(docid);
@@ -758,7 +797,11 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
             new_must_regenerate,
         );
         match &mut self.kind {
-            ChunkType::DocumentTemplate { document_template, session } => {
+            ChunkType::DocumentTemplate {
+                document_template,
+                ignore_document_template_failures,
+                session,
+            } => {
                 let doc_alloc = session.doc_alloc();
                 let ex = DocumentTemplateExtractor::new(
                     document_template,
@@ -766,18 +809,33 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
                     new_fields_ids_map,
                 );
 
-                update_autogenerated(
-                    docid,
-                    external_docid,
-                    [ex],
-                    old_document,
-                    new_document,
-                    &external_docid,
-                    old_must_regenerate,
-                    old_is_user_provided,
-                    session,
-                    unused_vectors_distribution,
-                )?
+                if *ignore_document_template_failures {
+                    update_autogenerated(
+                        docid,
+                        external_docid,
+                        [ex.ignore_errors()],
+                        old_document,
+                        new_document,
+                        &external_docid,
+                        old_must_regenerate,
+                        old_is_user_provided,
+                        session,
+                        unused_vectors_distribution,
+                    )
+                } else {
+                    update_autogenerated(
+                        docid,
+                        external_docid,
+                        [ex],
+                        old_document,
+                        new_document,
+                        &external_docid,
+                        old_must_regenerate,
+                        old_is_user_provided,
+                        session,
+                        unused_vectors_distribution,
+                    )
+                }?
             }
             ChunkType::Fragments { fragments, session } => {
                 let doc_alloc = session.doc_alloc();
@@ -844,23 +902,38 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
         );
 
         match &mut self.kind {
-            ChunkType::DocumentTemplate { document_template, session } => {
+            ChunkType::DocumentTemplate {
+                document_template,
+                ignore_document_template_failures,
+                session,
+            } => {
                 let doc_alloc = session.doc_alloc();
                 let ex = DocumentTemplateExtractor::new(
                     document_template,
                     doc_alloc,
                     new_fields_ids_map,
                 );
-
-                insert_autogenerated(
-                    docid,
-                    external_docid,
-                    [ex],
-                    new_document,
-                    &external_docid,
-                    session,
-                    unused_vectors_distribution,
-                )?;
+                if *ignore_document_template_failures {
+                    insert_autogenerated(
+                        docid,
+                        external_docid,
+                        [ex.ignore_errors()],
+                        new_document,
+                        &external_docid,
+                        session,
+                        unused_vectors_distribution,
+                    )?;
+                } else {
+                    insert_autogenerated(
+                        docid,
+                        external_docid,
+                        [ex],
+                        new_document,
+                        &external_docid,
+                        session,
+                        unused_vectors_distribution,
+                    )?;
+                }
             }
             ChunkType::Fragments { fragments, session } => {
                 let doc_alloc = session.doc_alloc();
@@ -884,7 +957,11 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
 
     pub fn drain(self, unused_vectors_distribution: &UnusedVectorsDistributionBump) -> Result<()> {
         match self.kind {
-            ChunkType::DocumentTemplate { document_template: _, session } => {
+            ChunkType::DocumentTemplate {
+                document_template: _,
+                ignore_document_template_failures: _,
+                session,
+            } => {
                 session.drain(unused_vectors_distribution)?;
             }
             ChunkType::Fragments { fragments: _, session } => {
@@ -896,9 +973,11 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
 
     pub fn embedder_name(&self) -> &'a str {
         match &self.kind {
-            ChunkType::DocumentTemplate { document_template: _, session } => {
-                session.embedder_name()
-            }
+            ChunkType::DocumentTemplate {
+                document_template: _,
+                ignore_document_template_failures: _,
+                session,
+            } => session.embedder_name(),
             ChunkType::Fragments { fragments: _, session } => session.embedder_name(),
         }
     }
@@ -967,7 +1046,11 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
             }
         }
         match &mut self.kind {
-            ChunkType::DocumentTemplate { document_template: _, session } => {
+            ChunkType::DocumentTemplate {
+                document_template: _,
+                ignore_document_template_failures: _,
+                session,
+            } => {
                 session.on_embed_mut().process_embeddings(
                     Metadata { docid, external_docid, extractor_id: 0 },
                     embeddings,
@@ -1078,3 +1161,41 @@ where
 
     Ok(())
 }
+
+#[derive(Clone, Copy, PartialEq, Eq, Default)]
+struct EmbedderFailureModes {
+    pub ignore_document_template_failures: bool,
+    pub ignore_embedder_failures: bool,
+}
+
+impl EmbedderFailureModes {
+    fn from_env() -> Self {
+        match std::env::var("MEILI_EXPERIMENTAL_CONFIG_EMBEDDER_FAILURE_MODES") {
+            Ok(failure_modes) => Self::parse_from_str(
+                &failure_modes,
+                "`MEILI_EXPERIMENTAL_CONFIG_EMBEDDER_FAILURE_MODES`",
+            ),
+            Err(std::env::VarError::NotPresent) => Self::default(),
+            Err(std::env::VarError::NotUnicode(_)) => panic!(
+                "`MEILI_EXPERIMENTAL_CONFIG_EMBEDDER_FAILURE_MODES` contains a non-unicode value"
+            ),
+        }
+    }
+
+    fn parse_from_str(failure_modes: &str, provenance: &'static str) -> Self {
+        let Self { mut ignore_document_template_failures, mut ignore_embedder_failures } =
+            Default::default();
+        for segment in failure_modes.split(',') {
+            let segment = segment.trim();
+            match segment {
+                "ignore_document_template_failures" => {
+                    ignore_document_template_failures = true;
+                }
+                "ignore_embedder_failures" => ignore_embedder_failures = true,
+                "" => continue,
+                segment => panic!("Unrecognized segment value for {provenance}: {segment}"),
+            }
+        }
+        Self { ignore_document_template_failures, ignore_embedder_failures }
+    }
+}

crates/milli/src/update/settings.rs

@@ -1631,8 +1631,11 @@ impl<'a, 't, 'i> Settings<'a, 't, 'i> {
 
             // Update index settings
             let embedding_config_updates = self.update_embedding_configs()?;
+            self.update_user_defined_searchable_attributes()?;
 
-            let new_inner_settings = InnerIndexSettings::from_index(self.index, self.wtxn, None)?;
+            let mut new_inner_settings =
+                InnerIndexSettings::from_index(self.index, self.wtxn, None)?;
+            new_inner_settings.recompute_searchables(self.wtxn, self.index)?;
 
             let primary_key_id = self
                 .index

crates/milli/src/update/upgrade/mod.rs

@@ -42,6 +42,8 @@ const UPGRADE_FUNCTIONS: &[&dyn UpgradeIndex] = &[
     &ToTargetNoOp { target: (1, 22, 0) },
     &ToTargetNoOp { target: (1, 23, 0) },
     &ToTargetNoOp { target: (1, 24, 0) },
+    &ToTargetNoOp { target: (1, 25, 0) },
+    &ToTargetNoOp { target: (1, 26, 0) },
     // This is the last upgrade function, it will be called when the index is up to date.
     // any other upgrade function should be added before this one.
     &ToCurrentNoOp {},
@@ -77,6 +79,8 @@ const fn start(from: (u32, u32, u32)) -> Option<usize> {
         (1, 22, _) => function_index!(12),
         (1, 23, _) => function_index!(13),
         (1, 24, _) => function_index!(14),
+        (1, 25, _) => function_index!(15),
+        (1, 26, _) => function_index!(16),
         // We deliberately don't add a placeholder with (VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH) here to force manually
         // considering dumpless upgrade.
         (_major, _minor, _patch) => return None,

crates/milli/src/vector/embedder/hf.rs

@@ -1,9 +1,11 @@
 use candle_core::Tensor;
 use candle_nn::VarBuilder;
-use candle_transformers::models::bert::{BertModel, Config, DTYPE};
+use candle_transformers::models::bert::{BertModel, Config as BertConfig, DTYPE};
+use candle_transformers::models::modernbert::{Config as ModernConfig, ModernBert};
 // FIXME: currently we'll be using the hub to retrieve model, in the future we might want to embed it into Meilisearch itself
 use hf_hub::api::sync::Api;
 use hf_hub::{Repo, RepoType};
+use safetensors::SafeTensors;
 use tokenizers::{PaddingParams, Tokenizer};
 
 use super::EmbeddingCache;
@@ -84,14 +86,21 @@ impl Default for EmbedderOptions {
     }
 }
 
+enum ModelKind {
+    Bert(BertModel),
+    Modern(ModernBert),
+}
+
 /// Perform embedding of documents and queries
 pub struct Embedder {
-    model: BertModel,
+    model: ModelKind,
     tokenizer: Tokenizer,
     options: EmbedderOptions,
     dimensions: usize,
     pooling: Pooling,
     cache: EmbeddingCache,
+    device: candle_core::Device,
+    max_len: usize,
 }
 
 impl std::fmt::Debug for Embedder {
@@ -101,10 +110,60 @@ impl std::fmt::Debug for Embedder {
             .field("tokenizer", &self.tokenizer)
             .field("options", &self.options)
             .field("pooling", &self.pooling)
+            .field("device", &self.device)
+            .field("max_len", &self.max_len)
             .finish()
     }
 }
 
+// some models do not have the "model." prefix in their safetensors weights
+fn change_tensor_names(
+    weights_path: &std::path::Path,
+) -> Result<std::path::PathBuf, NewEmbedderError> {
+    let data = std::fs::read(weights_path)
+        .map_err(|e| NewEmbedderError::safetensor_weight(candle_core::Error::Io(e)))?;
+
+    let tensors = SafeTensors::deserialize(&data)
+        .map_err(|e| NewEmbedderError::safetensor_weight(candle_core::Error::Msg(e.to_string())))?;
+
+    let names = tensors.names();
+    let has_model_prefix = names.iter().any(|n| n.starts_with("model."));
+
+    if has_model_prefix {
+        return Ok(weights_path.to_path_buf());
+    }
+
+    let fixed_path = weights_path.with_extension("fixed.safetensors");
+
+    if fixed_path.exists() {
+        return Ok(fixed_path);
+    }
+
+    let mut new_tensors = vec![];
+    for name in names {
+        let tensor_view = tensors.tensor(name).map_err(|e| {
+            NewEmbedderError::safetensor_weight(candle_core::Error::Msg(e.to_string()))
+        })?;
+
+        let new_name = format!("model.{}", name);
+        let data_offset = tensor_view.data();
+        let shape = tensor_view.shape();
+        let dtype = tensor_view.dtype();
+
+        new_tensors.push((new_name, shape.to_vec(), dtype, data_offset));
+    }
+
+    use safetensors::tensor::TensorView;
+    let views = new_tensors.iter().map(|(name, shape, dtype, data)| {
+        (name.as_str(), TensorView::new(*dtype, shape.clone(), data).unwrap())
+    });
+
+    safetensors::serialize_to_file(views, None, &fixed_path)
+        .map_err(|e| NewEmbedderError::safetensor_weight(candle_core::Error::Msg(e.to_string())))?;
+
+    Ok(fixed_path)
+}
+
 #[derive(Clone, Copy, serde::Deserialize)]
 struct PoolingConfig {
     #[serde(default)]
@@ -220,19 +279,42 @@ impl Embedder {
             (config, tokenizer, weights, source, pooling)
         };
 
-        let config = std::fs::read_to_string(&config_filename)
+        let config_str = std::fs::read_to_string(&config_filename)
             .map_err(|inner| NewEmbedderError::open_config(config_filename.clone(), inner))?;
-        let config: Config = serde_json::from_str(&config).map_err(|inner| {
-            NewEmbedderError::deserialize_config(
-                options.model.clone(),
-                config,
-                config_filename,
-                inner,
-            )
-        })?;
+
+        let cfg_val: serde_json::Value = match serde_json::from_str(&config_str) {
+            Ok(v) => v,
+            Err(inner) => {
+                return Err(NewEmbedderError::deserialize_config(
+                    options.model.clone(),
+                    config_str.clone(),
+                    config_filename.clone(),
+                    inner,
+                ));
+            }
+        };
+
+        let model_type = cfg_val.get("model_type").and_then(|v| v.as_str()).unwrap_or_default();
+        let arch_arr = cfg_val.get("architectures").and_then(|v| v.as_array());
+        let has_arch = |needle: &str| {
+            model_type.eq_ignore_ascii_case(needle)
+                || arch_arr.is_some_and(|arr| {
+                    arr.iter().filter_map(|v| v.as_str()).any(|s| s.to_lowercase().contains(needle))
+                })
+        };
+
+        let is_modern = has_arch("modernbert");
+        tracing::debug!(is_modern, model_type, "detected HF architecture");
+
         let mut tokenizer = Tokenizer::from_file(&tokenizer_filename)
             .map_err(|inner| NewEmbedderError::open_tokenizer(tokenizer_filename, inner))?;
 
+        let weights_filename = if is_modern && weight_source == WeightSource::Safetensors {
+            change_tensor_names(&weights_filename)?
+        } else {
+            weights_filename
+        };
+
         let vb = match weight_source {
             WeightSource::Pytorch => VarBuilder::from_pth(&weights_filename, DTYPE, &device)
                 .map_err(NewEmbedderError::pytorch_weight)?,
@@ -244,7 +326,31 @@ impl Embedder {
 
         tracing::debug!(model = options.model, weight=?weight_source, pooling=?pooling, "model config");
 
-        let model = BertModel::load(vb, &config).map_err(NewEmbedderError::load_model)?;
+        // max length from config, fallback to 512
+        let max_len =
+            cfg_val.get("max_position_embeddings").and_then(|v| v.as_u64()).unwrap_or(512) as usize;
+
+        let model = if is_modern {
+            let config: ModernConfig = serde_json::from_str(&config_str).map_err(|inner| {
+                NewEmbedderError::deserialize_config(
+                    options.model.clone(),
+                    config_str.clone(),
+                    config_filename.clone(),
+                    inner,
+                )
+            })?;
+            ModelKind::Modern(ModernBert::load(vb, &config).map_err(NewEmbedderError::load_model)?)
+        } else {
+            let config: BertConfig = serde_json::from_str(&config_str).map_err(|inner| {
+                NewEmbedderError::deserialize_config(
+                    options.model.clone(),
+                    config_str.clone(),
+                    config_filename.clone(),
+                    inner,
+                )
+            })?;
+            ModelKind::Bert(BertModel::load(vb, &config).map_err(NewEmbedderError::load_model)?)
+        };
 
         if let Some(pp) = tokenizer.get_padding_mut() {
             pp.strategy = tokenizers::PaddingStrategy::BatchLongest
@@ -263,6 +369,8 @@ impl Embedder {
             dimensions: 0,
             pooling,
             cache: EmbeddingCache::new(cache_cap),
+            device,
+            max_len,
         };
 
         let embeddings = this
@@ -321,15 +429,29 @@ impl Embedder {
     pub fn embed_one(&self, text: &str) -> std::result::Result<Embedding, EmbedError> {
         let tokens = self.tokenizer.encode(text, true).map_err(EmbedError::tokenize)?;
         let token_ids = tokens.get_ids();
-        let token_ids = if token_ids.len() > 512 { &token_ids[..512] } else { token_ids };
         let token_ids =
-            Tensor::new(token_ids, &self.model.device).map_err(EmbedError::tensor_shape)?;
+            if token_ids.len() > self.max_len { &token_ids[..self.max_len] } else { token_ids };
+        let token_ids = Tensor::new(token_ids, &self.device).map_err(EmbedError::tensor_shape)?;
         let token_ids = Tensor::stack(&[token_ids], 0).map_err(EmbedError::tensor_shape)?;
-        let token_type_ids = token_ids.zeros_like().map_err(EmbedError::tensor_shape)?;
-        let embeddings = self
-            .model
-            .forward(&token_ids, &token_type_ids, None)
-            .map_err(EmbedError::model_forward)?;
+
+        let embeddings = match &self.model {
+            ModelKind::Bert(model) => {
+                let token_type_ids = token_ids.zeros_like().map_err(EmbedError::tensor_shape)?;
+                model
+                    .forward(&token_ids, &token_type_ids, None)
+                    .map_err(EmbedError::model_forward)?
+            }
+            ModelKind::Modern(model) => {
+                let mut mask_vec = tokens.get_attention_mask().to_vec();
+                if mask_vec.len() > self.max_len {
+                    mask_vec.truncate(self.max_len);
+                }
+                let mask = Tensor::new(mask_vec.as_slice(), &self.device)
+                    .map_err(EmbedError::tensor_shape)?;
+                let mask = Tensor::stack(&[mask], 0).map_err(EmbedError::tensor_shape)?;
+                model.forward(&token_ids, &mask).map_err(EmbedError::model_forward)?
+            }
+        };
 
         let embedding = Self::pooling(embeddings, self.pooling)?;
 

crates/milli/src/vector/embedder/rest.rs

@@ -91,6 +91,7 @@ struct EmbedderData {
     request: RequestData,
     response: Response,
     configuration_source: ConfigurationSource,
+    max_retry_duration: std::time::Duration,
 }
 
 #[derive(Debug)]
@@ -182,10 +183,15 @@ impl Embedder {
     ) -> Result<Self, NewEmbedderError> {
         let bearer = options.api_key.as_deref().map(|api_key| format!("Bearer {api_key}"));
 
+        let timeout = std::env::var("MEILI_EXPERIMENTAL_REST_EMBEDDER_TIMEOUT_SECONDS")
+            .ok()
+            .map(|p| p.parse().unwrap())
+            .unwrap_or(30);
+
         let client = ureq::AgentBuilder::new()
             .max_idle_connections(REQUEST_PARALLELISM * 2)
             .max_idle_connections_per_host(REQUEST_PARALLELISM * 2)
-            .timeout(std::time::Duration::from_secs(30))
+            .timeout(std::time::Duration::from_secs(timeout))
             .build();
 
         let request = RequestData::new(
@@ -196,6 +202,14 @@ impl Embedder {
 
         let response = Response::new(options.response, &request)?;
 
+        let max_retry_duration =
+            std::env::var("MEILI_EXPERIMENTAL_REST_EMBEDDER_MAX_RETRY_DURATION_SECONDS")
+                .ok()
+                .map(|p| p.parse().unwrap())
+                .unwrap_or(60);
+
+        let max_retry_duration = std::time::Duration::from_secs(max_retry_duration);
+
         let data = EmbedderData {
             client,
             bearer,
@@ -204,6 +218,7 @@ impl Embedder {
             response,
             configuration_source,
             headers: options.headers,
+            max_retry_duration,
         };
 
         let dimensions = if let Some(dimensions) = options.dimensions {
@@ -457,7 +472,7 @@ where
             }
         }?;
 
-        let retry_duration = retry_duration.min(std::time::Duration::from_secs(60)); // don't wait more than a minute
+        let retry_duration = retry_duration.min(data.max_retry_duration); // don't wait more than the max duration
 
         // randomly up to double the retry duration
         let retry_duration = retry_duration

crates/milli/src/vector/error.rs

@@ -550,9 +550,9 @@ pub struct DeserializePoolingConfig {
 #[derive(Debug, thiserror::Error)]
 #[error("model `{model_name}` appears to be unsupported{}\n  - inner error: {inner}",
 if architectures.is_empty() {
-    "\n  - Note: only models with architecture \"BertModel\" are supported.".to_string()
+    "\n  - Note: only models with architecture \"BertModel\" or \"ModernBert\" are supported.".to_string()
 } else {
-    format!("\n  - Note: model has declared architectures `{architectures:?}`, only models with architecture `\"BertModel\"` are supported.")
+    format!("\n  - Note: model has declared architectures `{architectures:?}`, only models with architecture `\"BertModel\"` or `\"ModernBert\"` are supported.")
 })]
 pub struct UnsupportedModel {
     pub model_name: String,

crates/milli/src/vector/session.rs

@@ -44,6 +44,7 @@ pub struct EmbedSession<'doc, C, I> {
     embedder_name: &'doc str,
 
     embedder_stats: &'doc EmbedderStats,
+    ignore_embedding_failures: bool,
 
     on_embed: C,
 }
@@ -87,6 +88,7 @@ impl<'doc, C: OnEmbed<'doc>, I: Input> EmbedSession<'doc, C, I> {
         threads: &'doc ThreadPoolNoAbort,
         doc_alloc: &'doc Bump,
         embedder_stats: &'doc EmbedderStats,
+        ignore_embedding_failures: bool,
         on_embed: C,
     ) -> Self {
         let capacity = embedder.prompt_count_in_chunk_hint() * embedder.chunk_count_hint();
@@ -99,6 +101,7 @@ impl<'doc, C: OnEmbed<'doc>, I: Input> EmbedSession<'doc, C, I> {
             threads,
             embedder_name,
             embedder_stats,
+            ignore_embedding_failures,
             on_embed,
         }
     }
@@ -109,13 +112,12 @@ impl<'doc, C: OnEmbed<'doc>, I: Input> EmbedSession<'doc, C, I> {
         rendered: I,
         unused_vectors_distribution: &C::ErrorMetadata,
     ) -> Result<()> {
-        if self.inputs.len() < self.inputs.capacity() {
-            self.inputs.push(rendered);
-            self.metadata.push(metadata);
-            return Ok(());
+        if self.inputs.len() >= self.inputs.capacity() {
+            self.embed_chunks(unused_vectors_distribution)?;
         }
-
-        self.embed_chunks(unused_vectors_distribution)
+        self.inputs.push(rendered);
+        self.metadata.push(metadata);
+        Ok(())
     }
 
     pub fn drain(mut self, unused_vectors_distribution: &C::ErrorMetadata) -> Result<C> {
@@ -144,24 +146,33 @@ impl<'doc, C: OnEmbed<'doc>, I: Input> EmbedSession<'doc, C, I> {
                 Ok(())
             }
             Err(error) => {
-                // reset metadata and inputs, and send metadata to the error processing.
+                // send metadata to the error processing.
                 let doc_alloc = self.metadata.bump();
                 let metadata = std::mem::replace(
                     &mut self.metadata,
                     BVec::with_capacity_in(self.inputs.capacity(), doc_alloc),
                 );
-                self.inputs.clear();
-                return Err(self.on_embed.process_embedding_error(
+                Err(self.on_embed.process_embedding_error(
                     error,
                     self.embedder_name,
                     unused_vectors_distribution,
                     metadata,
-                ));
+                ))
             }
         };
         self.inputs.clear();
         self.metadata.clear();
-        res
+        if self.ignore_embedding_failures {
+            if let Err(err) = res {
+                tracing::warn!(
+                    %err,
+                    "ignored error embedding batch of documents due to failure policy"
+                );
+            }
+            Ok(())
+        } else {
+            res
+        }
     }
 
     pub(crate) fn embedder_name(&self) -> &'doc str {

crates/openapi-generator/Cargo.toml

@@ -7,6 +7,6 @@ publish = false
 [dependencies]
 meilisearch = { path = "../meilisearch" , default-features = false}
 serde_json = "1.0"
-clap = { version = "4.5.40", features = ["derive"] }
-anyhow = "1.0.98"
+clap = { version = "4.5.52", features = ["derive"] }
+anyhow = "1.0.100"
 utoipa = "5.4.0"

crates/tracing-trace/Cargo.toml

@@ -8,8 +8,8 @@ edition = "2021"
 [dependencies]
 color-spantrace = "0.3.0"
 fxprof-processed-profile = "0.7.0"
-serde = { version = "1.0.219", features = ["derive"] }
-serde_json = "1.0.140"
+serde = { version = "1.0.228", features = ["derive"] }
+serde_json = "1.0.145"
 tracing = "0.1.41"
 tracing-error = "0.2.1"
 tracing-subscriber = "0.3.20"
@@ -18,7 +18,7 @@ byte-unit = { version = "5.1.6", default-features = false, features = [
     "byte",
     "serde",
 ] }
-tokio = { version = "1.45.1", features = ["sync"] }
+tokio = { version = "1.48.0", features = ["sync"] }
 
 [target.'cfg(any(target_os = "linux", target_os = "macos"))'.dependencies]
-libproc = "0.14.10"
+libproc = "0.14.11"

crates/xtask/Cargo.toml

@@ -11,27 +11,27 @@ license.workspace = true
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-anyhow = "1.0.98"
+anyhow = "1.0.100"
 build-info = { version = "1.7.0", path = "../build-info" }
-cargo_metadata = "0.20.0"
-clap = { version = "4.5.40", features = ["derive"] }
+cargo_metadata = "0.23.1"
+clap = { version = "4.5.52", features = ["derive"] }
 futures-core = "0.3.31"
 futures-util = "0.3.31"
-reqwest = { version = "0.12.20", features = [
+reqwest = { version = "0.12.24", features = [
     "stream",
     "json",
     "rustls-tls",
 ], default-features = false }
-serde = { version = "1.0.219", features = ["derive"] }
-serde_json = "1.0.140"
+serde = { version = "1.0.228", features = ["derive"] }
+serde_json = "1.0.145"
 sha2 = "0.10.9"
-sysinfo = "0.35.2"
-time = { version = "0.3.41", features = [
+sysinfo = "0.37.2"
+time = { version = "0.3.44", features = [
     "serde",
     "serde-human-readable",
     "macros",
 ] }
-tokio = { version = "1.45.1", features = [
+tokio = { version = "1.48.0", features = [
     "rt",
     "net",
     "time",
@@ -41,4 +41,4 @@ tokio = { version = "1.45.1", features = [
 tracing = "0.1.41"
 tracing-subscriber = "0.3.20"
 tracing-trace = { version = "0.1.0", path = "../tracing-trace" }
-uuid = { version = "1.17.0", features = ["v7", "serde"] }
+uuid = { version = "1.18.1", features = ["v7", "serde"] }