Merge #4408

4408: Disable incremental facet update as a stop gap r=Kerollmops a=dureuill

# Pull Request

Temporarily disable incremental facet update to fix issue related with out-of-order sorts.

## Related issue

Workaround for #4409 

## What does this PR do?
- Always use bulk insert to insert facet values


Co-authored-by: Louis Dureuil <louis@meilisearch.com>

.github/workflows/benchmarks-pr.yml

@@ -90,7 +90,8 @@ jobs:
           set -x
           export base_ref=$(git merge-base origin/main ${{ steps.comment-branch.outputs.head_ref }} | head -c8)
           export base_filename=$(echo ${{ steps.command.outputs.command-arguments }}_main_${base_ref}.json)
-          echo 'Here are your benchmarks diff 👊' >> body.txt
+          export bench_name=$(echo ${{ steps.command.outputs.command-arguments }})
+          echo "Here are your $bench_name benchmarks diff 👊" >> body.txt
           echo '```' >> body.txt
           ./benchmarks/scripts/compare.sh $base_filename ${{ steps.file.outputs.basename }}.json >> body.txt
           echo '```' >> body.txt

.github/workflows/publish-apt-brew-pkg.yml

@@ -50,7 +50,7 @@ jobs:
     needs: check-version
     steps:
       - name: Create PR to Homebrew
-        uses: mislav/bump-homebrew-formula-action@v2
+        uses: mislav/bump-homebrew-formula-action@v3
         with:
           formula-name: meilisearch
           formula-path: Formula/m/meilisearch.rb

.github/workflows/publish-docker-images.yml

@@ -63,7 +63,7 @@ jobs:
         uses: docker/setup-buildx-action@v3
 
       - name: Login to Docker Hub
-        uses: docker/login-action@v2
+        uses: docker/login-action@v3
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}

.github/workflows/sdks-tests.yml

@@ -160,7 +160,7 @@ jobs:
         with:
           repository: meilisearch/meilisearch-js
       - name: Setup node
-        uses: actions/setup-node@v3
+        uses: actions/setup-node@v4
         with:
           cache: 'yarn'
       - name: Install dependencies
@@ -318,7 +318,7 @@ jobs:
         with:
           repository: meilisearch/meilisearch-js-plugins
       - name: Setup node
-        uses: actions/setup-node@v3
+        uses: actions/setup-node@v4
         with:
           cache: yarn
       - name: Install dependencies

.github/workflows/test-suite.yml

@@ -43,7 +43,7 @@ jobs:
           toolchain: nightly
           override: true
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run cargo check without any default features
         uses: actions-rs/cargo@v1
         with:
@@ -65,7 +65,7 @@ jobs:
     steps:
       - uses: actions/checkout@v3
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run cargo check without any default features
         uses: actions-rs/cargo@v1
         with:
@@ -149,7 +149,7 @@ jobs:
           toolchain: stable
           override: true
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run tests in debug
         uses: actions-rs/cargo@v1
         with:
@@ -168,7 +168,7 @@ jobs:
           override: true
           components: clippy
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run cargo clippy
         uses: actions-rs/cargo@v1
         with:
@@ -187,7 +187,7 @@ jobs:
           override: true
           components: rustfmt
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run cargo fmt
         # Since we never ran the `build.rs` script in the benchmark directory we are missing one auto-generated import file.
         # Since we want to trigger (and fail) this action as fast as possible, instead of building the benchmark crate

Cargo.toml

@@ -2,6 +2,7 @@
 resolver = "2"
 members = [
     "meilisearch",
+    "meilitool",
     "meilisearch-types",
     "meilisearch-auth",
     "meili-snap",
@@ -18,7 +19,7 @@ members = [
 ]
 
 [workspace.package]
-version = "1.4.1"
+version = "1.6.2"
 authors = ["Quentin de Quelen <quentin@dequelen.me>", "Clément Renault <clement@meilisearch.com>"]
 description = "Meilisearch HTTP server"
 homepage = "https://meilisearch.com"

Dockerfile

@@ -1,9 +1,9 @@
 # Compile
-FROM    rust:alpine3.16 AS compiler
+FROM    rust:1.71.1-alpine3.18 AS compiler
 
 RUN     apk add -q --update-cache --no-cache build-base openssl-dev
 
-WORKDIR /meilisearch
+WORKDIR /
 
 ARG     COMMIT_SHA
 ARG     COMMIT_DATE
@@ -17,7 +17,7 @@ RUN     set -eux; \
         if [ "$apkArch" = "aarch64" ]; then \
             export JEMALLOC_SYS_WITH_LG_PAGE=16; \
         fi && \
-        cargo build --release
+        cargo build --release -p meilisearch -p meilitool
 
 # Run
 FROM    alpine:3.16
@@ -28,9 +28,10 @@ ENV     MEILI_SERVER_PROVIDER docker
 RUN     apk update --quiet \
         && apk add -q --no-cache libgcc tini curl
 
-# add meilisearch to the `/bin` so you can run it from anywhere and it's easy
-# to find.
-COPY    --from=compiler /meilisearch/target/release/meilisearch /bin/meilisearch
+# add meilisearch and meilitool to the `/bin` so you can run it from anywhere
+# and it's easy to find.
+COPY    --from=compiler /target/release/meilisearch /bin/meilisearch
+COPY    --from=compiler /target/release/meilitool /bin/meilitool
 # To stay compatible with the older version of the container (pre v0.27.0) we're
 # going to symlink the meilisearch binary in the path to `/meilisearch`
 RUN     ln -s /bin/meilisearch /meilisearch

README.md

@@ -25,12 +25,6 @@
 
 <p align="center">⚡ A lightning-fast search engine that fits effortlessly into your apps, websites, and workflow 🔍</p>
 
----
-
-### 🔥 On November 2nd, we are hosting our first-ever live demo and product updates for [Meilisearch Cloud](https://www.meilisearch.com/cloud?utm_campaign=oss&utm_source=github&utm_medium=meilisearch). Make sure to [register here](https://us06web.zoom.us/meeting/register/tZMlc-mqrjIsH912-HTRe-AaT-pp41bDe81a#/registration) and bring your questions for live Q&A!
-
----
-
 Meilisearch helps you shape a delightful search experience in a snap, offering features that work out-of-the-box to speed up your workflow.
 
 <p align="center" name="demo">

benchmarks/benches/indexing.rs

@@ -36,7 +36,7 @@ fn setup_index() -> Index {
 }
 
 fn setup_settings<'t>(
-    wtxn: &mut RwTxn<'t, '_>,
+    wtxn: &mut RwTxn<'t>,
     index: &'t Index,
     primary_key: &str,
     searchable_fields: &[&str],
@@ -864,22 +864,12 @@ fn delete_documents_from_ids(index: Index, document_ids_to_delete: Vec<RoaringBi
 
     let indexer_config = IndexerConfig::default();
     for ids in document_ids_to_delete {
-        let external_documents_ids = index.external_documents_ids();
-        // FIXME: for filters matching a lot of documents, this will allocate a huge vec of external docids (strings).
-        // Since what we have is an iterator, it would be better to delete in chunks
-        let external_to_internal: std::result::Result<Vec<_>, RoaringBitmap> =
-            external_documents_ids
-                .find_external_id_of(&wtxn, ids)
-                .unwrap()
-                .only_external_ids()
-                .collect();
-        let ids = external_to_internal.unwrap();
         let config = IndexDocumentsConfig::default();
 
         let mut builder =
             IndexDocuments::new(&mut wtxn, &index, &indexer_config, config, |_| (), || false)
                 .unwrap();
-        (builder, _) = builder.remove_documents(ids).unwrap();
+        (builder, _) = builder.remove_documents_from_db_no_batch(&ids).unwrap();
         builder.execute().unwrap();
     }
 

config.toml

@@ -129,3 +129,6 @@ experimental_enable_metrics = false
 
 # Experimental RAM reduction during indexing, do not use in production, see: <https://github.com/meilisearch/product/discussions/652>
 experimental_reduce_indexing_memory_usage = false
+
+# Experimentally reduces the maximum number of tasks that will be processed at once, see: <https://github.com/orgs/meilisearch/discussions/713>
+# experimental_max_number_of_batched_tasks = 100

dump/src/lib.rs

@@ -267,6 +267,7 @@ pub(crate) mod test {
             dictionary: Setting::NotSet,
             synonyms: Setting::NotSet,
             distinct_attribute: Setting::NotSet,
+            proximity_precision: Setting::NotSet,
             typo_tolerance: Setting::NotSet,
             faceting: Setting::Set(FacetingSettings {
                 max_values_per_facet: Setting::Set(111),
@@ -275,6 +276,7 @@ pub(crate) mod test {
                 ),
             }),
             pagination: Setting::NotSet,
+            embedders: Setting::NotSet,
             _kind: std::marker::PhantomData,
         };
         settings.check()

dump/src/reader/compat/v5_to_v6.rs

@@ -345,6 +345,7 @@ impl<T> From<v5::Settings<T>> for v6::Settings<v6::Unchecked> {
             dictionary: v6::Setting::NotSet,
             synonyms: settings.synonyms.into(),
             distinct_attribute: settings.distinct_attribute.into(),
+            proximity_precision: v6::Setting::NotSet,
             typo_tolerance: match settings.typo_tolerance {
                 v5::Setting::Set(typo) => v6::Setting::Set(v6::TypoTolerance {
                     enabled: typo.enabled.into(),
@@ -377,6 +378,7 @@ impl<T> From<v5::Settings<T>> for v6::Settings<v6::Unchecked> {
                 v5::Setting::Reset => v6::Setting::Reset,
                 v5::Setting::NotSet => v6::Setting::NotSet,
             },
+            embedders: v6::Setting::NotSet,
             _kind: std::marker::PhantomData,
         }
     }

dump/src/reader/mod.rs

@@ -13,12 +13,12 @@ use crate::{Result, Version};
 
 mod compat;
 
-pub(self) mod v1;
-pub(self) mod v2;
-pub(self) mod v3;
-pub(self) mod v4;
-pub(self) mod v5;
-pub(self) mod v6;
+mod v1;
+mod v2;
+mod v3;
+mod v4;
+mod v5;
+mod v6;
 
 pub type Document = serde_json::Map<String, serde_json::Value>;
 pub type UpdateFile = dyn Iterator<Item = Result<Document>>;

dump/src/reader/v1/settings.rs

@@ -56,8 +56,7 @@ pub enum RankingRule {
     Desc(String),
 }
 
-static ASC_DESC_REGEX: Lazy<Regex> =
-    Lazy::new(|| Regex::new(r#"(asc|desc)\(([\w_-]+)\)"#).unwrap());
+static ASC_DESC_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(asc|desc)\(([\w_-]+)\)").unwrap());
 
 impl FromStr for RankingRule {
     type Err = ();

filter-parser/src/lib.rs

@@ -564,10 +564,10 @@ pub mod tests {
 
     #[test]
     fn parse_escaped() {
-        insta::assert_display_snapshot!(p(r#"title = 'foo\\'"#), @r#"{title} = {foo\}"#);
-        insta::assert_display_snapshot!(p(r#"title = 'foo\\\\'"#), @r#"{title} = {foo\\}"#);
-        insta::assert_display_snapshot!(p(r#"title = 'foo\\\\\\'"#), @r#"{title} = {foo\\\}"#);
-        insta::assert_display_snapshot!(p(r#"title = 'foo\\\\\\\\'"#), @r#"{title} = {foo\\\\}"#);
+        insta::assert_display_snapshot!(p(r"title = 'foo\\'"), @r#"{title} = {foo\}"#);
+        insta::assert_display_snapshot!(p(r"title = 'foo\\\\'"), @r#"{title} = {foo\\}"#);
+        insta::assert_display_snapshot!(p(r"title = 'foo\\\\\\'"), @r#"{title} = {foo\\\}"#);
+        insta::assert_display_snapshot!(p(r"title = 'foo\\\\\\\\'"), @r#"{title} = {foo\\\\}"#);
         // but it also works with other sequencies
         insta::assert_display_snapshot!(p(r#"title = 'foo\x20\n\t\"\'"'"#), @"{title} = {foo \n\t\"\'\"}");
     }

filter-parser/src/value.rs

@@ -270,8 +270,8 @@ pub mod test {
             ("aaaa", "", rtok("", "aaaa"), "aaaa"),
             (r#"aa"aa"#, r#""aa"#, rtok("", "aa"), "aa"),
             (r#"aa\"aa"#, r#""#, rtok("", r#"aa\"aa"#), r#"aa"aa"#),
-            (r#"aa\\\aa"#, r#""#, rtok("", r#"aa\\\aa"#), r#"aa\\\aa"#),
-            (r#"aa\\"\aa"#, r#""\aa"#, rtok("", r#"aa\\"#), r#"aa\\"#),
+            (r"aa\\\aa", r#""#, rtok("", r"aa\\\aa"), r"aa\\\aa"),
+            (r#"aa\\"\aa"#, r#""\aa"#, rtok("", r"aa\\"), r"aa\\"),
             (r#"aa\\\"\aa"#, r#""#, rtok("", r#"aa\\\"\aa"#), r#"aa\\"\aa"#),
             (r#"\"\""#, r#""#, rtok("", r#"\"\""#), r#""""#),
         ];
@@ -301,12 +301,12 @@ pub mod test {
         );
         // simple quote
         assert_eq!(
-            unescape(Span::new_extra(r#"Hello \'World\'"#, ""), '\''),
+            unescape(Span::new_extra(r"Hello \'World\'", ""), '\''),
             r#"Hello 'World'"#.to_string()
         );
         assert_eq!(
-            unescape(Span::new_extra(r#"Hello \\\'World\\\'"#, ""), '\''),
-            r#"Hello \\'World\\'"#.to_string()
+            unescape(Span::new_extra(r"Hello \\\'World\\\'", ""), '\''),
+            r"Hello \\'World\\'".to_string()
         );
     }
 
@@ -335,19 +335,19 @@ pub mod test {
             ("\"cha'nnel\"", "cha'nnel", false),
             ("I'm tamo", "I", false),
             // escaped thing but not quote
-            (r#""\\""#, r#"\"#, true),
-            (r#""\\\\\\""#, r#"\\\"#, true),
-            (r#""aa\\aa""#, r#"aa\aa"#, true),
+            (r#""\\""#, r"\", true),
+            (r#""\\\\\\""#, r"\\\", true),
+            (r#""aa\\aa""#, r"aa\aa", true),
             // with double quote
             (r#""Hello \"world\"""#, r#"Hello "world""#, true),
             (r#""Hello \\\"world\\\"""#, r#"Hello \"world\""#, true),
             (r#""I'm \"super\" tamo""#, r#"I'm "super" tamo"#, true),
             (r#""\"\"""#, r#""""#, true),
             // with simple quote
-            (r#"'Hello \'world\''"#, r#"Hello 'world'"#, true),
-            (r#"'Hello \\\'world\\\''"#, r#"Hello \'world\'"#, true),
+            (r"'Hello \'world\''", r#"Hello 'world'"#, true),
+            (r"'Hello \\\'world\\\''", r"Hello \'world\'", true),
             (r#"'I\'m "super" tamo'"#, r#"I'm "super" tamo"#, true),
-            (r#"'\'\''"#, r#"''"#, true),
+            (r"'\'\''", r#"''"#, true),
         ];
 
         for (input, expected, escaped) in test_case {

fuzzers/src/bin/fuzz-indexing.rs

@@ -113,7 +113,7 @@ fn main() {
                             index.documents(&wtxn, res.documents_ids).unwrap();
                             progression.fetch_add(1, Ordering::Relaxed);
                         }
-                        wtxn.abort().unwrap();
+                        wtxn.abort();
                     });
                     if let err @ Err(_) = handle.join() {
                         stop.store(true, Ordering::Relaxed);

index-scheduler/Cargo.toml

@@ -18,11 +18,12 @@ derive_builder = "0.12.0"
 dump = { path = "../dump" }
 enum-iterator = "1.4.0"
 file-store = { path = "../file-store" }
+flate2 = "1.0.28"
 log = "0.4.17"
 meilisearch-auth = { path = "../meilisearch-auth" }
 meilisearch-types = { path = "../meilisearch-types" }
 page_size = "0.5.0"
-puffin = "0.16.0"
+puffin = { version = "0.16.0", features = ["serialization"] }
 roaring = { version = "0.10.1", features = ["serde"] }
 serde = { version = "1.0.160", features = ["derive"] }
 serde_json = { version = "1.0.95", features = ["preserve_order"] }
@@ -30,6 +31,7 @@ synchronoise = "1.0.1"
 tempfile = "3.5.0"
 thiserror = "1.0.40"
 time = { version = "0.3.20", features = ["serde-well-known", "formatting", "parsing", "macros"] }
+ureq = "2.9.1"
 uuid = { version = "1.3.1", features = ["serde", "v4"] }
 
 [dev-dependencies]

index-scheduler/src/batch.rs

@@ -32,7 +32,7 @@ use meilisearch_types::milli::heed::CompactionOption;
 use meilisearch_types::milli::update::{
     IndexDocumentsConfig, IndexDocumentsMethod, IndexerConfig, Settings as MilliSettings,
 };
-use meilisearch_types::milli::{self, Filter, BEU32};
+use meilisearch_types::milli::{self, Filter};
 use meilisearch_types::settings::{apply_settings_to_builder, Settings, Unchecked};
 use meilisearch_types::tasks::{Details, IndexSwap, Kind, KindWithContent, Status, Task};
 use meilisearch_types::{compression, Index, VERSION_FILE_NAME};
@@ -104,12 +104,6 @@ pub(crate) enum IndexOperation {
         operations: Vec<DocumentOperation>,
         tasks: Vec<Task>,
     },
-    DocumentDeletion {
-        index_uid: String,
-        // The vec associated with each document deletion tasks.
-        documents: Vec<Vec<String>>,
-        tasks: Vec<Task>,
-    },
     IndexDocumentDeletionByFilter {
         index_uid: String,
         task: Task,
@@ -161,7 +155,6 @@ impl Batch {
             }
             Batch::IndexOperation { op, .. } => match op {
                 IndexOperation::DocumentOperation { tasks, .. }
-                | IndexOperation::DocumentDeletion { tasks, .. }
                 | IndexOperation::Settings { tasks, .. }
                 | IndexOperation::DocumentClear { tasks, .. } => {
                     tasks.iter().map(|task| task.uid).collect()
@@ -226,7 +219,6 @@ impl IndexOperation {
     pub fn index_uid(&self) -> &str {
         match self {
             IndexOperation::DocumentOperation { index_uid, .. }
-            | IndexOperation::DocumentDeletion { index_uid, .. }
             | IndexOperation::IndexDocumentDeletionByFilter { index_uid, .. }
             | IndexOperation::DocumentClear { index_uid, .. }
             | IndexOperation::Settings { index_uid, .. }
@@ -242,9 +234,6 @@ impl fmt::Display for IndexOperation {
             IndexOperation::DocumentOperation { .. } => {
                 f.write_str("IndexOperation::DocumentOperation")
             }
-            IndexOperation::DocumentDeletion { .. } => {
-                f.write_str("IndexOperation::DocumentDeletion")
-            }
             IndexOperation::IndexDocumentDeletionByFilter { .. } => {
                 f.write_str("IndexOperation::IndexDocumentDeletionByFilter")
             }
@@ -347,18 +336,27 @@ impl IndexScheduler {
             BatchKind::DocumentDeletion { deletion_ids } => {
                 let tasks = self.get_existing_tasks(rtxn, deletion_ids)?;
 
-                let mut documents = Vec::new();
+                let mut operations = Vec::with_capacity(tasks.len());
+                let mut documents_counts = Vec::with_capacity(tasks.len());
                 for task in &tasks {
                     match task.kind {
                         KindWithContent::DocumentDeletion { ref documents_ids, .. } => {
-                            documents.push(documents_ids.clone())
+                            operations.push(DocumentOperation::Delete(documents_ids.clone()));
+                            documents_counts.push(documents_ids.len() as u64);
                         }
                         _ => unreachable!(),
                     }
                 }
 
                 Ok(Some(Batch::IndexOperation {
-                    op: IndexOperation::DocumentDeletion { index_uid, documents, tasks },
+                    op: IndexOperation::DocumentOperation {
+                        index_uid,
+                        primary_key: None,
+                        method: IndexDocumentsMethod::ReplaceDocuments,
+                        documents_counts,
+                        operations,
+                        tasks,
+                    },
                     must_create_index,
                 }))
             }
@@ -586,7 +584,9 @@ impl IndexScheduler {
         let index_tasks = self.index_tasks(rtxn, index_name)? & enqueued;
 
         // If autobatching is disabled we only take one task at a time.
-        let tasks_limit = if self.autobatching_enabled { usize::MAX } else { 1 };
+        // Otherwise, we take only a maximum of tasks to create batches.
+        let tasks_limit =
+            if self.autobatching_enabled { self.max_number_of_batched_tasks } else { 1 };
 
         let enqueued = index_tasks
             .into_iter()
@@ -717,7 +717,7 @@ impl IndexScheduler {
 
                 // 2. Snapshot the index-scheduler LMDB env
                 //
-                // When we call copy_to_path, LMDB opens a read transaction by itself,
+                // When we call copy_to_file, LMDB opens a read transaction by itself,
                 // we can't provide our own. It is an issue as we would like to know
                 // the update files to copy but new ones can be enqueued between the copy
                 // of the env and the new transaction we open to retrieve the enqueued tasks.
@@ -730,7 +730,7 @@ impl IndexScheduler {
                 // 2.1 First copy the LMDB env of the index-scheduler
                 let dst = temp_snapshot_dir.path().join("tasks");
                 fs::create_dir_all(&dst)?;
-                self.env.copy_to_path(dst.join("data.mdb"), CompactionOption::Enabled)?;
+                self.env.copy_to_file(dst.join("data.mdb"), CompactionOption::Enabled)?;
 
                 // 2.2 Create a read transaction on the index-scheduler
                 let rtxn = self.env.read_txn()?;
@@ -755,7 +755,7 @@ impl IndexScheduler {
                     let index = self.index_mapper.index(&rtxn, name)?;
                     let dst = temp_snapshot_dir.path().join("indexes").join(uuid.to_string());
                     fs::create_dir_all(&dst)?;
-                    index.copy_to_path(dst.join("data.mdb"), CompactionOption::Enabled)?;
+                    index.copy_to_file(dst.join("data.mdb"), CompactionOption::Enabled)?;
                 }
 
                 drop(rtxn);
@@ -768,7 +768,7 @@ impl IndexScheduler {
                     .map_size(1024 * 1024 * 1024) // 1 GiB
                     .max_dbs(2)
                     .open(&self.auth_path)?;
-                auth.copy_to_path(dst.join("data.mdb"), CompactionOption::Enabled)?;
+                auth.copy_to_file(dst.join("data.mdb"), CompactionOption::Enabled)?;
 
                 // 5. Copy and tarball the flat snapshot
                 // 5.1 Find the original name of the database
@@ -824,6 +824,10 @@ impl IndexScheduler {
                 // 2. dump the tasks
                 let mut dump_tasks = dump.create_tasks_queue()?;
                 for ret in self.all_tasks.iter(&rtxn)? {
+                    if self.must_stop_processing.get() {
+                        return Err(Error::AbortedTask);
+                    }
+
                     let (_, mut t) = ret?;
                     let status = t.status;
                     let content_file = t.content_uuid();
@@ -844,6 +848,9 @@ impl IndexScheduler {
 
                     // 2.1. Dump the `content_file` associated with the task if there is one and the task is not finished yet.
                     if let Some(content_file) = content_file {
+                        if self.must_stop_processing.get() {
+                            return Err(Error::AbortedTask);
+                        }
                         if status == Status::Enqueued {
                             let content_file = self.file_store.get_update(content_file)?;
 
@@ -883,6 +890,9 @@ impl IndexScheduler {
 
                     // 3.1. Dump the documents
                     for ret in index.all_documents(&rtxn)? {
+                        if self.must_stop_processing.get() {
+                            return Err(Error::AbortedTask);
+                        }
                         let (_id, doc) = ret?;
                         let document = milli::obkv_to_json(&all_fields, &fields_ids_map, doc)?;
                         index_dumper.push_document(&document)?;
@@ -902,6 +912,9 @@ impl IndexScheduler {
                     "[year repr:full][month repr:numerical][day padding:zero]-[hour padding:zero][minute padding:zero][second padding:zero][subsecond digits:3]"
                 )).unwrap();
 
+                if self.must_stop_processing.get() {
+                    return Err(Error::AbortedTask);
+                }
                 let path = self.dumps_path.join(format!("{}.dump", dump_uid));
                 let file = File::create(path)?;
                 dump.persist_to(BufWriter::new(file))?;
@@ -922,6 +935,10 @@ impl IndexScheduler {
                     self.index_mapper.index(&rtxn, &index_uid)?
                 };
 
+                // the index operation can take a long time, so save this handle to make it available to the search for the duration of the tick
+                self.index_mapper
+                    .set_currently_updating_index(Some((index_uid.clone(), index.clone())));
+
                 let mut index_wtxn = index.write_txn()?;
                 let tasks = self.apply_index_operation(&mut index_wtxn, &index, op)?;
                 index_wtxn.commit()?;
@@ -1091,7 +1108,7 @@ impl IndexScheduler {
         for task_id in &index_lhs_task_ids | &index_rhs_task_ids {
             let mut task = self.get_task(wtxn, task_id)?.ok_or(Error::CorruptedTaskQueue)?;
             swap_index_uid_in_task(&mut task, (lhs, rhs));
-            self.all_tasks.put(wtxn, &BEU32::new(task_id), &task)?;
+            self.all_tasks.put(wtxn, &task_id, &task)?;
         }
 
         // 4. remove the task from indexuid = before_name
@@ -1117,7 +1134,7 @@ impl IndexScheduler {
     /// The list of processed tasks.
     fn apply_index_operation<'i>(
         &self,
-        index_wtxn: &mut RwTxn<'i, '_>,
+        index_wtxn: &mut RwTxn<'i>,
         index: &'i Index,
         operation: IndexOperation,
     ) -> Result<Vec<Task>> {
@@ -1185,6 +1202,10 @@ impl IndexScheduler {
 
                 let config = IndexDocumentsConfig { update_method: method, ..Default::default() };
 
+                let embedder_configs = index.embedding_configs(index_wtxn)?;
+                // TODO: consider Arc'ing the map too (we only need read access + we'll be cloning it multiple times, so really makes sense)
+                let embedders = self.embedders(embedder_configs)?;
+
                 let mut builder = milli::update::IndexDocuments::new(
                     index_wtxn,
                     index,
@@ -1203,6 +1224,8 @@ impl IndexScheduler {
                             let (new_builder, user_result) = builder.add_documents(reader)?;
                             builder = new_builder;
 
+                            builder = builder.with_embedders(embedders.clone());
+
                             let received_documents =
                                 if let Some(Details::DocumentAdditionOrUpdate {
                                     received_documents,
@@ -1275,45 +1298,6 @@ impl IndexScheduler {
 
                 Ok(tasks)
             }
-            IndexOperation::DocumentDeletion { index_uid: _, documents, mut tasks } => {
-                let indexer_config = self.index_mapper.indexer_config();
-                let config = IndexDocumentsConfig {
-                    update_method: IndexDocumentsMethod::ReplaceDocuments,
-                    ..Default::default()
-                };
-                let must_stop_processing = self.must_stop_processing.clone();
-
-                let mut builder = milli::update::IndexDocuments::new(
-                    index_wtxn,
-                    index,
-                    indexer_config,
-                    config,
-                    |indexing_step| trace!("update: {:?}", indexing_step),
-                    || must_stop_processing.get(),
-                )?;
-
-                let document_ids = documents.iter().flatten().cloned().collect();
-
-                let (new_builder, user_result) = builder.remove_documents(document_ids)?;
-                builder = new_builder;
-                // Uses Invariant: remove documents actually always returns Ok for the inner result
-                let count = user_result.unwrap();
-
-                for (task, documents) in tasks.iter_mut().zip(documents) {
-                    task.status = Status::Succeeded;
-                    task.details = Some(Details::DocumentDeletion {
-                        provided_ids: documents.len(),
-                        deleted_documents: Some(count.min(documents.len() as u64)),
-                    });
-                }
-
-                if !tasks.iter().all(|res| res.error.is_some()) {
-                    let addition = builder.execute()?;
-                    info!("document deletion done: {:?}", addition);
-                }
-
-                Ok(tasks)
-            }
             IndexOperation::IndexDocumentDeletionByFilter { mut task, index_uid: _ } => {
                 let filter =
                     if let KindWithContent::DocumentDeletionByFilter { filter_expr, .. } =
@@ -1503,10 +1487,9 @@ impl IndexScheduler {
         }
 
         for task in to_delete_tasks.iter() {
-            self.all_tasks.delete(wtxn, &BEU32::new(task))?;
+            self.all_tasks.delete(wtxn, &task)?;
         }
         for canceled_by in affected_canceled_by {
-            let canceled_by = BEU32::new(canceled_by);
             if let Some(mut tasks) = self.canceled_by.get(wtxn, &canceled_by)? {
                 tasks -= &to_delete_tasks;
                 if tasks.is_empty() {
@@ -1554,14 +1537,14 @@ impl IndexScheduler {
             task.details = task.details.map(|d| d.to_failed());
             self.update_task(wtxn, &task)?;
         }
-        self.canceled_by.put(wtxn, &BEU32::new(cancel_task_id), &tasks_to_cancel)?;
+        self.canceled_by.put(wtxn, &cancel_task_id, &tasks_to_cancel)?;
 
         Ok(content_files_to_delete)
     }
 }
 
 fn delete_document_by_filter<'a>(
-    wtxn: &mut RwTxn<'a, '_>,
+    wtxn: &mut RwTxn<'a>,
     filter: &serde_json::Value,
     indexer_config: &IndexerConfig,
     must_stop_processing: MustStopProcessing,
@@ -1575,18 +1558,6 @@ fn delete_document_by_filter<'a>(
             }
             e => e.into(),
         })?;
-        let external_documents_ids = index.external_documents_ids();
-        // FIXME: for filters matching a lot of documents, this will allocate a huge vec of external docids (strings).
-        // Since what we have is an iterator, it would be better to delete in chunks
-        let external_to_internal: std::result::Result<Vec<_>, RoaringBitmap> =
-            external_documents_ids
-                .find_external_id_of(wtxn, candidates)?
-                .only_external_ids()
-                .collect();
-        let document_ids = match external_to_internal {
-            Ok(external_ids) => external_ids,
-            Err(remaining_ids) => panic!("Couldn't find some external ids {:?}", remaining_ids),
-        };
 
         let config = IndexDocumentsConfig {
             update_method: IndexDocumentsMethod::ReplaceDocuments,
@@ -1602,13 +1573,10 @@ fn delete_document_by_filter<'a>(
             || must_stop_processing.get(),
         )?;
 
-        let (new_builder, user_result) = builder.remove_documents(document_ids)?;
+        let (new_builder, count) = builder.remove_documents_from_db_no_batch(&candidates)?;
         builder = new_builder;
-        // Uses Invariant: remove documents actually always returns Ok for the inner result
-        let count = user_result.unwrap();
 
         let _ = builder.execute()?;
-
         count
     } else {
         0

index-scheduler/src/error.rs

@@ -108,6 +108,8 @@ pub enum Error {
     TaskDeletionWithEmptyQuery,
     #[error("Query parameters to filter the tasks to cancel are missing. Available query parameters are: `uids`, `indexUids`, `statuses`, `types`, `canceledBy`, `beforeEnqueuedAt`, `afterEnqueuedAt`, `beforeStartedAt`, `afterStartedAt`, `beforeFinishedAt`, `afterFinishedAt`.")]
     TaskCancelationWithEmptyQuery,
+    #[error("Aborted task")]
+    AbortedTask,
 
     #[error(transparent)]
     Dump(#[from] dump::Error),
@@ -175,6 +177,7 @@ impl Error {
             | Error::TaskNotFound(_)
             | Error::TaskDeletionWithEmptyQuery
             | Error::TaskCancelationWithEmptyQuery
+            | Error::AbortedTask
             | Error::Dump(_)
             | Error::Heed(_)
             | Error::Milli(_)
@@ -236,6 +239,9 @@ impl ErrorCode for Error {
             Error::TaskDatabaseUpdate(_) => Code::Internal,
             Error::CreateBatch(_) => Code::Internal,
 
+            // This one should never be seen by the end user
+            Error::AbortedTask => Code::Internal,
+
             #[cfg(test)]
             Error::PlannedFailure => Code::Internal,
         }

index-scheduler/src/features.rs

@@ -56,12 +56,12 @@ impl RoFeatures {
         }
     }
 
-    pub fn check_vector(&self) -> Result<()> {
+    pub fn check_vector(&self, disabled_action: &'static str) -> Result<()> {
         if self.runtime.vector_store {
             Ok(())
         } else {
             Err(FeatureNotEnabledError {
-                disabled_action: "Passing `vector` as a query parameter",
+                disabled_action,
                 feature: "vector store",
                 issue_link: "https://github.com/meilisearch/product/discussions/677",
             }

index-scheduler/src/index_mapper/index_map.rs

@@ -1,12 +1,8 @@
-/// the map size to use when we don't succeed in reading it in indexes.
-const DEFAULT_MAP_SIZE: usize = 10 * 1024 * 1024 * 1024; // 10 GiB
-
 use std::collections::BTreeMap;
 use std::path::Path;
 use std::time::Duration;
 
-use meilisearch_types::heed::flags::Flags;
-use meilisearch_types::heed::{EnvClosingEvent, EnvOpenOptions};
+use meilisearch_types::heed::{EnvClosingEvent, EnvFlags, EnvOpenOptions};
 use meilisearch_types::milli::Index;
 use time::OffsetDateTime;
 use uuid::Uuid;
@@ -236,7 +232,7 @@ impl IndexMap {
         enable_mdb_writemap: bool,
         map_size_growth: usize,
     ) {
-        let map_size = index.map_size().unwrap_or(DEFAULT_MAP_SIZE) + map_size_growth;
+        let map_size = index.map_size() + map_size_growth;
         let closing_event = index.prepare_for_closing();
         let generation = self.next_generation();
         self.unavailable.insert(
@@ -309,7 +305,7 @@ fn create_or_open_index(
     options.map_size(clamp_to_page_size(map_size));
     options.max_readers(1024);
     if enable_mdb_writemap {
-        unsafe { options.flag(Flags::MdbWriteMap) };
+        unsafe { options.flags(EnvFlags::WRITE_MAP) };
     }
 
     if let Some((created, updated)) = date {
@@ -388,7 +384,7 @@ mod tests {
 
     fn assert_index_size(index: Index, expected: usize) {
         let expected = clamp_to_page_size(expected);
-        let index_map_size = index.map_size().unwrap();
+        let index_map_size = index.map_size();
         assert_eq!(index_map_size, expected);
     }
 }

index-scheduler/src/index_mapper/mod.rs

@@ -69,6 +69,10 @@ pub struct IndexMapper {
     /// Whether we open a meilisearch index with the MDB_WRITEMAP option or not.
     enable_mdb_writemap: bool,
     pub indexer_config: Arc<IndexerConfig>,
+
+    /// A few types of long running batches of tasks that act on a single index set this field
+    /// so that a handle to the index is available from other threads (search) in an optimized manner.
+    currently_updating_index: Arc<RwLock<Option<(String, Index)>>>,
 }
 
 /// Whether the index is available for use or is forbidden to be inserted back in the index map
@@ -151,6 +155,7 @@ impl IndexMapper {
             index_growth_amount,
             enable_mdb_writemap,
             indexer_config: Arc::new(indexer_config),
+            currently_updating_index: Default::default(),
         })
     }
 
@@ -303,6 +308,14 @@ impl IndexMapper {
 
     /// Return an index, may open it if it wasn't already opened.
     pub fn index(&self, rtxn: &RoTxn, name: &str) -> Result<Index> {
+        if let Some((current_name, current_index)) =
+            self.currently_updating_index.read().unwrap().as_ref()
+        {
+            if current_name == name {
+                return Ok(current_index.clone());
+            }
+        }
+
         let uuid = self
             .index_mapping
             .get(rtxn, name)?
@@ -474,4 +487,8 @@ impl IndexMapper {
     pub fn indexer_config(&self) -> &IndexerConfig {
         &self.indexer_config
     }
+
+    pub fn set_currently_updating_index(&self, index: Option<(String, Index)>) {
+        *self.currently_updating_index.write().unwrap() = index;
+    }
 }

index-scheduler/src/insta_snapshot.rs

@@ -1,7 +1,7 @@
 use std::collections::BTreeSet;
 use std::fmt::Write;
 
-use meilisearch_types::heed::types::{OwnedType, SerdeBincode, SerdeJson, Str};
+use meilisearch_types::heed::types::{SerdeBincode, SerdeJson, Str};
 use meilisearch_types::heed::{Database, RoTxn};
 use meilisearch_types::milli::{CboRoaringBitmapCodec, RoaringBitmapCodec, BEU32};
 use meilisearch_types::tasks::{Details, Task};
@@ -30,15 +30,19 @@ pub fn snapshot_index_scheduler(scheduler: &IndexScheduler) -> String {
         index_mapper,
         features: _,
         max_number_of_tasks: _,
+        max_number_of_batched_tasks: _,
         puffin_frame: _,
         wake_up: _,
         dumps_path: _,
         snapshots_path: _,
         auth_path: _,
         version_file_path: _,
+        webhook_url: _,
+        webhook_authorization_header: _,
         test_breakpoint_sdr: _,
         planned_failures: _,
         run_loop_iteration: _,
+        embedders: _,
     } = scheduler;
 
     let rtxn = env.read_txn().unwrap();
@@ -114,7 +118,7 @@ pub fn snapshot_bitmap(r: &RoaringBitmap) -> String {
     snap
 }
 
-pub fn snapshot_all_tasks(rtxn: &RoTxn, db: Database<OwnedType<BEU32>, SerdeJson<Task>>) -> String {
+pub fn snapshot_all_tasks(rtxn: &RoTxn, db: Database<BEU32, SerdeJson<Task>>) -> String {
     let mut snap = String::new();
     let iter = db.iter(rtxn).unwrap();
     for next in iter {
@@ -124,10 +128,7 @@ pub fn snapshot_all_tasks(rtxn: &RoTxn, db: Database<OwnedType<BEU32>, SerdeJson
     snap
 }
 
-pub fn snapshot_date_db(
-    rtxn: &RoTxn,
-    db: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
-) -> String {
+pub fn snapshot_date_db(rtxn: &RoTxn, db: Database<BEI128, CboRoaringBitmapCodec>) -> String {
     let mut snap = String::new();
     let iter = db.iter(rtxn).unwrap();
     for next in iter {
@@ -247,10 +248,7 @@ pub fn snapshot_index_tasks(rtxn: &RoTxn, db: Database<Str, RoaringBitmapCodec>)
     }
     snap
 }
-pub fn snapshot_canceled_by(
-    rtxn: &RoTxn,
-    db: Database<OwnedType<BEU32>, RoaringBitmapCodec>,
-) -> String {
+pub fn snapshot_canceled_by(rtxn: &RoTxn, db: Database<BEU32, RoaringBitmapCodec>) -> String {
     let mut snap = String::new();
     let iter = db.iter(rtxn).unwrap();
     for next in iter {

index-scheduler/src/lib.rs

@@ -27,13 +27,14 @@ mod index_mapper;
 mod insta_snapshot;
 mod lru;
 mod utils;
-mod uuid_codec;
+pub mod uuid_codec;
 
 pub type Result<T> = std::result::Result<T, Error>;
 pub type TaskId = u32;
 
 use std::collections::{BTreeMap, HashMap};
 use std::fs::File;
+use std::io::{self, BufReader, Read};
 use std::ops::{Bound, RangeBounds};
 use std::path::{Path, PathBuf};
 use std::sync::atomic::AtomicBool;
@@ -45,13 +46,18 @@ use dump::{KindDump, TaskDump, UpdateFile};
 pub use error::Error;
 pub use features::RoFeatures;
 use file_store::FileStore;
+use flate2::bufread::GzEncoder;
+use flate2::Compression;
 use meilisearch_types::error::ResponseError;
 use meilisearch_types::features::{InstanceTogglableFeatures, RuntimeTogglableFeatures};
-use meilisearch_types::heed::types::{OwnedType, SerdeBincode, SerdeJson, Str};
-use meilisearch_types::heed::{self, Database, Env, RoTxn, RwTxn};
+use meilisearch_types::heed::byteorder::BE;
+use meilisearch_types::heed::types::{SerdeBincode, SerdeJson, Str, I128};
+use meilisearch_types::heed::{self, Database, Env, PutFlags, RoTxn, RwTxn};
 use meilisearch_types::milli::documents::DocumentsBatchBuilder;
 use meilisearch_types::milli::update::IndexerConfig;
+use meilisearch_types::milli::vector::{Embedder, EmbedderOptions, EmbeddingConfigs};
 use meilisearch_types::milli::{self, CboRoaringBitmapCodec, Index, RoaringBitmapCodec, BEU32};
+use meilisearch_types::task_view::TaskView;
 use meilisearch_types::tasks::{Kind, KindWithContent, Status, Task};
 use puffin::FrameView;
 use roaring::RoaringBitmap;
@@ -64,8 +70,7 @@ use uuid::Uuid;
 use crate::index_mapper::IndexMapper;
 use crate::utils::{check_index_swap_validity, clamp_to_page_size};
 
-pub(crate) type BEI128 =
-    meilisearch_types::heed::zerocopy::I128<meilisearch_types::heed::byteorder::BE>;
+pub(crate) type BEI128 = I128<BE>;
 
 /// Defines a subset of tasks to be retrieved from the [`IndexScheduler`].
 ///
@@ -169,8 +174,8 @@ impl ProcessingTasks {
     }
 
     /// Set the processing tasks to an empty list
-    fn stop_processing(&mut self) {
-        self.processing = RoaringBitmap::new();
+    fn stop_processing(&mut self) -> RoaringBitmap {
+        std::mem::take(&mut self.processing)
     }
 
     /// Returns `true` if there, at least, is one task that is currently processing that we must stop.
@@ -240,6 +245,10 @@ pub struct IndexSchedulerOptions {
     pub snapshots_path: PathBuf,
     /// The path to the folder containing the dumps.
     pub dumps_path: PathBuf,
+    /// The URL on which we must send the tasks statuses
+    pub webhook_url: Option<String>,
+    /// The value we will send into the Authorization HTTP header on the webhook URL
+    pub webhook_authorization_header: Option<String>,
     /// The maximum size, in bytes, of the task index.
     pub task_db_size: usize,
     /// The size, in bytes, with which a meilisearch index is opened the first time of each meilisearch index.
@@ -258,6 +267,9 @@ pub struct IndexSchedulerOptions {
     /// The maximum number of tasks stored in the task queue before starting
     /// to auto schedule task deletions.
     pub max_number_of_tasks: usize,
+    /// If the autobatcher is allowed to automatically batch tasks
+    /// it will only batch this defined number of tasks at once.
+    pub max_number_of_batched_tasks: usize,
     /// The experimental features enabled for this instance.
     pub instance_features: InstanceTogglableFeatures,
 }
@@ -278,7 +290,7 @@ pub struct IndexScheduler {
     pub(crate) file_store: FileStore,
 
     // The main database, it contains all the tasks accessible by their Id.
-    pub(crate) all_tasks: Database<OwnedType<BEU32>, SerdeJson<Task>>,
+    pub(crate) all_tasks: Database<BEU32, SerdeJson<Task>>,
 
     /// All the tasks ids grouped by their status.
     // TODO we should not be able to serialize a `Status::Processing` in this database.
@@ -289,16 +301,16 @@ pub struct IndexScheduler {
     pub(crate) index_tasks: Database<Str, RoaringBitmapCodec>,
 
     /// Store the tasks that were canceled by a task uid
-    pub(crate) canceled_by: Database<OwnedType<BEU32>, RoaringBitmapCodec>,
+    pub(crate) canceled_by: Database<BEU32, RoaringBitmapCodec>,
 
     /// Store the task ids of tasks which were enqueued at a specific date
-    pub(crate) enqueued_at: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    pub(crate) enqueued_at: Database<BEI128, CboRoaringBitmapCodec>,
 
     /// Store the task ids of finished tasks which started being processed at a specific date
-    pub(crate) started_at: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    pub(crate) started_at: Database<BEI128, CboRoaringBitmapCodec>,
 
     /// Store the task ids of tasks which finished at a specific date
-    pub(crate) finished_at: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    pub(crate) finished_at: Database<BEI128, CboRoaringBitmapCodec>,
 
     /// In charge of creating, opening, storing and returning indexes.
     pub(crate) index_mapper: IndexMapper,
@@ -316,6 +328,14 @@ pub struct IndexScheduler {
     /// the finished tasks automatically.
     pub(crate) max_number_of_tasks: usize,
 
+    /// The maximum number of tasks that will be batched together.
+    pub(crate) max_number_of_batched_tasks: usize,
+
+    /// The webhook url we should send tasks to after processing every batches.
+    pub(crate) webhook_url: Option<String>,
+    /// The Authorization header to send to the webhook URL.
+    pub(crate) webhook_authorization_header: Option<String>,
+
     /// A frame to output the indexation profiling files to disk.
     pub(crate) puffin_frame: Arc<puffin::GlobalFrameView>,
 
@@ -331,6 +351,8 @@ pub struct IndexScheduler {
     /// The path to the version file of Meilisearch.
     pub(crate) version_file_path: PathBuf,
 
+    embedders: Arc<RwLock<HashMap<EmbedderOptions, Arc<Embedder>>>>,
+
     // ================= test
     // The next entry is dedicated to the tests.
     /// Provide a way to set a breakpoint in multiple part of the scheduler.
@@ -369,11 +391,15 @@ impl IndexScheduler {
             wake_up: self.wake_up.clone(),
             autobatching_enabled: self.autobatching_enabled,
             max_number_of_tasks: self.max_number_of_tasks,
+            max_number_of_batched_tasks: self.max_number_of_batched_tasks,
             puffin_frame: self.puffin_frame.clone(),
             snapshots_path: self.snapshots_path.clone(),
             dumps_path: self.dumps_path.clone(),
             auth_path: self.auth_path.clone(),
             version_file_path: self.version_file_path.clone(),
+            webhook_url: self.webhook_url.clone(),
+            webhook_authorization_header: self.webhook_authorization_header.clone(),
+            embedders: self.embedders.clone(),
             #[cfg(test)]
             test_breakpoint_sdr: self.test_breakpoint_sdr.clone(),
             #[cfg(test)]
@@ -466,10 +492,14 @@ impl IndexScheduler {
             puffin_frame: Arc::new(puffin::GlobalFrameView::default()),
             autobatching_enabled: options.autobatching_enabled,
             max_number_of_tasks: options.max_number_of_tasks,
+            max_number_of_batched_tasks: options.max_number_of_batched_tasks,
             dumps_path: options.dumps_path,
             snapshots_path: options.snapshots_path,
             auth_path: options.auth_path,
             version_file_path: options.version_file_path,
+            webhook_url: options.webhook_url,
+            webhook_authorization_header: options.webhook_authorization_header,
+            embedders: Default::default(),
 
             #[cfg(test)]
             test_breakpoint_sdr,
@@ -717,9 +747,7 @@ impl IndexScheduler {
         if let Some(canceled_by) = &query.canceled_by {
             let mut all_canceled_tasks = RoaringBitmap::new();
             for cancel_task_uid in canceled_by {
-                if let Some(canceled_by_uid) =
-                    self.canceled_by.get(rtxn, &BEU32::new(*cancel_task_uid))?
-                {
+                if let Some(canceled_by_uid) = self.canceled_by.get(rtxn, cancel_task_uid)? {
                     all_canceled_tasks |= canceled_by_uid;
                 }
             }
@@ -970,7 +998,7 @@ impl IndexScheduler {
 
         // if the task doesn't delete anything and 50% of the task queue is full, we must refuse to enqueue the incomming task
         if !matches!(&kind, KindWithContent::TaskDeletion { tasks, .. } if !tasks.is_empty())
-            && (self.env.non_free_pages_size()? * 100) / self.env.map_size()? as u64 > 50
+            && (self.env.non_free_pages_size()? * 100) / self.env.info().map_size as u64 > 50
         {
             return Err(Error::NoSpaceLeftInTaskQueue);
         }
@@ -996,7 +1024,7 @@ impl IndexScheduler {
         // Get rid of the mutability.
         let task = task;
 
-        self.all_tasks.append(&mut wtxn, &BEU32::new(task.uid), &task)?;
+        self.all_tasks.put_with_flags(&mut wtxn, PutFlags::APPEND, &task.uid, &task)?;
 
         for index in task.indexes() {
             self.update_index(&mut wtxn, index, |bitmap| {
@@ -1133,6 +1161,9 @@ impl IndexScheduler {
             handle.join().unwrap_or(Err(Error::ProcessBatchPanicked))
         };
 
+        // Reset the currently updating index to relinquish the index handle
+        self.index_mapper.set_currently_updating_index(None);
+
         #[cfg(test)]
         self.maybe_fail(tests::FailureLocation::AcquiringWtxn)?;
 
@@ -1167,10 +1198,11 @@ impl IndexScheduler {
             // If we have an abortion error we must stop the tick here and re-schedule tasks.
             Err(Error::Milli(milli::Error::InternalError(
                 milli::InternalError::AbortedIndexation,
-            ))) => {
+            )))
+            | Err(Error::AbortedTask) => {
                 #[cfg(test)]
                 self.breakpoint(Breakpoint::AbortedIndexation);
-                wtxn.abort().map_err(Error::HeedTransaction)?;
+                wtxn.abort();
 
                 // We make sure that we don't call `stop_processing` on the `processing_tasks`,
                 // this is because we want to let the next tick call `create_next_batch` and keep
@@ -1191,7 +1223,7 @@ impl IndexScheduler {
                 let index_uid = index_uid.unwrap();
                 // fixme: handle error more gracefully? not sure when this could happen
                 self.index_mapper.resize_index(&wtxn, &index_uid)?;
-                wtxn.abort().map_err(Error::HeedTransaction)?;
+                wtxn.abort();
 
                 return Ok(TickOutcome::TickAgain(0));
             }
@@ -1223,19 +1255,99 @@ impl IndexScheduler {
             }
         }
 
-        self.processing_tasks.write().unwrap().stop_processing();
+        let processed = self.processing_tasks.write().unwrap().stop_processing();
 
         #[cfg(test)]
         self.maybe_fail(tests::FailureLocation::CommittingWtxn)?;
 
         wtxn.commit().map_err(Error::HeedTransaction)?;
 
+        // We shouldn't crash the tick function if we can't send data to the webhook.
+        let _ = self.notify_webhook(&processed);
+
         #[cfg(test)]
         self.breakpoint(Breakpoint::AfterProcessing);
 
         Ok(TickOutcome::TickAgain(processed_tasks))
     }
 
+    /// Once the tasks changes have been commited we must send all the tasks that were updated to our webhook if there is one.
+    fn notify_webhook(&self, updated: &RoaringBitmap) -> Result<()> {
+        if let Some(ref url) = self.webhook_url {
+            struct TaskReader<'a, 'b> {
+                rtxn: &'a RoTxn<'a>,
+                index_scheduler: &'a IndexScheduler,
+                tasks: &'b mut roaring::bitmap::Iter<'b>,
+                buffer: Vec<u8>,
+                written: usize,
+            }
+
+            impl<'a, 'b> Read for TaskReader<'a, 'b> {
+                fn read(&mut self, mut buf: &mut [u8]) -> std::io::Result<usize> {
+                    if self.buffer.is_empty() {
+                        match self.tasks.next() {
+                            None => return Ok(0),
+                            Some(task_id) => {
+                                let task = self
+                                    .index_scheduler
+                                    .get_task(self.rtxn, task_id)
+                                    .map_err(|err| io::Error::new(io::ErrorKind::Other, err))?
+                                    .ok_or_else(|| {
+                                        io::Error::new(
+                                            io::ErrorKind::Other,
+                                            Error::CorruptedTaskQueue,
+                                        )
+                                    })?;
+
+                                serde_json::to_writer(
+                                    &mut self.buffer,
+                                    &TaskView::from_task(&task),
+                                )?;
+                                self.buffer.push(b'\n');
+                            }
+                        }
+                    }
+
+                    let mut to_write = &self.buffer[self.written..];
+                    let wrote = io::copy(&mut to_write, &mut buf)?;
+                    self.written += wrote as usize;
+
+                    // we wrote everything and must refresh our buffer on the next call
+                    if self.written == self.buffer.len() {
+                        self.written = 0;
+                        self.buffer.clear();
+                    }
+
+                    Ok(wrote as usize)
+                }
+            }
+
+            let rtxn = self.env.read_txn()?;
+
+            let task_reader = TaskReader {
+                rtxn: &rtxn,
+                index_scheduler: self,
+                tasks: &mut updated.into_iter(),
+                buffer: Vec::with_capacity(50), // on average a task is around ~100 bytes
+                written: 0,
+            };
+
+            // let reader = GzEncoder::new(BufReader::new(task_reader), Compression::default());
+            let reader = GzEncoder::new(BufReader::new(task_reader), Compression::default());
+            let request = ureq::post(url).set("Content-Encoding", "gzip");
+            let request = match &self.webhook_authorization_header {
+                Some(header) => request.set("Authorization", header),
+                None => request,
+            };
+
+            if let Err(e) = request.send(reader) {
+                log::error!("While sending data to the webhook: {e}");
+            }
+        }
+
+        Ok(())
+    }
+
     /// Register a task to cleanup the task queue if needed
     fn cleanup_task_queue(&self) -> Result<()> {
         let rtxn = self.env.read_txn().map_err(Error::HeedTransaction)?;
@@ -1310,6 +1422,40 @@ impl IndexScheduler {
         }
     }
 
+    // TODO: consider using a type alias or a struct embedder/template
+    pub fn embedders(
+        &self,
+        embedding_configs: Vec<(String, milli::vector::EmbeddingConfig)>,
+    ) -> Result<EmbeddingConfigs> {
+        let res: Result<_> = embedding_configs
+            .into_iter()
+            .map(|(name, milli::vector::EmbeddingConfig { embedder_options, prompt })| {
+                let prompt =
+                    Arc::new(prompt.try_into().map_err(meilisearch_types::milli::Error::from)?);
+                // optimistically return existing embedder
+                {
+                    let embedders = self.embedders.read().unwrap();
+                    if let Some(embedder) = embedders.get(&embedder_options) {
+                        return Ok((name, (embedder.clone(), prompt)));
+                    }
+                }
+
+                // add missing embedder
+                let embedder = Arc::new(
+                    Embedder::new(embedder_options.clone())
+                        .map_err(meilisearch_types::milli::vector::Error::from)
+                        .map_err(meilisearch_types::milli::Error::from)?,
+                );
+                {
+                    let mut embedders = self.embedders.write().unwrap();
+                    embedders.insert(embedder_options, embedder.clone());
+                }
+                Ok((name, (embedder, prompt)))
+            })
+            .collect();
+        res.map(EmbeddingConfigs::new)
+    }
+
     /// Blocks the thread until the test handle asks to progress to/through this breakpoint.
     ///
     /// Two messages are sent through the channel for each breakpoint.
@@ -1337,7 +1483,7 @@ impl IndexScheduler {
 
 pub struct Dump<'a> {
     index_scheduler: &'a IndexScheduler,
-    wtxn: RwTxn<'a, 'a>,
+    wtxn: RwTxn<'a>,
 
     indexes: HashMap<String, RoaringBitmap>,
     statuses: HashMap<Status, RoaringBitmap>,
@@ -1452,7 +1598,7 @@ impl<'a> Dump<'a> {
             },
         };
 
-        self.index_scheduler.all_tasks.put(&mut self.wtxn, &BEU32::new(task.uid), &task)?;
+        self.index_scheduler.all_tasks.put(&mut self.wtxn, &task.uid, &task)?;
 
         for index in task.indexes() {
             match self.indexes.get_mut(index) {
@@ -1494,8 +1640,8 @@ impl<'a> Dump<'a> {
             }
         }
 
-        self.statuses.entry(task.status).or_insert(RoaringBitmap::new()).insert(task.uid);
-        self.kinds.entry(task.kind.as_kind()).or_insert(RoaringBitmap::new()).insert(task.uid);
+        self.statuses.entry(task.status).or_default().insert(task.uid);
+        self.kinds.entry(task.kind.as_kind()).or_default().insert(task.uid);
 
         Ok(task)
     }
@@ -1615,6 +1761,8 @@ mod tests {
                 indexes_path: tempdir.path().join("indexes"),
                 snapshots_path: tempdir.path().join("snapshots"),
                 dumps_path: tempdir.path().join("dumps"),
+                webhook_url: None,
+                webhook_authorization_header: None,
                 task_db_size: 1000 * 1000, // 1 MB, we don't use MiB on purpose.
                 index_base_map_size: 1000 * 1000, // 1 MB, we don't use MiB on purpose.
                 enable_mdb_writemap: false,
@@ -1623,6 +1771,7 @@ mod tests {
                 indexer_config,
                 autobatching_enabled: true,
                 max_number_of_tasks: 1_000_000,
+                max_number_of_batched_tasks: usize::MAX,
                 instance_features: Default::default(),
             };
             configuration(&mut options);
@@ -4323,4 +4472,26 @@ mod tests {
         }
         "###);
     }
+
+    #[test]
+    fn cancel_processing_dump() {
+        let (index_scheduler, mut handle) = IndexScheduler::test(true, vec![]);
+
+        let dump_creation = KindWithContent::DumpCreation { keys: Vec::new(), instance_uid: None };
+        let dump_cancellation = KindWithContent::TaskCancelation {
+            query: "cancel dump".to_owned(),
+            tasks: RoaringBitmap::from_iter([0]),
+        };
+        let _ = index_scheduler.register(dump_creation).unwrap();
+        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "after_dump_register");
+        handle.advance_till([Start, BatchCreated, InsideProcessBatch]);
+
+        let _ = index_scheduler.register(dump_cancellation).unwrap();
+        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "cancel_registered");
+
+        snapshot!(format!("{:?}", handle.advance()), @"AbortedIndexation");
+
+        handle.advance_one_successful_batch();
+        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "cancel_processed");
+    }
 }

index-scheduler/src/snapshots/lib.rs/cancel_processing_dump/after_dump_register.snap

@@ -0,0 +1,35 @@
+---
+source: index-scheduler/src/lib.rs
+---
+### Autobatching Enabled = true
+### Processing Tasks:
+[]
+----------------------------------------------------------------------
+### All Tasks:
+0 {uid: 0, status: enqueued, details: { dump_uid: None }, kind: DumpCreation { keys: [], instance_uid: None }}
+----------------------------------------------------------------------
+### Status:
+enqueued [0,]
+----------------------------------------------------------------------
+### Kind:
+"dumpCreation" [0,]
+----------------------------------------------------------------------
+### Index Tasks:
+----------------------------------------------------------------------
+### Index Mapper:
+
+----------------------------------------------------------------------
+### Canceled By:
+
+----------------------------------------------------------------------
+### Enqueued At:
+[timestamp] [0,]
+----------------------------------------------------------------------
+### Started At:
+----------------------------------------------------------------------
+### Finished At:
+----------------------------------------------------------------------
+### File Store:
+
+----------------------------------------------------------------------
+

index-scheduler/src/snapshots/lib.rs/cancel_processing_dump/cancel_processed.snap

@@ -0,0 +1,45 @@
+---
+source: index-scheduler/src/lib.rs
+---
+### Autobatching Enabled = true
+### Processing Tasks:
+[]
+----------------------------------------------------------------------
+### All Tasks:
+0 {uid: 0, status: canceled, canceled_by: 1, details: { dump_uid: None }, kind: DumpCreation { keys: [], instance_uid: None }}
+1 {uid: 1, status: succeeded, details: { matched_tasks: 1, canceled_tasks: Some(0), original_filter: "cancel dump" }, kind: TaskCancelation { query: "cancel dump", tasks: RoaringBitmap<[0]> }}
+----------------------------------------------------------------------
+### Status:
+enqueued []
+succeeded [1,]
+canceled [0,]
+----------------------------------------------------------------------
+### Kind:
+"taskCancelation" [1,]
+"dumpCreation" [0,]
+----------------------------------------------------------------------
+### Index Tasks:
+----------------------------------------------------------------------
+### Index Mapper:
+
+----------------------------------------------------------------------
+### Canceled By:
+1 [0,]
+
+----------------------------------------------------------------------
+### Enqueued At:
+[timestamp] [0,]
+[timestamp] [1,]
+----------------------------------------------------------------------
+### Started At:
+[timestamp] [0,]
+[timestamp] [1,]
+----------------------------------------------------------------------
+### Finished At:
+[timestamp] [0,]
+[timestamp] [1,]
+----------------------------------------------------------------------
+### File Store:
+
+----------------------------------------------------------------------
+

index-scheduler/src/snapshots/lib.rs/cancel_processing_dump/cancel_registered.snap

@@ -0,0 +1,38 @@
+---
+source: index-scheduler/src/lib.rs
+---
+### Autobatching Enabled = true
+### Processing Tasks:
+[0,]
+----------------------------------------------------------------------
+### All Tasks:
+0 {uid: 0, status: enqueued, details: { dump_uid: None }, kind: DumpCreation { keys: [], instance_uid: None }}
+1 {uid: 1, status: enqueued, details: { matched_tasks: 1, canceled_tasks: None, original_filter: "cancel dump" }, kind: TaskCancelation { query: "cancel dump", tasks: RoaringBitmap<[0]> }}
+----------------------------------------------------------------------
+### Status:
+enqueued [0,1,]
+----------------------------------------------------------------------
+### Kind:
+"taskCancelation" [1,]
+"dumpCreation" [0,]
+----------------------------------------------------------------------
+### Index Tasks:
+----------------------------------------------------------------------
+### Index Mapper:
+
+----------------------------------------------------------------------
+### Canceled By:
+
+----------------------------------------------------------------------
+### Enqueued At:
+[timestamp] [0,]
+[timestamp] [1,]
+----------------------------------------------------------------------
+### Started At:
+----------------------------------------------------------------------
+### Finished At:
+----------------------------------------------------------------------
+### File Store:
+
+----------------------------------------------------------------------
+

index-scheduler/src/utils.rs

@@ -3,9 +3,9 @@
 use std::collections::{BTreeSet, HashSet};
 use std::ops::Bound;
 
-use meilisearch_types::heed::types::{DecodeIgnore, OwnedType};
+use meilisearch_types::heed::types::DecodeIgnore;
 use meilisearch_types::heed::{Database, RoTxn, RwTxn};
-use meilisearch_types::milli::{CboRoaringBitmapCodec, BEU32};
+use meilisearch_types::milli::CboRoaringBitmapCodec;
 use meilisearch_types::tasks::{Details, IndexSwap, Kind, KindWithContent, Status};
 use roaring::{MultiOps, RoaringBitmap};
 use time::OffsetDateTime;
@@ -18,7 +18,7 @@ impl IndexScheduler {
     }
 
     pub(crate) fn last_task_id(&self, rtxn: &RoTxn) -> Result<Option<TaskId>> {
-        Ok(self.all_tasks.remap_data_type::<DecodeIgnore>().last(rtxn)?.map(|(k, _)| k.get() + 1))
+        Ok(self.all_tasks.remap_data_type::<DecodeIgnore>().last(rtxn)?.map(|(k, _)| k + 1))
     }
 
     pub(crate) fn next_task_id(&self, rtxn: &RoTxn) -> Result<TaskId> {
@@ -26,7 +26,7 @@ impl IndexScheduler {
     }
 
     pub(crate) fn get_task(&self, rtxn: &RoTxn, task_id: TaskId) -> Result<Option<Task>> {
-        Ok(self.all_tasks.get(rtxn, &BEU32::new(task_id))?)
+        Ok(self.all_tasks.get(rtxn, &task_id)?)
     }
 
     /// Convert an iterator to a `Vec` of tasks. The tasks MUST exist or a
@@ -88,7 +88,7 @@ impl IndexScheduler {
             }
         }
 
-        self.all_tasks.put(wtxn, &BEU32::new(task.uid), task)?;
+        self.all_tasks.put(wtxn, &task.uid, task)?;
         Ok(())
     }
 
@@ -169,11 +169,11 @@ impl IndexScheduler {
 
 pub(crate) fn insert_task_datetime(
     wtxn: &mut RwTxn,
-    database: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    database: Database<BEI128, CboRoaringBitmapCodec>,
     time: OffsetDateTime,
     task_id: TaskId,
 ) -> Result<()> {
-    let timestamp = BEI128::new(time.unix_timestamp_nanos());
+    let timestamp = time.unix_timestamp_nanos();
     let mut task_ids = database.get(wtxn, &timestamp)?.unwrap_or_default();
     task_ids.insert(task_id);
     database.put(wtxn, &timestamp, &RoaringBitmap::from_iter(task_ids))?;
@@ -182,11 +182,11 @@ pub(crate) fn insert_task_datetime(
 
 pub(crate) fn remove_task_datetime(
     wtxn: &mut RwTxn,
-    database: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    database: Database<BEI128, CboRoaringBitmapCodec>,
     time: OffsetDateTime,
     task_id: TaskId,
 ) -> Result<()> {
-    let timestamp = BEI128::new(time.unix_timestamp_nanos());
+    let timestamp = time.unix_timestamp_nanos();
     if let Some(mut existing) = database.get(wtxn, &timestamp)? {
         existing.remove(task_id);
         if existing.is_empty() {
@@ -202,7 +202,7 @@ pub(crate) fn remove_task_datetime(
 pub(crate) fn keep_tasks_within_datetimes(
     rtxn: &RoTxn,
     tasks: &mut RoaringBitmap,
-    database: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    database: Database<BEI128, CboRoaringBitmapCodec>,
     after: Option<OffsetDateTime>,
     before: Option<OffsetDateTime>,
 ) -> Result<()> {
@@ -213,8 +213,8 @@ pub(crate) fn keep_tasks_within_datetimes(
         (Some(after), Some(before)) => (Bound::Excluded(*after), Bound::Excluded(*before)),
     };
     let mut collected_task_ids = RoaringBitmap::new();
-    let start = map_bound(start, |b| BEI128::new(b.unix_timestamp_nanos()));
-    let end = map_bound(end, |b| BEI128::new(b.unix_timestamp_nanos()));
+    let start = map_bound(start, |b| b.unix_timestamp_nanos());
+    let end = map_bound(end, |b| b.unix_timestamp_nanos());
     let iter = database.range(rtxn, &(start, end))?;
     for r in iter {
         let (_timestamp, task_ids) = r?;
@@ -337,8 +337,6 @@ impl IndexScheduler {
         let rtxn = self.env.read_txn().unwrap();
         for task in self.all_tasks.iter(&rtxn).unwrap() {
             let (task_id, task) = task.unwrap();
-            let task_id = task_id.get();
-
             let task_index_uid = task.index_uid().map(ToOwned::to_owned);
 
             let Task {
@@ -361,16 +359,13 @@ impl IndexScheduler {
                     .unwrap()
                     .contains(task.uid));
             }
-            let db_enqueued_at = self
-                .enqueued_at
-                .get(&rtxn, &BEI128::new(enqueued_at.unix_timestamp_nanos()))
-                .unwrap()
-                .unwrap();
+            let db_enqueued_at =
+                self.enqueued_at.get(&rtxn, &enqueued_at.unix_timestamp_nanos()).unwrap().unwrap();
             assert!(db_enqueued_at.contains(task_id));
             if let Some(started_at) = started_at {
                 let db_started_at = self
                     .started_at
-                    .get(&rtxn, &BEI128::new(started_at.unix_timestamp_nanos()))
+                    .get(&rtxn, &started_at.unix_timestamp_nanos())
                     .unwrap()
                     .unwrap();
                 assert!(db_started_at.contains(task_id));
@@ -378,7 +373,7 @@ impl IndexScheduler {
             if let Some(finished_at) = finished_at {
                 let db_finished_at = self
                     .finished_at
-                    .get(&rtxn, &BEI128::new(finished_at.unix_timestamp_nanos()))
+                    .get(&rtxn, &finished_at.unix_timestamp_nanos())
                     .unwrap()
                     .unwrap();
                 assert!(db_finished_at.contains(task_id));

index-scheduler/src/uuid_codec.rs

@@ -1,7 +1,7 @@
 use std::borrow::Cow;
 use std::convert::TryInto;
 
-use meilisearch_types::heed::{BytesDecode, BytesEncode};
+use meilisearch_types::heed::{BoxedError, BytesDecode, BytesEncode};
 use uuid::Uuid;
 
 /// A heed codec for value of struct Uuid.
@@ -10,15 +10,15 @@ pub struct UuidCodec;
 impl<'a> BytesDecode<'a> for UuidCodec {
     type DItem = Uuid;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        bytes.try_into().ok().map(Uuid::from_bytes)
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        bytes.try_into().map(Uuid::from_bytes).map_err(Into::into)
     }
 }
 
 impl BytesEncode<'_> for UuidCodec {
     type EItem = Uuid;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
-        Some(Cow::Borrowed(item.as_bytes()))
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item.as_bytes()))
     }
 }

meilisearch-auth/src/store.rs

@@ -4,17 +4,20 @@ use std::collections::HashSet;
 use std::convert::{TryFrom, TryInto};
 use std::fs::create_dir_all;
 use std::path::Path;
+use std::result::Result as StdResult;
 use std::str;
 use std::str::FromStr;
 use std::sync::Arc;
 
 use hmac::{Hmac, Mac};
+use meilisearch_types::heed::BoxedError;
 use meilisearch_types::index_uid_pattern::IndexUidPattern;
 use meilisearch_types::keys::KeyId;
 use meilisearch_types::milli;
-use meilisearch_types::milli::heed::types::{ByteSlice, DecodeIgnore, SerdeJson};
+use meilisearch_types::milli::heed::types::{Bytes, DecodeIgnore, SerdeJson};
 use meilisearch_types::milli::heed::{Database, Env, EnvOpenOptions, RwTxn};
 use sha2::Sha256;
+use thiserror::Error;
 use time::OffsetDateTime;
 use uuid::fmt::Hyphenated;
 use uuid::Uuid;
@@ -30,7 +33,7 @@ const KEY_ID_ACTION_INDEX_EXPIRATION_DB_NAME: &str = "keyid-action-index-expirat
 #[derive(Clone)]
 pub struct HeedAuthStore {
     env: Arc<Env>,
-    keys: Database<ByteSlice, SerdeJson<Key>>,
+    keys: Database<Bytes, SerdeJson<Key>>,
     action_keyid_index_expiration: Database<KeyIdActionCodec, SerdeJson<Option<OffsetDateTime>>>,
     should_close_on_drop: bool,
 }
@@ -276,7 +279,7 @@ impl HeedAuthStore {
     fn delete_key_from_inverted_db(&self, wtxn: &mut RwTxn, key: &KeyId) -> Result<()> {
         let mut iter = self
             .action_keyid_index_expiration
-            .remap_types::<ByteSlice, DecodeIgnore>()
+            .remap_types::<Bytes, DecodeIgnore>()
             .prefix_iter_mut(wtxn, key.as_bytes())?;
         while iter.next().transpose()?.is_some() {
             // safety: we don't keep references from inside the LMDB database.
@@ -294,23 +297,24 @@ pub struct KeyIdActionCodec;
 impl<'a> milli::heed::BytesDecode<'a> for KeyIdActionCodec {
     type DItem = (KeyId, Action, Option<&'a [u8]>);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (key_id_bytes, action_bytes) = try_split_array_at(bytes)?;
-        let (action_bytes, index) = match try_split_array_at(action_bytes)? {
-            (action, []) => (action, None),
-            (action, index) => (action, Some(index)),
-        };
+    fn bytes_decode(bytes: &'a [u8]) -> StdResult<Self::DItem, BoxedError> {
+        let (key_id_bytes, action_bytes) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
+        let (&action_byte, index) =
+            match try_split_array_at(action_bytes).ok_or(SliceTooShortError)? {
+                ([action], []) => (action, None),
+                ([action], index) => (action, Some(index)),
+            };
         let key_id = Uuid::from_bytes(*key_id_bytes);
-        let action = Action::from_repr(u8::from_be_bytes(*action_bytes))?;
+        let action = Action::from_repr(action_byte).ok_or(InvalidActionError { action_byte })?;
 
-        Some((key_id, action, index))
+        Ok((key_id, action, index))
     }
 }
 
 impl<'a> milli::heed::BytesEncode<'a> for KeyIdActionCodec {
     type EItem = (&'a KeyId, &'a Action, Option<&'a [u8]>);
 
-    fn bytes_encode((key_id, action, index): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((key_id, action, index): &Self::EItem) -> StdResult<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::new();
 
         bytes.extend_from_slice(key_id.as_bytes());
@@ -320,10 +324,20 @@ impl<'a> milli::heed::BytesEncode<'a> for KeyIdActionCodec {
             bytes.extend_from_slice(index);
         }
 
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }
 
+#[derive(Error, Debug)]
+#[error("the slice is too short")]
+pub struct SliceTooShortError;
+
+#[derive(Error, Debug)]
+#[error("cannot construct a valid Action from {action_byte}")]
+pub struct InvalidActionError {
+    pub action_byte: u8,
+}
+
 pub fn generate_key_as_hexa(uid: Uuid, master_key: &[u8]) -> String {
     // format uid as hyphenated allowing user to generate their own keys.
     let mut uid_buffer = [0; Hyphenated::LENGTH];

meilisearch-types/Cargo.toml

@@ -15,7 +15,7 @@ actix-web = { version = "4.3.1", default-features = false }
 anyhow = "1.0.70"
 convert_case = "0.6.0"
 csv = "1.2.1"
-deserr = { version = "0.6.0", features = ["actix-web"]}
+deserr = { version = "0.6.0", features = ["actix-web"] }
 either = { version = "1.8.1", features = ["serde"] }
 enum-iterator = "1.4.0"
 file-store = { path = "../file-store" }
@@ -50,6 +50,7 @@ hebrew = ["milli/hebrew"]
 japanese = ["milli/japanese"]
 # thai specialized tokenization
 thai = ["milli/thai"]
-
 # allow greek specialized tokenization
 greek = ["milli/greek"]
+# allow khmer specialized tokenization
+khmer = ["milli/khmer"]

meilisearch-types/src/deserr/mod.rs

@@ -188,3 +188,4 @@ merge_with_error_impl_take_error_message!(ParseOffsetDateTimeError);
 merge_with_error_impl_take_error_message!(ParseTaskKindError);
 merge_with_error_impl_take_error_message!(ParseTaskStatusError);
 merge_with_error_impl_take_error_message!(IndexUidFormatError);
+merge_with_error_impl_take_error_message!(InvalidSearchSemanticRatio);

meilisearch-types/src/error.rs

@@ -222,6 +222,8 @@ InvalidVectorsType                    , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentId                     , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentLimit                  , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentOffset                 , InvalidRequest       , BAD_REQUEST ;
+InvalidEmbedder                       , InvalidRequest       , BAD_REQUEST ;
+InvalidHybridQuery                    , InvalidRequest       , BAD_REQUEST ;
 InvalidIndexLimit                     , InvalidRequest       , BAD_REQUEST ;
 InvalidIndexOffset                    , InvalidRequest       , BAD_REQUEST ;
 InvalidIndexPrimaryKey                , InvalidRequest       , BAD_REQUEST ;
@@ -233,6 +235,7 @@ InvalidSearchAttributesToRetrieve     , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchCropLength               , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchCropMarker               , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchFacets                   , InvalidRequest       , BAD_REQUEST ;
+InvalidSearchSemanticRatio            , InvalidRequest       , BAD_REQUEST ;
 InvalidFacetSearchFacetName           , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchFilter                   , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchHighlightPostTag         , InvalidRequest       , BAD_REQUEST ;
@@ -252,9 +255,11 @@ InvalidSearchShowRankingScoreDetails  , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchSort                     , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsDisplayedAttributes    , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsDistinctAttribute      , InvalidRequest       , BAD_REQUEST ;
+InvalidSettingsProximityPrecision     , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsFaceting               , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsFilterableAttributes   , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsPagination             , InvalidRequest       , BAD_REQUEST ;
+InvalidSettingsEmbedders              , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsRankingRules           , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsSearchableAttributes   , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsSortableAttributes     , InvalidRequest       , BAD_REQUEST ;
@@ -294,15 +299,18 @@ MissingFacetSearchFacetName           , InvalidRequest       , BAD_REQUEST ;
 MissingIndexUid                       , InvalidRequest       , BAD_REQUEST ;
 MissingMasterKey                      , Auth                 , UNAUTHORIZED ;
 MissingPayload                        , InvalidRequest       , BAD_REQUEST ;
+MissingSearchHybrid                   , InvalidRequest       , BAD_REQUEST ;
 MissingSwapIndexes                    , InvalidRequest       , BAD_REQUEST ;
 MissingTaskFilters                    , InvalidRequest       , BAD_REQUEST ;
 NoSpaceLeftOnDevice                   , System               , UNPROCESSABLE_ENTITY;
 PayloadTooLarge                       , InvalidRequest       , PAYLOAD_TOO_LARGE ;
 TaskNotFound                          , InvalidRequest       , NOT_FOUND ;
 TooManyOpenFiles                      , System               , UNPROCESSABLE_ENTITY ;
+TooManyVectors                        , InvalidRequest       , BAD_REQUEST ;
 UnretrievableDocument                 , Internal             , BAD_REQUEST ;
 UnretrievableErrorCode                , InvalidRequest       , BAD_REQUEST ;
-UnsupportedMediaType                  , InvalidRequest       , UNSUPPORTED_MEDIA_TYPE
+UnsupportedMediaType                  , InvalidRequest       , UNSUPPORTED_MEDIA_TYPE ;
+VectorEmbeddingError                  , InvalidRequest       , BAD_REQUEST
 }
 
 impl ErrorCode for JoinError {
@@ -335,6 +343,13 @@ impl ErrorCode for milli::Error {
                     UserError::InvalidDocumentId { .. } | UserError::TooManyDocumentIds { .. } => {
                         Code::InvalidDocumentId
                     }
+                    UserError::MissingDocumentField(_) => Code::InvalidDocumentFields,
+                    UserError::InvalidFieldForSource { .. }
+                    | UserError::MissingFieldForSource { .. }
+                    | UserError::InvalidOpenAiModel { .. }
+                    | UserError::InvalidPrompt(_) => Code::InvalidSettingsEmbedders,
+                    UserError::TooManyEmbedders(_) => Code::InvalidSettingsEmbedders,
+                    UserError::InvalidPromptForEmbeddings(..) => Code::InvalidSettingsEmbedders,
                     UserError::NoPrimaryKeyCandidateFound => Code::IndexPrimaryKeyNoCandidateFound,
                     UserError::MultiplePrimaryKeyCandidatesFound { .. } => {
                         Code::IndexPrimaryKeyMultipleCandidatesFound
@@ -352,11 +367,15 @@ impl ErrorCode for milli::Error {
                     UserError::CriterionError(_) => Code::InvalidSettingsRankingRules,
                     UserError::InvalidGeoField { .. } => Code::InvalidDocumentGeoField,
                     UserError::InvalidVectorDimensions { .. } => Code::InvalidVectorDimensions,
+                    UserError::InvalidVectorsMapType { .. } => Code::InvalidVectorsType,
                     UserError::InvalidVectorsType { .. } => Code::InvalidVectorsType,
+                    UserError::TooManyVectors(_, _) => Code::TooManyVectors,
                     UserError::SortError(_) => Code::InvalidSearchSort,
                     UserError::InvalidMinTypoWordLenSetting(_, _) => {
                         Code::InvalidSettingsTypoTolerance
                     }
+                    UserError::InvalidEmbedder(_) => Code::InvalidEmbedder,
+                    UserError::VectorEmbeddingError(_) => Code::VectorEmbeddingError,
                 }
             }
         }
@@ -386,11 +405,11 @@ impl ErrorCode for HeedError {
             HeedError::Mdb(MdbError::Invalid) => Code::InvalidStoreFile,
             HeedError::Io(e) => e.error_code(),
             HeedError::Mdb(_)
-            | HeedError::Encoding
-            | HeedError::Decoding
+            | HeedError::Encoding(_)
+            | HeedError::Decoding(_)
             | HeedError::InvalidDatabaseTyping
             | HeedError::DatabaseClosing
-            | HeedError::BadOpenOptions => Code::Internal,
+            | HeedError::BadOpenOptions { .. } => Code::Internal,
         }
     }
 }
@@ -444,6 +463,15 @@ impl fmt::Display for DeserrParseIntError {
     }
 }
 
+impl fmt::Display for deserr_codes::InvalidSearchSemanticRatio {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(
+            f,
+            "the value of `semanticRatio` is invalid, expected a float between `0.0` and `1.0`."
+        )
+    }
+}
+
 #[macro_export]
 macro_rules! internal_error {
     ($target:ty : $($other:path), *) => {

meilisearch-types/src/lib.rs

@@ -9,6 +9,7 @@ pub mod index_uid_pattern;
 pub mod keys;
 pub mod settings;
 pub mod star_or;
+pub mod task_view;
 pub mod tasks;
 pub mod versioning;
 pub use milli::{heed, Index};

meilisearch-types/src/settings.rs

@@ -8,6 +8,7 @@ use std::str::FromStr;
 
 use deserr::{DeserializeError, Deserr, ErrorKind, MergeWithError, ValuePointerRef};
 use fst::IntoStreamer;
+use milli::proximity::ProximityPrecision;
 use milli::update::Setting;
 use milli::{Criterion, CriterionError, Index, DEFAULT_VALUES_PER_FACET};
 use serde::{Deserialize, Serialize, Serializer};
@@ -186,6 +187,9 @@ pub struct Settings<T> {
     #[deserr(default, error = DeserrJsonError<InvalidSettingsDistinctAttribute>)]
     pub distinct_attribute: Setting<String>,
     #[serde(default, skip_serializing_if = "Setting::is_not_set")]
+    #[deserr(default, error = DeserrJsonError<InvalidSettingsProximityPrecision>)]
+    pub proximity_precision: Setting<ProximityPrecisionView>,
+    #[serde(default, skip_serializing_if = "Setting::is_not_set")]
     #[deserr(default, error = DeserrJsonError<InvalidSettingsTypoTolerance>)]
     pub typo_tolerance: Setting<TypoSettings>,
     #[serde(default, skip_serializing_if = "Setting::is_not_set")]
@@ -195,6 +199,10 @@ pub struct Settings<T> {
     #[deserr(default, error = DeserrJsonError<InvalidSettingsPagination>)]
     pub pagination: Setting<PaginationSettings>,
 
+    #[serde(default, skip_serializing_if = "Setting::is_not_set")]
+    #[deserr(default, error = DeserrJsonError<InvalidSettingsEmbedders>)]
+    pub embedders: Setting<BTreeMap<String, Setting<milli::vector::settings::EmbeddingSettings>>>,
+
     #[serde(skip)]
     #[deserr(skip)]
     pub _kind: PhantomData<T>,
@@ -214,9 +222,11 @@ impl Settings<Checked> {
             separator_tokens: Setting::Reset,
             dictionary: Setting::Reset,
             distinct_attribute: Setting::Reset,
+            proximity_precision: Setting::Reset,
             typo_tolerance: Setting::Reset,
             faceting: Setting::Reset,
             pagination: Setting::Reset,
+            embedders: Setting::Reset,
             _kind: PhantomData,
         }
     }
@@ -234,9 +244,11 @@ impl Settings<Checked> {
             dictionary,
             synonyms,
             distinct_attribute,
+            proximity_precision,
             typo_tolerance,
             faceting,
             pagination,
+            embedders,
             ..
         } = self;
 
@@ -252,9 +264,11 @@ impl Settings<Checked> {
             dictionary,
             synonyms,
             distinct_attribute,
+            proximity_precision,
             typo_tolerance,
             faceting,
             pagination,
+            embedders,
             _kind: PhantomData,
         }
     }
@@ -296,12 +310,29 @@ impl Settings<Unchecked> {
             separator_tokens: self.separator_tokens,
             dictionary: self.dictionary,
             distinct_attribute: self.distinct_attribute,
+            proximity_precision: self.proximity_precision,
             typo_tolerance: self.typo_tolerance,
             faceting: self.faceting,
             pagination: self.pagination,
+            embedders: self.embedders,
             _kind: PhantomData,
         }
     }
+
+    pub fn validate(self) -> Result<Self, milli::Error> {
+        self.validate_embedding_settings()
+    }
+
+    fn validate_embedding_settings(mut self) -> Result<Self, milli::Error> {
+        let Setting::Set(mut configs) = self.embedders else { return Ok(self) };
+        for (name, config) in configs.iter_mut() {
+            let config_to_check = std::mem::take(config);
+            let checked_config = milli::update::validate_embedding_settings(config_to_check, name)?;
+            *config = checked_config
+        }
+        self.embedders = Setting::Set(configs);
+        Ok(self)
+    }
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -390,6 +421,12 @@ pub fn apply_settings_to_builder(
         Setting::NotSet => (),
     }
 
+    match settings.proximity_precision {
+        Setting::Set(ref precision) => builder.set_proximity_precision((*precision).into()),
+        Setting::Reset => builder.reset_proximity_precision(),
+        Setting::NotSet => (),
+    }
+
     match settings.typo_tolerance {
         Setting::Set(ref value) => {
             match value.enabled {
@@ -476,6 +513,12 @@ pub fn apply_settings_to_builder(
         Setting::Reset => builder.reset_pagination_max_total_hits(),
         Setting::NotSet => (),
     }
+
+    match settings.embedders.clone() {
+        Setting::Set(value) => builder.set_embedder_settings(value),
+        Setting::Reset => builder.reset_embedder_settings(),
+        Setting::NotSet => (),
+    }
 }
 
 pub fn settings(
@@ -509,6 +552,8 @@ pub fn settings(
 
     let distinct_field = index.distinct_field(rtxn)?.map(String::from);
 
+    let proximity_precision = index.proximity_precision(rtxn)?.map(ProximityPrecisionView::from);
+
     let synonyms = index.user_defined_synonyms(rtxn)?;
 
     let min_typo_word_len = MinWordSizeTyposSetting {
@@ -532,7 +577,10 @@ pub fn settings(
 
     let faceting = FacetingSettings {
         max_values_per_facet: Setting::Set(
-            index.max_values_per_facet(rtxn)?.unwrap_or(DEFAULT_VALUES_PER_FACET),
+            index
+                .max_values_per_facet(rtxn)?
+                .map(|x| x as usize)
+                .unwrap_or(DEFAULT_VALUES_PER_FACET),
         ),
         sort_facet_values_by: Setting::Set(
             index
@@ -545,10 +593,20 @@ pub fn settings(
 
     let pagination = PaginationSettings {
         max_total_hits: Setting::Set(
-            index.pagination_max_total_hits(rtxn)?.unwrap_or(DEFAULT_PAGINATION_MAX_TOTAL_HITS),
+            index
+                .pagination_max_total_hits(rtxn)?
+                .map(|x| x as usize)
+                .unwrap_or(DEFAULT_PAGINATION_MAX_TOTAL_HITS),
         ),
     };
 
+    let embedders: BTreeMap<_, _> = index
+        .embedding_configs(rtxn)?
+        .into_iter()
+        .map(|(name, config)| (name, Setting::Set(config.into())))
+        .collect();
+    let embedders = if embedders.is_empty() { Setting::NotSet } else { Setting::Set(embedders) };
+
     Ok(Settings {
         displayed_attributes: match displayed_attributes {
             Some(attrs) => Setting::Set(attrs),
@@ -569,10 +627,12 @@ pub fn settings(
             Some(field) => Setting::Set(field),
             None => Setting::Reset,
         },
+        proximity_precision: Setting::Set(proximity_precision.unwrap_or_default()),
         synonyms: Setting::Set(synonyms),
         typo_tolerance: Setting::Set(typo_tolerance),
         faceting: Setting::Set(faceting),
         pagination: Setting::Set(pagination),
+        embedders,
         _kind: PhantomData,
     })
 }
@@ -673,6 +733,32 @@ impl From<RankingRuleView> for Criterion {
     }
 }
 
+#[derive(Default, Debug, Clone, Copy, PartialEq, Eq, Deserr, Serialize, Deserialize)]
+#[serde(deny_unknown_fields, rename_all = "camelCase")]
+#[deserr(error = DeserrJsonError<InvalidSettingsProximityPrecision>, rename_all = camelCase, deny_unknown_fields)]
+pub enum ProximityPrecisionView {
+    #[default]
+    ByWord,
+    ByAttribute,
+}
+
+impl From<ProximityPrecision> for ProximityPrecisionView {
+    fn from(value: ProximityPrecision) -> Self {
+        match value {
+            ProximityPrecision::ByWord => ProximityPrecisionView::ByWord,
+            ProximityPrecision::ByAttribute => ProximityPrecisionView::ByAttribute,
+        }
+    }
+}
+impl From<ProximityPrecisionView> for ProximityPrecision {
+    fn from(value: ProximityPrecisionView) -> Self {
+        match value {
+            ProximityPrecisionView::ByWord => ProximityPrecision::ByWord,
+            ProximityPrecisionView::ByAttribute => ProximityPrecision::ByAttribute,
+        }
+    }
+}
+
 #[cfg(test)]
 pub(crate) mod test {
     use super::*;
@@ -692,9 +778,11 @@ pub(crate) mod test {
             dictionary: Setting::NotSet,
             synonyms: Setting::NotSet,
             distinct_attribute: Setting::NotSet,
+            proximity_precision: Setting::NotSet,
             typo_tolerance: Setting::NotSet,
             faceting: Setting::NotSet,
             pagination: Setting::NotSet,
+            embedders: Setting::NotSet,
             _kind: PhantomData::<Unchecked>,
         };
 
@@ -716,9 +804,11 @@ pub(crate) mod test {
             dictionary: Setting::NotSet,
             synonyms: Setting::NotSet,
             distinct_attribute: Setting::NotSet,
+            proximity_precision: Setting::NotSet,
             typo_tolerance: Setting::NotSet,
             faceting: Setting::NotSet,
             pagination: Setting::NotSet,
+            embedders: Setting::NotSet,
             _kind: PhantomData::<Unchecked>,
         };
 

meilisearch-types/src/task_view.rs

@@ -0,0 +1,139 @@
+use serde::Serialize;
+use time::{Duration, OffsetDateTime};
+
+use crate::error::ResponseError;
+use crate::settings::{Settings, Unchecked};
+use crate::tasks::{serialize_duration, Details, IndexSwap, Kind, Status, Task, TaskId};
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize)]
+#[serde(rename_all = "camelCase")]
+pub struct TaskView {
+    pub uid: TaskId,
+    #[serde(default)]
+    pub index_uid: Option<String>,
+    pub status: Status,
+    #[serde(rename = "type")]
+    pub kind: Kind,
+    pub canceled_by: Option<TaskId>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub details: Option<DetailsView>,
+    pub error: Option<ResponseError>,
+    #[serde(serialize_with = "serialize_duration", default)]
+    pub duration: Option<Duration>,
+    #[serde(with = "time::serde::rfc3339")]
+    pub enqueued_at: OffsetDateTime,
+    #[serde(with = "time::serde::rfc3339::option", default)]
+    pub started_at: Option<OffsetDateTime>,
+    #[serde(with = "time::serde::rfc3339::option", default)]
+    pub finished_at: Option<OffsetDateTime>,
+}
+
+impl TaskView {
+    pub fn from_task(task: &Task) -> TaskView {
+        TaskView {
+            uid: task.uid,
+            index_uid: task.index_uid().map(ToOwned::to_owned),
+            status: task.status,
+            kind: task.kind.as_kind(),
+            canceled_by: task.canceled_by,
+            details: task.details.clone().map(DetailsView::from),
+            error: task.error.clone(),
+            duration: task.started_at.zip(task.finished_at).map(|(start, end)| end - start),
+            enqueued_at: task.enqueued_at,
+            started_at: task.started_at,
+            finished_at: task.finished_at,
+        }
+    }
+}
+
+#[derive(Default, Debug, PartialEq, Eq, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+pub struct DetailsView {
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub received_documents: Option<u64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub indexed_documents: Option<Option<u64>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub primary_key: Option<Option<String>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub provided_ids: Option<usize>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub deleted_documents: Option<Option<u64>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub matched_tasks: Option<u64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub canceled_tasks: Option<Option<u64>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub deleted_tasks: Option<Option<u64>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub original_filter: Option<Option<String>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub dump_uid: Option<Option<String>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    #[serde(flatten)]
+    pub settings: Option<Box<Settings<Unchecked>>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub swaps: Option<Vec<IndexSwap>>,
+}
+
+impl From<Details> for DetailsView {
+    fn from(details: Details) -> Self {
+        match details {
+            Details::DocumentAdditionOrUpdate { received_documents, indexed_documents } => {
+                DetailsView {
+                    received_documents: Some(received_documents),
+                    indexed_documents: Some(indexed_documents),
+                    ..DetailsView::default()
+                }
+            }
+            Details::SettingsUpdate { settings } => {
+                DetailsView { settings: Some(settings), ..DetailsView::default() }
+            }
+            Details::IndexInfo { primary_key } => {
+                DetailsView { primary_key: Some(primary_key), ..DetailsView::default() }
+            }
+            Details::DocumentDeletion {
+                provided_ids: received_document_ids,
+                deleted_documents,
+            } => DetailsView {
+                provided_ids: Some(received_document_ids),
+                deleted_documents: Some(deleted_documents),
+                original_filter: Some(None),
+                ..DetailsView::default()
+            },
+            Details::DocumentDeletionByFilter { original_filter, deleted_documents } => {
+                DetailsView {
+                    provided_ids: Some(0),
+                    original_filter: Some(Some(original_filter)),
+                    deleted_documents: Some(deleted_documents),
+                    ..DetailsView::default()
+                }
+            }
+            Details::ClearAll { deleted_documents } => {
+                DetailsView { deleted_documents: Some(deleted_documents), ..DetailsView::default() }
+            }
+            Details::TaskCancelation { matched_tasks, canceled_tasks, original_filter } => {
+                DetailsView {
+                    matched_tasks: Some(matched_tasks),
+                    canceled_tasks: Some(canceled_tasks),
+                    original_filter: Some(Some(original_filter)),
+                    ..DetailsView::default()
+                }
+            }
+            Details::TaskDeletion { matched_tasks, deleted_tasks, original_filter } => {
+                DetailsView {
+                    matched_tasks: Some(matched_tasks),
+                    deleted_tasks: Some(deleted_tasks),
+                    original_filter: Some(Some(original_filter)),
+                    ..DetailsView::default()
+                }
+            }
+            Details::Dump { dump_uid } => {
+                DetailsView { dump_uid: Some(dump_uid), ..DetailsView::default() }
+            }
+            Details::IndexSwap { swaps } => {
+                DetailsView { swaps: Some(swaps), ..Default::default() }
+            }
+        }
+    }
+}

meilisearch/Cargo.toml

@@ -39,7 +39,7 @@ byte-unit = { version = "4.0.19", default-features = false, features = [
 bytes = "1.4.0"
 clap = { version = "4.2.1", features = ["derive", "env"] }
 crossbeam-channel = "0.5.8"
-deserr = { version = "0.6.0", features = ["actix-web"]}
+deserr = { version = "0.6.0", features = ["actix-web"] }
 dump = { path = "../dump" }
 either = "1.8.1"
 env_logger = "0.10.0"
@@ -104,6 +104,7 @@ walkdir = "2.3.3"
 yaup = "0.2.1"
 serde_urlencoded = "0.7.1"
 termcolor = "1.2.0"
+url = { version = "2.5.0", features = ["serde"] }
 
 [dev-dependencies]
 actix-rt = "2.8.0"
@@ -150,7 +151,8 @@ hebrew = ["meilisearch-types/hebrew"]
 japanese = ["meilisearch-types/japanese"]
 thai = ["meilisearch-types/thai"]
 greek = ["meilisearch-types/greek"]
+khmer = ["meilisearch-types/khmer"]
 
 [package.metadata.mini-dashboard]
-assets-url = "https://github.com/meilisearch/mini-dashboard/releases/download/v0.2.11/build.zip"
-sha1 = "83cd44ed1e5f97ecb581dc9f958a63f4ccc982d9"
+assets-url = "https://github.com/meilisearch/mini-dashboard/releases/download/v0.2.13/build.zip"
+sha1 = "e20cc9b390003c6c844f4b8bcc5c5013191a77ff"

meilisearch/src/analytics/segment_analytics.rs

@@ -36,7 +36,7 @@ use crate::routes::{create_all_stats, Stats};
 use crate::search::{
     FacetSearchResult, MatchingStrategy, SearchQuery, SearchQueryWithIndex, SearchResult,
     DEFAULT_CROP_LENGTH, DEFAULT_CROP_MARKER, DEFAULT_HIGHLIGHT_POST_TAG,
-    DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_SEARCH_LIMIT,
+    DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_SEARCH_LIMIT, DEFAULT_SEMANTIC_RATIO,
 };
 use crate::Opt;
 
@@ -251,6 +251,7 @@ struct Infos {
     env: String,
     experimental_enable_metrics: bool,
     experimental_reduce_indexing_memory_usage: bool,
+    experimental_max_number_of_batched_tasks: usize,
     db_path: bool,
     import_dump: bool,
     dump_dir: bool,
@@ -263,6 +264,8 @@ struct Infos {
     ignore_snapshot_if_db_exists: bool,
     http_addr: bool,
     http_payload_size_limit: Byte,
+    task_queue_webhook: bool,
+    task_webhook_authorization_header: bool,
     log_level: String,
     max_indexing_memory: MaxMemory,
     max_indexing_threads: MaxThreads,
@@ -285,9 +288,12 @@ impl From<Opt> for Infos {
             db_path,
             experimental_enable_metrics,
             experimental_reduce_indexing_memory_usage,
+            experimental_max_number_of_batched_tasks,
             http_addr,
             master_key: _,
             env,
+            task_webhook_url,
+            task_webhook_authorization_header,
             max_index_size: _,
             max_task_db_size: _,
             http_payload_size_limit,
@@ -340,6 +346,9 @@ impl From<Opt> for Infos {
             ignore_snapshot_if_db_exists,
             http_addr: http_addr != default_http_addr(),
             http_payload_size_limit,
+            experimental_max_number_of_batched_tasks,
+            task_queue_webhook: task_webhook_url.is_some(),
+            task_webhook_authorization_header: task_webhook_authorization_header.is_some(),
             log_level: log_level.to_string(),
             max_indexing_memory,
             max_indexing_threads,
@@ -583,6 +592,11 @@ pub struct SearchAggregator {
     // vector
     // The maximum number of floats in a vector request
     max_vector_size: usize,
+    // Whether the semantic ratio passed to a hybrid search equals the default ratio.
+    semantic_ratio: bool,
+    // Whether a non-default embedder was specified
+    embedder: bool,
+    hybrid: bool,
 
     // every time a search is done, we increment the counter linked to the used settings
     matching_strategy: HashMap<String, usize>,
@@ -636,6 +650,7 @@ impl SearchAggregator {
             crop_marker,
             matching_strategy,
             attributes_to_search_on,
+            hybrid,
         } = query;
 
         let mut ret = Self::default();
@@ -709,6 +724,12 @@ impl SearchAggregator {
         ret.show_ranking_score = *show_ranking_score;
         ret.show_ranking_score_details = *show_ranking_score_details;
 
+        if let Some(hybrid) = hybrid {
+            ret.semantic_ratio = hybrid.semantic_ratio != DEFAULT_SEMANTIC_RATIO();
+            ret.embedder = hybrid.embedder.is_some();
+            ret.hybrid = true;
+        }
+
         ret
     }
 
@@ -762,6 +783,9 @@ impl SearchAggregator {
             facets_total_number_of_facets,
             show_ranking_score,
             show_ranking_score_details,
+            semantic_ratio,
+            embedder,
+            hybrid,
         } = other;
 
         if self.timestamp.is_none() {
@@ -807,6 +831,9 @@ impl SearchAggregator {
 
         // vector
         self.max_vector_size = self.max_vector_size.max(max_vector_size);
+        self.semantic_ratio |= semantic_ratio;
+        self.hybrid |= hybrid;
+        self.embedder |= embedder;
 
         // pagination
         self.max_limit = self.max_limit.max(max_limit);
@@ -875,6 +902,9 @@ impl SearchAggregator {
             facets_total_number_of_facets,
             show_ranking_score,
             show_ranking_score_details,
+            semantic_ratio,
+            embedder,
+            hybrid,
         } = self;
 
         if total_received == 0 {
@@ -914,6 +944,11 @@ impl SearchAggregator {
                 "vector": {
                     "max_vector_size": max_vector_size,
                 },
+                "hybrid": {
+                    "enabled": hybrid,
+                    "semantic_ratio": semantic_ratio,
+                    "embedder": embedder,
+                },
                 "pagination": {
                    "max_limit": max_limit,
                    "max_offset": max_offset,
@@ -1009,6 +1044,7 @@ impl MultiSearchAggregator {
                     crop_marker: _,
                     matching_strategy: _,
                     attributes_to_search_on: _,
+                    hybrid: _,
                 } = query;
 
                 index_uid.as_str()
@@ -1155,6 +1191,7 @@ impl FacetSearchAggregator {
             filter,
             matching_strategy,
             attributes_to_search_on,
+            hybrid,
         } = query;
 
         let mut ret = Self::default();
@@ -1168,7 +1205,8 @@ impl FacetSearchAggregator {
             || vector.is_some()
             || filter.is_some()
             || *matching_strategy != MatchingStrategy::default()
-            || attributes_to_search_on.is_some();
+            || attributes_to_search_on.is_some()
+            || hybrid.is_some();
 
         ret
     }

meilisearch/src/error.rs

@@ -51,6 +51,8 @@ pub enum MeilisearchHttpError {
     DocumentFormat(#[from] DocumentFormatError),
     #[error(transparent)]
     Join(#[from] JoinError),
+    #[error("Invalid request: missing `hybrid` parameter when both `q` and `vector` are present.")]
+    MissingSearchHybrid,
 }
 
 impl ErrorCode for MeilisearchHttpError {
@@ -74,6 +76,7 @@ impl ErrorCode for MeilisearchHttpError {
             MeilisearchHttpError::FileStore(_) => Code::Internal,
             MeilisearchHttpError::DocumentFormat(e) => e.error_code(),
             MeilisearchHttpError::Join(_) => Code::Internal,
+            MeilisearchHttpError::MissingSearchHybrid => Code::MissingSearchHybrid,
         }
     }
 }

meilisearch/src/lib.rs

@@ -228,12 +228,15 @@ fn open_or_create_database_unchecked(
             indexes_path: opt.db_path.join("indexes"),
             snapshots_path: opt.snapshot_dir.clone(),
             dumps_path: opt.dump_dir.clone(),
+            webhook_url: opt.task_webhook_url.as_ref().map(|url| url.to_string()),
+            webhook_authorization_header: opt.task_webhook_authorization_header.clone(),
             task_db_size: opt.max_task_db_size.get_bytes() as usize,
             index_base_map_size: opt.max_index_size.get_bytes() as usize,
             enable_mdb_writemap: opt.experimental_reduce_indexing_memory_usage,
             indexer_config: (&opt.indexer_options).try_into()?,
             autobatching_enabled: true,
             max_number_of_tasks: 1_000_000,
+            max_number_of_batched_tasks: opt.experimental_max_number_of_batched_tasks,
             index_growth_amount: byte_unit::Byte::from_str("10GiB").unwrap().get_bytes() as usize,
             index_count: DEFAULT_INDEX_COUNT,
             instance_features,
@@ -397,6 +400,7 @@ pub fn configure_data(
         .app_data(web::Data::from(analytics))
         .app_data(
             web::JsonConfig::default()
+                .limit(http_payload_size_limit)
                 .content_type(|mime| mime == mime::APPLICATION_JSON)
                 .error_handler(|err, req: &HttpRequest| match err {
                     JsonPayloadError::ContentType => match req.headers().get(CONTENT_TYPE) {

meilisearch/src/main.rs

@@ -19,7 +19,11 @@ static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;
 /// does all the setup before meilisearch is launched
 fn setup(opt: &Opt) -> anyhow::Result<()> {
     let mut log_builder = env_logger::Builder::new();
-    log_builder.parse_filters(&opt.log_level.to_string());
+    let log_filters = format!(
+        "{},h2=warn,hyper=warn,tokio_util=warn,tracing=warn,rustls=warn,mio=warn,reqwest=warn",
+        opt.log_level
+    );
+    log_builder.parse_filters(&log_filters);
 
     log_builder.init();
 

meilisearch/src/option.rs

@@ -21,6 +21,7 @@ use rustls::RootCertStore;
 use rustls_pemfile::{certs, pkcs8_private_keys, rsa_private_keys};
 use serde::{Deserialize, Serialize};
 use sysinfo::{RefreshKind, System, SystemExt};
+use url::Url;
 
 const POSSIBLE_ENV: [&str; 2] = ["development", "production"];
 
@@ -28,6 +29,8 @@ const MEILI_DB_PATH: &str = "MEILI_DB_PATH";
 const MEILI_HTTP_ADDR: &str = "MEILI_HTTP_ADDR";
 const MEILI_MASTER_KEY: &str = "MEILI_MASTER_KEY";
 const MEILI_ENV: &str = "MEILI_ENV";
+const MEILI_TASK_WEBHOOK_URL: &str = "MEILI_TASK_WEBHOOK_URL";
+const MEILI_TASK_WEBHOOK_AUTHORIZATION_HEADER: &str = "MEILI_TASK_WEBHOOK_AUTHORIZATION_HEADER";
 #[cfg(feature = "analytics")]
 const MEILI_NO_ANALYTICS: &str = "MEILI_NO_ANALYTICS";
 const MEILI_HTTP_PAYLOAD_SIZE_LIMIT: &str = "MEILI_HTTP_PAYLOAD_SIZE_LIMIT";
@@ -51,6 +54,8 @@ const MEILI_LOG_LEVEL: &str = "MEILI_LOG_LEVEL";
 const MEILI_EXPERIMENTAL_ENABLE_METRICS: &str = "MEILI_EXPERIMENTAL_ENABLE_METRICS";
 const MEILI_EXPERIMENTAL_REDUCE_INDEXING_MEMORY_USAGE: &str =
     "MEILI_EXPERIMENTAL_REDUCE_INDEXING_MEMORY_USAGE";
+const MEILI_EXPERIMENTAL_MAX_NUMBER_OF_BATCHED_TASKS: &str =
+    "MEILI_EXPERIMENTAL_MAX_NUMBER_OF_BATCHED_TASKS";
 
 const DEFAULT_CONFIG_FILE_PATH: &str = "./config.toml";
 const DEFAULT_DB_PATH: &str = "./data.ms";
@@ -154,6 +159,14 @@ pub struct Opt {
     #[serde(default = "default_env")]
     pub env: String,
 
+    /// Called whenever a task finishes so a third party can be notified.
+    #[clap(long, env = MEILI_TASK_WEBHOOK_URL)]
+    pub task_webhook_url: Option<Url>,
+
+    /// The Authorization header to send on the webhook URL whenever a task finishes so a third party can be notified.
+    #[clap(long, env = MEILI_TASK_WEBHOOK_AUTHORIZATION_HEADER)]
+    pub task_webhook_authorization_header: Option<String>,
+
     /// Deactivates Meilisearch's built-in telemetry when provided.
     ///
     /// Meilisearch automatically collects data from all instances that do not opt out using this flag.
@@ -301,6 +314,11 @@ pub struct Opt {
     #[serde(default)]
     pub experimental_reduce_indexing_memory_usage: bool,
 
+    /// Experimentally reduces the maximum number of tasks that will be processed at once, see: <https://github.com/orgs/meilisearch/discussions/713>
+    #[clap(long, env = MEILI_EXPERIMENTAL_MAX_NUMBER_OF_BATCHED_TASKS, default_value_t = default_limit_batched_tasks())]
+    #[serde(default = "default_limit_batched_tasks")]
+    pub experimental_max_number_of_batched_tasks: usize,
+
     #[serde(flatten)]
     #[clap(flatten)]
     pub indexer_options: IndexerOpts,
@@ -368,9 +386,12 @@ impl Opt {
             http_addr,
             master_key,
             env,
+            task_webhook_url,
+            task_webhook_authorization_header,
             max_index_size: _,
             max_task_db_size: _,
             http_payload_size_limit,
+            experimental_max_number_of_batched_tasks,
             ssl_cert_path,
             ssl_key_path,
             ssl_auth_path,
@@ -392,8 +413,8 @@ impl Opt {
             config_file_path: _,
             #[cfg(feature = "analytics")]
             no_analytics,
-            experimental_enable_metrics: enable_metrics_route,
-            experimental_reduce_indexing_memory_usage: reduce_indexing_memory_usage,
+            experimental_enable_metrics,
+            experimental_reduce_indexing_memory_usage,
         } = self;
         export_to_env_if_not_present(MEILI_DB_PATH, db_path);
         export_to_env_if_not_present(MEILI_HTTP_ADDR, http_addr);
@@ -401,6 +422,16 @@ impl Opt {
             export_to_env_if_not_present(MEILI_MASTER_KEY, master_key);
         }
         export_to_env_if_not_present(MEILI_ENV, env);
+        if let Some(task_webhook_url) = task_webhook_url {
+            export_to_env_if_not_present(MEILI_TASK_WEBHOOK_URL, task_webhook_url.to_string());
+        }
+        if let Some(task_webhook_authorization_header) = task_webhook_authorization_header {
+            export_to_env_if_not_present(
+                MEILI_TASK_WEBHOOK_AUTHORIZATION_HEADER,
+                task_webhook_authorization_header,
+            );
+        }
+
         #[cfg(feature = "analytics")]
         {
             export_to_env_if_not_present(MEILI_NO_ANALYTICS, no_analytics.to_string());
@@ -409,6 +440,10 @@ impl Opt {
             MEILI_HTTP_PAYLOAD_SIZE_LIMIT,
             http_payload_size_limit.to_string(),
         );
+        export_to_env_if_not_present(
+            MEILI_EXPERIMENTAL_MAX_NUMBER_OF_BATCHED_TASKS,
+            experimental_max_number_of_batched_tasks.to_string(),
+        );
         if let Some(ssl_cert_path) = ssl_cert_path {
             export_to_env_if_not_present(MEILI_SSL_CERT_PATH, ssl_cert_path);
         }
@@ -433,11 +468,11 @@ impl Opt {
         export_to_env_if_not_present(MEILI_LOG_LEVEL, log_level.to_string());
         export_to_env_if_not_present(
             MEILI_EXPERIMENTAL_ENABLE_METRICS,
-            enable_metrics_route.to_string(),
+            experimental_enable_metrics.to_string(),
         );
         export_to_env_if_not_present(
             MEILI_EXPERIMENTAL_REDUCE_INDEXING_MEMORY_USAGE,
-            reduce_indexing_memory_usage.to_string(),
+            experimental_reduce_indexing_memory_usage.to_string(),
         );
         indexer_options.export_to_env();
     }
@@ -727,6 +762,10 @@ fn default_http_payload_size_limit() -> Byte {
     Byte::from_str(DEFAULT_HTTP_PAYLOAD_SIZE_LIMIT).unwrap()
 }
 
+fn default_limit_batched_tasks() -> usize {
+    usize::MAX
+}
+
 fn default_snapshot_dir() -> PathBuf {
     PathBuf::from(DEFAULT_SNAPSHOT_DIR)
 }

meilisearch/src/routes/indexes/documents.rs

@@ -3,7 +3,7 @@ use std::io::ErrorKind;
 use actix_web::http::header::CONTENT_TYPE;
 use actix_web::web::Data;
 use actix_web::{web, HttpMessage, HttpRequest, HttpResponse};
-use bstr::ByteSlice;
+use bstr::ByteSlice as _;
 use deserr::actix_web::{AwebJson, AwebQueryParameter};
 use deserr::Deserr;
 use futures::StreamExt;

meilisearch/src/routes/indexes/facet_search.rs

@@ -13,9 +13,9 @@ use crate::analytics::{Analytics, FacetSearchAggregator};
 use crate::extractors::authentication::policies::*;
 use crate::extractors::authentication::GuardedData;
 use crate::search::{
-    add_search_rules, perform_facet_search, MatchingStrategy, SearchQuery, DEFAULT_CROP_LENGTH,
-    DEFAULT_CROP_MARKER, DEFAULT_HIGHLIGHT_POST_TAG, DEFAULT_HIGHLIGHT_PRE_TAG,
-    DEFAULT_SEARCH_LIMIT, DEFAULT_SEARCH_OFFSET,
+    add_search_rules, perform_facet_search, HybridQuery, MatchingStrategy, SearchQuery,
+    DEFAULT_CROP_LENGTH, DEFAULT_CROP_MARKER, DEFAULT_HIGHLIGHT_POST_TAG,
+    DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_SEARCH_LIMIT, DEFAULT_SEARCH_OFFSET,
 };
 
 pub fn configure(cfg: &mut web::ServiceConfig) {
@@ -36,6 +36,8 @@ pub struct FacetSearchQuery {
     pub q: Option<String>,
     #[deserr(default, error = DeserrJsonError<InvalidSearchVector>)]
     pub vector: Option<Vec<f32>>,
+    #[deserr(default, error = DeserrJsonError<InvalidHybridQuery>)]
+    pub hybrid: Option<HybridQuery>,
     #[deserr(default, error = DeserrJsonError<InvalidSearchFilter>)]
     pub filter: Option<Value>,
     #[deserr(default, error = DeserrJsonError<InvalidSearchMatchingStrategy>, default)]
@@ -95,6 +97,7 @@ impl From<FacetSearchQuery> for SearchQuery {
             filter,
             matching_strategy,
             attributes_to_search_on,
+            hybrid,
         } = value;
 
         SearchQuery {
@@ -119,6 +122,7 @@ impl From<FacetSearchQuery> for SearchQuery {
             matching_strategy,
             vector,
             attributes_to_search_on,
+            hybrid,
         }
     }
 }

meilisearch/src/routes/indexes/search.rs

@@ -2,12 +2,14 @@ use actix_web::web::Data;
 use actix_web::{web, HttpRequest, HttpResponse};
 use deserr::actix_web::{AwebJson, AwebQueryParameter};
 use index_scheduler::IndexScheduler;
-use log::debug;
+use log::{debug, warn};
 use meilisearch_types::deserr::query_params::Param;
 use meilisearch_types::deserr::{DeserrJsonError, DeserrQueryParamError};
 use meilisearch_types::error::deserr_codes::*;
 use meilisearch_types::error::ResponseError;
 use meilisearch_types::index_uid::IndexUid;
+use meilisearch_types::milli;
+use meilisearch_types::milli::vector::DistributionShift;
 use meilisearch_types::serde_cs::vec::CS;
 use serde_json::Value;
 
@@ -16,9 +18,9 @@ use crate::extractors::authentication::policies::*;
 use crate::extractors::authentication::GuardedData;
 use crate::extractors::sequential_extractor::SeqHandler;
 use crate::search::{
-    add_search_rules, perform_search, MatchingStrategy, SearchQuery, DEFAULT_CROP_LENGTH,
-    DEFAULT_CROP_MARKER, DEFAULT_HIGHLIGHT_POST_TAG, DEFAULT_HIGHLIGHT_PRE_TAG,
-    DEFAULT_SEARCH_LIMIT, DEFAULT_SEARCH_OFFSET,
+    add_search_rules, perform_search, HybridQuery, MatchingStrategy, SearchQuery, SemanticRatio,
+    DEFAULT_CROP_LENGTH, DEFAULT_CROP_MARKER, DEFAULT_HIGHLIGHT_POST_TAG,
+    DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_SEARCH_LIMIT, DEFAULT_SEARCH_OFFSET, DEFAULT_SEMANTIC_RATIO,
 };
 
 pub fn configure(cfg: &mut web::ServiceConfig) {
@@ -74,6 +76,31 @@ pub struct SearchQueryGet {
     matching_strategy: MatchingStrategy,
     #[deserr(default, error = DeserrQueryParamError<InvalidSearchAttributesToSearchOn>)]
     pub attributes_to_search_on: Option<CS<String>>,
+    #[deserr(default, error = DeserrQueryParamError<InvalidEmbedder>)]
+    pub hybrid_embedder: Option<String>,
+    #[deserr(default, error = DeserrQueryParamError<InvalidSearchSemanticRatio>)]
+    pub hybrid_semantic_ratio: Option<SemanticRatioGet>,
+}
+
+#[derive(Debug, Clone, Copy, Default, PartialEq, deserr::Deserr)]
+#[deserr(try_from(String) = TryFrom::try_from -> InvalidSearchSemanticRatio)]
+pub struct SemanticRatioGet(SemanticRatio);
+
+impl std::convert::TryFrom<String> for SemanticRatioGet {
+    type Error = InvalidSearchSemanticRatio;
+
+    fn try_from(s: String) -> Result<Self, Self::Error> {
+        let f: f32 = s.parse().map_err(|_| InvalidSearchSemanticRatio)?;
+        Ok(SemanticRatioGet(SemanticRatio::try_from(f)?))
+    }
+}
+
+impl std::ops::Deref for SemanticRatioGet {
+    type Target = SemanticRatio;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
 }
 
 impl From<SearchQueryGet> for SearchQuery {
@@ -86,6 +113,20 @@ impl From<SearchQueryGet> for SearchQuery {
             None => None,
         };
 
+        let hybrid = match (other.hybrid_embedder, other.hybrid_semantic_ratio) {
+            (None, None) => None,
+            (None, Some(semantic_ratio)) => {
+                Some(HybridQuery { semantic_ratio: *semantic_ratio, embedder: None })
+            }
+            (Some(embedder), None) => Some(HybridQuery {
+                semantic_ratio: DEFAULT_SEMANTIC_RATIO(),
+                embedder: Some(embedder),
+            }),
+            (Some(embedder), Some(semantic_ratio)) => {
+                Some(HybridQuery { semantic_ratio: *semantic_ratio, embedder: Some(embedder) })
+            }
+        };
+
         Self {
             q: other.q,
             vector: other.vector.map(CS::into_inner),
@@ -108,6 +149,7 @@ impl From<SearchQueryGet> for SearchQuery {
             crop_marker: other.crop_marker,
             matching_strategy: other.matching_strategy,
             attributes_to_search_on: other.attributes_to_search_on.map(|o| o.into_iter().collect()),
+            hybrid,
         }
     }
 }
@@ -158,8 +200,12 @@ pub async fn search_with_url_query(
 
     let index = index_scheduler.index(&index_uid)?;
     let features = index_scheduler.features();
+
+    let distribution = embed(&mut query, index_scheduler.get_ref(), &index).await?;
+
     let search_result =
-        tokio::task::spawn_blocking(move || perform_search(&index, query, features)).await?;
+        tokio::task::spawn_blocking(move || perform_search(&index, query, features, distribution))
+            .await?;
     if let Ok(ref search_result) = search_result {
         aggregate.succeed(search_result);
     }
@@ -193,8 +239,12 @@ pub async fn search_with_post(
     let index = index_scheduler.index(&index_uid)?;
 
     let features = index_scheduler.features();
+
+    let distribution = embed(&mut query, index_scheduler.get_ref(), &index).await?;
+
     let search_result =
-        tokio::task::spawn_blocking(move || perform_search(&index, query, features)).await?;
+        tokio::task::spawn_blocking(move || perform_search(&index, query, features, distribution))
+            .await?;
     if let Ok(ref search_result) = search_result {
         aggregate.succeed(search_result);
     }
@@ -206,6 +256,80 @@ pub async fn search_with_post(
     Ok(HttpResponse::Ok().json(search_result))
 }
 
+pub async fn embed(
+    query: &mut SearchQuery,
+    index_scheduler: &IndexScheduler,
+    index: &milli::Index,
+) -> Result<Option<DistributionShift>, ResponseError> {
+    match (&query.hybrid, &query.vector, &query.q) {
+        (Some(HybridQuery { semantic_ratio: _, embedder }), None, Some(q))
+            if !q.trim().is_empty() =>
+        {
+            let embedder_configs = index.embedding_configs(&index.read_txn()?)?;
+            let embedders = index_scheduler.embedders(embedder_configs)?;
+
+            let embedder = if let Some(embedder_name) = embedder {
+                embedders.get(embedder_name)
+            } else {
+                embedders.get_default()
+            };
+
+            let embedder = embedder
+                .ok_or(milli::UserError::InvalidEmbedder("default".to_owned()))
+                .map_err(milli::Error::from)?
+                .0;
+
+            let distribution = embedder.distribution();
+
+            let embeddings = embedder
+                .embed(vec![q.to_owned()])
+                .await
+                .map_err(milli::vector::Error::from)
+                .map_err(milli::Error::from)?
+                .pop()
+                .expect("No vector returned from embedding");
+
+            if embeddings.iter().nth(1).is_some() {
+                warn!("Ignoring embeddings past the first one in long search query");
+                query.vector = Some(embeddings.iter().next().unwrap().to_vec());
+            } else {
+                query.vector = Some(embeddings.into_inner());
+            }
+            Ok(distribution)
+        }
+        (Some(hybrid), vector, _) => {
+            let embedder_configs = index.embedding_configs(&index.read_txn()?)?;
+            let embedders = index_scheduler.embedders(embedder_configs)?;
+
+            let embedder = if let Some(embedder_name) = &hybrid.embedder {
+                embedders.get(embedder_name)
+            } else {
+                embedders.get_default()
+            };
+
+            let embedder = embedder
+                .ok_or(milli::UserError::InvalidEmbedder("default".to_owned()))
+                .map_err(milli::Error::from)?
+                .0;
+
+            if let Some(vector) = vector {
+                if vector.len() != embedder.dimensions() {
+                    return Err(meilisearch_types::milli::Error::UserError(
+                        meilisearch_types::milli::UserError::InvalidVectorDimensions {
+                            expected: embedder.dimensions(),
+                            found: vector.len(),
+                        },
+                    )
+                    .into());
+                }
+            }
+
+            Ok(embedder.distribution())
+        }
+        _ => Ok(None),
+    }
+}
+
 #[cfg(test)]
 mod test {
     use super::*;

meilisearch/src/routes/indexes/settings.rs

@@ -7,6 +7,7 @@ use meilisearch_types::deserr::DeserrJsonError;
 use meilisearch_types::error::ResponseError;
 use meilisearch_types::facet_values_sort::FacetValuesSort;
 use meilisearch_types::index_uid::IndexUid;
+use meilisearch_types::milli::update::Setting;
 use meilisearch_types::settings::{settings, RankingRuleView, Settings, Unchecked};
 use meilisearch_types::tasks::KindWithContent;
 use serde_json::json;
@@ -78,6 +79,7 @@ macro_rules! make_setting_route {
 
                 let body = body.into_inner();
 
+                #[allow(clippy::redundant_closure_call)]
                 $analytics(&body, &req);
 
                 let new_settings = Settings {
@@ -88,6 +90,11 @@ macro_rules! make_setting_route {
                     ..Default::default()
                 };
 
+                let new_settings = $crate::routes::indexes::settings::validate_settings(
+                    new_settings,
+                    &index_scheduler,
+                )?;
+
                 let allow_index_creation =
                     index_scheduler.filters().allow_index_creation(&index_uid);
 
@@ -434,6 +441,31 @@ make_setting_route!(
     }
 );
 
+make_setting_route!(
+    "/proximity-precision",
+    put,
+    meilisearch_types::settings::ProximityPrecisionView,
+    meilisearch_types::deserr::DeserrJsonError<
+        meilisearch_types::error::deserr_codes::InvalidSettingsProximityPrecision,
+    >,
+    proximity_precision,
+    "proximityPrecision",
+    analytics,
+    |precision: &Option<meilisearch_types::settings::ProximityPrecisionView>, req: &HttpRequest| {
+        use serde_json::json;
+        analytics.publish(
+            "ProximityPrecision Updated".to_string(),
+            json!({
+                "proximity_precision": {
+                    "set": precision.is_some(),
+                    "value": precision.unwrap_or_default(),
+                }
+            }),
+            Some(req),
+        );
+    }
+);
+
 make_setting_route!(
     "/ranking-rules",
     put,
@@ -520,6 +552,67 @@ make_setting_route!(
     }
 );
 
+make_setting_route!(
+    "/embedders",
+    patch,
+    std::collections::BTreeMap<String, Setting<meilisearch_types::milli::vector::settings::EmbeddingSettings>>,
+    meilisearch_types::deserr::DeserrJsonError<
+        meilisearch_types::error::deserr_codes::InvalidSettingsEmbedders,
+    >,
+    embedders,
+    "embedders",
+    analytics,
+    |setting: &Option<std::collections::BTreeMap<String, Setting<meilisearch_types::milli::vector::settings::EmbeddingSettings>>>, req: &HttpRequest| {
+
+
+        analytics.publish(
+            "Embedders Updated".to_string(),
+            serde_json::json!({"embedders": crate::routes::indexes::settings::embedder_analytics(setting.as_ref())}),
+            Some(req),
+        );
+    }
+);
+
+fn embedder_analytics(
+    setting: Option<
+        &std::collections::BTreeMap<
+            String,
+            Setting<meilisearch_types::milli::vector::settings::EmbeddingSettings>,
+        >,
+    >,
+) -> serde_json::Value {
+    let mut sources = std::collections::HashSet::new();
+
+    if let Some(s) = &setting {
+        for source in s
+            .values()
+            .filter_map(|config| config.clone().set())
+            .filter_map(|config| config.source.set())
+        {
+            use meilisearch_types::milli::vector::settings::EmbedderSource;
+            match source {
+                EmbedderSource::OpenAi => sources.insert("openAi"),
+                EmbedderSource::HuggingFace => sources.insert("huggingFace"),
+                EmbedderSource::UserProvided => sources.insert("userProvided"),
+            };
+        }
+    };
+
+    let document_template_used = setting.as_ref().map(|map| {
+        map.values()
+            .filter_map(|config| config.clone().set())
+            .any(|config| config.document_template.set().is_some())
+    });
+
+    json!(
+        {
+            "total": setting.as_ref().map(|s| s.len()),
+            "sources": sources,
+            "document_template_used": document_template_used,
+        }
+    )
+}
+
 macro_rules! generate_configure {
     ($($mod:ident),*) => {
         pub fn configure(cfg: &mut web::ServiceConfig) {
@@ -540,6 +633,7 @@ generate_configure!(
     displayed_attributes,
     searchable_attributes,
     distinct_attribute,
+    proximity_precision,
     stop_words,
     separator_tokens,
     non_separator_tokens,
@@ -548,7 +642,8 @@ generate_configure!(
     ranking_rules,
     typo_tolerance,
     pagination,
-    faceting
+    faceting,
+    embedders
 );
 
 pub async fn update_all(
@@ -561,6 +656,7 @@ pub async fn update_all(
     let index_uid = IndexUid::try_from(index_uid.into_inner())?;
 
     let new_settings = body.into_inner();
+    let new_settings = validate_settings(new_settings, &index_scheduler)?;
 
     analytics.publish(
         "Settings Updated".to_string(),
@@ -593,6 +689,10 @@ pub async fn update_all(
             "distinct_attribute": {
                 "set": new_settings.distinct_attribute.as_ref().set().is_some()
             },
+            "proximity_precision": {
+                "set": new_settings.proximity_precision.as_ref().set().is_some(),
+                "value": new_settings.proximity_precision.as_ref().set().copied().unwrap_or_default()
+            },
             "typo_tolerance": {
                 "enabled": new_settings.typo_tolerance
                     .as_ref()
@@ -652,6 +752,7 @@ pub async fn update_all(
             "synonyms": {
                 "total": new_settings.synonyms.as_ref().set().map(|synonyms| synonyms.len()),
             },
+            "embedders": crate::routes::indexes::settings::embedder_analytics(new_settings.embedders.as_ref().set())
         }),
         Some(&req),
     );
@@ -706,3 +807,13 @@ pub async fn delete_all(
     debug!("returns: {:?}", task);
     Ok(HttpResponse::Accepted().json(task))
 }
+
+fn validate_settings(
+    settings: Settings<Unchecked>,
+    index_scheduler: &IndexScheduler,
+) -> Result<Settings<Unchecked>, ResponseError> {
+    if matches!(settings.embedders, Setting::Set(_)) {
+        index_scheduler.features().check_vector("Passing `embedders` in settings")?
+    }
+    Ok(settings.validate()?)
+}

meilisearch/src/routes/multi_search.rs

@@ -13,6 +13,7 @@ use crate::analytics::{Analytics, MultiSearchAggregator};
 use crate::extractors::authentication::policies::ActionPolicy;
 use crate::extractors::authentication::{AuthenticationError, GuardedData};
 use crate::extractors::sequential_extractor::SeqHandler;
+use crate::routes::indexes::search::embed;
 use crate::search::{
     add_search_rules, perform_search, SearchQueryWithIndex, SearchResultWithIndex,
 };
@@ -46,49 +47,51 @@ pub async fn multi_search_with_post(
     // Explicitly expect a `(ResponseError, usize)` for the error type rather than `ResponseError` only,
     // so that `?` doesn't work if it doesn't use `with_index`, ensuring that it is not forgotten in case of code
     // changes.
-    let search_results: Result<_, (ResponseError, usize)> = (|| {
-        async {
-            let mut search_results = Vec::with_capacity(queries.len());
-            for (query_index, (index_uid, mut query)) in
-                queries.into_iter().map(SearchQueryWithIndex::into_index_query).enumerate()
-            {
-                debug!("multi-search #{query_index}: called with params: {:?}", query);
+    let search_results: Result<_, (ResponseError, usize)> = async {
+        let mut search_results = Vec::with_capacity(queries.len());
+        for (query_index, (index_uid, mut query)) in
+            queries.into_iter().map(SearchQueryWithIndex::into_index_query).enumerate()
+        {
+            debug!("multi-search #{query_index}: called with params: {:?}", query);
 
-                // Check index from API key
-                if !index_scheduler.filters().is_index_authorized(&index_uid) {
-                    return Err(AuthenticationError::InvalidToken).with_index(query_index);
-                }
-                // Apply search rules from tenant token
-                if let Some(search_rules) =
-                    index_scheduler.filters().get_index_search_rules(&index_uid)
-                {
-                    add_search_rules(&mut query, search_rules);
-                }
-
-                let index = index_scheduler
-                    .index(&index_uid)
-                    .map_err(|err| {
-                        let mut err = ResponseError::from(err);
-                        // Patch the HTTP status code to 400 as it defaults to 404 for `index_not_found`, but
-                        // here the resource not found is not part of the URL.
-                        err.code = StatusCode::BAD_REQUEST;
-                        err
-                    })
-                    .with_index(query_index)?;
-
-                let search_result =
-                    tokio::task::spawn_blocking(move || perform_search(&index, query, features))
-                        .await
-                        .with_index(query_index)?;
-
-                search_results.push(SearchResultWithIndex {
-                    index_uid: index_uid.into_inner(),
-                    result: search_result.with_index(query_index)?,
-                });
+            // Check index from API key
+            if !index_scheduler.filters().is_index_authorized(&index_uid) {
+                return Err(AuthenticationError::InvalidToken).with_index(query_index);
             }
-            Ok(search_results)
+            // Apply search rules from tenant token
+            if let Some(search_rules) = index_scheduler.filters().get_index_search_rules(&index_uid)
+            {
+                add_search_rules(&mut query, search_rules);
+            }
+
+            let index = index_scheduler
+                .index(&index_uid)
+                .map_err(|err| {
+                    let mut err = ResponseError::from(err);
+                    // Patch the HTTP status code to 400 as it defaults to 404 for `index_not_found`, but
+                    // here the resource not found is not part of the URL.
+                    err.code = StatusCode::BAD_REQUEST;
+                    err
+                })
+                .with_index(query_index)?;
+
+            let distribution = embed(&mut query, index_scheduler.get_ref(), &index)
+                .await
+                .with_index(query_index)?;
+
+            let search_result = tokio::task::spawn_blocking(move || {
+                perform_search(&index, query, features, distribution)
+            })
+            .await
+            .with_index(query_index)?;
+
+            search_results.push(SearchResultWithIndex {
+                index_uid: index_uid.into_inner(),
+                result: search_result.with_index(query_index)?,
+            });
         }
-    })()
+        Ok(search_results)
+    }
     .await;
 
     if search_results.is_ok() {

meilisearch/src/routes/tasks.rs

@@ -8,11 +8,9 @@ use meilisearch_types::deserr::DeserrQueryParamError;
 use meilisearch_types::error::deserr_codes::*;
 use meilisearch_types::error::{InvalidTaskDateError, ResponseError};
 use meilisearch_types::index_uid::IndexUid;
-use meilisearch_types::settings::{Settings, Unchecked};
 use meilisearch_types::star_or::{OptionStarOr, OptionStarOrList};
-use meilisearch_types::tasks::{
-    serialize_duration, Details, IndexSwap, Kind, KindWithContent, Status, Task,
-};
+use meilisearch_types::task_view::TaskView;
+use meilisearch_types::tasks::{Kind, KindWithContent, Status};
 use serde::Serialize;
 use serde_json::json;
 use time::format_description::well_known::Rfc3339;
@@ -37,140 +35,6 @@ pub fn configure(cfg: &mut web::ServiceConfig) {
     .service(web::resource("/cancel").route(web::post().to(SeqHandler(cancel_tasks))))
     .service(web::resource("/{task_id}").route(web::get().to(SeqHandler(get_task))));
 }
-
-#[derive(Debug, Clone, PartialEq, Eq, Serialize)]
-#[serde(rename_all = "camelCase")]
-pub struct TaskView {
-    pub uid: TaskId,
-    #[serde(default)]
-    pub index_uid: Option<String>,
-    pub status: Status,
-    #[serde(rename = "type")]
-    pub kind: Kind,
-    pub canceled_by: Option<TaskId>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub details: Option<DetailsView>,
-    pub error: Option<ResponseError>,
-    #[serde(serialize_with = "serialize_duration", default)]
-    pub duration: Option<Duration>,
-    #[serde(with = "time::serde::rfc3339")]
-    pub enqueued_at: OffsetDateTime,
-    #[serde(with = "time::serde::rfc3339::option", default)]
-    pub started_at: Option<OffsetDateTime>,
-    #[serde(with = "time::serde::rfc3339::option", default)]
-    pub finished_at: Option<OffsetDateTime>,
-}
-
-impl TaskView {
-    pub fn from_task(task: &Task) -> TaskView {
-        TaskView {
-            uid: task.uid,
-            index_uid: task.index_uid().map(ToOwned::to_owned),
-            status: task.status,
-            kind: task.kind.as_kind(),
-            canceled_by: task.canceled_by,
-            details: task.details.clone().map(DetailsView::from),
-            error: task.error.clone(),
-            duration: task.started_at.zip(task.finished_at).map(|(start, end)| end - start),
-            enqueued_at: task.enqueued_at,
-            started_at: task.started_at,
-            finished_at: task.finished_at,
-        }
-    }
-}
-
-#[derive(Default, Debug, PartialEq, Eq, Clone, Serialize)]
-#[serde(rename_all = "camelCase")]
-pub struct DetailsView {
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub received_documents: Option<u64>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub indexed_documents: Option<Option<u64>>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub primary_key: Option<Option<String>>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub provided_ids: Option<usize>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub deleted_documents: Option<Option<u64>>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub matched_tasks: Option<u64>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub canceled_tasks: Option<Option<u64>>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub deleted_tasks: Option<Option<u64>>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub original_filter: Option<Option<String>>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub dump_uid: Option<Option<String>>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    #[serde(flatten)]
-    pub settings: Option<Box<Settings<Unchecked>>>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub swaps: Option<Vec<IndexSwap>>,
-}
-
-impl From<Details> for DetailsView {
-    fn from(details: Details) -> Self {
-        match details {
-            Details::DocumentAdditionOrUpdate { received_documents, indexed_documents } => {
-                DetailsView {
-                    received_documents: Some(received_documents),
-                    indexed_documents: Some(indexed_documents),
-                    ..DetailsView::default()
-                }
-            }
-            Details::SettingsUpdate { settings } => {
-                DetailsView { settings: Some(settings), ..DetailsView::default() }
-            }
-            Details::IndexInfo { primary_key } => {
-                DetailsView { primary_key: Some(primary_key), ..DetailsView::default() }
-            }
-            Details::DocumentDeletion {
-                provided_ids: received_document_ids,
-                deleted_documents,
-            } => DetailsView {
-                provided_ids: Some(received_document_ids),
-                deleted_documents: Some(deleted_documents),
-                original_filter: Some(None),
-                ..DetailsView::default()
-            },
-            Details::DocumentDeletionByFilter { original_filter, deleted_documents } => {
-                DetailsView {
-                    provided_ids: Some(0),
-                    original_filter: Some(Some(original_filter)),
-                    deleted_documents: Some(deleted_documents),
-                    ..DetailsView::default()
-                }
-            }
-            Details::ClearAll { deleted_documents } => {
-                DetailsView { deleted_documents: Some(deleted_documents), ..DetailsView::default() }
-            }
-            Details::TaskCancelation { matched_tasks, canceled_tasks, original_filter } => {
-                DetailsView {
-                    matched_tasks: Some(matched_tasks),
-                    canceled_tasks: Some(canceled_tasks),
-                    original_filter: Some(Some(original_filter)),
-                    ..DetailsView::default()
-                }
-            }
-            Details::TaskDeletion { matched_tasks, deleted_tasks, original_filter } => {
-                DetailsView {
-                    matched_tasks: Some(matched_tasks),
-                    deleted_tasks: Some(deleted_tasks),
-                    original_filter: Some(Some(original_filter)),
-                    ..DetailsView::default()
-                }
-            }
-            Details::Dump { dump_uid } => {
-                DetailsView { dump_uid: Some(dump_uid), ..DetailsView::default() }
-            }
-            Details::IndexSwap { swaps } => {
-                DetailsView { swaps: Some(swaps), ..Default::default() }
-            }
-        }
-    }
-}
-
 #[derive(Debug, Deserr)]
 #[deserr(error = DeserrQueryParamError, rename_all = camelCase, deny_unknown_fields)]
 pub struct TasksFilterQuery {

meilisearch/src/search.rs

@@ -7,24 +7,21 @@ use deserr::Deserr;
 use either::Either;
 use index_scheduler::RoFeatures;
 use indexmap::IndexMap;
-use log::warn;
 use meilisearch_auth::IndexSearchRules;
 use meilisearch_types::deserr::DeserrJsonError;
 use meilisearch_types::error::deserr_codes::*;
 use meilisearch_types::heed::RoTxn;
 use meilisearch_types::index_uid::IndexUid;
-use meilisearch_types::milli::score_details::{ScoreDetails, ScoringStrategy};
-use meilisearch_types::milli::{
-    dot_product_similarity, FacetValueHit, InternalError, OrderBy, SearchForFacetValues,
-};
+use meilisearch_types::milli::score_details::{self, ScoreDetails, ScoringStrategy};
+use meilisearch_types::milli::vector::DistributionShift;
+use meilisearch_types::milli::{FacetValueHit, OrderBy, SearchForFacetValues};
 use meilisearch_types::settings::DEFAULT_PAGINATION_MAX_TOTAL_HITS;
 use meilisearch_types::{milli, Document};
 use milli::tokenizer::TokenizerBuilder;
 use milli::{
     AscDesc, FieldId, FieldsIdsMap, Filter, FormatOptions, Index, MatchBounds, MatcherBuilder,
-    SortError, TermsMatchingStrategy, VectorOrArrayOfVectors, DEFAULT_VALUES_PER_FACET,
+    SortError, TermsMatchingStrategy, DEFAULT_VALUES_PER_FACET,
 };
-use ordered_float::OrderedFloat;
 use regex::Regex;
 use serde::Serialize;
 use serde_json::{json, Value};
@@ -39,6 +36,7 @@ pub const DEFAULT_CROP_LENGTH: fn() -> usize = || 10;
 pub const DEFAULT_CROP_MARKER: fn() -> String = || "…".to_string();
 pub const DEFAULT_HIGHLIGHT_PRE_TAG: fn() -> String = || "<em>".to_string();
 pub const DEFAULT_HIGHLIGHT_POST_TAG: fn() -> String = || "</em>".to_string();
+pub const DEFAULT_SEMANTIC_RATIO: fn() -> SemanticRatio = || SemanticRatio(0.5);
 
 #[derive(Debug, Clone, Default, PartialEq, Deserr)]
 #[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
@@ -47,6 +45,8 @@ pub struct SearchQuery {
     pub q: Option<String>,
     #[deserr(default, error = DeserrJsonError<InvalidSearchVector>)]
     pub vector: Option<Vec<f32>>,
+    #[deserr(default, error = DeserrJsonError<InvalidHybridQuery>)]
+    pub hybrid: Option<HybridQuery>,
     #[deserr(default = DEFAULT_SEARCH_OFFSET(), error = DeserrJsonError<InvalidSearchOffset>)]
     pub offset: usize,
     #[deserr(default = DEFAULT_SEARCH_LIMIT(), error = DeserrJsonError<InvalidSearchLimit>)]
@@ -87,6 +87,48 @@ pub struct SearchQuery {
     pub attributes_to_search_on: Option<Vec<String>>,
 }
 
+#[derive(Debug, Clone, Default, PartialEq, Deserr)]
+#[deserr(error = DeserrJsonError<InvalidHybridQuery>, rename_all = camelCase, deny_unknown_fields)]
+pub struct HybridQuery {
+    /// TODO validate that sementic ratio is between 0.0 and 1,0
+    #[deserr(default, error = DeserrJsonError<InvalidSearchSemanticRatio>, default)]
+    pub semantic_ratio: SemanticRatio,
+    #[deserr(default, error = DeserrJsonError<InvalidEmbedder>, default)]
+    pub embedder: Option<String>,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Deserr)]
+#[deserr(try_from(f32) = TryFrom::try_from -> InvalidSearchSemanticRatio)]
+pub struct SemanticRatio(f32);
+
+impl Default for SemanticRatio {
+    fn default() -> Self {
+        DEFAULT_SEMANTIC_RATIO()
+    }
+}
+
+impl std::convert::TryFrom<f32> for SemanticRatio {
+    type Error = InvalidSearchSemanticRatio;
+
+    fn try_from(f: f32) -> Result<Self, Self::Error> {
+        // the suggested "fix" is: `!(0.0..=1.0).contains(&f)`` which is allegedly less readable
+        #[allow(clippy::manual_range_contains)]
+        if f > 1.0 || f < 0.0 {
+            Err(InvalidSearchSemanticRatio)
+        } else {
+            Ok(SemanticRatio(f))
+        }
+    }
+}
+
+impl std::ops::Deref for SemanticRatio {
+    type Target = f32;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
 impl SearchQuery {
     pub fn is_finite_pagination(&self) -> bool {
         self.page.or(self.hits_per_page).is_some()
@@ -106,6 +148,8 @@ pub struct SearchQueryWithIndex {
     pub q: Option<String>,
     #[deserr(default, error = DeserrJsonError<InvalidSearchQ>)]
     pub vector: Option<Vec<f32>>,
+    #[deserr(default, error = DeserrJsonError<InvalidHybridQuery>)]
+    pub hybrid: Option<HybridQuery>,
     #[deserr(default = DEFAULT_SEARCH_OFFSET(), error = DeserrJsonError<InvalidSearchOffset>)]
     pub offset: usize,
     #[deserr(default = DEFAULT_SEARCH_LIMIT(), error = DeserrJsonError<InvalidSearchLimit>)]
@@ -171,6 +215,7 @@ impl SearchQueryWithIndex {
             crop_marker,
             matching_strategy,
             attributes_to_search_on,
+            hybrid,
         } = self;
         (
             index_uid,
@@ -196,6 +241,7 @@ impl SearchQueryWithIndex {
                 crop_marker,
                 matching_strategy,
                 attributes_to_search_on,
+                hybrid,
                 // do not use ..Default::default() here,
                 // rather add any missing field from `SearchQuery` to `SearchQueryWithIndex`
             },
@@ -335,19 +381,44 @@ fn prepare_search<'t>(
     rtxn: &'t RoTxn,
     query: &'t SearchQuery,
     features: RoFeatures,
+    distribution: Option<DistributionShift>,
 ) -> Result<(milli::Search<'t>, bool, usize, usize), MeilisearchHttpError> {
     let mut search = index.search(rtxn);
 
-    if query.vector.is_some() && query.q.is_some() {
-        warn!("Ignoring the query string `q` when used with the `vector` parameter.");
+    if query.vector.is_some() {
+        features.check_vector("Passing `vector` as a query parameter")?;
     }
 
+    if query.hybrid.is_some() {
+        features.check_vector("Passing `hybrid` as a query parameter")?;
+    }
+
+    if query.hybrid.is_none() && query.q.is_some() && query.vector.is_some() {
+        return Err(MeilisearchHttpError::MissingSearchHybrid);
+    }
+
+    search.distribution_shift(distribution);
+
     if let Some(ref vector) = query.vector {
-        search.vector(vector.clone());
+        match &query.hybrid {
+            // If semantic ratio is 0.0, only the query search will impact the search results,
+            // skip the vector
+            Some(hybrid) if *hybrid.semantic_ratio == 0.0 => (),
+            _otherwise => {
+                search.vector(vector.clone());
+            }
+        }
     }
 
-    if let Some(ref query) = query.q {
-        search.query(query);
+    if let Some(ref q) = query.q {
+        match &query.hybrid {
+            // If semantic ratio is 1.0, only the vector search will impact the search results,
+            // skip the query
+            Some(hybrid) if *hybrid.semantic_ratio == 1.0 => (),
+            _otherwise => {
+                search.query(q);
+            }
+        }
     }
 
     if let Some(ref searchable) = query.attributes_to_search_on {
@@ -360,6 +431,7 @@ fn prepare_search<'t>(
     let max_total_hits = index
         .pagination_max_total_hits(rtxn)
         .map_err(milli::Error::from)?
+        .map(|x| x as usize)
         .unwrap_or(DEFAULT_PAGINATION_MAX_TOTAL_HITS);
 
     search.exhaustive_number_hits(is_finite_pagination);
@@ -373,8 +445,8 @@ fn prepare_search<'t>(
         features.check_score_details()?;
     }
 
-    if query.vector.is_some() {
-        features.check_vector()?;
+    if let Some(HybridQuery { embedder: Some(embedder), .. }) = &query.hybrid {
+        search.embedder_name(embedder);
     }
 
     // compute the offset on the limit depending on the pagination mode.
@@ -420,15 +492,22 @@ pub fn perform_search(
     index: &Index,
     query: SearchQuery,
     features: RoFeatures,
+    distribution: Option<DistributionShift>,
 ) -> Result<SearchResult, MeilisearchHttpError> {
     let before_search = Instant::now();
     let rtxn = index.read_txn()?;
 
     let (search, is_finite_pagination, max_total_hits, offset) =
-        prepare_search(index, &rtxn, &query, features)?;
+        prepare_search(index, &rtxn, &query, features, distribution)?;
 
     let milli::SearchResult { documents_ids, matching_words, candidates, document_scores, .. } =
-        search.execute()?;
+        match &query.hybrid {
+            Some(hybrid) => match *hybrid.semantic_ratio {
+                ratio if ratio == 0.0 || ratio == 1.0 => search.execute()?,
+                ratio => search.execute_hybrid(ratio)?,
+            },
+            None => search.execute()?,
+        };
 
     let fields_ids_map = index.fields_ids_map(&rtxn).unwrap();
 
@@ -537,13 +616,17 @@ pub fn perform_search(
             insert_geo_distance(sort, &mut document);
         }
 
-        let semantic_score = match query.vector.as_ref() {
-            Some(vector) => match extract_field("_vectors", &fields_ids_map, obkv)? {
-                Some(vectors) => compute_semantic_score(vector, vectors)?,
-                None => None,
-            },
-            None => None,
-        };
+        let mut semantic_score = None;
+        for details in &score {
+            if let ScoreDetails::Vector(score_details::Vector {
+                target_vector: _,
+                value_similarity: Some((_matching_vector, similarity)),
+            }) = details
+            {
+                semantic_score = Some(*similarity);
+                break;
+            }
+        }
 
         let ranking_score =
             query.show_ranking_score.then(|| ScoreDetails::global_score(score.iter()));
@@ -586,6 +669,7 @@ pub fn perform_search(
             let max_values_by_facet = index
                 .max_values_per_facet(&rtxn)
                 .map_err(milli::Error::from)?
+                .map(|x| x as usize)
                 .unwrap_or(DEFAULT_VALUES_PER_FACET);
             facet_distribution.max_values_per_facet(max_values_by_facet);
 
@@ -645,11 +729,15 @@ pub fn perform_facet_search(
     let before_search = Instant::now();
     let rtxn = index.read_txn()?;
 
-    let (search, _, _, _) = prepare_search(index, &rtxn, &search_query, features)?;
-    let mut facet_search = SearchForFacetValues::new(facet_name, search);
+    let (search, _, _, _) = prepare_search(index, &rtxn, &search_query, features, None)?;
+    let mut facet_search =
+        SearchForFacetValues::new(facet_name, search, search_query.hybrid.is_some());
     if let Some(facet_query) = &facet_query {
         facet_search.query(facet_query);
     }
+    if let Some(max_facets) = index.max_values_per_facet(&rtxn)? {
+        facet_search.max_values(max_facets as usize);
+    }
 
     Ok(FacetSearchResult {
         facet_hits: facet_search.execute()?,
@@ -674,18 +762,6 @@ fn insert_geo_distance(sorts: &[String], document: &mut Document) {
     }
 }
 
-fn compute_semantic_score(query: &[f32], vectors: Value) -> milli::Result<Option<f32>> {
-    let vectors = serde_json::from_value(vectors)
-        .map(VectorOrArrayOfVectors::into_array_of_vectors)
-        .map_err(InternalError::SerdeJson)?;
-    Ok(vectors
-        .into_iter()
-        .flatten()
-        .map(|v| OrderedFloat(dot_product_similarity(query, &v)))
-        .max()
-        .map(OrderedFloat::into_inner))
-}
-
 fn compute_formatted_options(
     attr_to_highlight: &HashSet<String>,
     attr_to_crop: &[String],
@@ -813,22 +889,6 @@ fn make_document(
     Ok(document)
 }
 
-/// Extract the JSON value under the field name specified
-/// but doesn't support nested objects.
-fn extract_field(
-    field_name: &str,
-    field_ids_map: &FieldsIdsMap,
-    obkv: obkv::KvReaderU16,
-) -> Result<Option<serde_json::Value>, MeilisearchHttpError> {
-    match field_ids_map.id(field_name) {
-        Some(fid) => match obkv.get(fid) {
-            Some(value) => Ok(serde_json::from_slice(value).map(Some)?),
-            None => Ok(None),
-        },
-        None => Ok(None),
-    }
-}
-
 fn format_fields<'a>(
     document: &Document,
     field_ids_map: &FieldsIdsMap,
@@ -840,6 +900,14 @@ fn format_fields<'a>(
     let mut matches_position = compute_matches.then(BTreeMap::new);
     let mut document = document.clone();
 
+    // reduce the formatted option list to the attributes that should be formatted,
+    // instead of all the attributes to display.
+    let formatting_fields_options: Vec<_> = formatted_options
+        .iter()
+        .filter(|(_, option)| option.should_format())
+        .map(|(fid, option)| (field_ids_map.name(*fid).unwrap(), option))
+        .collect();
+
     // select the attributes to retrieve
     let displayable_names =
         displayable_ids.iter().map(|&fid| field_ids_map.name(fid).expect("Missing field name"));
@@ -848,13 +916,15 @@ fn format_fields<'a>(
         // to the value and merge them together. eg. If a user said he wanted to highlight `doggo`
         // and crop `doggo.name`. `doggo.name` needs to be highlighted + cropped while `doggo.age` is only
         // highlighted.
-        let format = formatted_options
+        // Warn: The time to compute the format list scales with the number of fields to format;
+        // cumulated with map_leaf_values that iterates over all the nested fields, it gives a quadratic complexity:
+        // d*f where d is the total number of fields to display and f is the total number of fields to format.
+        let format = formatting_fields_options
             .iter()
-            .filter(|(field, _option)| {
-                let name = field_ids_map.name(**field).unwrap();
+            .filter(|(name, _option)| {
                 milli::is_faceted_by(name, key) || milli::is_faceted_by(key, name)
             })
-            .map(|(_, option)| *option)
+            .map(|(_, option)| **option)
             .reduce(|acc, option| acc.merge(option));
         let mut infos = Vec::new();
 
@@ -951,7 +1021,7 @@ fn format_value<'a>(
                     let value = matcher.format(format_options);
                     Value::String(value.into_owned())
                 }
-                None => Value::Number(number),
+                None => Value::String(s),
             }
         }
         value => value,

meilisearch/tests/common/mod.rs

@@ -5,9 +5,11 @@ pub mod service;
 
 use std::fmt::{self, Display};
 
+#[allow(unused)]
 pub use index::{GetAllDocumentsOptions, GetDocumentOptions};
 use meili_snap::json_string;
 use serde::{Deserialize, Serialize};
+#[allow(unused)]
 pub use server::{default_settings, Server};
 
 #[derive(Debug, Clone, Default, Serialize, Deserialize, PartialEq, Eq)]
@@ -62,7 +64,7 @@ impl Display for Value {
         write!(
             f,
             "{}",
-            json_string!(self, { ".enqueuedAt" => "[date]", ".processedAt" => "[date]", ".finishedAt" => "[date]", ".duration" => "[duration]" })
+            json_string!(self, { ".enqueuedAt" => "[date]", ".startedAt" => "[date]", ".finishedAt" => "[date]", ".duration" => "[duration]" })
         )
     }
 }

meilisearch/tests/documents/add_documents.rs

@@ -1760,6 +1760,181 @@ async fn add_documents_invalid_geo_field() {
       "finishedAt": "[date]"
     }
     "###);
+
+    // The three next tests are related to #4333
+
+    // _geo has a lat and lng but set to `null`
+    let documents = json!([
+        {
+            "id": "12",
+            "_geo": { "lng": null, "lat": 67}
+        }
+    ]);
+
+    let (response, code) = index.add_documents(documents, None).await;
+    snapshot!(code, @"202 Accepted");
+    let response = index.wait_task(response.uid()).await;
+    snapshot!(json_string!(response, { ".duration" => "[duration]", ".enqueuedAt" => "[date]", ".startedAt" => "[date]", ".finishedAt" => "[date]" }),
+        @r###"
+    {
+      "uid": 14,
+      "indexUid": "test",
+      "status": "failed",
+      "type": "documentAdditionOrUpdate",
+      "canceledBy": null,
+      "details": {
+        "receivedDocuments": 1,
+        "indexedDocuments": 0
+      },
+      "error": {
+        "message": "Could not parse longitude in the document with the id: `12`. Was expecting a finite number but instead got `null`.",
+        "code": "invalid_document_geo_field",
+        "type": "invalid_request",
+        "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
+      },
+      "duration": "[duration]",
+      "enqueuedAt": "[date]",
+      "startedAt": "[date]",
+      "finishedAt": "[date]"
+    }
+    "###);
+
+    // _geo has a lat and lng but set to `null`
+    let documents = json!([
+        {
+            "id": "12",
+            "_geo": { "lng": 35, "lat": null }
+        }
+    ]);
+
+    let (response, code) = index.add_documents(documents, None).await;
+    snapshot!(code, @"202 Accepted");
+    let response = index.wait_task(response.uid()).await;
+    snapshot!(json_string!(response, { ".duration" => "[duration]", ".enqueuedAt" => "[date]", ".startedAt" => "[date]", ".finishedAt" => "[date]" }),
+        @r###"
+    {
+      "uid": 15,
+      "indexUid": "test",
+      "status": "failed",
+      "type": "documentAdditionOrUpdate",
+      "canceledBy": null,
+      "details": {
+        "receivedDocuments": 1,
+        "indexedDocuments": 0
+      },
+      "error": {
+        "message": "Could not parse latitude in the document with the id: `12`. Was expecting a finite number but instead got `null`.",
+        "code": "invalid_document_geo_field",
+        "type": "invalid_request",
+        "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
+      },
+      "duration": "[duration]",
+      "enqueuedAt": "[date]",
+      "startedAt": "[date]",
+      "finishedAt": "[date]"
+    }
+    "###);
+
+    // _geo has a lat and lng but set to `null`
+    let documents = json!([
+        {
+            "id": "13",
+            "_geo": { "lng": null, "lat": null }
+        }
+    ]);
+
+    let (response, code) = index.add_documents(documents, None).await;
+    snapshot!(code, @"202 Accepted");
+    let response = index.wait_task(response.uid()).await;
+    snapshot!(json_string!(response, { ".duration" => "[duration]", ".enqueuedAt" => "[date]", ".startedAt" => "[date]", ".finishedAt" => "[date]" }),
+        @r###"
+    {
+      "uid": 16,
+      "indexUid": "test",
+      "status": "failed",
+      "type": "documentAdditionOrUpdate",
+      "canceledBy": null,
+      "details": {
+        "receivedDocuments": 1,
+        "indexedDocuments": 0
+      },
+      "error": {
+        "message": "Could not parse latitude nor longitude in the document with the id: `13`. Was expecting finite numbers but instead got `null` and `null`.",
+        "code": "invalid_document_geo_field",
+        "type": "invalid_request",
+        "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
+      },
+      "duration": "[duration]",
+      "enqueuedAt": "[date]",
+      "startedAt": "[date]",
+      "finishedAt": "[date]"
+    }
+    "###);
+}
+
+// Related to #4333
+#[actix_rt::test]
+async fn add_invalid_geo_and_then_settings() {
+    let server = Server::new().await;
+    let index = server.index("test");
+    index.create(Some("id")).await;
+
+    // _geo is not an object
+    let documents = json!([
+        {
+            "id": "11",
+            "_geo": { "lat": null, "lng": null },
+        }
+    ]);
+    let (ret, code) = index.add_documents(documents, None).await;
+    snapshot!(code, @"202 Accepted");
+    let ret = index.wait_task(ret.uid()).await;
+    snapshot!(ret, @r###"
+    {
+      "uid": 1,
+      "indexUid": "test",
+      "status": "succeeded",
+      "type": "documentAdditionOrUpdate",
+      "canceledBy": null,
+      "details": {
+        "receivedDocuments": 1,
+        "indexedDocuments": 1
+      },
+      "error": null,
+      "duration": "[duration]",
+      "enqueuedAt": "[date]",
+      "startedAt": "[date]",
+      "finishedAt": "[date]"
+    }
+    "###);
+
+    let (ret, code) = index.update_settings(json!({"sortableAttributes": ["_geo"]})).await;
+    snapshot!(code, @"202 Accepted");
+    let ret = index.wait_task(ret.uid()).await;
+    snapshot!(ret, @r###"
+    {
+      "uid": 2,
+      "indexUid": "test",
+      "status": "failed",
+      "type": "settingsUpdate",
+      "canceledBy": null,
+      "details": {
+        "sortableAttributes": [
+          "_geo"
+        ]
+      },
+      "error": {
+        "message": "Could not parse latitude in the document with the id: `\"11\"`. Was expecting a finite number but instead got `null`.",
+        "code": "invalid_document_geo_field",
+        "type": "invalid_request",
+        "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
+      },
+      "duration": "[duration]",
+      "enqueuedAt": "[date]",
+      "startedAt": "[date]",
+      "finishedAt": "[date]"
+    }
+    "###);
 }
 
 #[actix_rt::test]

meilisearch/tests/dumps/data.rs

@@ -20,6 +20,8 @@ pub enum GetDump {
     RubyGemsWithSettingsV4,
 
     TestV5,
+
+    TestV6WithExperimental,
 }
 
 impl GetDump {
@@ -68,6 +70,10 @@ impl GetDump {
             GetDump::TestV5 => {
                 exist_relative_path!("tests/assets/v5_v0.28.0_test_dump.dump").into()
             }
+            GetDump::TestV6WithExperimental => exist_relative_path!(
+                "tests/assets/v6_v1.6.0_use_deactivated_experimental_setting.dump"
+            )
+            .into(),
         }
     }
 }

meilisearch/tests/dumps/mod.rs

@@ -59,6 +59,7 @@ async fn import_dump_v1_movie_raw() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -219,6 +220,7 @@ async fn import_dump_v1_movie_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -365,6 +367,7 @@ async fn import_dump_v1_rubygems_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -497,6 +500,7 @@ async fn import_dump_v2_movie_raw() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -641,6 +645,7 @@ async fn import_dump_v2_movie_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -784,6 +789,7 @@ async fn import_dump_v2_rubygems_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -916,6 +922,7 @@ async fn import_dump_v3_movie_raw() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1060,6 +1067,7 @@ async fn import_dump_v3_movie_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1203,6 +1211,7 @@ async fn import_dump_v3_rubygems_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1335,6 +1344,7 @@ async fn import_dump_v4_movie_raw() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1479,6 +1489,7 @@ async fn import_dump_v4_movie_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1622,6 +1633,7 @@ async fn import_dump_v4_rubygems_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": "byWord",
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1810,3 +1822,107 @@ async fn import_dump_v5() {
         json_string!(tasks, { ".results[].details.dumpUid" => "[uid]",  ".results[].duration" => "[duration]" ,  ".results[].startedAt" => "[date]" ,  ".results[].finishedAt" => "[date]"  })
     );
 }
+
+#[actix_rt::test]
+async fn import_dump_v6_containing_experimental_features() {
+    let temp = tempfile::tempdir().unwrap();
+
+    let options = Opt {
+        import_dump: Some(GetDump::TestV6WithExperimental.path()),
+        ..default_settings(temp.path())
+    };
+    let mut server = Server::new_auth_with_options(options, temp).await;
+    server.use_api_key("MASTER_KEY");
+
+    let (indexes, code) = server.list_indexes(None, None).await;
+    assert_eq!(code, 200, "{indexes}");
+
+    assert_eq!(indexes["results"].as_array().unwrap().len(), 1);
+    assert_eq!(indexes["results"][0]["uid"], json!("movies"));
+    assert_eq!(indexes["results"][0]["primaryKey"], json!("id"));
+
+    let (response, code) = server.get_features().await;
+    meili_snap::snapshot!(code, @"200 OK");
+    meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
+    {
+      "scoreDetails": false,
+      "vectorStore": false,
+      "metrics": false,
+      "exportPuffinReports": false
+    }
+    "###);
+
+    let index = server.index("movies");
+
+    let (response, code) = index.settings().await;
+    meili_snap::snapshot!(code, @"200 OK");
+    meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
+    {
+      "displayedAttributes": [
+        "*"
+      ],
+      "searchableAttributes": [
+        "*"
+      ],
+      "filterableAttributes": [],
+      "sortableAttributes": [],
+      "rankingRules": [
+        "words",
+        "typo",
+        "proximity"
+      ],
+      "stopWords": [],
+      "nonSeparatorTokens": [],
+      "separatorTokens": [],
+      "dictionary": [],
+      "synonyms": {},
+      "distinctAttribute": null,
+      "proximityPrecision": "byAttribute",
+      "typoTolerance": {
+        "enabled": true,
+        "minWordSizeForTypos": {
+          "oneTypo": 5,
+          "twoTypos": 9
+        },
+        "disableOnWords": [],
+        "disableOnAttributes": []
+      },
+      "faceting": {
+        "maxValuesPerFacet": 100,
+        "sortFacetValuesBy": {
+          "*": "alpha"
+        }
+      },
+      "pagination": {
+        "maxTotalHits": 1000
+      }
+    }
+    "###);
+
+    // the expected order is [1, 3, 2] instead of [3, 1, 2]
+    // because the attribute scale doesn't make the difference between 1 and 3.
+    index
+        .search(json!({"q": "the soup of day"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}

meilisearch/tests/search/distinct.rs

@@ -4,23 +4,111 @@ use once_cell::sync::Lazy;
 use crate::common::{Server, Value};
 use crate::json;
 
-pub(self) static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
-        {"productId": 1, "shopId": 1},
-        {"productId": 2, "shopId": 1},
-        {"productId": 3, "shopId": 2},
-        {"productId": 4, "shopId": 2},
-        {"productId": 5, "shopId": 3},
-        {"productId": 6, "shopId": 3},
-        {"productId": 7, "shopId": 4},
-        {"productId": 8, "shopId": 4},
-        {"productId": 9, "shopId": 5},
-        {"productId": 10, "shopId": 5}
+      {
+        "id": 1,
+        "description": "Leather Jacket",
+        "brand": "Lee Jeans",
+        "product_id": "123456",
+        "color": "Brown"
+      },
+      {
+        "id": 2,
+        "description": "Leather Jacket",
+        "brand": "Lee Jeans",
+        "product_id": "123456",
+        "color": "Black"
+      },
+      {
+        "id": 3,
+        "description": "Leather Jacket",
+        "brand": "Lee Jeans",
+        "product_id": "123456",
+        "color": "Blue"
+      },
+      {
+        "id": 4,
+        "description": "T-Shirt",
+        "brand": "Nike",
+        "product_id": "789012",
+        "color": "Red"
+      },
+      {
+        "id": 5,
+        "description": "T-Shirt",
+        "brand": "Nike",
+        "product_id": "789012",
+        "color": "Blue"
+      },
+      {
+        "id": 6,
+        "description": "Running Shoes",
+        "brand": "Adidas",
+        "product_id": "456789",
+        "color": "Black"
+      },
+      {
+        "id": 7,
+        "description": "Running Shoes",
+        "brand": "Adidas",
+        "product_id": "456789",
+        "color": "White"
+      },
+      {
+        "id": 8,
+        "description": "Hoodie",
+        "brand": "Puma",
+        "product_id": "987654",
+        "color": "Gray"
+      },
+      {
+        "id": 9,
+        "description": "Sweater",
+        "brand": "Gap",
+        "product_id": "234567",
+        "color": "Green"
+      },
+      {
+        "id": 10,
+        "description": "Sweater",
+        "brand": "Gap",
+        "product_id": "234567",
+        "color": "Red"
+      },
+      {
+        "id": 11,
+        "description": "Sweater",
+        "brand": "Gap",
+        "product_id": "234567",
+        "color": "Blue"
+      },
+      {
+        "id": 12,
+        "description": "Jeans",
+        "brand": "Levi's",
+        "product_id": "345678",
+        "color": "Indigo"
+      },
+      {
+        "id": 13,
+        "description": "Jeans",
+        "brand": "Levi's",
+        "product_id": "345678",
+        "color": "Black"
+      },
+      {
+        "id": 14,
+        "description": "Jeans",
+        "brand": "Levi's",
+        "product_id": "345678",
+        "color": "Stone Wash"
+      }
     ])
 });
 
-pub(self) static DOCUMENT_PRIMARY_KEY: &str = "productId";
-pub(self) static DOCUMENT_DISTINCT_KEY: &str = "shopId";
+static DOCUMENT_PRIMARY_KEY: &str = "id";
+static DOCUMENT_DISTINCT_KEY: &str = "product_id";
 
 /// testing: https://github.com/meilisearch/meilisearch/issues/4078
 #[actix_rt::test]
@@ -33,31 +121,121 @@ async fn distinct_search_with_offset_no_ranking() {
     index.update_distinct_attribute(json!(DOCUMENT_DISTINCT_KEY)).await;
     index.wait_task(1).await;
 
-    fn get_hits(Value(response): Value) -> Vec<i64> {
+    fn get_hits(response: &Value) -> Vec<&str> {
         let hits_array = response["hits"].as_array().unwrap();
-        hits_array.iter().map(|h| h[DOCUMENT_DISTINCT_KEY].as_i64().unwrap()).collect::<Vec<_>>()
+        hits_array.iter().map(|h| h[DOCUMENT_DISTINCT_KEY].as_str().unwrap()).collect::<Vec<_>>()
     }
 
-    let (response, code) = index.search_post(json!({"limit": 2, "offset": 0})).await;
-    let hits = get_hits(response);
+    let (response, code) = index.search_post(json!({"offset": 0, "limit": 2})).await;
+    let hits = get_hits(&response);
     snapshot!(code, @"200 OK");
     snapshot!(hits.len(), @"2");
-    snapshot!(format!("{:?}", hits), @"[1, 2]");
+    snapshot!(format!("{:?}", hits), @r#"["123456", "789012"]"#);
+    snapshot!(response["estimatedTotalHits"] , @"11");
 
-    let (response, code) = index.search_post(json!({"limit": 2, "offset": 2})).await;
-    let hits = get_hits(response);
+    let (response, code) = index.search_post(json!({"offset": 2, "limit": 2})).await;
+    let hits = get_hits(&response);
     snapshot!(code, @"200 OK");
     snapshot!(hits.len(), @"2");
-    snapshot!(format!("{:?}", hits), @"[3, 4]");
+    snapshot!(format!("{:?}", hits), @r#"["456789", "987654"]"#);
+    snapshot!(response["estimatedTotalHits"], @"10");
 
-    let (response, code) = index.search_post(json!({"limit": 10, "offset": 4})).await;
-    let hits = get_hits(response);
+    let (response, code) = index.search_post(json!({"offset": 4, "limit": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"2");
+    snapshot!(format!("{:?}", hits), @r#"["234567", "345678"]"#);
+    snapshot!(response["estimatedTotalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"offset": 5, "limit": 2})).await;
+    let hits = get_hits(&response);
     snapshot!(code, @"200 OK");
     snapshot!(hits.len(), @"1");
-    snapshot!(format!("{:?}", hits), @"[5]");
+    snapshot!(format!("{:?}", hits), @r#"["345678"]"#);
+    snapshot!(response["estimatedTotalHits"], @"6");
 
-    let (response, code) = index.search_post(json!({"limit": 10, "offset": 5})).await;
-    let hits = get_hits(response);
+    let (response, code) = index.search_post(json!({"offset": 6, "limit": 2})).await;
+    let hits = get_hits(&response);
     snapshot!(code, @"200 OK");
     snapshot!(hits.len(), @"0");
+    snapshot!(format!("{:?}", hits), @r#"[]"#);
+    snapshot!(response["estimatedTotalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"offset": 7, "limit": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"0");
+    snapshot!(format!("{:?}", hits), @r#"[]"#);
+    snapshot!(response["estimatedTotalHits"], @"6");
+}
+
+/// testing: https://github.com/meilisearch/meilisearch/issues/4130
+#[actix_rt::test]
+async fn distinct_search_with_pagination_no_ranking() {
+    let server = Server::new().await;
+    let index = server.index("test");
+
+    let documents = DOCUMENTS.clone();
+    index.add_documents(documents, Some(DOCUMENT_PRIMARY_KEY)).await;
+    index.update_distinct_attribute(json!(DOCUMENT_DISTINCT_KEY)).await;
+    index.wait_task(1).await;
+
+    fn get_hits(response: &Value) -> Vec<&str> {
+        let hits_array = response["hits"].as_array().unwrap();
+        hits_array.iter().map(|h| h[DOCUMENT_DISTINCT_KEY].as_str().unwrap()).collect::<Vec<_>>()
+    }
+
+    let (response, code) = index.search_post(json!({"page": 0, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"0");
+    snapshot!(format!("{:?}", hits), @r#"[]"#);
+    snapshot!(response["page"], @"0");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 1, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"2");
+    snapshot!(format!("{:?}", hits), @r#"["123456", "789012"]"#);
+    snapshot!(response["page"], @"1");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 2, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"2");
+    snapshot!(format!("{:?}", hits), @r#"["456789", "987654"]"#);
+    snapshot!(response["page"], @"2");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 3, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"2");
+    snapshot!(format!("{:?}", hits), @r#"["234567", "345678"]"#);
+    snapshot!(response["page"], @"3");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 4, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"0");
+    snapshot!(format!("{:?}", hits), @r#"[]"#);
+    snapshot!(response["page"], @"4");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 2, "hitsPerPage": 3})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"3");
+    snapshot!(format!("{:?}", hits), @r#"["987654", "234567", "345678"]"#);
+    snapshot!(response["page"], @"2");
+    snapshot!(response["totalPages"], @"2");
+    snapshot!(response["totalHits"], @"6");
 }

meilisearch/tests/search/facet_search.rs

@@ -4,7 +4,7 @@ use once_cell::sync::Lazy;
 use crate::common::{Server, Value};
 use crate::json;
 
-pub(self) static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
         {
             "title": "Shazam!",
@@ -105,6 +105,24 @@ async fn more_advanced_facet_search() {
     snapshot!(response["facetHits"].as_array().unwrap().len(), @"1");
 }
 
+#[actix_rt::test]
+async fn simple_facet_search_with_max_values() {
+    let server = Server::new().await;
+    let index = server.index("test");
+
+    let documents = DOCUMENTS.clone();
+    index.update_settings_faceting(json!({ "maxValuesPerFacet": 1 })).await;
+    index.update_settings_filterable_attributes(json!(["genres"])).await;
+    index.add_documents(documents, None).await;
+    index.wait_task(2).await;
+
+    let (response, code) =
+        index.facet_search(json!({"facetName": "genres", "facetQuery": "a"})).await;
+
+    assert_eq!(code, 200, "{}", response);
+    assert_eq!(dbg!(response)["facetHits"].as_array().unwrap().len(), 1);
+}
+
 #[actix_rt::test]
 async fn non_filterable_facet_search_error() {
     let server = Server::new().await;

meilisearch/tests/search/geo.rs

@@ -4,7 +4,7 @@ use once_cell::sync::Lazy;
 use crate::common::{Server, Value};
 use crate::json;
 
-pub(self) static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
         {
             "id": 1,

meilisearch/tests/search/hybrid.rs

@@ -0,0 +1,221 @@
+use meili_snap::snapshot;
+use once_cell::sync::Lazy;
+
+use crate::common::index::Index;
+use crate::common::{Server, Value};
+use crate::json;
+
+async fn index_with_documents<'a>(server: &'a Server, documents: &Value) -> Index<'a> {
+    let index = server.index("test");
+
+    let (response, code) = server.set_features(json!({"vectorStore": true})).await;
+
+    meili_snap::snapshot!(code, @"200 OK");
+    meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
+    {
+      "scoreDetails": false,
+      "vectorStore": true,
+      "metrics": false,
+      "exportPuffinReports": false
+    }
+    "###);
+
+    let (response, code) = index
+        .update_settings(json!({ "embedders": {"default": {
+                "source": "userProvided",
+                "dimensions": 2}}} ))
+        .await;
+    assert_eq!(202, code, "{:?}", response);
+    index.wait_task(response.uid()).await;
+
+    let (response, code) = index.add_documents(documents.clone(), None).await;
+    assert_eq!(202, code, "{:?}", response);
+    index.wait_task(response.uid()).await;
+    index
+}
+
+static SIMPLE_SEARCH_DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+    json!([
+    {
+        "title": "Shazam!",
+        "desc": "a Captain Marvel ersatz",
+        "id": "1",
+        "_vectors": {"default": [1.0, 3.0]},
+    },
+    {
+        "title": "Captain Planet",
+        "desc": "He's not part of the Marvel Cinematic Universe",
+        "id": "2",
+        "_vectors": {"default": [1.0, 2.0]},
+    },
+    {
+        "title": "Captain Marvel",
+        "desc": "a Shazam ersatz",
+        "id": "3",
+        "_vectors": {"default": [2.0, 3.0]},
+    }])
+});
+
+static SINGLE_DOCUMENT: Lazy<Value> = Lazy::new(|| {
+    json!([{
+            "title": "Shazam!",
+            "desc": "a Captain Marvel ersatz",
+            "id": "1",
+            "_vectors": {"default": [1.0, 3.0]},
+    }])
+});
+
+#[actix_rt::test]
+async fn simple_search() {
+    let server = Server::new().await;
+    let index = index_with_documents(&server, &SIMPLE_SEARCH_DOCUMENTS).await;
+
+    let (response, code) = index
+        .search_post(
+            json!({"q": "Captain", "vector": [1.0, 1.0], "hybrid": {"semanticRatio": 0.2}}),
+        )
+        .await;
+    snapshot!(code, @"200 OK");
+    snapshot!(response["hits"], @r###"[{"title":"Captain Planet","desc":"He's not part of the Marvel Cinematic Universe","id":"2","_vectors":{"default":[1.0,2.0]}},{"title":"Captain Marvel","desc":"a Shazam ersatz","id":"3","_vectors":{"default":[2.0,3.0]}},{"title":"Shazam!","desc":"a Captain Marvel ersatz","id":"1","_vectors":{"default":[1.0,3.0]}}]"###);
+
+    let (response, code) = index
+        .search_post(
+            json!({"q": "Captain", "vector": [1.0, 1.0], "hybrid": {"semanticRatio": 0.8}}),
+        )
+        .await;
+    snapshot!(code, @"200 OK");
+    snapshot!(response["hits"], @r###"[{"title":"Captain Marvel","desc":"a Shazam ersatz","id":"3","_vectors":{"default":[2.0,3.0]},"_semanticScore":0.99029034},{"title":"Captain Planet","desc":"He's not part of the Marvel Cinematic Universe","id":"2","_vectors":{"default":[1.0,2.0]},"_semanticScore":0.97434163},{"title":"Shazam!","desc":"a Captain Marvel ersatz","id":"1","_vectors":{"default":[1.0,3.0]},"_semanticScore":0.9472136}]"###);
+}
+
+#[actix_rt::test]
+async fn highlighter() {
+    let server = Server::new().await;
+    let index = index_with_documents(&server, &SIMPLE_SEARCH_DOCUMENTS).await;
+
+    let (response, code) = index
+        .search_post(json!({"q": "Captain Marvel", "vector": [1.0, 1.0],
+            "hybrid": {"semanticRatio": 0.2},
+            "attributesToHighlight": [
+                     "desc"
+                   ],
+                   "highlightPreTag": "**BEGIN**",
+                   "highlightPostTag": "**END**"
+        }))
+        .await;
+    snapshot!(code, @"200 OK");
+    snapshot!(response["hits"], @r###"[{"title":"Captain Marvel","desc":"a Shazam ersatz","id":"3","_vectors":{"default":[2.0,3.0]},"_formatted":{"title":"Captain Marvel","desc":"a Shazam ersatz","id":"3","_vectors":{"default":["2.0","3.0"]}}},{"title":"Shazam!","desc":"a Captain Marvel ersatz","id":"1","_vectors":{"default":[1.0,3.0]},"_formatted":{"title":"Shazam!","desc":"a **BEGIN**Captain**END** **BEGIN**Marvel**END** ersatz","id":"1","_vectors":{"default":["1.0","3.0"]}}},{"title":"Captain Planet","desc":"He's not part of the Marvel Cinematic Universe","id":"2","_vectors":{"default":[1.0,2.0]},"_formatted":{"title":"Captain Planet","desc":"He's not part of the **BEGIN**Marvel**END** Cinematic Universe","id":"2","_vectors":{"default":["1.0","2.0"]}}}]"###);
+
+    let (response, code) = index
+        .search_post(json!({"q": "Captain Marvel", "vector": [1.0, 1.0],
+            "hybrid": {"semanticRatio": 0.8},
+            "attributesToHighlight": [
+                     "desc"
+                   ],
+                   "highlightPreTag": "**BEGIN**",
+                   "highlightPostTag": "**END**"
+        }))
+        .await;
+    snapshot!(code, @"200 OK");
+    snapshot!(response["hits"], @r###"[{"title":"Captain Marvel","desc":"a Shazam ersatz","id":"3","_vectors":{"default":[2.0,3.0]},"_formatted":{"title":"Captain Marvel","desc":"a Shazam ersatz","id":"3","_vectors":{"default":["2.0","3.0"]}},"_semanticScore":0.99029034},{"title":"Captain Planet","desc":"He's not part of the Marvel Cinematic Universe","id":"2","_vectors":{"default":[1.0,2.0]},"_formatted":{"title":"Captain Planet","desc":"He's not part of the **BEGIN**Marvel**END** Cinematic Universe","id":"2","_vectors":{"default":["1.0","2.0"]}},"_semanticScore":0.97434163},{"title":"Shazam!","desc":"a Captain Marvel ersatz","id":"1","_vectors":{"default":[1.0,3.0]},"_formatted":{"title":"Shazam!","desc":"a **BEGIN**Captain**END** **BEGIN**Marvel**END** ersatz","id":"1","_vectors":{"default":["1.0","3.0"]}},"_semanticScore":0.9472136}]"###);
+
+    // no highlighting on full semantic
+    let (response, code) = index
+        .search_post(json!({"q": "Captain Marvel", "vector": [1.0, 1.0],
+            "hybrid": {"semanticRatio": 1.0},
+            "attributesToHighlight": [
+                     "desc"
+                   ],
+                   "highlightPreTag": "**BEGIN**",
+                   "highlightPostTag": "**END**"
+        }))
+        .await;
+    snapshot!(code, @"200 OK");
+    snapshot!(response["hits"], @r###"[{"title":"Captain Marvel","desc":"a Shazam ersatz","id":"3","_vectors":{"default":[2.0,3.0]},"_formatted":{"title":"Captain Marvel","desc":"a Shazam ersatz","id":"3","_vectors":{"default":["2.0","3.0"]}},"_semanticScore":0.99029034},{"title":"Captain Planet","desc":"He's not part of the Marvel Cinematic Universe","id":"2","_vectors":{"default":[1.0,2.0]},"_formatted":{"title":"Captain Planet","desc":"He's not part of the Marvel Cinematic Universe","id":"2","_vectors":{"default":["1.0","2.0"]}},"_semanticScore":0.97434163},{"title":"Shazam!","desc":"a Captain Marvel ersatz","id":"1","_vectors":{"default":[1.0,3.0]},"_formatted":{"title":"Shazam!","desc":"a Captain Marvel ersatz","id":"1","_vectors":{"default":["1.0","3.0"]}}}]"###);
+}
+
+#[actix_rt::test]
+async fn invalid_semantic_ratio() {
+    let server = Server::new().await;
+    let index = index_with_documents(&server, &SIMPLE_SEARCH_DOCUMENTS).await;
+
+    let (response, code) = index
+        .search_post(
+            json!({"q": "Captain", "vector": [1.0, 1.0], "hybrid": {"semanticRatio": 1.2}}),
+        )
+        .await;
+    snapshot!(code, @"400 Bad Request");
+    snapshot!(response, @r###"
+    {
+      "message": "Invalid value at `.hybrid.semanticRatio`: the value of `semanticRatio` is invalid, expected a float between `0.0` and `1.0`.",
+      "code": "invalid_search_semantic_ratio",
+      "type": "invalid_request",
+      "link": "https://docs.meilisearch.com/errors#invalid_search_semantic_ratio"
+    }
+    "###);
+
+    let (response, code) = index
+        .search_post(
+            json!({"q": "Captain", "vector": [1.0, 1.0], "hybrid": {"semanticRatio": -0.8}}),
+        )
+        .await;
+    snapshot!(code, @"400 Bad Request");
+    snapshot!(response, @r###"
+    {
+      "message": "Invalid value at `.hybrid.semanticRatio`: the value of `semanticRatio` is invalid, expected a float between `0.0` and `1.0`.",
+      "code": "invalid_search_semantic_ratio",
+      "type": "invalid_request",
+      "link": "https://docs.meilisearch.com/errors#invalid_search_semantic_ratio"
+    }
+    "###);
+
+    let (response, code) = index
+        .search_get(
+            &yaup::to_string(
+                &json!({"q": "Captain", "vector": [1.0, 1.0], "hybridSemanticRatio": 1.2}),
+            )
+            .unwrap(),
+        )
+        .await;
+    snapshot!(code, @"400 Bad Request");
+    snapshot!(response, @r###"
+    {
+      "message": "Invalid value in parameter `hybridSemanticRatio`: the value of `semanticRatio` is invalid, expected a float between `0.0` and `1.0`.",
+      "code": "invalid_search_semantic_ratio",
+      "type": "invalid_request",
+      "link": "https://docs.meilisearch.com/errors#invalid_search_semantic_ratio"
+    }
+    "###);
+
+    let (response, code) = index
+        .search_get(
+            &yaup::to_string(
+                &json!({"q": "Captain", "vector": [1.0, 1.0], "hybridSemanticRatio": -0.2}),
+            )
+            .unwrap(),
+        )
+        .await;
+    snapshot!(code, @"400 Bad Request");
+    snapshot!(response, @r###"
+    {
+      "message": "Invalid value in parameter `hybridSemanticRatio`: the value of `semanticRatio` is invalid, expected a float between `0.0` and `1.0`.",
+      "code": "invalid_search_semantic_ratio",
+      "type": "invalid_request",
+      "link": "https://docs.meilisearch.com/errors#invalid_search_semantic_ratio"
+    }
+    "###);
+}
+
+#[actix_rt::test]
+async fn single_document() {
+    let server = Server::new().await;
+    let index = index_with_documents(&server, &SINGLE_DOCUMENT).await;
+
+    let (response, code) = index
+    .search_post(
+        json!({"vector": [1.0, 3.0], "hybrid": {"semanticRatio": 1.0}, "showRankingScore": true}),
+    )
+    .await;
+
+    snapshot!(code, @"200 OK");
+    snapshot!(response["hits"][0], @r###"{"title":"Shazam!","desc":"a Captain Marvel ersatz","id":"1","_vectors":{"default":[1.0,3.0]},"_rankingScore":1.0,"_semanticScore":1.0}"###);
+}

meilisearch/tests/search/mod.rs

@@ -6,6 +6,7 @@ mod errors;
 mod facet_search;
 mod formatted;
 mod geo;
+mod hybrid;
 mod multi;
 mod pagination;
 mod restrict_searchable;
@@ -15,32 +16,37 @@ use once_cell::sync::Lazy;
 use crate::common::{Server, Value};
 use crate::json;
 
-pub(self) static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
         {
             "title": "Shazam!",
             "id": "287947",
+            "_vectors": { "manual": [1, 2, 3]},
         },
         {
             "title": "Captain Marvel",
             "id": "299537",
+            "_vectors": { "manual": [1, 2, 54] },
         },
         {
             "title": "Escape Room",
             "id": "522681",
+            "_vectors": { "manual": [10, -23, 32] },
         },
         {
             "title": "How to Train Your Dragon: The Hidden World",
             "id": "166428",
+            "_vectors": { "manual": [-100, 231, 32] },
         },
         {
             "title": "Gläss",
             "id": "450465",
+            "_vectors": { "manual": [-100, 340, 90] },
         }
     ])
 });
 
-pub(self) static NESTED_DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static NESTED_DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
         {
             "id": 852,
@@ -57,6 +63,7 @@ pub(self) static NESTED_DOCUMENTS: Lazy<Value> = Lazy::new(|| {
                 },
             ],
             "cattos": "pésti",
+            "_vectors": { "manual": [1, 2, 3]},
         },
         {
             "id": 654,
@@ -69,12 +76,14 @@ pub(self) static NESTED_DOCUMENTS: Lazy<Value> = Lazy::new(|| {
                 },
             ],
             "cattos": ["simba", "pestiféré"],
+            "_vectors": { "manual": [1, 2, 54] },
         },
         {
             "id": 750,
             "father": "romain",
             "mother": "michelle",
             "cattos": ["enigma"],
+            "_vectors": { "manual": [10, 23, 32] },
         },
         {
             "id": 951,
@@ -91,6 +100,7 @@ pub(self) static NESTED_DOCUMENTS: Lazy<Value> = Lazy::new(|| {
                 },
             ],
             "cattos": ["moumoute", "gomez"],
+            "_vectors": { "manual": [10, 23, 32] },
         },
     ])
 });
@@ -802,6 +812,13 @@ async fn experimental_feature_score_details() {
                   {
                     "title": "How to Train Your Dragon: The Hidden World",
                     "id": "166428",
+                    "_vectors": {
+                      "manual": [
+                        -100,
+                        231,
+                        32
+                      ]
+                    },
                     "_rankingScoreDetails": {
                       "words": {
                         "order": 0,
@@ -823,7 +840,7 @@ async fn experimental_feature_score_details() {
                         "order": 3,
                         "attributeRankingOrderScore": 1.0,
                         "queryWordDistanceScore": 0.8095238095238095,
-                        "score": 0.9365079365079364
+                        "score": 0.9727891156462584
                       },
                       "exactness": {
                         "order": 4,
@@ -870,13 +887,100 @@ async fn experimental_feature_vector_store() {
     meili_snap::snapshot!(code, @"200 OK");
     meili_snap::snapshot!(response["vectorStore"], @"true");
 
+    let (response, code) = index
+        .update_settings(json!({"embedders": {
+            "manual": {
+                "source": "userProvided",
+                "dimensions": 3,
+            }
+        }}))
+        .await;
+
+    meili_snap::snapshot!(response, @r###"
+    {
+      "taskUid": 1,
+      "indexUid": "test",
+      "status": "enqueued",
+      "type": "settingsUpdate",
+      "enqueuedAt": "[date]"
+    }
+    "###);
+    meili_snap::snapshot!(code, @"202 Accepted");
+    let response = index.wait_task(response.uid()).await;
+
+    meili_snap::snapshot!(meili_snap::json_string!(response["status"]), @"\"succeeded\"");
+
     let (response, code) = index
         .search_post(json!({
             "vector": [1.0, 2.0, 3.0],
         }))
         .await;
+
     meili_snap::snapshot!(code, @"200 OK");
-    meili_snap::snapshot!(meili_snap::json_string!(response["hits"]), @"[]");
+    // vector search returns all documents that don't have vectors in the last bucket, like all sorts
+    meili_snap::snapshot!(meili_snap::json_string!(response["hits"]), @r###"
+    [
+      {
+        "title": "Shazam!",
+        "id": "287947",
+        "_vectors": {
+          "manual": [
+            1,
+            2,
+            3
+          ]
+        },
+        "_semanticScore": 1.0
+      },
+      {
+        "title": "Captain Marvel",
+        "id": "299537",
+        "_vectors": {
+          "manual": [
+            1,
+            2,
+            54
+          ]
+        },
+        "_semanticScore": 0.9129112
+      },
+      {
+        "title": "Gläss",
+        "id": "450465",
+        "_vectors": {
+          "manual": [
+            -100,
+            340,
+            90
+          ]
+        },
+        "_semanticScore": 0.8106413
+      },
+      {
+        "title": "How to Train Your Dragon: The Hidden World",
+        "id": "166428",
+        "_vectors": {
+          "manual": [
+            -100,
+            231,
+            32
+          ]
+        },
+        "_semanticScore": 0.74120104
+      },
+      {
+        "title": "Escape Room",
+        "id": "522681",
+        "_vectors": {
+          "manual": [
+            10,
+            -23,
+            32
+          ]
+        }
+      }
+    ]
+    "###);
 }
 
 #[cfg(feature = "default")]
@@ -1126,7 +1230,14 @@ async fn simple_search_with_strange_synonyms() {
             [
               {
                 "title": "How to Train Your Dragon: The Hidden World",
-                "id": "166428"
+                "id": "166428",
+                "_vectors": {
+                  "manual": [
+                    -100,
+                    231,
+                    32
+                  ]
+                }
               }
             ]
             "###);
@@ -1140,7 +1251,14 @@ async fn simple_search_with_strange_synonyms() {
             [
               {
                 "title": "How to Train Your Dragon: The Hidden World",
-                "id": "166428"
+                "id": "166428",
+                "_vectors": {
+                  "manual": [
+                    -100,
+                    231,
+                    32
+                  ]
+                }
               }
             ]
             "###);
@@ -1154,7 +1272,14 @@ async fn simple_search_with_strange_synonyms() {
             [
               {
                 "title": "How to Train Your Dragon: The Hidden World",
-                "id": "166428"
+                "id": "166428",
+                "_vectors": {
+                  "manual": [
+                    -100,
+                    231,
+                    32
+                  ]
+                }
               }
             ]
             "###);

meilisearch/tests/search/multi.rs

@@ -72,7 +72,14 @@ async fn simple_search_single_index() {
         "hits": [
           {
             "title": "Gläss",
-            "id": "450465"
+            "id": "450465",
+            "_vectors": {
+              "manual": [
+                -100,
+                340,
+                90
+              ]
+            }
           }
         ],
         "query": "glass",
@@ -86,7 +93,14 @@ async fn simple_search_single_index() {
         "hits": [
           {
             "title": "Captain Marvel",
-            "id": "299537"
+            "id": "299537",
+            "_vectors": {
+              "manual": [
+                1,
+                2,
+                54
+              ]
+            }
           }
         ],
         "query": "captain",
@@ -177,7 +191,14 @@ async fn simple_search_two_indexes() {
         "hits": [
           {
             "title": "Gläss",
-            "id": "450465"
+            "id": "450465",
+            "_vectors": {
+              "manual": [
+                -100,
+                340,
+                90
+              ]
+            }
           }
         ],
         "query": "glass",
@@ -203,7 +224,14 @@ async fn simple_search_two_indexes() {
                 "age": 4
               }
             ],
-            "cattos": "pésti"
+            "cattos": "pésti",
+            "_vectors": {
+              "manual": [
+                1,
+                2,
+                3
+              ]
+            }
           },
           {
             "id": 654,
@@ -218,7 +246,14 @@ async fn simple_search_two_indexes() {
             "cattos": [
               "simba",
               "pestiféré"
-            ]
+            ],
+            "_vectors": {
+              "manual": [
+                1,
+                2,
+                54
+              ]
+            }
           }
         ],
         "query": "pésti",

meilisearch/tests/search/restrict_searchable.rs

@@ -335,3 +335,35 @@ async fn exactness_ranking_rule_order() {
         })
         .await;
 }
+
+#[actix_rt::test]
+async fn search_on_exact_field() {
+    let server = Server::new().await;
+    let index = index_with_documents(
+        &server,
+        &json!([
+        {
+            "title": "Captain Marvel",
+            "exact": "Captain Marivel",
+            "id": "1",
+        },
+        {
+            "title": "Captain Marivel",
+            "exact": "Captain the Marvel",
+            "id": "2",
+        }]),
+    )
+    .await;
+
+    let (response, code) =
+        index.update_settings_typo_tolerance(json!({ "disableOnAttributes": ["exact"] })).await;
+    assert_eq!(202, code, "{:?}", response);
+    index.wait_task(1).await;
+    // Searching on an exact attribute should only return the document matching without typo.
+    index
+        .search(json!({"q": "Marvel", "attributesToSearchOn": ["exact"]}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(response["hits"].as_array().unwrap().len(), @"1");
+        })
+        .await;
+}

meilisearch/tests/settings/get_settings.rs

@@ -54,7 +54,7 @@ async fn get_settings() {
     let (response, code) = index.settings().await;
     assert_eq!(code, 200);
     let settings = response.as_object().unwrap();
-    assert_eq!(settings.keys().len(), 14);
+    assert_eq!(settings.keys().len(), 15);
     assert_eq!(settings["displayedAttributes"], json!(["*"]));
     assert_eq!(settings["searchableAttributes"], json!(["*"]));
     assert_eq!(settings["filterableAttributes"], json!([]));
@@ -83,6 +83,7 @@ async fn get_settings() {
             "maxTotalHits": 1000,
         })
     );
+    assert_eq!(settings["proximityPrecision"], json!("byWord"));
 }
 
 #[actix_rt::test]

meilisearch/tests/settings/mod.rs

@@ -1,4 +1,5 @@
 mod distinct;
 mod errors;
 mod get_settings;
+mod proximity_settings;
 mod tokenizer_customization;

meilisearch/tests/settings/proximity_settings.rs

@@ -0,0 +1,352 @@
+use meili_snap::{json_string, snapshot};
+use once_cell::sync::Lazy;
+
+use crate::common::Server;
+use crate::json;
+
+static DOCUMENTS: Lazy<crate::common::Value> = Lazy::new(|| {
+    json!([
+        {
+            "id": 1,
+            "a": "Soup of the day",
+            "b": "many the fish",
+        },
+        {
+            "id": 2,
+            "a": "Soup of day",
+            "b": "many the lazy fish",
+        },
+        {
+            "id": 3,
+            "a": "the Soup of day",
+            "b": "many the fish",
+        },
+    ])
+});
+
+#[actix_rt::test]
+async fn attribute_scale_search() {
+    let server = Server::new().await;
+    let index = server.index("test");
+
+    index.add_documents(DOCUMENTS.clone(), None).await;
+    index.wait_task(0).await;
+
+    let (response, code) = index
+        .update_settings(json!({
+            "proximityPrecision": "byAttribute",
+            "rankingRules": ["words", "typo", "proximity"],
+        }))
+        .await;
+    assert_eq!("202", code.as_str(), "{:?}", response);
+    index.wait_task(1).await;
+
+    // the expected order is [1, 3, 2] instead of [3, 1, 2]
+    // because the attribute scale doesn't make the difference between 1 and 3.
+    index
+        .search(json!({"q": "the soup of day"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // the expected order is [1, 2, 3] instead of [1, 3, 2]
+    // because the attribute scale sees all the word in the same attribute
+    // and so doesn't make the difference between the documents.
+    index
+        .search(json!({"q": "many the fish"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}
+
+#[actix_rt::test]
+async fn attribute_scale_phrase_search() {
+    let server = Server::new().await;
+    let index = server.index("test");
+
+    index.add_documents(DOCUMENTS.clone(), None).await;
+    index.wait_task(0).await;
+
+    let (_response, _code) = index
+        .update_settings(json!({
+            "proximityPrecision": "byAttribute",
+            "rankingRules": ["words", "typo", "proximity"],
+        }))
+        .await;
+    index.wait_task(1).await;
+
+    // the expected order is [1, 3] instead of [3, 1]
+    // because the attribute scale doesn't make the difference between 1 and 3.
+    // But 2 shouldn't be returned because "the" is not in the same attribute.
+    index
+        .search(json!({"q": "\"the soup of day\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // the expected order is [1, 2, 3] instead of [1, 3]
+    // because the attribute scale sees all the word in the same attribute
+    // and so doesn't make the difference between the documents.
+    index
+        .search(json!({"q": "\"many the fish\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}
+
+#[actix_rt::test]
+async fn word_scale_set_and_reset() {
+    let server = Server::new().await;
+    let index = server.index("test");
+
+    index.add_documents(DOCUMENTS.clone(), None).await;
+    index.wait_task(0).await;
+
+    // Set and reset the setting ensuring the swap between the 2 settings is applied.
+    let (_response, _code) = index
+        .update_settings(json!({
+            "proximityPrecision": "byAttribute",
+            "rankingRules": ["words", "typo", "proximity"],
+        }))
+        .await;
+    index.wait_task(1).await;
+
+    let (_response, _code) = index
+        .update_settings(json!({
+            "proximityPrecision": "byWord",
+            "rankingRules": ["words", "typo", "proximity"],
+        }))
+        .await;
+    index.wait_task(2).await;
+
+    // [3, 1, 2]
+    index
+        .search(json!({"q": "the soup of day"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // [1, 3, 2]
+    index
+        .search(json!({"q": "many the fish"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // [3]
+    index
+        .search(json!({"q": "\"the soup of day\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // [1, 3]
+    index
+        .search(json!({"q": "\"many the fish\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}
+
+#[actix_rt::test]
+async fn attribute_scale_default_ranking_rules() {
+    let server = Server::new().await;
+    let index = server.index("test");
+
+    index.add_documents(DOCUMENTS.clone(), None).await;
+    index.wait_task(0).await;
+
+    let (response, code) = index
+        .update_settings(json!({
+            "proximityPrecision": "byAttribute"
+        }))
+        .await;
+    assert_eq!("202", code.as_str(), "{:?}", response);
+    index.wait_task(1).await;
+
+    // the expected order is [3, 1, 2]
+    index
+        .search(json!({"q": "the soup of day"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // the expected order is [1, 3, 2] instead of [1, 3]
+    // because the attribute scale sees all the word in the same attribute
+    // and so doesn't remove the document 2.
+    index
+        .search(json!({"q": "\"many the fish\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}

meilisearch/tests/tasks/mod.rs

@@ -1,4 +1,5 @@
 mod errors;
+mod webhook;
 
 use meili_snap::insta::assert_json_snapshot;
 use time::format_description::well_known::Rfc3339;

meilisearch/tests/tasks/webhook.rs

@@ -0,0 +1,123 @@
+//! To test the webhook, we need to spawn a new server with a URL listening for
+//! post requests. The webhook handle starts a server and forwards all the
+//! received requests into a channel for you to handle.
+
+use std::sync::Arc;
+
+use actix_http::body::MessageBody;
+use actix_web::dev::{ServiceFactory, ServiceResponse};
+use actix_web::web::{Bytes, Data};
+use actix_web::{post, App, HttpResponse, HttpServer};
+use meili_snap::{json_string, snapshot};
+use meilisearch::Opt;
+use tokio::sync::mpsc;
+use url::Url;
+
+use crate::common::{default_settings, Server};
+use crate::json;
+
+#[post("/")]
+async fn forward_body(sender: Data<mpsc::UnboundedSender<Vec<u8>>>, body: Bytes) -> HttpResponse {
+    let body = body.to_vec();
+    sender.send(body).unwrap();
+    HttpResponse::Ok().into()
+}
+
+fn create_app(
+    sender: Arc<mpsc::UnboundedSender<Vec<u8>>>,
+) -> actix_web::App<
+    impl ServiceFactory<
+        actix_web::dev::ServiceRequest,
+        Config = (),
+        Response = ServiceResponse<impl MessageBody>,
+        Error = actix_web::Error,
+        InitError = (),
+    >,
+> {
+    App::new().service(forward_body).app_data(Data::from(sender))
+}
+
+struct WebhookHandle {
+    pub server_handle: tokio::task::JoinHandle<Result<(), std::io::Error>>,
+    pub url: String,
+    pub receiver: mpsc::UnboundedReceiver<Vec<u8>>,
+}
+
+async fn create_webhook_server() -> WebhookHandle {
+    let mut log_builder = env_logger::Builder::new();
+    log_builder.parse_filters("info");
+    log_builder.init();
+
+    let (sender, receiver) = mpsc::unbounded_channel();
+    let sender = Arc::new(sender);
+
+    // By listening on the port 0, the system will give us any available port.
+    let server =
+        HttpServer::new(move || create_app(sender.clone())).bind(("127.0.0.1", 0)).unwrap();
+    let (ip, scheme) = server.addrs_with_scheme()[0];
+    let url = format!("{scheme}://{ip}/");
+
+    let server_handle = tokio::spawn(server.run());
+    WebhookHandle { server_handle, url, receiver }
+}
+
+#[actix_web::test]
+async fn test_basic_webhook() {
+    let WebhookHandle { server_handle, url, mut receiver } = create_webhook_server().await;
+
+    let db_path = tempfile::tempdir().unwrap();
+    let server = Server::new_with_options(Opt {
+        task_webhook_url: Some(Url::parse(&url).unwrap()),
+        ..default_settings(db_path.path())
+    })
+    .await
+    .unwrap();
+
+    let index = server.index("tamo");
+    // May be flaky: we're relying on the fact that while the first document addition is processed, the other
+    // operations will be received and will be batched together. If it doesn't happen it's not a problem
+    // the rest of the test won't assume anything about the number of tasks per batch.
+    for i in 0..5 {
+        let (_, _status) = index.add_documents(json!({ "id": i, "doggo": "bone" }), None).await;
+    }
+
+    let mut nb_tasks = 0;
+    while let Some(payload) = receiver.recv().await {
+        let payload = String::from_utf8(payload).unwrap();
+        let jsonl = payload.split('\n');
+        for json in jsonl {
+            if json.is_empty() {
+                break; // we reached EOF
+            }
+            nb_tasks += 1;
+            let json: serde_json::Value = serde_json::from_str(json).unwrap();
+            snapshot!(
+                json_string!(json, { ".uid" => "[uid]", ".duration" => "[duration]", ".enqueuedAt" => "[date]", ".startedAt" => "[date]", ".finishedAt" => "[date]" }),
+            @r###"
+            {
+              "uid": "[uid]",
+              "indexUid": "tamo",
+              "status": "succeeded",
+              "type": "documentAdditionOrUpdate",
+              "canceledBy": null,
+              "details": {
+                "receivedDocuments": 1,
+                "indexedDocuments": 1
+              },
+              "error": null,
+              "duration": "[duration]",
+              "enqueuedAt": "[date]",
+              "startedAt": "[date]",
+              "finishedAt": "[date]"
+            }
+            "###);
+        }
+        if nb_tasks == 5 {
+            break;
+        }
+    }
+
+    assert!(nb_tasks == 5, "We should have received the 5 tasks but only received {nb_tasks}");
+
+    server_handle.abort();
+}

meilitool/Cargo.toml

@@ -0,0 +1,19 @@
+[package]
+name = "meilitool"
+description = "A CLI to edit a Meilisearch database from the command line"
+version.workspace = true
+authors.workspace = true
+homepage.workspace = true
+readme.workspace = true
+edition.workspace = true
+license.workspace = true
+
+[dependencies]
+anyhow = "1.0.75"
+clap = { version = "4.2.1", features = ["derive"] }
+dump = { path = "../dump" }
+file-store = { path = "../file-store" }
+meilisearch-auth = { path = "../meilisearch-auth" }
+meilisearch-types = { path = "../meilisearch-types" }
+time = { version = "0.3.30", features = ["formatting"] }
+uuid = { version = "1.5.0", features = ["v4"], default-features = false }

meilitool/src/main.rs

@@ -0,0 +1,314 @@
+use std::fs::{read_dir, read_to_string, remove_file, File};
+use std::io::BufWriter;
+use std::path::PathBuf;
+
+use anyhow::Context;
+use clap::{Parser, Subcommand};
+use dump::{DumpWriter, IndexMetadata};
+use file_store::FileStore;
+use meilisearch_auth::AuthController;
+use meilisearch_types::heed::types::{SerdeJson, Str};
+use meilisearch_types::heed::{Database, Env, EnvOpenOptions, RoTxn, RwTxn, Unspecified};
+use meilisearch_types::milli::documents::{obkv_to_object, DocumentsBatchReader};
+use meilisearch_types::milli::{obkv_to_json, BEU32};
+use meilisearch_types::tasks::{Status, Task};
+use meilisearch_types::versioning::check_version_file;
+use meilisearch_types::Index;
+use time::macros::format_description;
+use time::OffsetDateTime;
+use uuid_codec::UuidCodec;
+
+mod uuid_codec;
+
+#[derive(Parser)]
+#[command(author, version, about, long_about = None)]
+struct Cli {
+    /// The database path where the Meilisearch is running.
+    #[arg(long, default_value = "data.ms/")]
+    db_path: PathBuf,
+
+    #[command(subcommand)]
+    command: Command,
+}
+
+#[derive(Subcommand)]
+enum Command {
+    /// Clears the task queue and make it empty.
+    ///
+    /// This command can be safely executed even if Meilisearch is running and processing tasks.
+    /// Once the task queue is empty you can restart Meilisearch and no more tasks must be visible,
+    /// even the ones that were processing. However, it's highly possible that you see the processing
+    /// tasks in the queue again with an associated internal error message.
+    ClearTaskQueue,
+
+    /// Exports a dump from the Meilisearch database.
+    ///
+    /// Make sure to run this command when Meilisearch is not running or running but not processing tasks.
+    /// If tasks are being processed while a dump is being exported there are chances for the dump to be
+    /// malformed with missing tasks.
+    ///
+    /// TODO Verify this claim or make sure it cannot happen and we can export dumps
+    ///      without caring about killing Meilisearch first!
+    ExportADump {
+        /// The directory in which the dump will be created.
+        #[arg(long, default_value = "dumps/")]
+        dump_dir: PathBuf,
+
+        /// Skip dumping the enqueued or processing tasks.
+        ///
+        /// Can be useful when there are a lot of them and it is not particularly useful
+        /// to keep them. Note that only the enqueued tasks takes up space so skipping
+        /// the processed ones is not particularly interesting.
+        #[arg(long)]
+        skip_enqueued_tasks: bool,
+    },
+}
+
+fn main() -> anyhow::Result<()> {
+    let Cli { db_path, command } = Cli::parse();
+
+    check_version_file(&db_path).context("While checking the version file")?;
+
+    match command {
+        Command::ClearTaskQueue => clear_task_queue(db_path),
+        Command::ExportADump { dump_dir, skip_enqueued_tasks } => {
+            export_a_dump(db_path, dump_dir, skip_enqueued_tasks)
+        }
+    }
+}
+
+/// Clears the task queue located at `db_path`.
+fn clear_task_queue(db_path: PathBuf) -> anyhow::Result<()> {
+    let path = db_path.join("tasks");
+    let env = EnvOpenOptions::new()
+        .max_dbs(100)
+        .open(&path)
+        .with_context(|| format!("While trying to open {:?}", path.display()))?;
+
+    eprintln!("Deleting tasks from the database...");
+
+    let mut wtxn = env.write_txn()?;
+    let all_tasks = try_opening_poly_database(&env, &wtxn, "all-tasks")?;
+    let total = all_tasks.len(&wtxn)?;
+    let status = try_opening_poly_database(&env, &wtxn, "status")?;
+    let kind = try_opening_poly_database(&env, &wtxn, "kind")?;
+    let index_tasks = try_opening_poly_database(&env, &wtxn, "index-tasks")?;
+    let canceled_by = try_opening_poly_database(&env, &wtxn, "canceled_by")?;
+    let enqueued_at = try_opening_poly_database(&env, &wtxn, "enqueued-at")?;
+    let started_at = try_opening_poly_database(&env, &wtxn, "started-at")?;
+    let finished_at = try_opening_poly_database(&env, &wtxn, "finished-at")?;
+
+    try_clearing_poly_database(&mut wtxn, all_tasks, "all-tasks")?;
+    try_clearing_poly_database(&mut wtxn, status, "status")?;
+    try_clearing_poly_database(&mut wtxn, kind, "kind")?;
+    try_clearing_poly_database(&mut wtxn, index_tasks, "index-tasks")?;
+    try_clearing_poly_database(&mut wtxn, canceled_by, "canceled_by")?;
+    try_clearing_poly_database(&mut wtxn, enqueued_at, "enqueued-at")?;
+    try_clearing_poly_database(&mut wtxn, started_at, "started-at")?;
+    try_clearing_poly_database(&mut wtxn, finished_at, "finished-at")?;
+
+    wtxn.commit().context("While committing the transaction")?;
+
+    eprintln!("Successfully deleted {total} tasks from the tasks database!");
+    eprintln!("Deleting the content files from disk...");
+
+    let mut count = 0usize;
+    let update_files = db_path.join("update_files");
+    let entries = read_dir(&update_files).with_context(|| {
+        format!("While trying to read the content of {:?}", update_files.display())
+    })?;
+    for result in entries {
+        match result {
+            Ok(ent) => match remove_file(ent.path()) {
+                Ok(_) => count += 1,
+                Err(e) => eprintln!("Error while deleting {:?}: {}", ent.path().display(), e),
+            },
+            Err(e) => {
+                eprintln!("Error while reading a file in {:?}: {}", update_files.display(), e)
+            }
+        }
+    }
+
+    eprintln!("Sucessfully deleted {count} content files from disk!");
+
+    Ok(())
+}
+
+fn try_opening_database<KC: 'static, DC: 'static>(
+    env: &Env,
+    rtxn: &RoTxn,
+    db_name: &str,
+) -> anyhow::Result<Database<KC, DC>> {
+    env.open_database(rtxn, Some(db_name))
+        .with_context(|| format!("While opening the {db_name:?} database"))?
+        .with_context(|| format!("Missing the {db_name:?} database"))
+}
+
+fn try_opening_poly_database(
+    env: &Env,
+    rtxn: &RoTxn,
+    db_name: &str,
+) -> anyhow::Result<Database<Unspecified, Unspecified>> {
+    env.database_options()
+        .name(db_name)
+        .open(rtxn)
+        .with_context(|| format!("While opening the {db_name:?} poly database"))?
+        .with_context(|| format!("Missing the {db_name:?} poly database"))
+}
+
+fn try_clearing_poly_database(
+    wtxn: &mut RwTxn,
+    database: Database<Unspecified, Unspecified>,
+    db_name: &str,
+) -> anyhow::Result<()> {
+    database.clear(wtxn).with_context(|| format!("While clearing the {db_name:?} database"))
+}
+
+/// Exports a dump into the dump directory.
+fn export_a_dump(
+    db_path: PathBuf,
+    dump_dir: PathBuf,
+    skip_enqueued_tasks: bool,
+) -> Result<(), anyhow::Error> {
+    let started_at = OffsetDateTime::now_utc();
+
+    // 1. Extracts the instance UID from disk
+    let instance_uid_path = db_path.join("instance-uid");
+    let instance_uid = match read_to_string(&instance_uid_path) {
+        Ok(content) => match content.trim().parse() {
+            Ok(uuid) => Some(uuid),
+            Err(e) => {
+                eprintln!("Impossible to parse instance-uid: {e}");
+                None
+            }
+        },
+        Err(e) => {
+            eprintln!("Impossible to read {}: {}", instance_uid_path.display(), e);
+            None
+        }
+    };
+
+    let dump = DumpWriter::new(instance_uid).context("While creating a new dump")?;
+    let file_store =
+        FileStore::new(db_path.join("update_files")).context("While opening the FileStore")?;
+
+    let index_scheduler_path = db_path.join("tasks");
+    let env = EnvOpenOptions::new()
+        .max_dbs(100)
+        .open(&index_scheduler_path)
+        .with_context(|| format!("While trying to open {:?}", index_scheduler_path.display()))?;
+
+    eprintln!("Dumping the keys...");
+
+    // 2. dump the keys
+    let auth_store = AuthController::new(&db_path, &None)
+        .with_context(|| format!("While opening the auth store at {}", db_path.display()))?;
+    let mut dump_keys = dump.create_keys()?;
+    let mut count = 0;
+    for key in auth_store.list_keys()? {
+        dump_keys.push_key(&key)?;
+        count += 1;
+    }
+    dump_keys.flush()?;
+
+    eprintln!("Successfully dumped {count} keys!");
+
+    let rtxn = env.read_txn()?;
+    let all_tasks: Database<BEU32, SerdeJson<Task>> =
+        try_opening_database(&env, &rtxn, "all-tasks")?;
+    let index_mapping: Database<Str, UuidCodec> =
+        try_opening_database(&env, &rtxn, "index-mapping")?;
+
+    if skip_enqueued_tasks {
+        eprintln!("Skip dumping the enqueued tasks...");
+    } else {
+        eprintln!("Dumping the enqueued tasks...");
+
+        // 3. dump the tasks
+        let mut dump_tasks = dump.create_tasks_queue()?;
+        let mut count = 0;
+        for ret in all_tasks.iter(&rtxn)? {
+            let (_, t) = ret?;
+            let status = t.status;
+            let content_file = t.content_uuid();
+            let mut dump_content_file = dump_tasks.push_task(&t.into())?;
+
+            // 3.1. Dump the `content_file` associated with the task if there is one and the task is not finished yet.
+            if let Some(content_file_uuid) = content_file {
+                if status == Status::Enqueued {
+                    let content_file = file_store.get_update(content_file_uuid)?;
+
+                    let reader =
+                        DocumentsBatchReader::from_reader(content_file).with_context(|| {
+                            format!("While reading content file {:?}", content_file_uuid)
+                        })?;
+
+                    let (mut cursor, documents_batch_index) = reader.into_cursor_and_fields_index();
+                    while let Some(doc) = cursor.next_document().with_context(|| {
+                        format!("While iterating on content file {:?}", content_file_uuid)
+                    })? {
+                        dump_content_file
+                            .push_document(&obkv_to_object(&doc, &documents_batch_index)?)?;
+                    }
+                    dump_content_file.flush()?;
+                    count += 1;
+                }
+            }
+        }
+        dump_tasks.flush()?;
+
+        eprintln!("Successfully dumped {count} enqueued tasks!");
+    }
+
+    eprintln!("Dumping the indexes...");
+
+    // 4. Dump the indexes
+    let mut count = 0;
+    for result in index_mapping.iter(&rtxn)? {
+        let (uid, uuid) = result?;
+        let index_path = db_path.join("indexes").join(uuid.to_string());
+        let index = Index::new(EnvOpenOptions::new(), &index_path).with_context(|| {
+            format!("While trying to open the index at path {:?}", index_path.display())
+        })?;
+
+        let rtxn = index.read_txn()?;
+        let metadata = IndexMetadata {
+            uid: uid.to_owned(),
+            primary_key: index.primary_key(&rtxn)?.map(String::from),
+            created_at: index.created_at(&rtxn)?,
+            updated_at: index.updated_at(&rtxn)?,
+        };
+        let mut index_dumper = dump.create_index(uid, &metadata)?;
+
+        let fields_ids_map = index.fields_ids_map(&rtxn)?;
+        let all_fields: Vec<_> = fields_ids_map.iter().map(|(id, _)| id).collect();
+
+        // 4.1. Dump the documents
+        for ret in index.all_documents(&rtxn)? {
+            let (_id, doc) = ret?;
+            let document = obkv_to_json(&all_fields, &fields_ids_map, doc)?;
+            index_dumper.push_document(&document)?;
+        }
+
+        // 4.2. Dump the settings
+        let settings = meilisearch_types::settings::settings(&index, &rtxn)?;
+        index_dumper.settings(&settings)?;
+        count += 1;
+    }
+
+    eprintln!("Successfully dumped {count} indexes!");
+    // We will not dump experimental feature settings
+    eprintln!("The tool is not dumping experimental features, please set them by hand afterward");
+
+    let dump_uid = started_at.format(format_description!(
+        "[year repr:full][month repr:numerical][day padding:zero]-[hour padding:zero][minute padding:zero][second padding:zero][subsecond digits:3]"
+    )).unwrap();
+
+    let path = dump_dir.join(format!("{}.dump", dump_uid));
+    let file = File::create(&path)?;
+    dump.persist_to(BufWriter::new(file))?;
+
+    eprintln!("Dump exported at path {:?}", path.display());
+
+    Ok(())
+}

meilitool/src/uuid_codec.rs

@@ -0,0 +1,24 @@
+use std::borrow::Cow;
+use std::convert::TryInto;
+
+use meilisearch_types::heed::{BoxedError, BytesDecode, BytesEncode};
+use uuid::Uuid;
+
+/// A heed codec for value of struct Uuid.
+pub struct UuidCodec;
+
+impl<'a> BytesDecode<'a> for UuidCodec {
+    type DItem = Uuid;
+
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        bytes.try_into().map(Uuid::from_bytes).map_err(Into::into)
+    }
+}
+
+impl BytesEncode<'_> for UuidCodec {
+    type EItem = Uuid;
+
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item.as_bytes()))
+    }
+}

milli/Cargo.toml

@@ -17,23 +17,25 @@ bincode = "1.3.3"
 bstr = "1.4.0"
 bytemuck = { version = "1.13.1", features = ["extern_crate_alloc"] }
 byteorder = "1.4.3"
-charabia = { version = "0.8.3", default-features = false }
+charabia = { version = "0.8.5", default-features = false }
 concat-arrays = "0.1.2"
 crossbeam-channel = "0.5.8"
-deserr = { version = "0.6.0", features = ["actix-web"]}
+deserr = "0.6.0"
 either = { version = "1.8.1", features = ["serde"] }
 flatten-serde-json = { path = "../flatten-serde-json" }
 fst = "0.4.7"
 fxhash = "0.2.1"
 geoutils = "0.5.1"
 grenad = { version = "0.4.5", default-features = false, features = [
-    "rayon", "tempfile"
+    "rayon",
+    "tempfile",
 ] }
-heed = { git = "https://github.com/meilisearch/heed", tag = "v0.12.7", default-features = false, features = [
-    "lmdb", "read-txn-no-tls"
+heed = { version = "0.20.0-alpha.9", default-features = false, features = [
+    "serde-json",
+    "serde-bincode",
+    "read-txn-no-tls",
 ] }
 indexmap = { version = "2.0.0", features = ["serde"] }
-instant-distance = { version = "0.6.1", features = ["with-serde"] }
 json-depth-checker = { path = "../json-depth-checker" }
 levenshtein_automata = { version = "0.2.1", features = ["fst_automaton"] }
 memmap2 = "0.7.1"
@@ -72,6 +74,23 @@ puffin = "0.16.0"
 log = "0.4.17"
 logging_timer = "1.1.0"
 csv = "1.2.1"
+candle-core = { git = "https://github.com/huggingface/candle.git", version = "0.3.1" }
+candle-transformers = { git = "https://github.com/huggingface/candle.git", version = "0.3.1" }
+candle-nn = { git = "https://github.com/huggingface/candle.git", version = "0.3.1" }
+tokenizers = { git = "https://github.com/huggingface/tokenizers.git", tag = "v0.14.1", version = "0.14.1", default_features = false, features = ["onig"] }
+hf-hub = { git = "https://github.com/dureuill/hf-hub.git", branch = "rust_tls", default_features = false, features = [
+    "online",
+] }
+tokio = { version = "1.34.0", features = ["rt"] }
+futures = "0.3.29"
+reqwest = { version = "0.11.16", features = [
+    "rustls-tls",
+    "json",
+], default-features = false }
+tiktoken-rs = "0.5.7"
+liquid = "0.26.4"
+arroy = { git = "https://github.com/meilisearch/arroy.git", version = "0.1.0" }
+rand = "0.8.5"
 
 [dev-dependencies]
 mimalloc = { version = "0.1.37", default-features = false }
@@ -83,7 +102,15 @@ meili-snap = { path = "../meili-snap" }
 rand = { version = "0.8.5", features = ["small_rng"] }
 
 [features]
-all-tokenizations = ["charabia/chinese", "charabia/hebrew", "charabia/japanese", "charabia/thai", "charabia/korean", "charabia/greek"]
+all-tokenizations = [
+    "charabia/chinese",
+    "charabia/hebrew",
+    "charabia/japanese",
+    "charabia/thai",
+    "charabia/korean",
+    "charabia/greek",
+    "charabia/khmer",
+]
 
 # Use POSIX semaphores instead of SysV semaphores in LMDB
 # For more information on this feature, see heed's Cargo.toml
@@ -107,3 +134,6 @@ thai = ["charabia/thai"]
 
 # allow greek specialized tokenization
 greek = ["charabia/greek"]
+
+# allow khmer specialized tokenization
+khmer = ["charabia/khmer"]

milli/examples/search.rs

@@ -5,8 +5,8 @@ use std::time::Instant;
 
 use heed::EnvOpenOptions;
 use milli::{
-    execute_search, DefaultSearchLogger, GeoSortStrategy, Index, SearchContext, SearchLogger,
-    TermsMatchingStrategy,
+    execute_search, filtered_universe, DefaultSearchLogger, GeoSortStrategy, Index, SearchContext,
+    SearchLogger, TermsMatchingStrategy,
 };
 
 #[global_allocator]
@@ -49,14 +49,15 @@ fn main() -> Result<(), Box<dyn Error>> {
             let start = Instant::now();
 
             let mut ctx = SearchContext::new(&index, &txn);
+            let universe = filtered_universe(&ctx, &None)?;
+
             let docs = execute_search(
                 &mut ctx,
-                &(!query.trim().is_empty()).then(|| query.trim().to_owned()),
-                &None,
+                (!query.trim().is_empty()).then(|| query.trim()),
                 TermsMatchingStrategy::Last,
                 milli::score_details::ScoringStrategy::Skip,
                 false,
-                &None,
+                universe,
                 &None,
                 GeoSortStrategy::default(),
                 0,

milli/src/distance.rs

@@ -1,41 +0,0 @@
-use std::ops;
-
-use instant_distance::Point;
-use serde::{Deserialize, Serialize};
-
-use crate::normalize_vector;
-
-#[derive(Debug, Default, Clone, Serialize, Deserialize)]
-pub struct NDotProductPoint(Vec<f32>);
-
-impl NDotProductPoint {
-    pub fn new(point: Vec<f32>) -> Self {
-        NDotProductPoint(normalize_vector(point))
-    }
-
-    pub fn into_inner(self) -> Vec<f32> {
-        self.0
-    }
-}
-
-impl ops::Deref for NDotProductPoint {
-    type Target = [f32];
-
-    fn deref(&self) -> &Self::Target {
-        self.0.as_slice()
-    }
-}
-
-impl Point for NDotProductPoint {
-    fn distance(&self, other: &Self) -> f32 {
-        let dist = 1.0 - dot_product_similarity(&self.0, &other.0);
-        debug_assert!(!dist.is_nan());
-        dist
-    }
-}
-
-/// Returns the dot product similarity score that will between 0.0 and 1.0
-/// if both vectors are normalized. The higher the more similar the vectors are.
-pub fn dot_product_similarity(a: &[f32], b: &[f32]) -> f32 {
-    a.iter().zip(b).map(|(a, b)| a * b).sum()
-}

milli/src/documents/mod.rs

@@ -1,5 +1,6 @@
 mod builder;
 mod enriched;
+mod primary_key;
 mod reader;
 mod serde_impl;
 
@@ -11,6 +12,7 @@ use bimap::BiHashMap;
 pub use builder::DocumentsBatchBuilder;
 pub use enriched::{EnrichedDocument, EnrichedDocumentsBatchCursor, EnrichedDocumentsBatchReader};
 use obkv::KvReader;
+pub use primary_key::{DocumentIdExtractionError, FieldIdMapper, PrimaryKey, DEFAULT_PRIMARY_KEY};
 pub use reader::{DocumentsBatchCursor, DocumentsBatchCursorError, DocumentsBatchReader};
 use serde::{Deserialize, Serialize};
 
@@ -87,6 +89,12 @@ impl DocumentsBatchIndex {
     }
 }
 
+impl FieldIdMapper for DocumentsBatchIndex {
+    fn id(&self, name: &str) -> Option<FieldId> {
+        self.id(name)
+    }
+}
+
 #[derive(Debug, thiserror::Error)]
 pub enum Error {
     #[error("Error parsing number {value:?} at line {line}: {error}")]

milli/src/documents/primary_key.rs

@@ -0,0 +1,172 @@
+use std::iter;
+use std::result::Result as StdResult;
+
+use serde_json::Value;
+
+use crate::{FieldId, InternalError, Object, Result, UserError};
+
+/// The symbol used to define levels in a nested primary key.
+const PRIMARY_KEY_SPLIT_SYMBOL: char = '.';
+
+/// The default primary that is used when not specified.
+pub const DEFAULT_PRIMARY_KEY: &str = "id";
+
+/// Trait for objects that can map the name of a field to its [`FieldId`].
+pub trait FieldIdMapper {
+    /// Attempts to map the passed name to its [`FieldId`].
+    ///
+    /// `None` if the field with this name was not found.
+    fn id(&self, name: &str) -> Option<FieldId>;
+}
+
+/// A type that represent the type of primary key that has been set
+/// for this index, a classic flat one or a nested one.
+#[derive(Debug, Clone, Copy)]
+pub enum PrimaryKey<'a> {
+    Flat { name: &'a str, field_id: FieldId },
+    Nested { name: &'a str },
+}
+
+pub enum DocumentIdExtractionError {
+    InvalidDocumentId(UserError),
+    MissingDocumentId,
+    TooManyDocumentIds(usize),
+}
+
+impl<'a> PrimaryKey<'a> {
+    pub fn new(path: &'a str, fields: &impl FieldIdMapper) -> Option<Self> {
+        Some(if path.contains(PRIMARY_KEY_SPLIT_SYMBOL) {
+            Self::Nested { name: path }
+        } else {
+            let field_id = fields.id(path)?;
+            Self::Flat { name: path, field_id }
+        })
+    }
+
+    pub fn name(&self) -> &str {
+        match self {
+            PrimaryKey::Flat { name, .. } => name,
+            PrimaryKey::Nested { name } => name,
+        }
+    }
+
+    pub fn document_id(
+        &self,
+        document: &obkv::KvReader<FieldId>,
+        fields: &impl FieldIdMapper,
+    ) -> Result<StdResult<String, DocumentIdExtractionError>> {
+        match self {
+            PrimaryKey::Flat { name: _, field_id } => match document.get(*field_id) {
+                Some(document_id_bytes) => {
+                    let document_id = serde_json::from_slice(document_id_bytes)
+                        .map_err(InternalError::SerdeJson)?;
+                    match validate_document_id_value(document_id)? {
+                        Ok(document_id) => Ok(Ok(document_id)),
+                        Err(user_error) => {
+                            Ok(Err(DocumentIdExtractionError::InvalidDocumentId(user_error)))
+                        }
+                    }
+                }
+                None => Ok(Err(DocumentIdExtractionError::MissingDocumentId)),
+            },
+            nested @ PrimaryKey::Nested { .. } => {
+                let mut matching_documents_ids = Vec::new();
+                for (first_level_name, right) in nested.possible_level_names() {
+                    if let Some(field_id) = fields.id(first_level_name) {
+                        if let Some(value_bytes) = document.get(field_id) {
+                            let object = serde_json::from_slice(value_bytes)
+                                .map_err(InternalError::SerdeJson)?;
+                            fetch_matching_values(object, right, &mut matching_documents_ids);
+
+                            if matching_documents_ids.len() >= 2 {
+                                return Ok(Err(DocumentIdExtractionError::TooManyDocumentIds(
+                                    matching_documents_ids.len(),
+                                )));
+                            }
+                        }
+                    }
+                }
+
+                match matching_documents_ids.pop() {
+                    Some(document_id) => match validate_document_id_value(document_id)? {
+                        Ok(document_id) => Ok(Ok(document_id)),
+                        Err(user_error) => {
+                            Ok(Err(DocumentIdExtractionError::InvalidDocumentId(user_error)))
+                        }
+                    },
+                    None => Ok(Err(DocumentIdExtractionError::MissingDocumentId)),
+                }
+            }
+        }
+    }
+
+    /// Returns an `Iterator` that gives all the possible fields names the primary key
+    /// can have depending of the first level name and depth of the objects.
+    pub fn possible_level_names(&self) -> impl Iterator<Item = (&str, &str)> + '_ {
+        let name = self.name();
+        name.match_indices(PRIMARY_KEY_SPLIT_SYMBOL)
+            .map(move |(i, _)| (&name[..i], &name[i + PRIMARY_KEY_SPLIT_SYMBOL.len_utf8()..]))
+            .chain(iter::once((name, "")))
+    }
+}
+
+fn fetch_matching_values(value: Value, selector: &str, output: &mut Vec<Value>) {
+    match value {
+        Value::Object(object) => fetch_matching_values_in_object(object, selector, "", output),
+        otherwise => output.push(otherwise),
+    }
+}
+
+fn fetch_matching_values_in_object(
+    object: Object,
+    selector: &str,
+    base_key: &str,
+    output: &mut Vec<Value>,
+) {
+    for (key, value) in object {
+        let base_key = if base_key.is_empty() {
+            key.to_string()
+        } else {
+            format!("{}{}{}", base_key, PRIMARY_KEY_SPLIT_SYMBOL, key)
+        };
+
+        if starts_with(selector, &base_key) {
+            match value {
+                Value::Object(object) => {
+                    fetch_matching_values_in_object(object, selector, &base_key, output)
+                }
+                value => output.push(value),
+            }
+        }
+    }
+}
+
+fn starts_with(selector: &str, key: &str) -> bool {
+    selector.strip_prefix(key).map_or(false, |tail| {
+        tail.chars().next().map(|c| c == PRIMARY_KEY_SPLIT_SYMBOL).unwrap_or(true)
+    })
+}
+
+// FIXME: move to a DocumentId struct
+
+fn validate_document_id(document_id: &str) -> Option<&str> {
+    if !document_id.is_empty()
+        && document_id.chars().all(|c| matches!(c, 'a'..='z' | 'A'..='Z' | '0'..='9' | '-' | '_'))
+    {
+        Some(document_id)
+    } else {
+        None
+    }
+}
+
+pub fn validate_document_id_value(document_id: Value) -> Result<StdResult<String, UserError>> {
+    match document_id {
+        Value::String(string) => match validate_document_id(&string) {
+            Some(s) if s.len() == string.len() => Ok(Ok(string)),
+            Some(s) => Ok(Ok(s.to_string())),
+            None => Ok(Err(UserError::InvalidDocumentId { document_id: Value::String(string) })),
+        },
+        Value::Number(number) if number.is_i64() => Ok(Ok(number.to_string())),
+        content => Ok(Err(UserError::InvalidDocumentId { document_id: content })),
+    }
+}

milli/src/error.rs

@@ -61,6 +61,10 @@ pub enum InternalError {
     AbortedIndexation,
     #[error("The matching words list contains at least one invalid member.")]
     InvalidMatchingWords,
+    #[error(transparent)]
+    ArroyError(#[from] arroy::Error),
+    #[error(transparent)]
+    VectorEmbeddingError(#[from] crate::vector::Error),
 }
 
 #[derive(Error, Debug)]
@@ -110,8 +114,10 @@ only composed of alphanumeric characters (a-z A-Z 0-9), hyphens (-) and undersco
     InvalidGeoField(#[from] GeoError),
     #[error("Invalid vector dimensions: expected: `{}`, found: `{}`.", .expected, .found)]
     InvalidVectorDimensions { expected: usize, found: usize },
-    #[error("The `_vectors` field in the document with the id: `{document_id}` is not an array. Was expecting an array of floats or an array of arrays of floats but instead got `{value}`.")]
-    InvalidVectorsType { document_id: Value, value: Value },
+    #[error("The `_vectors.{subfield}` field in the document with id: `{document_id}` is not an array. Was expecting an array of floats or an array of arrays of floats but instead got `{value}`.")]
+    InvalidVectorsType { document_id: Value, value: Value, subfield: String },
+    #[error("The `_vectors` field in the document with id: `{document_id}` is not an object. Was expecting an object with a key for each embedder with manually provided vectors, but instead got `{value}`")]
+    InvalidVectorsMapType { document_id: Value, value: Value },
     #[error("{0}")]
     InvalidFilter(String),
     #[error("Invalid type for filter subexpression: expected: {}, found: {1}.", .0.join(", "))]
@@ -152,7 +158,7 @@ only composed of alphanumeric characters (a-z A-Z 0-9), hyphens (-) and undersco
         valid_fields: BTreeSet<String>,
         hidden_fields: bool,
     },
-    #[error("{}", HeedError::BadOpenOptions)]
+    #[error("an environment is already opened with different options")]
     InvalidLmdbOpenOptions,
     #[error("You must specify where `sort` is listed in the rankingRules setting to use the sort parameter at search time.")]
     SortRankingRuleMissing,
@@ -180,6 +186,76 @@ only composed of alphanumeric characters (a-z A-Z 0-9), hyphens (-) and undersco
     UnknownInternalDocumentId { document_id: DocumentId },
     #[error("`minWordSizeForTypos` setting is invalid. `oneTypo` and `twoTypos` fields should be between `0` and `255`, and `twoTypos` should be greater or equals to `oneTypo` but found `oneTypo: {0}` and twoTypos: {1}`.")]
     InvalidMinTypoWordLenSetting(u8, u8),
+    #[error(transparent)]
+    VectorEmbeddingError(#[from] crate::vector::Error),
+    #[error(transparent)]
+    MissingDocumentField(#[from] crate::prompt::error::RenderPromptError),
+    #[error(transparent)]
+    InvalidPrompt(#[from] crate::prompt::error::NewPromptError),
+    #[error("`.embedders.{0}.documentTemplate`: Invalid template: {1}.")]
+    InvalidPromptForEmbeddings(String, crate::prompt::error::NewPromptError),
+    #[error("Too many embedders in the configuration. Found {0}, but limited to 256.")]
+    TooManyEmbedders(usize),
+    #[error("Cannot find embedder with name {0}.")]
+    InvalidEmbedder(String),
+    #[error("Too many vectors for document with id {0}: found {1}, but limited to 256.")]
+    TooManyVectors(String, usize),
+    #[error("`.embedders.{embedder_name}`: Field `{field}` unavailable for source `{source_}` (only available for sources: {}). Available fields: {}",
+        allowed_sources_for_field
+         .iter()
+         .map(|accepted| format!("`{}`", accepted))
+         .collect::<Vec<String>>()
+         .join(", "),
+        allowed_fields_for_source
+         .iter()
+         .map(|accepted| format!("`{}`", accepted))
+         .collect::<Vec<String>>()
+         .join(", ")
+    )]
+    InvalidFieldForSource {
+        embedder_name: String,
+        source_: crate::vector::settings::EmbedderSource,
+        field: &'static str,
+        allowed_fields_for_source: &'static [&'static str],
+        allowed_sources_for_field: &'static [crate::vector::settings::EmbedderSource],
+    },
+    #[error("`.embedders.{embedder_name}.model`: Invalid model `{model}` for OpenAI. Supported models: {:?}", crate::vector::openai::EmbeddingModel::supported_models())]
+    InvalidOpenAiModel { embedder_name: String, model: String },
+    #[error("`.embedders.{embedder_name}`: Missing field `{field}` (note: this field is mandatory for source {source_})")]
+    MissingFieldForSource {
+        field: &'static str,
+        source_: crate::vector::settings::EmbedderSource,
+        embedder_name: String,
+    },
+}
+
+impl From<crate::vector::Error> for Error {
+    fn from(value: crate::vector::Error) -> Self {
+        match value.fault() {
+            FaultSource::User => Error::UserError(value.into()),
+            FaultSource::Runtime => Error::InternalError(value.into()),
+            FaultSource::Bug => Error::InternalError(value.into()),
+            FaultSource::Undecided => Error::InternalError(value.into()),
+        }
+    }
+}
+
+impl From<arroy::Error> for Error {
+    fn from(value: arroy::Error) -> Self {
+        match value {
+            arroy::Error::Heed(heed) => heed.into(),
+            arroy::Error::Io(io) => io.into(),
+            arroy::Error::InvalidVecDimension { expected, received } => {
+                Error::UserError(UserError::InvalidVectorDimensions { expected, found: received })
+            }
+            arroy::Error::DatabaseFull
+            | arroy::Error::InvalidItemAppend
+            | arroy::Error::UnmatchingDistance { .. }
+            | arroy::Error::MissingMetadata => {
+                Error::InternalError(InternalError::ArroyError(value))
+            }
+        }
+    }
 }
 
 #[derive(Error, Debug)]
@@ -326,15 +402,36 @@ impl From<HeedError> for Error {
             HeedError::Mdb(MdbError::MapFull) => UserError(MaxDatabaseSizeReached),
             HeedError::Mdb(MdbError::Invalid) => UserError(InvalidStoreFile),
             HeedError::Mdb(error) => InternalError(Store(error)),
-            HeedError::Encoding => InternalError(Serialization(Encoding { db_name: None })),
-            HeedError::Decoding => InternalError(Serialization(Decoding { db_name: None })),
+            // TODO use the encoding
+            HeedError::Encoding(_) => InternalError(Serialization(Encoding { db_name: None })),
+            HeedError::Decoding(_) => InternalError(Serialization(Decoding { db_name: None })),
             HeedError::InvalidDatabaseTyping => InternalError(InvalidDatabaseTyping),
             HeedError::DatabaseClosing => InternalError(DatabaseClosing),
-            HeedError::BadOpenOptions => UserError(InvalidLmdbOpenOptions),
+            HeedError::BadOpenOptions { .. } => UserError(InvalidLmdbOpenOptions),
         }
     }
 }
 
+#[derive(Debug, Clone, Copy)]
+pub enum FaultSource {
+    User,
+    Runtime,
+    Bug,
+    Undecided,
+}
+
+impl std::fmt::Display for FaultSource {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        let s = match self {
+            FaultSource::User => "user error",
+            FaultSource::Runtime => "runtime error",
+            FaultSource::Bug => "coding error",
+            FaultSource::Undecided => "error",
+        };
+        f.write_str(s)
+    }
+}
+
 #[test]
 fn conditionally_lookup_for_error_message() {
     let prefix = "Attribute `name` is not sortable.";

milli/src/external_documents_ids.rs

@@ -1,8 +1,7 @@
 use std::collections::HashMap;
 
-use heed::types::{OwnedType, Str};
+use heed::types::Str;
 use heed::{Database, RoIter, RoTxn, RwTxn};
-use roaring::RoaringBitmap;
 
 use crate::{DocumentId, BEU32};
 
@@ -17,10 +16,10 @@ pub struct DocumentOperation {
     pub kind: DocumentOperationKind,
 }
 
-pub struct ExternalDocumentsIds(Database<Str, OwnedType<BEU32>>);
+pub struct ExternalDocumentsIds(Database<Str, BEU32>);
 
 impl ExternalDocumentsIds {
-    pub fn new(db: Database<Str, OwnedType<BEU32>>) -> ExternalDocumentsIds {
+    pub fn new(db: Database<Str, BEU32>) -> ExternalDocumentsIds {
         ExternalDocumentsIds(db)
     }
 
@@ -30,7 +29,7 @@ impl ExternalDocumentsIds {
     }
 
     pub fn get<A: AsRef<str>>(&self, rtxn: &RoTxn, external_id: A) -> heed::Result<Option<u32>> {
-        Ok(self.0.get(rtxn, external_id.as_ref())?.map(|x| x.get()))
+        self.0.get(rtxn, external_id.as_ref())
     }
 
     /// An helper function to debug this type, returns an `HashMap` of both,
@@ -39,28 +38,11 @@ impl ExternalDocumentsIds {
         let mut map = HashMap::default();
         for result in self.0.iter(rtxn)? {
             let (external, internal) = result?;
-            map.insert(external.to_owned(), internal.get());
+            map.insert(external.to_owned(), internal);
         }
         Ok(map)
     }
 
-    /// Looks for the internal ids in the passed bitmap, and returns an iterator over the mapping between
-    /// these internal ids and their external id.
-    ///
-    /// The returned iterator has `Result<(String, DocumentId), RoaringBitmap>` as `Item`,
-    /// where the returned values can be:
-    /// - `Ok((external_id, internal_id))`: if a mapping was found
-    /// - `Err(remaining_ids)`: if the external ids for some of the requested internal ids weren't found.
-    ///   In that case the returned bitmap contains the internal ids whose external ids were not found after traversing
-    ///   the entire fst.
-    pub fn find_external_id_of<'t>(
-        &self,
-        rtxn: &'t RoTxn,
-        internal_ids: RoaringBitmap,
-    ) -> heed::Result<ExternalToInternalOwnedIterator<'t>> {
-        self.0.iter(rtxn).map(|iter| ExternalToInternalOwnedIterator { iter, internal_ids })
-    }
-
     /// Applies the list of operations passed as argument, modifying the current external to internal id mapping.
     ///
     /// If the list contains multiple operations on the same external id, then the result is unspecified.
@@ -73,7 +55,7 @@ impl ExternalDocumentsIds {
         for DocumentOperation { external_id, internal_id, kind } in operations {
             match kind {
                 DocumentOperationKind::Create => {
-                    self.0.put(wtxn, &external_id, &BEU32::new(internal_id))?;
+                    self.0.put(wtxn, &external_id, &internal_id)?;
                 }
                 DocumentOperationKind::Delete => {
                     if !self.0.delete(wtxn, &external_id)? {
@@ -87,60 +69,7 @@ impl ExternalDocumentsIds {
     }
 
     /// Returns an iterator over all the external ids.
-    pub fn iter<'t>(&self, rtxn: &'t RoTxn) -> heed::Result<RoIter<'t, Str, OwnedType<BEU32>>> {
+    pub fn iter<'t>(&self, rtxn: &'t RoTxn) -> heed::Result<RoIter<'t, Str, BEU32>> {
         self.0.iter(rtxn)
     }
 }
-
-/// An iterator over mappings between requested internal ids and external ids.
-///
-/// See [`ExternalDocumentsIds::find_external_id_of`] for details.
-pub struct ExternalToInternalOwnedIterator<'t> {
-    iter: RoIter<'t, Str, OwnedType<BEU32>>,
-    internal_ids: RoaringBitmap,
-}
-
-impl<'t> Iterator for ExternalToInternalOwnedIterator<'t> {
-    /// A result indicating if a mapping was found, or if the stream was exhausted without finding all internal ids.
-    type Item = Result<(&'t str, DocumentId), RoaringBitmap>;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        // if all requested ids were found, we won't find any other, so short-circuit
-        if self.internal_ids.is_empty() {
-            return None;
-        }
-        loop {
-            let (external, internal) = match self.iter.next() {
-                Some(Ok((external, internal))) => (external, internal),
-                // TODO manage this better, remove panic
-                Some(Err(e)) => panic!("{}", e),
-                _ => {
-                    // we exhausted the stream but we still have some internal ids to find
-                    let remaining_ids = std::mem::take(&mut self.internal_ids);
-                    return Some(Err(remaining_ids));
-                    // note: next calls to `next` will return `None` since we replaced the internal_ids
-                    // with the default empty bitmap
-                }
-            };
-            let internal = internal.get();
-            let was_contained = self.internal_ids.remove(internal);
-            if was_contained {
-                return Some(Ok((external, internal)));
-            }
-        }
-    }
-}
-
-impl<'t> ExternalToInternalOwnedIterator<'t> {
-    /// Returns the bitmap of internal ids whose external id are yet to be found
-    pub fn remaining_internal_ids(&self) -> &RoaringBitmap {
-        &self.internal_ids
-    }
-
-    /// Consumes this iterator and returns an iterator over only the external ids, ignoring the internal ids.
-    ///
-    /// Use this when you don't need the mapping between the external and the internal ids.
-    pub fn only_external_ids(self) -> impl Iterator<Item = Result<String, RoaringBitmap>> + 't {
-        self.map(|res| res.map(|(external, _internal)| external.to_owned()))
-    }
-}

milli/src/fields_ids_map.rs

@@ -81,6 +81,12 @@ impl Default for FieldsIdsMap {
     }
 }
 
+impl crate::documents::FieldIdMapper for FieldsIdsMap {
+    fn id(&self, name: &str) -> Option<FieldId> {
+        self.id(name)
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

milli/src/heed_codec/beu16_str_codec.rs

@@ -2,26 +2,28 @@ use std::borrow::Cow;
 use std::convert::TryInto;
 use std::str;
 
+use heed::BoxedError;
+
 pub struct BEU16StrCodec;
 
 impl<'a> heed::BytesDecode<'a> for BEU16StrCodec {
     type DItem = (u16, &'a str);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let (n_bytes, str_bytes) = bytes.split_at(2);
-        let n = n_bytes.try_into().map(u16::from_be_bytes).ok()?;
-        let s = str::from_utf8(str_bytes).ok()?;
-        Some((n, s))
+        let n = n_bytes.try_into().map(u16::from_be_bytes)?;
+        let s = str::from_utf8(str_bytes)?;
+        Ok((n, s))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for BEU16StrCodec {
     type EItem = (u16, &'a str);
 
-    fn bytes_encode((n, s): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((n, s): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(s.len() + 2);
         bytes.extend_from_slice(&n.to_be_bytes());
         bytes.extend_from_slice(s.as_bytes());
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/beu32_str_codec.rs

@@ -2,26 +2,28 @@ use std::borrow::Cow;
 use std::convert::TryInto;
 use std::str;
 
+use heed::BoxedError;
+
 pub struct BEU32StrCodec;
 
 impl<'a> heed::BytesDecode<'a> for BEU32StrCodec {
     type DItem = (u32, &'a str);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let (n_bytes, str_bytes) = bytes.split_at(4);
-        let n = n_bytes.try_into().map(u32::from_be_bytes).ok()?;
-        let s = str::from_utf8(str_bytes).ok()?;
-        Some((n, s))
+        let n = n_bytes.try_into().map(u32::from_be_bytes)?;
+        let s = str::from_utf8(str_bytes)?;
+        Ok((n, s))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for BEU32StrCodec {
     type EItem = (u32, &'a str);
 
-    fn bytes_encode((n, s): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((n, s): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(s.len() + 4);
         bytes.extend_from_slice(&n.to_be_bytes());
         bytes.extend_from_slice(s.as_bytes());
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/byte_slice_ref.rs

@@ -1,23 +1,23 @@
 use std::borrow::Cow;
 
-use heed::{BytesDecode, BytesEncode};
+use heed::{BoxedError, BytesDecode, BytesEncode};
 
-/// A codec for values of type `&[u8]`. Unlike `ByteSlice`, its `EItem` and `DItem` associated
+/// A codec for values of type `&[u8]`. Unlike `Bytes`, its `EItem` and `DItem` associated
 /// types are equivalent (= `&'a [u8]`) and these values can reside within another structure.
-pub struct ByteSliceRefCodec;
+pub struct BytesRefCodec;
 
-impl<'a> BytesEncode<'a> for ByteSliceRefCodec {
+impl<'a> BytesEncode<'a> for BytesRefCodec {
     type EItem = &'a [u8];
 
-    fn bytes_encode(item: &'a Self::EItem) -> Option<Cow<'a, [u8]>> {
-        Some(Cow::Borrowed(item))
+    fn bytes_encode(item: &'a Self::EItem) -> Result<Cow<'a, [u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item))
     }
 }
 
-impl<'a> BytesDecode<'a> for ByteSliceRefCodec {
+impl<'a> BytesDecode<'a> for BytesRefCodec {
     type DItem = &'a [u8];
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        Some(bytes)
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        Ok(bytes)
     }
 }

milli/src/heed_codec/facet/field_doc_id_facet_codec.rs

@@ -1,8 +1,9 @@
 use std::borrow::Cow;
 use std::marker::PhantomData;
 
-use heed::{BytesDecode, BytesEncode};
+use heed::{BoxedError, BytesDecode, BytesEncode};
 
+use crate::heed_codec::SliceTooShortError;
 use crate::{try_split_array_at, DocumentId, FieldId};
 
 pub struct FieldDocIdFacetCodec<C>(PhantomData<C>);
@@ -13,16 +14,16 @@ where
 {
     type DItem = (FieldId, DocumentId, C::DItem);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (field_id_bytes, bytes) = try_split_array_at(bytes)?;
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let (field_id_bytes, bytes) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
         let field_id = u16::from_be_bytes(field_id_bytes);
 
-        let (document_id_bytes, bytes) = try_split_array_at(bytes)?;
+        let (document_id_bytes, bytes) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
         let document_id = u32::from_be_bytes(document_id_bytes);
 
         let value = C::bytes_decode(bytes)?;
 
-        Some((field_id, document_id, value))
+        Ok((field_id, document_id, value))
     }
 }
 
@@ -32,13 +33,15 @@ where
 {
     type EItem = (FieldId, DocumentId, C::EItem);
 
-    fn bytes_encode((field_id, document_id, value): &'a Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(
+        (field_id, document_id, value): &'a Self::EItem,
+    ) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(32);
         bytes.extend_from_slice(&field_id.to_be_bytes()); // 2 bytes
         bytes.extend_from_slice(&document_id.to_be_bytes()); // 4 bytes
         let value_bytes = C::bytes_encode(value)?;
         // variable length, if f64 -> 16 bytes, if string -> large, potentially
         bytes.extend_from_slice(&value_bytes);
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/facet/mod.rs

@@ -5,8 +5,8 @@ use std::borrow::Cow;
 use std::convert::TryFrom;
 use std::marker::PhantomData;
 
-use heed::types::{DecodeIgnore, OwnedType};
-use heed::{BytesDecode, BytesEncode};
+use heed::types::DecodeIgnore;
+use heed::{BoxedError, BytesDecode, BytesEncode};
 use roaring::RoaringBitmap;
 
 pub use self::field_doc_id_facet_codec::FieldDocIdFacetCodec;
@@ -18,7 +18,7 @@ pub type FieldDocIdFacetF64Codec = FieldDocIdFacetCodec<OrderedF64Codec>;
 pub type FieldDocIdFacetStringCodec = FieldDocIdFacetCodec<StrRefCodec>;
 pub type FieldDocIdFacetIgnoreCodec = FieldDocIdFacetCodec<DecodeIgnore>;
 
-pub type FieldIdCodec = OwnedType<BEU16>;
+pub type FieldIdCodec = BEU16;
 
 /// Tries to split a slice in half at the given middle point,
 /// `None` if the slice is too short.
@@ -58,7 +58,7 @@ where
 {
     type EItem = FacetGroupKey<T::EItem>;
 
-    fn bytes_encode(value: &'a Self::EItem) -> Option<Cow<'a, [u8]>> {
+    fn bytes_encode(value: &'a Self::EItem) -> Result<Cow<'a, [u8]>, BoxedError> {
         let mut v = vec![];
         v.extend_from_slice(&value.field_id.to_be_bytes());
         v.extend_from_slice(&[value.level]);
@@ -66,7 +66,7 @@ where
         let bound = T::bytes_encode(&value.left_bound)?;
         v.extend_from_slice(&bound);
 
-        Some(Cow::Owned(v))
+        Ok(Cow::Owned(v))
     }
 }
 impl<'a, T> heed::BytesDecode<'a> for FacetGroupKeyCodec<T>
@@ -75,11 +75,11 @@ where
 {
     type DItem = FacetGroupKey<T::DItem>;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let fid = u16::from_be_bytes(<[u8; 2]>::try_from(&bytes[0..=1]).ok()?);
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let fid = u16::from_be_bytes(<[u8; 2]>::try_from(&bytes[0..=1])?);
         let level = bytes[2];
         let bound = T::bytes_decode(&bytes[3..])?;
-        Some(FacetGroupKey { field_id: fid, level, left_bound: bound })
+        Ok(FacetGroupKey { field_id: fid, level, left_bound: bound })
     }
 }
 
@@ -87,17 +87,17 @@ pub struct FacetGroupValueCodec;
 impl<'a> heed::BytesEncode<'a> for FacetGroupValueCodec {
     type EItem = FacetGroupValue;
 
-    fn bytes_encode(value: &'a Self::EItem) -> Option<Cow<'a, [u8]>> {
+    fn bytes_encode(value: &'a Self::EItem) -> Result<Cow<'a, [u8]>, BoxedError> {
         let mut v = vec![value.size];
         CboRoaringBitmapCodec::serialize_into(&value.bitmap, &mut v);
-        Some(Cow::Owned(v))
+        Ok(Cow::Owned(v))
     }
 }
 impl<'a> heed::BytesDecode<'a> for FacetGroupValueCodec {
     type DItem = FacetGroupValue;
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let size = bytes[0];
-        let bitmap = CboRoaringBitmapCodec::deserialize_from(&bytes[1..]).ok()?;
-        Some(FacetGroupValue { size, bitmap })
+        let bitmap = CboRoaringBitmapCodec::deserialize_from(&bytes[1..])?;
+        Ok(FacetGroupValue { size, bitmap })
     }
 }

milli/src/heed_codec/facet/ordered_f64_codec.rs

@@ -1,37 +1,45 @@
 use std::borrow::Cow;
 use std::convert::TryInto;
 
-use heed::BytesDecode;
+use heed::{BoxedError, BytesDecode};
+use thiserror::Error;
 
 use crate::facet::value_encoding::f64_into_bytes;
+use crate::heed_codec::SliceTooShortError;
 
 pub struct OrderedF64Codec;
 
 impl<'a> BytesDecode<'a> for OrderedF64Codec {
     type DItem = f64;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         if bytes.len() < 16 {
-            return None;
+            Err(SliceTooShortError.into())
+        } else {
+            bytes[8..].try_into().map(f64::from_be_bytes).map_err(Into::into)
         }
-        let f = bytes[8..].try_into().ok().map(f64::from_be_bytes)?;
-        Some(f)
     }
 }
 
 impl heed::BytesEncode<'_> for OrderedF64Codec {
     type EItem = f64;
 
-    fn bytes_encode(f: &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(f: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut buffer = [0u8; 16];
 
         // write the globally ordered float
-        let bytes = f64_into_bytes(*f)?;
+        let bytes = f64_into_bytes(*f).ok_or(InvalidGloballyOrderedFloatError { float: *f })?;
         buffer[..8].copy_from_slice(&bytes[..]);
         // Then the f64 value just to be able to read it back
         let bytes = f.to_be_bytes();
         buffer[8..16].copy_from_slice(&bytes[..]);
 
-        Some(Cow::Owned(buffer.to_vec()))
+        Ok(Cow::Owned(buffer.to_vec()))
     }
 }
+
+#[derive(Error, Debug)]
+#[error("the float {float} cannot be converted to a globally ordered representation")]
+pub struct InvalidGloballyOrderedFloatError {
+    float: f64,
+}

milli/src/heed_codec/field_id_word_count_codec.rs

@@ -1,5 +1,8 @@
 use std::borrow::Cow;
 
+use heed::BoxedError;
+
+use super::SliceTooShortError;
 use crate::{try_split_array_at, FieldId};
 
 pub struct FieldIdWordCountCodec;
@@ -7,21 +10,21 @@ pub struct FieldIdWordCountCodec;
 impl<'a> heed::BytesDecode<'a> for FieldIdWordCountCodec {
     type DItem = (FieldId, u8);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (field_id_bytes, bytes) = try_split_array_at(bytes)?;
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let (field_id_bytes, bytes) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
         let field_id = u16::from_be_bytes(field_id_bytes);
-        let ([word_count], _nothing) = try_split_array_at(bytes)?;
-        Some((field_id, word_count))
+        let ([word_count], _nothing) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
+        Ok((field_id, word_count))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for FieldIdWordCountCodec {
     type EItem = (FieldId, u8);
 
-    fn bytes_encode((field_id, word_count): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((field_id, word_count): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(2 + 1);
         bytes.extend_from_slice(&field_id.to_be_bytes());
         bytes.push(*word_count);
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/fst_set_codec.rs

@@ -1,7 +1,7 @@
 use std::borrow::Cow;
 
 use fst::Set;
-use heed::{BytesDecode, BytesEncode};
+use heed::{BoxedError, BytesDecode, BytesEncode};
 
 /// A codec for values of type `Set<&[u8]>`.
 pub struct FstSetCodec;
@@ -9,15 +9,15 @@ pub struct FstSetCodec;
 impl<'a> BytesEncode<'a> for FstSetCodec {
     type EItem = Set<Vec<u8>>;
 
-    fn bytes_encode(item: &'a Self::EItem) -> Option<Cow<'a, [u8]>> {
-        Some(Cow::Borrowed(item.as_fst().as_bytes()))
+    fn bytes_encode(item: &'a Self::EItem) -> Result<Cow<'a, [u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item.as_fst().as_bytes()))
     }
 }
 
 impl<'a> BytesDecode<'a> for FstSetCodec {
     type DItem = Set<&'a [u8]>;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        Set::new(bytes).ok()
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        Set::new(bytes).map_err(Into::into)
     }
 }

milli/src/heed_codec/mod.rs

@@ -12,8 +12,10 @@ mod str_beu32_codec;
 mod str_ref;
 mod str_str_u8_codec;
 
-pub use byte_slice_ref::ByteSliceRefCodec;
+pub use byte_slice_ref::BytesRefCodec;
+use heed::BoxedError;
 pub use str_ref::StrRefCodec;
+use thiserror::Error;
 
 pub use self::beu16_str_codec::BEU16StrCodec;
 pub use self::beu32_str_codec::BEU32StrCodec;
@@ -31,5 +33,9 @@ pub use self::str_str_u8_codec::{U8StrStrCodec, UncheckedU8StrStrCodec};
 pub trait BytesDecodeOwned {
     type DItem;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem>;
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError>;
 }
+
+#[derive(Error, Debug)]
+#[error("the slice is too short")]
+pub struct SliceTooShortError;

milli/src/heed_codec/obkv_codec.rs

@@ -1,5 +1,6 @@
 use std::borrow::Cow;
 
+use heed::BoxedError;
 use obkv::{KvReaderU16, KvWriterU16};
 
 pub struct ObkvCodec;
@@ -7,15 +8,15 @@ pub struct ObkvCodec;
 impl<'a> heed::BytesDecode<'a> for ObkvCodec {
     type DItem = KvReaderU16<'a>;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        Some(KvReaderU16::new(bytes))
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        Ok(KvReaderU16::new(bytes))
     }
 }
 
 impl heed::BytesEncode<'_> for ObkvCodec {
     type EItem = KvWriterU16<Vec<u8>>;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
-        item.clone().into_inner().map(Cow::Owned).ok()
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
+        item.clone().into_inner().map(Cow::Owned).map_err(Into::into)
     }
 }

milli/src/heed_codec/roaring_bitmap/bo_roaring_bitmap_codec.rs

@@ -2,7 +2,7 @@ use std::borrow::Cow;
 use std::convert::TryInto;
 use std::mem::size_of;
 
-use heed::BytesDecode;
+use heed::{BoxedError, BytesDecode};
 use roaring::RoaringBitmap;
 
 use crate::heed_codec::BytesDecodeOwned;
@@ -19,22 +19,22 @@ impl BoRoaringBitmapCodec {
 impl BytesDecode<'_> for BoRoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         let mut bitmap = RoaringBitmap::new();
 
         for chunk in bytes.chunks(size_of::<u32>()) {
-            let bytes = chunk.try_into().ok()?;
+            let bytes = chunk.try_into()?;
             bitmap.push(u32::from_ne_bytes(bytes));
         }
 
-        Some(bitmap)
+        Ok(bitmap)
     }
 }
 
 impl BytesDecodeOwned for BoRoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         Self::bytes_decode(bytes)
     }
 }
@@ -42,9 +42,9 @@ impl BytesDecodeOwned for BoRoaringBitmapCodec {
 impl heed::BytesEncode<'_> for BoRoaringBitmapCodec {
     type EItem = RoaringBitmap;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut out = Vec::new();
         BoRoaringBitmapCodec::serialize_into(item, &mut out);
-        Some(Cow::Owned(out))
+        Ok(Cow::Owned(out))
     }
 }

milli/src/heed_codec/roaring_bitmap/cbo_roaring_bitmap_codec.rs

@@ -3,6 +3,7 @@ use std::io;
 use std::mem::size_of;
 
 use byteorder::{NativeEndian, ReadBytesExt, WriteBytesExt};
+use heed::BoxedError;
 use roaring::RoaringBitmap;
 
 use crate::heed_codec::BytesDecodeOwned;
@@ -132,26 +133,26 @@ impl CboRoaringBitmapCodec {
 impl heed::BytesDecode<'_> for CboRoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
-        Self::deserialize_from(bytes).ok()
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        Self::deserialize_from(bytes).map_err(Into::into)
     }
 }
 
 impl BytesDecodeOwned for CboRoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
-        Self::deserialize_from(bytes).ok()
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        Self::deserialize_from(bytes).map_err(Into::into)
     }
 }
 
 impl heed::BytesEncode<'_> for CboRoaringBitmapCodec {
     type EItem = RoaringBitmap;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut vec = Vec::with_capacity(Self::serialized_size(item));
         Self::serialize_into(item, &mut vec);
-        Some(Cow::Owned(vec))
+        Ok(Cow::Owned(vec))
     }
 }
 

milli/src/heed_codec/roaring_bitmap/roaring_bitmap_codec.rs

@@ -1,5 +1,6 @@
 use std::borrow::Cow;
 
+use heed::BoxedError;
 use roaring::RoaringBitmap;
 
 use crate::heed_codec::BytesDecodeOwned;
@@ -9,25 +10,25 @@ pub struct RoaringBitmapCodec;
 impl heed::BytesDecode<'_> for RoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
-        RoaringBitmap::deserialize_unchecked_from(bytes).ok()
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        RoaringBitmap::deserialize_unchecked_from(bytes).map_err(Into::into)
     }
 }
 
 impl BytesDecodeOwned for RoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
-        RoaringBitmap::deserialize_from(bytes).ok()
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        RoaringBitmap::deserialize_from(bytes).map_err(Into::into)
     }
 }
 
 impl heed::BytesEncode<'_> for RoaringBitmapCodec {
     type EItem = RoaringBitmap;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(item.serialized_size());
-        item.serialize_into(&mut bytes).ok()?;
-        Some(Cow::Owned(bytes))
+        item.serialize_into(&mut bytes)?;
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/roaring_bitmap_length/bo_roaring_bitmap_len_codec.rs

@@ -1,6 +1,6 @@
 use std::mem;
 
-use heed::BytesDecode;
+use heed::{BoxedError, BytesDecode};
 
 use crate::heed_codec::BytesDecodeOwned;
 
@@ -9,15 +9,15 @@ pub struct BoRoaringBitmapLenCodec;
 impl BytesDecode<'_> for BoRoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
-        Some((bytes.len() / mem::size_of::<u32>()) as u64)
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        Ok((bytes.len() / mem::size_of::<u32>()) as u64)
     }
 }
 
 impl BytesDecodeOwned for BoRoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         Self::bytes_decode(bytes)
     }
 }

milli/src/heed_codec/roaring_bitmap_length/cbo_roaring_bitmap_len_codec.rs

@@ -1,6 +1,6 @@
 use std::mem;
 
-use heed::BytesDecode;
+use heed::{BoxedError, BytesDecode};
 
 use super::{BoRoaringBitmapLenCodec, RoaringBitmapLenCodec};
 use crate::heed_codec::roaring_bitmap::cbo_roaring_bitmap_codec::THRESHOLD;
@@ -11,7 +11,7 @@ pub struct CboRoaringBitmapLenCodec;
 impl BytesDecode<'_> for CboRoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         if bytes.len() <= THRESHOLD * mem::size_of::<u32>() {
             // If there is threshold or less than threshold integers that can fit into this array
             // of bytes it means that we used the ByteOrder codec serializer.
@@ -27,7 +27,7 @@ impl BytesDecode<'_> for CboRoaringBitmapLenCodec {
 impl BytesDecodeOwned for CboRoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         Self::bytes_decode(bytes)
     }
 }

milli/src/heed_codec/roaring_bitmap_length/roaring_bitmap_len_codec.rs

@@ -2,6 +2,7 @@ use std::io::{self, BufRead, Read};
 use std::mem;
 
 use byteorder::{LittleEndian, ReadBytesExt};
+use heed::BoxedError;
 
 use crate::heed_codec::BytesDecodeOwned;
 
@@ -56,16 +57,16 @@ impl RoaringBitmapLenCodec {
 impl heed::BytesDecode<'_> for RoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
-        RoaringBitmapLenCodec::deserialize_from_slice(bytes).ok()
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        RoaringBitmapLenCodec::deserialize_from_slice(bytes).map_err(Into::into)
     }
 }
 
 impl BytesDecodeOwned for RoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
-        RoaringBitmapLenCodec::deserialize_from_slice(bytes).ok()
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        RoaringBitmapLenCodec::deserialize_from_slice(bytes).map_err(Into::into)
     }
 }
 

milli/src/heed_codec/script_language_codec.rs

@@ -1,30 +1,31 @@
 use std::borrow::Cow;
+use std::ffi::CStr;
 use std::str;
 
 use charabia::{Language, Script};
+use heed::BoxedError;
 
 pub struct ScriptLanguageCodec;
 
 impl<'a> heed::BytesDecode<'a> for ScriptLanguageCodec {
     type DItem = (Script, Language);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let sep = bytes.iter().position(|b| *b == 0)?;
-        let (s_bytes, l_bytes) = bytes.split_at(sep);
-        let script = str::from_utf8(s_bytes).ok()?;
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let cstr = CStr::from_bytes_until_nul(bytes)?;
+        let script = cstr.to_str()?;
         let script_name = Script::from_name(script);
-        let lan = str::from_utf8(l_bytes).ok()?;
         // skip '\0' byte between the two strings.
-        let lan_name = Language::from_name(&lan[1..]);
+        let lan = str::from_utf8(&bytes[script.len() + 1..])?;
+        let lan_name = Language::from_name(lan);
 
-        Some((script_name, lan_name))
+        Ok((script_name, lan_name))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for ScriptLanguageCodec {
     type EItem = (Script, Language);
 
-    fn bytes_encode((script, lan): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((script, lan): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let script_name = script.name().as_bytes();
         let lan_name = lan.name().as_bytes();
 
@@ -33,6 +34,6 @@ impl<'a> heed::BytesEncode<'a> for ScriptLanguageCodec {
         bytes.push(0);
         bytes.extend_from_slice(lan_name);
 
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/str_beu32_codec.rs

@@ -3,37 +3,41 @@ use std::convert::TryInto;
 use std::mem::size_of;
 use std::str;
 
+use heed::BoxedError;
+
+use super::SliceTooShortError;
+
 pub struct StrBEU32Codec;
 
 impl<'a> heed::BytesDecode<'a> for StrBEU32Codec {
     type DItem = (&'a str, u32);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let footer_len = size_of::<u32>();
 
         if bytes.len() < footer_len {
-            return None;
+            return Err(SliceTooShortError.into());
         }
 
         let (word, bytes) = bytes.split_at(bytes.len() - footer_len);
-        let word = str::from_utf8(word).ok()?;
-        let pos = bytes.try_into().map(u32::from_be_bytes).ok()?;
+        let word = str::from_utf8(word)?;
+        let pos = bytes.try_into().map(u32::from_be_bytes)?;
 
-        Some((word, pos))
+        Ok((word, pos))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for StrBEU32Codec {
     type EItem = (&'a str, u32);
 
-    fn bytes_encode((word, pos): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((word, pos): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let pos = pos.to_be_bytes();
 
         let mut bytes = Vec::with_capacity(word.len() + pos.len());
         bytes.extend_from_slice(word.as_bytes());
         bytes.extend_from_slice(&pos[..]);
 
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }
 
@@ -42,26 +46,27 @@ pub struct StrBEU16Codec;
 impl<'a> heed::BytesDecode<'a> for StrBEU16Codec {
     type DItem = (&'a str, u16);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let footer_len = size_of::<u16>();
 
         if bytes.len() < footer_len + 1 {
-            return None;
+            return Err(SliceTooShortError.into());
         }
 
         let (word_plus_nul_byte, bytes) = bytes.split_at(bytes.len() - footer_len);
-        let (_, word) = word_plus_nul_byte.split_last()?;
-        let word = str::from_utf8(word).ok()?;
-        let pos = bytes.try_into().map(u16::from_be_bytes).ok()?;
+        // unwrap: we just checked the footer + 1 above.
+        let (_, word) = word_plus_nul_byte.split_last().unwrap();
+        let word = str::from_utf8(word)?;
+        let pos = bytes.try_into().map(u16::from_be_bytes)?;
 
-        Some((word, pos))
+        Ok((word, pos))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for StrBEU16Codec {
     type EItem = (&'a str, u16);
 
-    fn bytes_encode((word, pos): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((word, pos): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let pos = pos.to_be_bytes();
 
         let mut bytes = Vec::with_capacity(word.len() + 1 + pos.len());
@@ -69,6 +74,6 @@ impl<'a> heed::BytesEncode<'a> for StrBEU16Codec {
         bytes.push(0);
         bytes.extend_from_slice(&pos[..]);
 
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/str_ref.rs

@@ -1,6 +1,6 @@
 use std::borrow::Cow;
 
-use heed::{BytesDecode, BytesEncode};
+use heed::{BoxedError, BytesDecode, BytesEncode};
 
 /// A codec for values of type `&str`. Unlike `Str`, its `EItem` and `DItem` associated
 /// types are equivalent (= `&'a str`) and these values can reside within another structure.
@@ -8,15 +8,14 @@ pub struct StrRefCodec;
 impl<'a> BytesEncode<'a> for StrRefCodec {
     type EItem = &'a str;
 
-    fn bytes_encode(item: &'a &'a str) -> Option<Cow<'a, [u8]>> {
-        Some(Cow::Borrowed(item.as_bytes()))
+    fn bytes_encode(item: &'a &'a str) -> Result<Cow<'a, [u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item.as_bytes()))
     }
 }
 impl<'a> BytesDecode<'a> for StrRefCodec {
     type DItem = &'a str;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let s = std::str::from_utf8(bytes).ok()?;
-        Some(s)
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        std::str::from_utf8(bytes).map_err(Into::into)
     }
 }

milli/src/heed_codec/str_str_u8_codec.rs

@@ -1,32 +1,36 @@
 use std::borrow::Cow;
+use std::ffi::CStr;
 use std::str;
 
+use heed::BoxedError;
+
+use super::SliceTooShortError;
+
 pub struct U8StrStrCodec;
 
 impl<'a> heed::BytesDecode<'a> for U8StrStrCodec {
     type DItem = (u8, &'a str, &'a str);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (n, bytes) = bytes.split_first()?;
-        let s1_end = bytes.iter().position(|b| *b == 0)?;
-        let (s1_bytes, rest) = bytes.split_at(s1_end);
-        let s2_bytes = &rest[1..];
-        let s1 = str::from_utf8(s1_bytes).ok()?;
-        let s2 = str::from_utf8(s2_bytes).ok()?;
-        Some((*n, s1, s2))
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let (n, bytes) = bytes.split_first().ok_or(SliceTooShortError)?;
+        let cstr = CStr::from_bytes_until_nul(bytes)?;
+        let s1 = cstr.to_str()?;
+        // skip '\0' byte between the two strings.
+        let s2 = str::from_utf8(&bytes[s1.len() + 1..])?;
+        Ok((*n, s1, s2))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for U8StrStrCodec {
     type EItem = (u8, &'a str, &'a str);
 
-    fn bytes_encode((n, s1, s2): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((n, s1, s2): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(s1.len() + s2.len() + 1);
         bytes.push(*n);
         bytes.extend_from_slice(s1.as_bytes());
         bytes.push(0);
         bytes.extend_from_slice(s2.as_bytes());
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }
 pub struct UncheckedU8StrStrCodec;
@@ -34,24 +38,25 @@ pub struct UncheckedU8StrStrCodec;
 impl<'a> heed::BytesDecode<'a> for UncheckedU8StrStrCodec {
     type DItem = (u8, &'a [u8], &'a [u8]);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (n, bytes) = bytes.split_first()?;
-        let s1_end = bytes.iter().position(|b| *b == 0)?;
-        let (s1_bytes, rest) = bytes.split_at(s1_end);
-        let s2_bytes = &rest[1..];
-        Some((*n, s1_bytes, s2_bytes))
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let (n, bytes) = bytes.split_first().ok_or(SliceTooShortError)?;
+        let cstr = CStr::from_bytes_until_nul(bytes)?;
+        let s1_bytes = cstr.to_bytes();
+        // skip '\0' byte between the two strings.
+        let s2_bytes = &bytes[s1_bytes.len() + 1..];
+        Ok((*n, s1_bytes, s2_bytes))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for UncheckedU8StrStrCodec {
     type EItem = (u8, &'a [u8], &'a [u8]);
 
-    fn bytes_encode((n, s1, s2): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((n, s1, s2): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(s1.len() + s2.len() + 1);
         bytes.push(*n);
         bytes.extend_from_slice(s1);
         bytes.push(0);
         bytes.extend_from_slice(s2);
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }