Make hannoy the default vector store when creating an index

Fix embedding skip

crates/meilisearch/tests/search/hybrid.rs

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

crates/milli/src/vector/session.rs

@@ -112,13 +112,12 @@ impl<'doc, C: OnEmbed<'doc>, I: Input> EmbedSession<'doc, C, I> {
         rendered: I,
         unused_vectors_distribution: &C::ErrorMetadata,
     ) -> Result<()> {
-        if self.inputs.len() < self.inputs.capacity() {
-            self.inputs.push(rendered);
-            self.metadata.push(metadata);
-            return Ok(());
+        if self.inputs.len() >= self.inputs.capacity() {
+            self.embed_chunks(unused_vectors_distribution)?;
         }
-
-        self.embed_chunks(unused_vectors_distribution)
+        self.inputs.push(rendered);
+        self.metadata.push(metadata);
+        Ok(())
     }
 
     pub fn drain(mut self, unused_vectors_distribution: &C::ErrorMetadata) -> Result<C> {

crates/milli/src/vector/store.rs

@@ -27,10 +27,10 @@ const HANNOY_M0: usize = 32;
     utoipa::ToSchema,
 )]
 pub enum VectorStoreBackend {
-    #[default]
     #[deserr(rename = "stable")]
     #[serde(rename = "stable")]
     Arroy,
+    #[default]
     #[deserr(rename = "experimental")]
     #[serde(rename = "experimental")]
     Hannoy,