first implementation

app/lib/features/geckoview/features/tabs/domain/repositories/gecko_inference.dart

@@ -134,6 +134,45 @@ class GeckoInferenceRepository extends _$GeckoInferenceRepository {
     return neighbors;
   }
 
+  Future<List<List<String>>?> suggestClusters({
+    required List<String> unassignedDocumentsInput,
+  }) async {
+    if (!ref.read(
+      generalSettingsWithDefaultsProvider.select(
+        (settings) => settings.enableLocalAiFeatures,
+      ),
+    )) {
+      return null;
+    }
+
+    final processedDocuments = <String, String>{};
+    final unassignedDocumentsProcessed = unassignedDocumentsInput.map((doc) {
+      final processed = preprocessText(doc);
+      if (processed != doc) {
+        processedDocuments[processed] = doc;
+      }
+
+      return processed;
+    }).toList();
+
+    final embeddings = await generateDocumentEmbeddings(
+      unassignedDocumentsProcessed,
+    );
+
+    final clusters = embeddings.mapNotNull(
+      (embeddings) => clusterEmbeddings(embeddings: embeddings.values.toList())
+          .map(
+            (cluster) =>
+                cluster.map((i) => embeddings.keys.elementAt(i)).toList(),
+          )
+          .toList(),
+    );
+
+    print(clusters);
+
+    return null;
+  }
+
   Future<Map<String, List<double>>?> generateDocumentEmbeddings(
     List<String> documents,
   ) async {
@@ -196,6 +235,32 @@ Future<String?> containerTopic(Ref ref, String containerId) async {
   return topic;
 }
 
+@Riverpod()
+Future<List<List<String>>?> suggestClusters(Ref ref) async {
+  final unassignedTitles = await ref.watch(
+    containerTabsDataProvider(null).selectAsync(
+      (tabData) => EquatableValue(
+        tabData
+            .where((tab) => tab.title.isNotEmpty)
+            .map((tab) => (tab.id, tab.title!))
+            .toSet(),
+      ),
+    ),
+  );
+
+  if (unassignedTitles.value.isNotEmpty) {
+    await ref
+        .read(geckoInferenceRepositoryProvider.notifier)
+        .suggestClusters(
+          unassignedDocumentsInput: unassignedTitles.value
+              .map((tab) => tab.$2)
+              .toList(),
+        );
+  }
+
+  return null;
+}
+
 @Riverpod()
 Future<List<String>?> containerTabSuggestions(
   Ref ref,

app/lib/features/geckoview/features/tabs/domain/repositories/gecko_inference.g.dart

@@ -42,7 +42,7 @@ final class GeckoInferenceRepositoryProvider
 }
 
 String _$geckoInferenceRepositoryHash() =>
-    r'6099998a3f49fc4faf650878ab4afd6268c593e2';
+    r'1f08047bee912a475bfd0851f8a9868a41bca836';
 
 abstract class _$GeckoInferenceRepository extends $Notifier<void> {
   void build();
@@ -132,6 +132,47 @@ final class ContainerTopicFamily extends $Family
   String toString() => r'containerTopicProvider';
 }
 
+@ProviderFor(suggestClusters)
+const suggestClustersProvider = SuggestClustersProvider._();
+
+final class SuggestClustersProvider
+    extends
+        $FunctionalProvider<
+          AsyncValue<List<List<String>>?>,
+          List<List<String>>?,
+          FutureOr<List<List<String>>?>
+        >
+    with
+        $FutureModifier<List<List<String>>?>,
+        $FutureProvider<List<List<String>>?> {
+  const SuggestClustersProvider._()
+    : super(
+        from: null,
+        argument: null,
+        retry: null,
+        name: r'suggestClustersProvider',
+        isAutoDispose: true,
+        dependencies: null,
+        $allTransitiveDependencies: null,
+      );
+
+  @override
+  String debugGetCreateSourceHash() => _$suggestClustersHash();
+
+  @$internal
+  @override
+  $FutureProviderElement<List<List<String>>?> $createElement(
+    $ProviderPointer pointer,
+  ) => $FutureProviderElement(pointer);
+
+  @override
+  FutureOr<List<List<String>>?> create(Ref ref) {
+    return suggestClusters(ref);
+  }
+}
+
+String _$suggestClustersHash() => r'9d212158b80ba53afe305db938ab399b1e931d88';
+
 @ProviderFor(containerTabSuggestions)
 const containerTabSuggestionsProvider = ContainerTabSuggestionsFamily._();