Fix: Ensure embedding indexes are created even when schemas are reused

When wikidata schemas already exist (from previous imports), they are reused
and new_schema() is not called, which means post_schema_add() is never called,
which means embedding indexes are never created.

Add ensure_embedding_indexes() method that explicitly creates embedding indexes
after schema initialization, regardless of whether schemas were newly created
or reused. This ensures embedding search works correctly for all imports.

The method:
- Checks if embedding service is available
- Creates embedding index for entity metadata description field
- Handles case where index already exists gracefully

This fixes the issue where 'similar' command in interactive mode fails with
'Index not found: CellDoesNotExisted' error.

Author: shisoft

src/apps/wikidata/importer.rs

@@ -556,6 +556,46 @@ impl WikidataImporter {
             println!("  TYPE_LIST field definition: {:?}", *ID_TYPE_LIST);
         }
 
+        // Ensure embedding indexes are created for entity metadata description field
+        // This is needed when schemas are reused from previous imports
+        self.ensure_embedding_indexes().await?;
+
+        Ok(())
+    }
+
+    /// Ensure embedding indexes exist for schema fields with IndexType::Embedding
+    /// This is needed because when schemas are reused (not newly created),
+    /// post_schema_add() is not called and indexes don't get created.
+    async fn ensure_embedding_indexes(&self) -> Result<(), ImportError> {
+        println!("\n🔍 Ensuring embedding indexes are created...");
+        
+        let embedding_client = &self.server.neb_server.indexer.as_ref()
+            .ok_or_else(|| ImportError::Schema("Indexer not available".to_string()))?
+            .clients
+            .embedding_client;
+
+        // Check if embedding service is available
+        if !embedding_client.is_embedding_index_core_set() {
+            println!("⚠️  Embedding service not initialized, skipping embedding index creation");
+            return Ok(());
+        }
+
+        // Create embedding index for entity metadata description field
+        let description_field_id = bifrost_hasher::hash_str("description");
+        let model = neb::index::embedding::EmbeddingModel::new("default");
+
+        println!("  Creating embedding index for entity metadata descriptions...");
+        match embedding_client
+            .new_index(ENTITY_METADATA_SCHEMA_ID, description_field_id, &model)
+            .await
+        {
+            Ok(_) => println!("  ✓ Embedding index created successfully"),
+            Err(e) => {
+                // Index might already exist, which is fine
+                println!("  ℹ️  Embedding index creation result: {:?}", e);
+            }
+        }
+
         Ok(())
     }