diff --git a/api/app/core/memory/read_services/content_search.py b/api/app/core/memory/read_services/content_search.py
index 58356e84..54d99060 100644
--- a/api/app/core/memory/read_services/content_search.py
+++ b/api/app/core/memory/read_services/content_search.py
@@ -12,8 +12,8 @@ from app.repositories.neo4j.neo4j_connector import Neo4jConnector
 
 logger = logging.getLogger(__name__)
 
-DEFAULT_ALPHA = 0.7
-DEFAULT_FULLTEXT_SCORE_THRESHOLD = 1
+DEFAULT_ALPHA = 0.6
+DEFAULT_FULLTEXT_SCORE_THRESHOLD = 1.5
 DEFAULT_COSINE_SCORE_THRESHOLD = 0.5
 DEFAULT_CONTENT_SCORE_THRESHOLD = 0.5
 
@@ -112,7 +112,7 @@ class Neo4jSearchService:
             kw = float(combined[item_id].get("kw_score", 0) or 0)
             emb = float(combined[item_id].get("embedding_score", 0) or 0)
             base = self.alpha * emb + (1 - self.alpha) * kw
-            combined[item_id]["content_score"] = base + min(1 - base, kw * emb)
+            combined[item_id]["content_score"] = base + min(1 - base, 0.1 * kw * emb)
         results = sorted(combined.values(), key=lambda x: x["content_score"], reverse=True)
         # results = [
         #     res for res in results
diff --git a/api/app/core/memory/read_services/result_builder.py b/api/app/core/memory/read_services/result_builder.py
index 949ff3ed..dd376c7c 100644
--- a/api/app/core/memory/read_services/result_builder.py
+++ b/api/app/core/memory/read_services/result_builder.py
@@ -61,14 +61,18 @@ class EntityBuilder(BaseBuilder):
     def data(self) -> dict:
         return {
             "id": self.record.get("id"),
-            "content": self.record.get("name"),
+            "name": self.record.get("name"),
+            "description": self.record.get("description"),
             "kw_score": self.record.get("kw_score", 0.0),
             "emb_score": self.record.get("embedding_score", 0.0)
         }
 
     @property
     def content(self) -> str:
-        return self.record.get("name")
+        return (f"<entity>"
+                f"<name>{self.record.get("name")}<name>"
+                f"<description>{self.record.get("description")}</description>"
+                f"</entity>")
 
 
 class SummaryBuilder(BaseBuilder):
diff --git a/api/app/repositories/neo4j/create_indexes.py b/api/app/repositories/neo4j/create_indexes.py
index 7caeea8a..0a9aaf71 100644
--- a/api/app/repositories/neo4j/create_indexes.py
+++ b/api/app/repositories/neo4j/create_indexes.py
@@ -19,7 +19,8 @@ async def create_fulltext_indexes():
         # """)
         # 创建 Entities 索引
         await connector.execute_query("""
-            CREATE FULLTEXT INDEX entitiesFulltext IF NOT EXISTS FOR (e:ExtractedEntity) ON EACH [e.name]
+            CREATE FULLTEXT INDEX entitiesFulltext IF NOT EXISTS 
+            FOR (e:ExtractedEntity) ON EACH [e.name, e.description, e.aliases]
             OPTIONS { indexConfig: { `fulltext.analyzer`: 'cjk' } }
         """)
 
@@ -139,6 +140,16 @@ async def create_vector_indexes():
         await connector.close()
 
 
+async def create_user_indexes():
+    connector = Neo4jConnector()
+    await connector.execute_query(
+        """
+        CREATE INDEX user_perceptual IF NOT EXISTS
+        FOR (p:Perceptual) ON (p.end_user_id);
+        """
+    )
+
+
 async def create_unique_constraints():
     """Create uniqueness constraints for core node identifiers.
     Ensures concurrent MERGE operations remain safe and prevents duplicates.
diff --git a/api/app/repositories/neo4j/graph_search.py b/api/app/repositories/neo4j/graph_search.py
index 354c0e23..70913267 100644
--- a/api/app/repositories/neo4j/graph_search.py
+++ b/api/app/repositories/neo4j/graph_search.py
@@ -45,14 +45,17 @@ def cosine_similarity_search(
     vectors: np.ndarray = np.array(vectors, dtype=np.float32)
     vectors_norm = vectors / np.linalg.norm(vectors, axis=1, keepdims=True)
     query: np.ndarray = np.array(query, dtype=np.float32)
-    query_norm = query / np.linalg.norm(query)
+    norm = np.linalg.norm(query)
+    if norm == 0:
+        return {}
+    query_norm = query / norm
 
     similarities = vectors_norm @ query_norm
     similarities = np.clip(similarities, 0, 1)
     top_k = min(limit, similarities.shape[0])
     if top_k <= 0:
         return {}
-    top_indices = np.argpartition(-similarities, top_k - 1)[-top_k:]
+    top_indices = np.argpartition(-similarities, top_k - 1)[:top_k]
     top_indices = top_indices[np.argsort(-similarities[top_indices])]
     result = {}
     for idx in top_indices:
@@ -510,7 +513,7 @@ async def search_graph_by_embedding(
     task_keys = []
 
     for node_type in include:
-        tasks.append(search_by_embedding(connector, node_type, end_user_id, embedding, limit))
+        tasks.append(search_by_embedding(connector, node_type, end_user_id, embedding, limit*2))
         task_keys.append(node_type.value)
 
     task_results = await asyncio.gather(*tasks, return_exceptions=True)