diff --git a/kag/builder/component/extractor/knowledge_unit_extractor.py b/kag/builder/component/extractor/knowledge_unit_extractor.py
index 6d99c03a6..57f787985 100644
--- a/kag/builder/component/extractor/knowledge_unit_extractor.py
+++ b/kag/builder/component/extractor/knowledge_unit_extractor.py
@@ -584,10 +584,15 @@ def triple_to_knowledge_unit(triple):
                 {"name": knowledge_id, "category": "KnowledgeUnit"}
             )
             core_entities = {}
-            for item in knowledge_value.get("core_entities", "").split(","):
-                if not item.strip():
-                    continue
-                core_entities[item.strip()] = "Others"
+            raw_core_entities = knowledge_value.get("core_entities", {})
+            if isinstance(raw_core_entities, dict):
+                for key, val in raw_core_entities.items():
+                    if key.strip():
+                        core_entities[key.strip()] = val if val else "Others"
+            else:
+                for item in str(raw_core_entities).split(","):
+                    if item.strip():
+                        core_entities[item.strip()] = "Others"
 
             for core_entity, ent_type in core_entities.items():
                 if core_entity == "":
diff --git a/knext/command/sub_command/project.py b/knext/command/sub_command/project.py
index d9e80c17a..1ac634a6d 100644
--- a/knext/command/sub_command/project.py
+++ b/knext/command/sub_command/project.py
@@ -165,7 +165,7 @@ def create_project(
 
     llm_config_checker = LLMConfigChecker()
     vectorize_model_config_checker = VectorizeModelConfigChecker()
-    llm_config = config.get("chat_llm", {})
+    llm_config = config.get("llm", config.get("chat_llm", {}))
     vectorize_model_config = config.get("vectorizer", {})
     try:
         llm_config_checker.check(json.dumps(llm_config))
@@ -260,7 +260,7 @@ def update_project(proj_path):
 
     llm_config_checker = LLMConfigChecker()
     vectorize_model_config_checker = VectorizeModelConfigChecker()
-    llm_config = env.config.get("chat_llm", {})
+    llm_config = env.config.get("llm", env.config.get("chat_llm", {}))
     vectorize_model_config = env.config.get("vectorizer", {})
     try:
         llm_config_checker.check(json.dumps(llm_config))