diff --git a/__init__.py b/__init__.py
index 081e814..4315115 100644
--- a/__init__.py
+++ b/__init__.py
@@ -83,6 +83,8 @@ class MZ_CustomizeInstruct:
             "required": {
                 "system": ("STRING", {"multiline": True, "default": mz_prompts.Long_prompt}),
                 "instruct": ("STRING", {"multiline": True, "default": "Short: {text}"}),
+                "prefix": ("STRING", {"default": "Long: "}),
+                "suffix": ("STRING", {"default": ""}),
             },
         }
 
diff --git a/mz_llama_cpp.py b/mz_llama_cpp.py
index 5840895..35e05c7 100644
--- a/mz_llama_cpp.py
+++ b/mz_llama_cpp.py
@@ -482,11 +482,7 @@ def base_query_beautify_prompt_text(args_dict):
             mz_prompt_utils.Utils.print_log(f"question: {question}")
 
 
-        
-
-
         if schema is not None:
-
             response_json = llama_cpp_simple_interrogator_to_json(
                 model_file=model_file, 
                 system=system_prompt,
@@ -494,9 +490,7 @@ def base_query_beautify_prompt_text(args_dict):
                 schema=schema,
                 options=options,
             ) 
-            mz_prompt_utils.Utils.print_log(f"response_json: {response_json}")
-            if keep_device is False:
-                freed_gpu_memory(model_file=model_file)
+            mz_prompt_utils.Utils.print_log(f"response_json: {response_json}") 
 
             response = json.loads(response_json)
             full_responses = []
@@ -535,8 +529,20 @@ def base_query_beautify_prompt_text(args_dict):
                 options=options,
             )
 
+            prefix = customize_instruct.get("prefix", "")
+            # 去除前缀
+            if prefix != "":
+                full_response = full_response.lstrip(prefix)
+
+            suffix = customize_instruct.get("suffix", "")
+            # 去除后缀
+            if suffix != "":
+                full_response = full_response.rstrip(suffix)
+
+
+        if keep_device is False:
+            freed_gpu_memory(model_file=model_file)
 
-        
         # 去除换行
         while full_response.find("\n") != -1:
             full_response = full_response.replace("\n", " ")