diff --git a/modelscope_agent/agents/role_play.py b/modelscope_agent/agents/role_play.py
index 31f302a9..f9a36998 100644
--- a/modelscope_agent/agents/role_play.py
+++ b/modelscope_agent/agents/role_play.py
@@ -6,6 +6,7 @@
 from modelscope_agent.agent_env_util import AgentEnvMixin
 from modelscope_agent.llm.base import BaseChatModel
 from modelscope_agent.tools.base import BaseTool
+from modelscope_agent.utils.logger import agent_logger as logger
 from modelscope_agent.utils.tokenization_utils import count_tokens
 from modelscope_agent.utils.utils import check_and_limit_input_length
 
@@ -260,11 +261,13 @@ def _run(self,
             planning_prompt = self.llm.build_raw_prompt(messages)
 
         max_turn = 10
+        call_llm_count = 0
         while True and max_turn > 0:
             # print('=====one input planning_prompt======')
             # print(planning_prompt)
             # print('=============Answer=================')
             max_turn -= 1
+            call_llm_count += 1
             if self.llm.support_function_calling():
                 output = self.llm.chat_with_functions(
                     messages=messages,
@@ -282,6 +285,7 @@ def _run(self,
                     **kwargs)
 
             llm_result = ''
+            logger.info(f'call llm {call_llm_count} times output: {output}')
             for s in output:
                 if isinstance(s, dict):
                     llm_result = s
diff --git a/modelscope_agent/llm/__init__.py b/modelscope_agent/llm/__init__.py
index 2a1f6169..d63714aa 100644
--- a/modelscope_agent/llm/__init__.py
+++ b/modelscope_agent/llm/__init__.py
@@ -5,7 +5,7 @@
 from .modelscope import ModelScopeChatGLM, ModelScopeLLM
 from .ollama import OllamaLLM
 from .openai import OpenAi
-from .zhipu import GLM4, ZhipuLLM
+from .zhipu import ZhipuLLM
 
 
 def get_chat_model(model: str, model_server: str, **kwargs) -> BaseChatModel:
@@ -26,5 +26,5 @@ def get_chat_model(model: str, model_server: str, **kwargs) -> BaseChatModel:
 
 __all__ = [
     'LLM_REGISTRY', 'BaseChatModel', 'OpenAi', 'DashScopeLLM', 'QwenChatAtDS',
-    'ModelScopeLLM', 'ModelScopeChatGLM', 'ZhipuLLM', 'GLM4', 'OllamaLLM'
+    'ModelScopeLLM', 'ModelScopeChatGLM', 'ZhipuLLM', 'OllamaLLM'
 ]
diff --git a/modelscope_agent/llm/base.py b/modelscope_agent/llm/base.py
index 06df1383..05c14da9 100644
--- a/modelscope_agent/llm/base.py
+++ b/modelscope_agent/llm/base.py
@@ -211,6 +211,9 @@ def support_function_calling(self) -> bool:
                 if response.get('function_call', None):
                     # logger.info('Support of function calling is detected.')
                     self._support_fn_call = True
+                if response.get('tool_calls', None):
+                    # logger.info('Support of function calling is detected.')
+                    self._support_fn_call = True
             except FnCallNotImplError:
                 pass
             except AttributeError:
diff --git a/modelscope_agent/llm/zhipu.py b/modelscope_agent/llm/zhipu.py
index 643ab30b..aed5a608 100644
--- a/modelscope_agent/llm/zhipu.py
+++ b/modelscope_agent/llm/zhipu.py
@@ -1,6 +1,7 @@
 import os
 from typing import Dict, Iterator, List, Optional
 
+from modelscope_agent.utils.logger import agent_logger as logger
 from zhipuai import ZhipuAI
 
 from .base import BaseChatModel, register_llm
@@ -31,9 +32,12 @@ class ZhipuLLM(BaseChatModel):
     Universal LLM model interface on zhipu
     """
 
-    def __init__(self, model: str, model_server: str, **kwargs):
-        super().__init__(model, model_server)
-        self._support_fn_call = True
+    def __init__(self,
+                 model: str,
+                 model_server: str,
+                 support_fn_call: bool = True,
+                 **kwargs):
+        super().__init__(model, model_server, support_fn_call=support_fn_call)
         api_key = kwargs.get('api_key', os.getenv('ZHIPU_API_KEY', '')).strip()
         assert api_key, 'ZHIPU_API_KEY is required.'
         self.client = ZhipuAI(api_key=api_key)
@@ -45,7 +49,8 @@ def _chat_stream(self,
                      **kwargs) -> Iterator[str]:
         if not functions or not len(functions):
             tool_choice = 'none'
-        print(f'====> stream messages: {messages}')
+        logger.info(
+            f'====> stream messages: {messages}, functions: {functions}')
         response = self.client.chat.completions.create(
             model=self.model,
             messages=messages,
@@ -62,18 +67,15 @@ def _chat_no_stream(self,
                         **kwargs) -> str:
         if not functions or not len(functions):
             tool_choice = 'none'
-        print(f'====> no stream messages: {messages}')
+        logger.info(
+            f'====> no stream messages: {messages}, functions: {functions}')
         response = self.client.chat.completions.create(
             model=self.model,
             messages=messages,
             tools=functions,
             tool_choice=tool_choice,
         )
-        return response.choices[0].message
+        message = response.choices[0].message
+        output = message.content if not functions else [message.model_dump()]
 
-
-@register_llm('glm-4')
-class GLM4(ZhipuLLM):
-    """
-    glm-4 from zhipu
-    """
+        return output
diff --git a/modelscope_agent/utils/retry.py b/modelscope_agent/utils/retry.py
index 314b7148..ea098d61 100644
--- a/modelscope_agent/utils/retry.py
+++ b/modelscope_agent/utils/retry.py
@@ -1,5 +1,6 @@
 import time
 from functools import wraps
+from traceback import format_exc
 
 from modelscope_agent.utils.logger import agent_logger as logger
 
@@ -26,9 +27,9 @@ def wrapper(*args, **kwargs):
                     return func(*args, **kwargs)
                 except AssertionError as e:
                     raise AssertionError(e)
-                except Exception as e:
+                except Exception:
                     logger.warning(
-                        f'Attempt to run {func.__name__} {attempts + 1} failed: {e}'
+                        f'Attempt to run {func.__name__} {attempts + 1} failed: {format_exc()}'
                     )
                     attempts += 1
                     time.sleep(delay_seconds)