diff --git a/metagpt/llm.py b/metagpt/llm.py
index 4c9993441..30ced25d2 100644
--- a/metagpt/llm.py
+++ b/metagpt/llm.py
@@ -5,13 +5,15 @@
 @Author  : alexanderwu
 @File    : llm.py
 """
+from typing import Optional
 
-
+from metagpt.configs.llm_config import LLMConfig
 from metagpt.context import CONTEXT
 from metagpt.provider.base_llm import BaseLLM
 
 
-def LLM() -> BaseLLM:
+def LLM(llm_config: Optional[LLMConfig] = None) -> BaseLLM:
     """get the default llm provider if name is None"""
-    # context.use_llm(name=name, provider=provider)
+    if llm_config is not None:
+        CONTEXT.llm_with_cost_manager_from_llm_config(llm_config)
     return CONTEXT.llm()
diff --git a/metagpt/roles/engineer.py b/metagpt/roles/engineer.py
index bc56ca813..8b0895a69 100644
--- a/metagpt/roles/engineer.py
+++ b/metagpt/roles/engineer.py
@@ -110,7 +110,7 @@ class Engineer(Role):
             # Code review
             if review:
                 action = WriteCodeReview(i_context=coding_context, context=self.context, llm=self.llm)
-                self._init_action_system_message(action)
+                self._init_action(action)
                 coding_context = await action.run()
             await src_file_repo.save(
                 coding_context.filename,
diff --git a/metagpt/roles/role.py b/metagpt/roles/role.py
index 10b60d30e..e7e5ead84 100644
--- a/metagpt/roles/role.py
+++ b/metagpt/roles/role.py
@@ -146,7 +146,7 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
         super().__init__(**data)
 
         if self.is_human:
-            self.llm = HumanProvider()
+            self.llm = HumanProvider(None)
 
         self.llm.system_prompt = self._get_prefix()
         self._watch(data.get("watch") or [UserRequirement])
@@ -222,7 +222,8 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
     def _setting(self):
         return f"{self.name}({self.profile})"
 
-    def _init_action_system_message(self, action: Action):
+    def _init_action(self, action: Action):
+        action.set_llm(self.llm, override=False)
         action.set_prefix(self._get_prefix())
 
     def set_action(self, action: Action):
@@ -238,7 +239,7 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
         self._reset()
         for action in actions:
             if not isinstance(action, Action):
-                i = action(name="", llm=self.llm)
+                i = action()
             else:
                 if self.is_human and not isinstance(action.llm, HumanProvider):
                     logger.warning(
@@ -247,7 +248,7 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
                         f"try passing in Action classes instead of initialized instances"
                     )
                 i = action
-            self._init_action_system_message(i)
+            self._init_action(i)
             self.actions.append(i)
             self.states.append(f"{len(self.actions)}. {action}")
 
diff --git a/tests/metagpt/provider/test_spark_api.py b/tests/metagpt/provider/test_spark_api.py
index 213c19676..2cb6bf559 100644
--- a/tests/metagpt/provider/test_spark_api.py
+++ b/tests/metagpt/provider/test_spark_api.py
@@ -1,9 +1,11 @@
 #!/usr/bin/env python
 # -*- coding: utf-8 -*-
 # @Desc   : the unittest of spark api
+from pathlib import Path
 
 import pytest
 
+from metagpt.config2 import Config
 from metagpt.provider.spark_api import GetMessageFromWeb, SparkLLM
 from tests.metagpt.provider.mock_llm_config import mock_llm_config
 
@@ -33,6 +35,14 @@ def mock_spark_get_msg_from_web_run(self) -> str:
     return resp_content
 
 
+@pytest.mark.asyncio
+async def test_spark_aask():
+    llm = SparkLLM(Config.model_validate_yaml(Path.home() / ".metagpt" / "spark.yaml").llm)
+
+    resp = await llm.aask("Hello!")
+    print(resp)
+
+
 @pytest.mark.asyncio
 async def test_spark_acompletion(mocker):
     mocker.patch("metagpt.provider.spark_api.GetMessageFromWeb.run", mock_spark_get_msg_from_web_run)
diff --git a/tests/metagpt/test_context_mixin.py b/tests/metagpt/test_context_mixin.py
index 2c237b2ec..cc202a473 100644
--- a/tests/metagpt/test_context_mixin.py
+++ b/tests/metagpt/test_context_mixin.py
@@ -5,10 +5,15 @@
 @Author  : alexanderwu
 @File    : test_context_mixin.py
 """
+import pytest
 from pydantic import BaseModel
 
+from metagpt.actions import Action
 from metagpt.config2 import Config
 from metagpt.context_mixin import ContextMixin
+from metagpt.environment import Environment
+from metagpt.roles import Role
+from metagpt.team import Team
 from tests.metagpt.provider.mock_llm_config import (
     mock_llm_config,
     mock_llm_config_proxy,
@@ -91,3 +96,27 @@ def test_config_mixin_4_multi_inheritance_override_config():
     print(obj.__dict__.keys())
     assert "private_config" in obj.__dict__.keys()
     assert obj.llm.model == "mock_zhipu_model"
+
+
+@pytest.mark.asyncio
+async def test_debate_two_roles():
+    config = Config.default()
+    config.llm.model = "gpt-4-1106-preview"
+    action1 = Action(config=config, name="AlexSay", instruction="Say your opinion with emotion and don't repeat it")
+    action2 = Action(name="BobSay", instruction="Say your opinion with emotion and don't repeat it")
+    biden = Role(
+        name="Alex", profile="Democratic candidate", goal="Win the election", actions=[action1], watch=[action2]
+    )
+    trump = Role(
+        name="Bob", profile="Republican candidate", goal="Win the election", actions=[action2], watch=[action1]
+    )
+    env = Environment(desc="US election live broadcast")
+    team = Team(investment=10.0, env=env, roles=[biden, trump])
+
+    print(action1.llm.system_prompt)
+    print(action2.llm.system_prompt)
+    print(biden.llm.system_prompt)
+    print(trump.llm.system_prompt)
+
+    history = await team.run(idea="Topic: climate change. Under 80 words per message.", send_to="Alex", n_round=3)
+    assert "Alex" in history