fix llm set

examples/debate_simple.py

@@ -13,7 +13,9 @@ from metagpt.roles import Role
 from metagpt.team import Team
 
 action1 = Action(name="AlexSay", instruction="Express your opinion with emotion and don't repeat it")
+action1.llm.model = "gpt-4-1106-preview"
 action2 = Action(name="BobSay", instruction="Express your opinion with emotion and don't repeat it")
+action2.llm.model = "gpt-3.5-turbo-1106"
 alex = Role(name="Alex", profile="Democratic candidate", goal="Win the election", actions=[action1], watch=[action2])
 bob = Role(name="Bob", profile="Republican candidate", goal="Win the election", actions=[action2], watch=[action1])
 env = Environment(desc="US election live broadcast")

metagpt/actions/action.py

@@ -25,7 +25,7 @@ from metagpt.utils.file_repository import FileRepository
 
 
 class Action(SerializationMixin, ContextMixin, BaseModel):
-    model_config = ConfigDict(arbitrary_types_allowed=True, exclude=["llm"])
+    model_config = ConfigDict(arbitrary_types_allowed=True)
 
     name: str = ""
     i_context: Union[dict, CodingContext, CodeSummarizeContext, TestingContext, RunCodeContext, str, None] = ""

metagpt/context_mixin.py

@@ -57,6 +57,8 @@ class ContextMixin(BaseModel):
     def set_config(self, config: Config, override=False):
         """Set config"""
         self.set("private_config", config, override)
+        if config is not None:
+            _ = self.llm  # init llm
 
     def set_llm(self, llm: BaseLLM, override=False):
         """Set llm"""

metagpt/provider/openai_api.py

@@ -220,10 +220,12 @@ class OpenAILLM(BaseLLM):
 
     @handle_exception
     def _update_costs(self, usage: CompletionUsage):
-        if self.config.calc_usage and usage:
+        if self.config.calc_usage and usage and self.cost_manager:
             self.cost_manager.update_cost(usage.prompt_tokens, usage.completion_tokens, self.model)
 
     def get_costs(self) -> Costs:
+        if not self.cost_manager:
+            return Costs()
         return self.cost_manager.get_costs()
 
     def _get_max_tokens(self, messages: list[dict]):

metagpt/roles/role.py

@@ -131,6 +131,13 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
 
     role_id: str = ""
     states: list[str] = []
+
+    # scenarios to set action system_prompt:
+    #   1. `__init__` while using Role(actions=[...])
+    #   2. add action to role while using `role.set_action(action)`
+    #   3. set_todo while using `role.set_todo(action)`
+    #   4. when role.system_prompt is being updated (e.g. by `role.system_prompt = "..."`)
+    # Additional, if llm is not set, we will use role's llm
     actions: list[SerializeAsAny[Action]] = Field(default=[], validate_default=True)
     rc: RoleContext = Field(default_factory=RoleContext)
     addresses: set[str] = set()
@@ -222,6 +229,12 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
     def _setting(self):
         return f"{self.name}({self.profile})"
 
+    @model_validator(mode="after")
+    def _check_actions(self):
+        """Check actions and set llm and prefix for each action."""
+        self.set_actions(self.actions)
+        return self
+
     def _init_action(self, action: Action):
         action.set_llm(self.llm, override=False)
         action.set_prefix(self._get_prefix())
@@ -306,6 +319,7 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
         if env:
             env.set_addresses(self, self.addresses)
             self.llm.system_prompt = self._get_prefix()
+            self.set_actions(self.actions)  # reset actions to update llm and prefix
 
     def _get_prefix(self):
         """Get the role prefix"""
@@ -318,7 +332,8 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
             prefix += CONSTRAINT_TEMPLATE.format(**{"constraints": self.constraints})
 
         if self.rc.env and self.rc.env.desc:
-            other_role_names = ", ".join(self.rc.env.role_names())
+            all_roles = self.rc.env.role_names()
+            other_role_names = ", ".join([r for r in all_roles if r != self.name])
             env_desc = f"You are in {self.rc.env.desc} with roles({other_role_names})."
             prefix += env_desc
         return prefix
@@ -478,7 +493,6 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
             if not msg.cause_by:
                 msg.cause_by = UserRequirement
             self.put_message(msg)
-
         if not await self._observe():
             # If there is no new information, suspend and wait
             logger.debug(f"{self._setting}: no news. waiting.")

tests/metagpt/actions/test_action_node.py

@@ -23,14 +23,12 @@ from metagpt.team import Team
 async def test_debate_two_roles():
     action1 = Action(name="AlexSay", instruction="Express your opinion with emotion and don't repeat it")
     action2 = Action(name="BobSay", instruction="Express your opinion with emotion and don't repeat it")
-    biden = Role(
+    alex = 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]
-    )
+    bob = 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])
+    team = Team(investment=10.0, env=env, roles=[alex, bob])
 
     history = await team.run(idea="Topic: climate change. Under 80 words per message.", send_to="Alex", n_round=3)
     assert "Alex" in history
@@ -39,9 +37,9 @@ async def test_debate_two_roles():
 @pytest.mark.asyncio
 async def test_debate_one_role_in_env():
     action = Action(name="Debate", instruction="Express your opinion with emotion and don't repeat it")
-    biden = Role(name="Alex", profile="Democratic candidate", goal="Win the election", actions=[action])
+    alex = Role(name="Alex", profile="Democratic candidate", goal="Win the election", actions=[action])
     env = Environment(desc="US election live broadcast")
-    team = Team(investment=10.0, env=env, roles=[biden])
+    team = Team(investment=10.0, env=env, roles=[alex])
     history = await team.run(idea="Topic: climate change. Under 80 words per message.", send_to="Alex", n_round=3)
     assert "Alex" in history
 
@@ -49,8 +47,8 @@ async def test_debate_one_role_in_env():
 @pytest.mark.asyncio
 async def test_debate_one_role():
     action = Action(name="Debate", instruction="Express your opinion with emotion and don't repeat it")
-    biden = Role(name="Alex", profile="Democratic candidate", goal="Win the election", actions=[action])
-    msg: Message = await biden.run("Topic: climate change. Under 80 words per message.")
+    alex = Role(name="Alex", profile="Democratic candidate", goal="Win the election", actions=[action])
+    msg: Message = await alex.run("Topic: climate change. Under 80 words per message.")
 
     assert len(msg.content) > 10
     assert msg.sent_from == "metagpt.roles.role.Role"

tests/metagpt/test_context_mixin.py

@@ -104,19 +104,12 @@ async def test_debate_two_roles():
     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(
+    alex = 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]
-    )
+    bob = 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)
+    team = Team(investment=10.0, env=env, roles=[alex, bob])
 
     history = await team.run(idea="Topic: climate change. Under 80 words per message.", send_to="Alex", n_round=3)
     assert "Alex" in history