FIX multiLLM bug

metagpt/actions/action.py

@@ -49,7 +49,6 @@ class Action(SerializationMixin, ContextMixin, BaseModel):
             llm = create_llm_instance(config)
             llm.cost_manager = data.llm.cost_manager
             data.llm = llm
-            data.config = config   # if not set self.config, self.llm will be reset when you call Role.set_actions function
         return data
 
     @property

metagpt/roles/role.py

@@ -247,7 +247,7 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
         return self
 
     def _init_action(self, action: Action):
-        if not action.private_config:
+        if not action.private_llm:
             action.set_llm(self.llm, override=True)
         else:
             action.set_llm(self.llm, override=False)

tests/data/config/config2_multi_llm.yaml

@@ -0,0 +1,27 @@
+llm:
+  api_type: "openai"  # or azure / ollama / groq etc.
+  base_url: "YOUR_gpt-3.5-turbo_BASE_URL"
+  api_key: "YOUR_gpt-3.5-turbo_API_KEY"
+  model: "gpt-3.5-turbo"  # or gpt-3.5-turbo
+  # proxy: "YOUR_gpt-3.5-turbo_PROXY"  # for LLM API requests
+  # timeout: 600 # Optional. If set to 0, default value is 300.
+  # Details: https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/
+  pricing_plan: "" # Optional. Use for Azure LLM when its model name is not the same as OpenAI's
+
+models:
+  "YOUR_MODEL_NAME_1": # model: "gpt-4-turbo"  # or gpt-3.5-turbo
+    api_type: "openai"  # or azure / ollama / groq etc.
+    base_url: "YOUR_MODEL_1_BASE_URL"
+    api_key: "YOUR_MODEL_1_API_KEY"
+    # proxy: "YOUR_MODEL_1_PROXY"  # for LLM API requests
+    # timeout: 600 # Optional. If set to 0, default value is 300.
+    # Details: https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/
+    pricing_plan: "" # Optional. Use for Azure LLM when its model name is not the same as OpenAI's
+  "YOUR_MODEL_NAME_2": # model: "gpt-4-turbo"  # or gpt-3.5-turbo
+    api_type: "openai"  # or azure / ollama / groq etc.
+    base_url: "YOUR_MODEL_2_BASE_URL"
+    api_key: "YOUR_MODEL_2_API_KEY"
+    proxy: "YOUR_MODEL_2_PROXY"  # for LLM API requests
+    # timeout: 600 # Optional. If set to 0, default value is 300.
+    # Details: https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/
+    pricing_plan: "" # Optional. Use for Azure LLM when its model name is not the same as OpenAI's

tests/metagpt/actions/test_action_multi_llm.py

@@ -0,0 +1,53 @@
+import pytest
+
+
+from metagpt.config2 import Config
+from metagpt.roles.role import Role, RoleReactMode
+from metagpt.actions.action import Action
+from metagpt.context import Context
+from metagpt.provider.llm_provider_registry import create_llm_instance
+from metagpt.const import TEST_DATA_PATH
+
+def test_set_llm():
+    config1 = Config.default()
+    config2 = Config.default()
+    config2.llm.model = "gpt-3.5-turbo"
+
+    context = Context(config=config1)
+    act = Action(context=context)
+    assert act.config.llm.model == config1.llm.model
+
+    llm2 = create_llm_instance(config2.llm)
+    act.llm = llm2
+    assert act.llm.model == llm2.model
+
+    role = Role(context=context)
+    role.set_actions([act])
+    assert act.llm.model == llm2.model
+
+    role1 = Role(context=context)
+    act1 = Action(context=context)
+    assert act1.config.llm.model == config1.llm.model
+    act1.config = config2
+    role1.set_actions([act1])
+    assert act1.llm.model == llm2.model
+    
+    # multiple LLM
+
+    config3_path = TEST_DATA_PATH / "config/config2_multi_llm.yaml"
+    dict3 = Config.read_yaml(config3_path)
+    config3 = Config(**dict3)
+    context3 = Context(config=config3)
+    role3 = Role(context=context3)
+    act3 = Action(context=context3, llm_name_or_type="YOUR_MODEL_NAME_1")
+    assert act3.config.llm.model == "gpt-3.5-turbo"
+    assert act3.llm.model == "gpt-4-turbo"
+    role3.set_actions([act3])
+    assert act3.config.llm.model == "gpt-3.5-turbo"
+    assert act3.llm.model == "gpt-4-turbo"
+    
+    
+
+
+
+