feat: upgrade MiniMax default model to M3

- Set MiniMax-M3 as the default model for the voice, task, and UI agent configs
- Keep MiniMax-M2.7 / MiniMax-M2.7-highspeed selectable via the *_LLM_MODEL env overrides
- Update MiniMax provider unit tests to assert the new default

Author: octo-patch

src/gradientbang/utils/llm_factory.py

@@ -464,7 +464,7 @@ def get_voice_llm_config() -> LLMServiceConfig:
         LLMProvider.GOOGLE: "gemini-2.5-flash",
         LLMProvider.ANTHROPIC: "claude-sonnet-4-5-20250929",
         LLMProvider.OPENAI: "gpt-4.1",
-        LLMProvider.MINIMAX: "MiniMax-M2.7",
+        LLMProvider.MINIMAX: "MiniMax-M3",
     }
 
     model = current.VOICE_LLM_MODEL or default_models[provider]
@@ -519,7 +519,7 @@ def get_task_agent_llm_config() -> LLMServiceConfig:
         LLMProvider.GOOGLE: "gemini-2.5-flash",
         LLMProvider.ANTHROPIC: "claude-sonnet-4-5-20250929",
         LLMProvider.OPENAI: "gpt-4.1",
-        LLMProvider.MINIMAX: "MiniMax-M2.7",
+        LLMProvider.MINIMAX: "MiniMax-M3",
     }
 
     model = current.TASK_LLM_MODEL or default_models[provider]
@@ -573,7 +573,7 @@ def get_ui_agent_llm_config() -> LLMServiceConfig:
         LLMProvider.GOOGLE: "gemini-2.5-flash",
         LLMProvider.ANTHROPIC: "claude-haiku-4-5-20251001",
         LLMProvider.OPENAI: "gpt-4.1",
-        LLMProvider.MINIMAX: "MiniMax-M2.7",
+        LLMProvider.MINIMAX: "MiniMax-M3",
     }
     model = (
         current.UI_AGENT_LLM_MODEL

tests/unit/test_llm_factory_minimax.py

@@ -62,7 +62,7 @@ def test_creates_openai_service_with_minimax_base_url(self, _mock):
             with patch("pipecat.services.openai.llm.OpenAILLMService") as mock_cls:
                 mock_cls.return_value = MagicMock()
                 mock_cls.Settings = OpenAILLMService.Settings
-                _create_minimax_service(api_key="test-key", model="MiniMax-M2.7")
+                _create_minimax_service(api_key="test-key", model="MiniMax-M3")
                 mock_cls.assert_called_once()
                 _, kwargs = mock_cls.call_args
                 assert kwargs["api_key"] == "test-key"
@@ -76,14 +76,14 @@ def test_respects_minimax_base_url_env_var(self, _mock):
             with patch("pipecat.services.openai.llm.OpenAILLMService") as mock_cls:
                 mock_cls.return_value = MagicMock()
                 mock_cls.Settings = OpenAILLMService.Settings
-                _create_minimax_service(api_key="test-key", model="MiniMax-M2.7")
+                _create_minimax_service(api_key="test-key", model="MiniMax-M3")
                 _, kwargs = mock_cls.call_args
                 assert kwargs["base_url"] == "https://custom.minimax.io/v1"
 
     def test_default_temperature_is_1(self):
         """MiniMax does not accept temperature=0; default must be 1.0."""
         with patch.dict(os.environ, {"MINIMAX_API_KEY": "test-key"}):
-            service = _create_minimax_service(api_key="test-key", model="MiniMax-M2.7")
+            service = _create_minimax_service(api_key="test-key", model="MiniMax-M3")
             assert service._settings.temperature == 1.0
 
     def test_passes_function_call_timeout(self):
@@ -93,7 +93,7 @@ def test_passes_function_call_timeout(self):
             mock_cls.return_value = MagicMock()
             mock_cls.Settings = OpenAILLMService.Settings
             _create_minimax_service(
-                api_key="test-key", model="MiniMax-M2.7", function_call_timeout_secs=30.0
+                api_key="test-key", model="MiniMax-M3", function_call_timeout_secs=30.0
             )
             _, kwargs = mock_cls.call_args
             assert kwargs["function_call_timeout_secs"] == 30.0
@@ -104,15 +104,15 @@ class TestCreateLLMServiceMiniMax:
     def test_create_llm_service_routes_minimax(self):
         config = LLMServiceConfig(
             provider=LLMProvider.MINIMAX,
-            model="MiniMax-M2.7",
+            model="MiniMax-M3",
             api_key="test-key",
         )
         with patch(
             "gradientbang.utils.llm_factory._create_minimax_service"
         ) as mock_minimax:
             mock_minimax.return_value = MagicMock()
             create_llm_service(config)
-            mock_minimax.assert_called_once_with("test-key", "MiniMax-M2.7", None)
+            mock_minimax.assert_called_once_with("test-key", "MiniMax-M3", None)
 
     def test_create_llm_service_minimax_with_timeout(self):
         config = LLMServiceConfig(
@@ -140,7 +140,7 @@ def test_voice_llm_config_minimax_default_model(self):
             os.environ.pop("VOICE_LLM_MODEL", None)
             config = get_voice_llm_config()
         assert config.provider == LLMProvider.MINIMAX
-        assert config.model == "MiniMax-M2.7"
+        assert config.model == "MiniMax-M3"
 
     def test_task_llm_config_minimax_default_model(self):
         with patch.dict(
@@ -151,7 +151,7 @@ def test_task_llm_config_minimax_default_model(self):
             os.environ.pop("TASK_LLM_MODEL", None)
             config = get_task_agent_llm_config()
         assert config.provider == LLMProvider.MINIMAX
-        assert config.model == "MiniMax-M2.7"
+        assert config.model == "MiniMax-M3"
 
     def test_ui_agent_llm_config_minimax_default_model(self):
         with patch.dict(
@@ -162,7 +162,7 @@ def test_ui_agent_llm_config_minimax_default_model(self):
             os.environ.pop("UI_AGENT_LLM_MODEL", None)
             config = get_ui_agent_llm_config()
         assert config.provider == LLMProvider.MINIMAX
-        assert config.model == "MiniMax-M2.7"
+        assert config.model == "MiniMax-M3"
 
     def test_voice_llm_config_minimax_custom_model(self):
         with patch.dict(