fix(ui): eliminate task cancellation state flickering and update model references

artemis/agents/flash/runner.py

@@ -113,7 +113,7 @@ async def run(self, state: State) -> dict:
         except Exception as e:
             logger.warning(f"Failed to get operator LLM from config, using default: {e}")
 
-            llm = RobustChatModelWrapper(get_google_llm(model_name="gemini-2.5-flash"), self.ctx)
+            llm = RobustChatModelWrapper(get_google_llm(model_name="gemini-3.7-flash"), self.ctx)
 
         tools_declaration = self._get_tools()
 

artemis/agents/flash/summarizer.py

@@ -54,7 +54,7 @@ def __init__(self, ctx: ArtemisContext, model_name: str | None = None):
         self._retry_delays = (0.0, 0.5, 1.0, 2.0, 3.0)
 
         # Initialize lightweight VLM: prioritize explicit model_name
-        target_model = model_name or "gemini-2.5-flash-lite"
+        target_model = model_name or "gemini-3.5-flash-lite"
         try:
             if model_name:
                 self._llm = get_google_llm(model_name=target_model, temperature=0.0)

artemis/config/agent.py

@@ -241,7 +241,7 @@ class StepSummarizerConfig(BaseModel):
         description="Whether to asynchronously summarize historical steps to replace pruned images.",
     )
     model: str = Field(
-        default="gemini-2.5-flash-lite",
+        default="gemini-3.5-flash-lite",
         description="Lightweight model used for background step state summarization.",
     )
     prune_history_xml: bool = Field(

artemis/config/constants.py

@@ -105,7 +105,7 @@
 DEFAULT_STREAM_PING_INTERVAL = 15.0
 
 DEFAULT_PROFILE = "pro"
-DEFAULT_MODEL = "gemini-2.5-flash"
+DEFAULT_MODEL = "gemini-3.7-flash"
 DEFAULT_EXPLORER_VERSION: Literal["flash", "pro", "ultra"] = "flash"
 
 

artemis/interfaces/cli/commands/init.py

@@ -125,7 +125,7 @@ def init_command() -> None:
             "",
             "# Execution Defaults",
             "ARTEMIS_DEFAULT_PROFILE=pro",
-            "ARTEMIS_DEFAULT_MODEL=gemini-2.5-flash",
+            "ARTEMIS_DEFAULT_MODEL=gemini-3.7-flash",
             "ARTEMIS_TRACES_DIR=./traces",
             "",
         ]

artemis/llm/router.py

@@ -72,7 +72,7 @@ class ModelEndpoint(BaseModel):
     """Configuration definition for an LLM/VLM model endpoint."""
 
     provider: ModelProvider = Field(default=ModelProvider.GOOGLE, description="Model provider")
-    model_name: str = Field(default="gemini-2.5-flash", description="Model name identifier")
+    model_name: str = Field(default="gemini-3.7-flash", description="Model name identifier")
     api_key: str | None = Field(default=None, description="API Key or secret")
     api_base: str | None = Field(default=None, description="Custom API endpoint base URL")
     temperature: float = Field(default=0.0, description="Sampling temperature")
@@ -370,7 +370,7 @@ def create_model(cls, endpoint: ModelEndpoint) -> BaseChatModel:
 class ModelRouter:
     """Dynamic router mapping agent roles to primary and fallback model endpoints."""
 
-    def __init__(self, default_model: str = "gemini-2.5-flash"):
+    def __init__(self, default_model: str = "gemini-3.7-flash"):
         self.default_endpoint = ModelEndpoint(
             provider=ModelProvider.GOOGLE,
             model_name=default_model,
@@ -383,26 +383,26 @@ def _setup_default_routes(self) -> None:
         """Initializes recommended role mappings."""
         self._role_routes["planner"] = ModelEndpoint(
             provider=ModelProvider.GOOGLE,
-            model_name="gemini-2.5-pro",
+            model_name="gemini-3.7-pro",
             temperature=0.2,
         )
         self._role_routes["operator"] = ModelEndpoint(
             provider=ModelProvider.GOOGLE,
-            model_name="gemini-2.5-flash",
+            model_name="gemini-3.7-flash",
             temperature=0.0,
         )
         self._role_routes["validator"] = ModelEndpoint(
             provider=ModelProvider.GOOGLE,
-            model_name="gemini-2.5-pro",
+            model_name="gemini-3.7-pro",
             temperature=0.1,
         )
         self._role_routes["summarizer"] = ModelEndpoint(
             provider=ModelProvider.GOOGLE,
-            model_name="gemini-2.5-flash",
+            model_name="gemini-3.7-flash",
             temperature=0.0,
         )
         self._fallback_routes["planner"] = [
-            ModelEndpoint(provider=ModelProvider.GOOGLE, model_name="gemini-2.5-flash"),
+            ModelEndpoint(provider=ModelProvider.GOOGLE, model_name="gemini-3.7-flash"),
         ]
 
     def set_role_route(

artemis/sdk/examples/video_transcription_example.py

@@ -39,11 +39,9 @@ def get_video_capable_llm_config() -> LLMConfig:
 
     The video_analyzer must use a video-capable Gemini model:
     - gemini-3.7-flash (recommended - fast and capable)
+    - gemini-3.7-pro
     - gemini-3.6-flash
     - gemini-3-pro-preview
-    - gemini-2.5-flash
-    - gemini-2.5-pro
-    - gemini-2.0-flash
     """
     return LLMConfig(
         planner=LLMWithFallback(

artemis/services/llm.py

@@ -776,7 +776,7 @@ def _get_val(obj, attr, expected_type):
         return val if isinstance(val, expected_type) else None
 
     provider_val = getattr(cfg, "provider", "google")
-    model_val = getattr(cfg, "model", "gemini-2.5-flash")
+    model_val = getattr(cfg, "model", "gemini-3.7-flash")
 
     return ModelEndpoint(
         provider=ModelProvider.from_string(provider_val),

config/artemis.jsonc

@@ -28,8 +28,8 @@
     },
     "gemini-flash": {
       "provider": "google",
-      "model": "gemini-2.5-flash",
-      "fallback": { "provider": "google", "model": "gemini-2.0-flash" }
+      "model": "gemini-3.7-flash",
+      "fallback": { "provider": "google", "model": "gemini-3.6-flash" }
     },
     "openai-gpt4o": {
       "provider": "openai",

tests/integration/agents/explorer/test_explorer_all_tools/run_live_tools_test.py

@@ -73,7 +73,7 @@ async def run_single_set(set_idx, query, search_query, labels, coords_nx, coords
     mock_ctx = create_mock_context()
     mock_ctx.llm_config.utils = MagicMock()
     mock_ctx.llm_config.utils.object_detector = MagicMock()
-    mock_ctx.llm_config.utils.object_detector.model = "gemini-2.5-flash"
+    mock_ctx.llm_config.utils.object_detector.model = "gemini-3.5-flash"
     mock_ctx.llm_config.utils.object_detector.timeout = 10.0
 
     mock_state = create_mock_state()

tests/integration/agents/explorer/test_explorer_all_tools/test_object_detector_concurrency.py

@@ -34,7 +34,7 @@ async def test_object_detector_concurrency_and_throttling(tmp_path):
     # Configure the object detector LLM config explicitly to avoid MagicMocks for timeout and model
     od_cfg = MagicMock()
     od_cfg.model = "gemini-robotics-er-early-access"
-    od_cfg.fix_model = "gemini-2.5-flash"
+    od_cfg.fix_model = "gemini-3.5-flash"
     od_cfg.timeout = 10.0
     mock_ctx.llm_config.utils.object_detector = od_cfg
 

tests/unit/agents/test_failure_analyzer.py

@@ -42,7 +42,7 @@ def mock_context(tmp_path):
     ctx = Mock(spec=ArtemisContext)
     ctx.llm_config = Mock()
     mock_llm_cfg = Mock()
-    mock_llm_cfg.model = "gemini-2.5-pro"
+    mock_llm_cfg.model = "gemini-3.7-pro"
     mock_llm_cfg.temperature = 0.1
     ctx.llm_config.get_agent.return_value = mock_llm_cfg
 

tests/unit/agents/test_flash_runner.py

@@ -27,7 +27,7 @@ def mock_context():
     ctx = Mock(spec=ArtemisContext)
     ctx.llm_config = Mock()
     mock_llm_cfg = Mock()
-    mock_llm_cfg.model = "gemini-2.5-flash"
+    mock_llm_cfg.model = "gemini-3.7-flash"
     mock_llm_cfg.temperature = 0.1
     ctx.llm_config.get_agent.return_value = mock_llm_cfg
 

tests/unit/agents/test_flash_step_summarizer.py

@@ -361,14 +361,14 @@ def test_flash_config_and_builder():
         max_turns=25,
         explorer_mode="flash",
         step_summarizer=True,
-        step_summarizer_model="gemini-2.5-flash-lite",
+        step_summarizer_model="gemini-3.5-flash-lite",
         prune_history_xml=True,
     ).build()
 
     assert cfg.flash.max_turns == 25
     assert cfg.flash.explorer_mode == "flash"
     assert cfg.flash.step_summarizer.enabled is True
-    assert cfg.flash.step_summarizer.model == "gemini-2.5-flash-lite"
+    assert cfg.flash.step_summarizer.model == "gemini-3.5-flash-lite"
     assert cfg.flash.step_summarizer.prune_history_xml is True
 
 

tests/unit/test_agents.py

@@ -95,13 +95,13 @@ def test_agent_registry():
 def test_agent_config():
     """Verify AgentConfig parameters can be customized."""
     config = AgentConfig(
-        model_name="gemini-2.5-pro",
+        model_name="gemini-3.7-pro",
         temperature=0.7,
         max_tokens=2048,
         enabled_tools=["click", "swipe"],
     )
     agent = DummyTestAgent(config=config)
-    assert agent.config.model_name == "gemini-2.5-pro"
+    assert agent.config.model_name == "gemini-3.7-pro"
     assert agent.config.temperature == 0.7
     assert agent.config.enabled_tools == ["click", "swipe"]
 

tests/unit/test_model_router.py

@@ -52,13 +52,13 @@ def test_model_router_fallbacks():
     """Verify fallback endpoint configuration."""
     router = ModelRouter()
     fallbacks = [
-        ModelEndpoint(provider=ModelProvider.GEMINI, model_name="gemini-2.5-flash"),
+        ModelEndpoint(provider=ModelProvider.GEMINI, model_name="gemini-3.7-flash"),
         ModelEndpoint(provider=ModelProvider.OPENAI, model_name="gpt-4o-mini"),
     ]
-    router.set_role_route("custom_worker", "gemini-2.5-pro", fallbacks=fallbacks)
+    router.set_role_route("custom_worker", "gemini-3.7-pro", fallbacks=fallbacks)
     fb = router.get_fallbacks("custom_worker")
     assert len(fb) == 2
-    assert fb[0].model_name == "gemini-2.5-flash"
+    assert fb[0].model_name == "gemini-3.7-flash"
     assert fb[1].model_name == "gpt-4o-mini"
 
 
@@ -113,7 +113,7 @@ def bind_tools(self, tools, *args, **kwargs):
     # Case 1: enable_grounding=True automatically attaches google_search and tool_config
     ep_google = ModelEndpoint(
         provider=ModelProvider.GOOGLE,
-        model_name="gemini-2.5-flash",
+        model_name="gemini-3.7-flash",
         enable_grounding=True,
     )
     wrapper = RobustChatModelWrapper(MockGeminiModel(), endpoint=ep_google)
@@ -127,7 +127,7 @@ def bind_tools(self, tools, *args, **kwargs):
     # Case 2: Explicit google_search in tools automatically ensures tool_config
     ep_google_no_ground = ModelEndpoint(
         provider=ModelProvider.GOOGLE,
-        model_name="gemini-2.5-flash",
+        model_name="gemini-3.7-flash",
         enable_grounding=False,
     )
     wrapper_explicit = RobustChatModelWrapper(MockGeminiModel(), endpoint=ep_google_no_ground)
@@ -187,7 +187,7 @@ def dummy_tool(query: str) -> str:
         return query
 
     # Trigger patching via Google model creation
-    ep = ModelEndpoint(provider=ModelProvider.GOOGLE, model_name="gemini-2.5-flash", api_key="fake")
+    ep = ModelEndpoint(provider=ModelProvider.GOOGLE, model_name="gemini-3.7-flash", api_key="fake")
     model = ModelFactory.create_model(ep)
     assert isinstance(model, ChatGoogleGenerativeAI)