minitap-mobile-use 2.0.0__py3-none-any.whl → 2.1.0__py3-none-any.whl

minitap/mobile_use/agents/orchestrator/types.py

@@ -1,14 +1,11 @@
-from enum import Enum
+from typing import Annotated
 
 from pydantic import BaseModel
 
 
-class OrchestratorStatus(Enum):
-    CONTINUE = "continue"
-    RESUME = "resume"
-    REPLAN = "replan"
-
-
 class OrchestratorOutput(BaseModel):
-    status: OrchestratorStatus
+    completed_subgoal_ids: Annotated[
+        list[str], "IDs of subgoals that can now be marked as complete"
+    ] = []
+    needs_replaning: Annotated[bool, "Whether the orchestrator needs to replan the subgoal plan"]
     reason: str

minitap/mobile_use/agents/outputter/outputter.py

@@ -1,6 +1,5 @@
 import json
 from pathlib import Path
-from typing import Dict, Type, Union
 
 from jinja2 import Template
 from langchain_core.messages import BaseMessage, HumanMessage, SystemMessage
@@ -49,7 +48,7 @@ async def outputter(
     structured_llm = llm
 
     if output_config.structured_output:
-        schema: Union[Dict, Type[BaseModel], None] = None
+        schema: dict | type[BaseModel] | None = None
         so = output_config.structured_output
 
         if isinstance(so, dict):

minitap/mobile_use/agents/planner/planner.md

@@ -12,7 +12,9 @@ You work like an agile tech lead: defining the key milestones without locking in
    - Subgoals should reflect real interactions with mobile UIs (e.g. "Open app", "Tap search bar", "Scroll to item", "Send message to Bob", etc).
    - Don't assume the full UI is visible yet. Plan based on how most mobile apps work, and keep flexibility.
    - List of agents thoughts is empty which is expected, since it is the first plan.
-   - Don't use precise UI actions when formulating subgoals like "copy", "paste", "tap", "swipe", ... unless explicitly asked in the initial goal.
+   - Avoid too granular UI actions based tasks (e.g. "tap", "swipe", "copy", "paste") unless explicitly required.
+   - The executor has the following available tools: **{{ executor_tools_list }}**.
+     When one of these tools offers a direct shortcut (e.g. `openLink` instead of manually launching a browser and typing a URL), prefer it over decomposed manual steps.
 
 2. **Replanning**
    If you're asked to **revise a previous plan**, you'll also receive:
@@ -25,7 +27,7 @@ You work like an agile tech lead: defining the key milestones without locking in
 
 ### Output
 
-You must output a **list of strings**, each representing a clear subgoal.
+You must output a **list of subgoals (description + optional subgoal ID)**, each representing a clear subgoal.
 Each subgoal should be:
 
 - Focused on **realistic mobile interactions**
@@ -33,32 +35,40 @@ Each subgoal should be:
 - Sequential (later steps may depend on earlier ones)
 - Don't use loop-like formulation unless necessary (e.g. don't say "repeat this X times", instead reuse the same steps X times as subgoals)
 
+If you're replaning and need to keep a previous subgoal, you **must keep the same subgoal ID**.
+
 ### Examples
 
 #### **Initial Goal**: "Open WhatsApp and send 'I’m running late' to Alice"
 
 **Plan**:
 
-- Open the WhatsApp app
-- Locate or search for Alice
-- Open the conversation with Alice
-- Type the message "I’m running late"
-- Send the message
+- Open the WhatsApp app (ID: None -> will be generated as a UUID like bc3c362d-f498-4f1a-991e-4a2d1f8c1226)
+- Locate or search for Alice (ID: None)
+- Open the conversation with Alice (ID: None)
+- Type the message "I’m running late" (ID: None)
+- Send the message (ID: None)
+
+#### **Initial Goal**: "Go on https://tesla.com, and tell me what is the first car being displayed"
+
+**Plan**:
+
+- Open the link https://tesla.com (ID: None)
+- Find the first car displayed on the home page (ID: None)
 
 #### **Replanning Example**
 
-**Original Plan**: same as above
+**Original Plan**: same as above with IDs set
 **Agent Thoughts**:
 
-- Couldn’t find Alice in recent chats
+- Couldn't find Alice in recent chats
 - Search bar was present on top of the chat screen
 - Keyboard appeared after tapping search
 
 **New Plan**:
 
-- Unlock the phone if needed
-- Open WhatsApp
-- Tap the search bar
-- Search for "Alice"
-- Select the correct chat
-- Type and send "I’m running late"
+- Open WhatsApp (ID: bc3c362d-f498-4f1a-991e-4a2d1f8c1226)
+- Tap the search bar (ID: None)
+- Search for "Alice" (ID: None)
+- Select the correct chat (ID: None)
+- Type and send "I’m running late" (ID: None)

minitap/mobile_use/agents/planner/planner.py

@@ -1,12 +1,15 @@
+import uuid
 from pathlib import Path
 
 from jinja2 import Template
 from langchain_core.messages import HumanMessage, SystemMessage
+
 from minitap.mobile_use.agents.planner.types import PlannerOutput, Subgoal, SubgoalStatus
 from minitap.mobile_use.agents.planner.utils import one_of_them_is_failure
 from minitap.mobile_use.context import MobileUseContext
 from minitap.mobile_use.graph.state import State
 from minitap.mobile_use.services.llm import get_llm
+from minitap.mobile_use.tools.index import EXECUTOR_WRAPPERS_TOOLS, format_tools_list
 from minitap.mobile_use.utils.decorators import wrap_with_callbacks
 from minitap.mobile_use.utils.logger import get_logger
 
@@ -35,6 +38,7 @@ class PlannerNode:
             initial_goal=state.initial_goal,
             previous_plan="\n".join(str(s) for s in state.subgoal_plan),
             agent_thoughts="\n".join(state.agents_thoughts),
+            executor_tools_list=format_tools_list(self.ctx, EXECUTOR_WRAPPERS_TOOLS),
         )
         messages = [
             SystemMessage(content=system_message),
@@ -47,7 +51,8 @@ class PlannerNode:
 
         subgoals_plan = [
             Subgoal(
-                description=subgoal,
+                id=subgoal.id or str(uuid.uuid4()),
+                description=subgoal.description,
                 status=SubgoalStatus.NOT_STARTED,
                 completion_reason=None,
             )
@@ -61,4 +66,5 @@ class PlannerNode:
             update={
                 "subgoal_plan": subgoals_plan,
             },
+            agent="planner",
         )

minitap/mobile_use/agents/planner/types.py

@@ -1,12 +1,16 @@
 from enum import Enum
-from typing import Optional
 
 from pydantic import BaseModel
-from typing_extensions import Annotated
+from typing import Annotated
+
+
+class PlannerSubgoalOutput(BaseModel):
+    id: Annotated[str | None, "If not provided, it will be generated"] = None
+    description: str
 
 
 class PlannerOutput(BaseModel):
-    subgoals: list[str]
+    subgoals: list[PlannerSubgoalOutput]
 
 
 class SubgoalStatus(Enum):
@@ -17,9 +21,10 @@ class SubgoalStatus(Enum):
 
 
 class Subgoal(BaseModel):
+    id: Annotated[str, "Unique identifier of the subgoal"]
     description: Annotated[str, "Description of the subgoal"]
     completion_reason: Annotated[
-        Optional[str], "Reason why the subgoal was completed (failure or success)"
+        str | None, "Reason why the subgoal was completed (failure or success)"
     ] = None
     status: SubgoalStatus
 
@@ -35,7 +40,7 @@ class Subgoal(BaseModel):
             case SubgoalStatus.NOT_STARTED:
                 status_emoji = "(not started yet)"
 
-        output = f"- {self.description} : {status_emoji}."
+        output = f"- [ID:{self.id}]: {self.description} : {status_emoji}."
         if self.completion_reason:
             output += f" Completion reason: {self.completion_reason}"
         return output

minitap/mobile_use/agents/planner/utils.py

@@ -5,6 +5,10 @@ def get_current_subgoal(subgoals: list[Subgoal]) -> Subgoal | None:
     return next((s for s in subgoals if s.status == SubgoalStatus.PENDING), None)
 
 
+def get_subgoals_by_ids(subgoals: list[Subgoal], ids: list[str]) -> list[Subgoal]:
+    return [s for s in subgoals if s.id in ids]
+
+
 def get_next_subgoal(subgoals: list[Subgoal]) -> Subgoal | None:
     return next((s for s in subgoals if s.status == SubgoalStatus.NOT_STARTED), None)
 
@@ -21,6 +25,13 @@ def complete_current_subgoal(subgoals: list[Subgoal]) -> list[Subgoal]:
     return subgoals
 
 
+def complete_subgoals_by_ids(subgoals: list[Subgoal], ids: list[str]) -> list[Subgoal]:
+    for subgoal in subgoals:
+        if subgoal.id in ids:
+            subgoal.status = SubgoalStatus.SUCCESS
+    return subgoals
+
+
 def fail_current_subgoal(subgoals: list[Subgoal]) -> list[Subgoal]:
     current_subgoal = get_current_subgoal(subgoals)
     if not current_subgoal:

minitap/mobile_use/agents/summarizer/summarizer.py

@@ -3,6 +3,7 @@ from langchain_core.messages import (
     RemoveMessage,
     ToolMessage,
 )
+
 from minitap.mobile_use.constants import MAX_MESSAGES_IN_HISTORY
 from minitap.mobile_use.context import MobileUseContext
 from minitap.mobile_use.graph.state import State
@@ -22,7 +23,7 @@ class SummarizerNode:
         start_removal = False
 
         for msg in reversed(state.messages[:nb_removal_candidates]):
-            if isinstance(msg, (ToolMessage, HumanMessage)):
+            if isinstance(msg, ToolMessage | HumanMessage):
                 start_removal = True
             if start_removal and msg.id:
                 remove_messages.append(RemoveMessage(id=msg.id))

minitap/mobile_use/clients/device_hardware_client.py

@@ -12,6 +12,9 @@ class DeviceHardwareClient:
         url = urljoin(self.base_url, f"/api/{path.lstrip('/')}")
         return self.session.get(url, **kwargs)
 
+    def get_rich_hierarchy(self) -> list[dict]:
+        return self.get("last-view-hierarchy").json().get("children", [])
+
     def post(self, path: str, **kwargs):
         url = urljoin(self.base_url, f"/api/{path.lstrip('/')}")
         return self.session.post(url, **kwargs)

minitap/mobile_use/config.py

@@ -1,7 +1,7 @@
 import json
 import os
 from pathlib import Path
-from typing import Annotated, Any, Literal, Optional, Union
+from typing import Annotated, Any, Literal
 
 from dotenv import load_dotenv
 from pydantic import BaseModel, Field, SecretStr, ValidationError, model_validator
@@ -17,15 +17,17 @@ logger = get_logger(__name__)
 
 
 class Settings(BaseSettings):
-    OPENAI_API_KEY: Optional[SecretStr] = None
-    GOOGLE_API_KEY: Optional[SecretStr] = None
-    XAI_API_KEY: Optional[SecretStr] = None
-    OPEN_ROUTER_API_KEY: Optional[SecretStr] = None
+    OPENAI_API_KEY: SecretStr | None = None
+    GOOGLE_API_KEY: SecretStr | None = None
+    XAI_API_KEY: SecretStr | None = None
+    OPEN_ROUTER_API_KEY: SecretStr | None = None
 
-    DEVICE_SCREEN_API_BASE_URL: Optional[str] = None
-    DEVICE_HARDWARE_BRIDGE_BASE_URL: Optional[str] = None
-    ADB_HOST: Optional[str] = None
-    ADB_PORT: Optional[int] = None
+    OPENAI_BASE_URL: str | None = None
+
+    DEVICE_SCREEN_API_BASE_URL: str | None = None
+    DEVICE_HARDWARE_BRIDGE_BASE_URL: str | None = None
+    ADB_HOST: str | None = None
+    ADB_PORT: int | None = None
 
     model_config = {"env_file": ".env", "extra": "ignore"}
 
@@ -69,7 +71,7 @@ def prepare_output_files() -> tuple[str | None, str | None]:
     return validated_events_path, validated_results_path
 
 
-def record_events(output_path: Path | None, events: Union[list[str], BaseModel, Any]):
+def record_events(output_path: Path | None, events: list[str] | BaseModel | Any):
     if not output_path:
         return
 
@@ -168,7 +170,7 @@ def get_default_llm_config() -> LLMConfig:
     try:
         if not os.path.exists(ROOT_DIR / DEFAULT_LLM_CONFIG_FILENAME):
             raise Exception("Default llm config not found")
-        with open(ROOT_DIR / DEFAULT_LLM_CONFIG_FILENAME, "r") as f:
+        with open(ROOT_DIR / DEFAULT_LLM_CONFIG_FILENAME) as f:
             default_config_dict = load_jsonc(f)
         return LLMConfig.model_validate(default_config_dict["default"])
     except Exception as e:
@@ -209,7 +211,7 @@ def parse_llm_config() -> LLMConfig:
     override_config_dict = {}
     if os.path.exists(ROOT_DIR / OVERRIDE_LLM_CONFIG_FILENAME):
         logger.info("Loading custom llm config...")
-        with open(ROOT_DIR / OVERRIDE_LLM_CONFIG_FILENAME, "r") as f:
+        with open(ROOT_DIR / OVERRIDE_LLM_CONFIG_FILENAME) as f:
             override_config_dict = load_jsonc(f)
     else:
         logger.warning("Custom llm config not found, loading default config")
@@ -235,7 +237,7 @@ def initialize_llm_config() -> LLMConfig:
 
 class OutputConfig(BaseModel):
     structured_output: Annotated[
-        Optional[Union[type[BaseModel], dict]],
+        type[BaseModel] | dict | None,
         Field(
             default=None,
             description=(
@@ -245,7 +247,7 @@ class OutputConfig(BaseModel):
         ),
     ]
     output_description: Annotated[
-        Optional[str],
+        str | None,
         Field(
             default=None,
             description=(

minitap/mobile_use/constants.py

@@ -1,2 +1,3 @@
 RECURSION_LIMIT = 400
 MAX_MESSAGES_IN_HISTORY = 25
+EXECUTOR_MESSAGES_KEY = "executor_messages"

minitap/mobile_use/context.py

@@ -6,12 +6,11 @@ Uses ContextVar to avoid prop drilling and maintain clean function signatures.
 
 from enum import Enum
 from pathlib import Path
-from typing import Optional
 
 from adbutils import AdbClient
 from openai import BaseModel
 from pydantic import ConfigDict
-from typing_extensions import Literal
+from typing import Literal
 
 from minitap.mobile_use.clients.device_hardware_client import DeviceHardwareClient
 from minitap.mobile_use.clients.screen_api_client import ScreenApiClient
@@ -56,8 +55,8 @@ class MobileUseContext(BaseModel):
     hw_bridge_client: DeviceHardwareClient
     screen_api_client: ScreenApiClient
     llm_config: LLMConfig
-    adb_client: Optional[AdbClient] = None
-    execution_setup: Optional[ExecutionSetup] = None
+    adb_client: AdbClient | None = None
+    execution_setup: ExecutionSetup | None = None
 
     def get_adb_client(self) -> AdbClient:
         if self.adb_client is None:

minitap/mobile_use/controllers/mobile_command_controller.py

@@ -1,6 +1,6 @@
 import uuid
 from enum import Enum
-from typing import Annotated, Literal, Optional, Union
+from typing import Annotated, Literal
 
 import yaml
 from langgraph.types import Command
@@ -9,6 +9,7 @@ from requests import JSONDecodeError
 
 from minitap.mobile_use.clients.device_hardware_client import DeviceHardwareClient
 from minitap.mobile_use.clients.screen_api_client import ScreenApiClient
+from minitap.mobile_use.config import initialize_llm_config
 from minitap.mobile_use.context import DeviceContext, DevicePlatform, MobileUseContext
 from minitap.mobile_use.utils.errors import ControllerErrors
 from minitap.mobile_use.utils.logger import get_logger
@@ -42,7 +43,7 @@ class RunFlowRequest(BaseModel):
     dry_run: bool = Field(default=False, alias="dryRun")
 
 
-def run_flow(ctx: MobileUseContext, flow_steps: list, dry_run: bool = False) -> Optional[dict]:
+def run_flow(ctx: MobileUseContext, flow_steps: list, dry_run: bool = False) -> dict | None:
     """
     Run a flow i.e, a sequence of commands.
     Returns None on success, or the response body of the failed command.
@@ -136,20 +137,20 @@ class SelectorRequestWithPercentages(BaseModel):
         return {"point": self.percentages.to_str()}
 
 
-SelectorRequest = Union[
-    IdSelectorRequest,
-    SelectorRequestWithCoordinates,
-    SelectorRequestWithPercentages,
-    TextSelectorRequest,
-    IdWithTextSelectorRequest,
-]
+SelectorRequest = (
+    IdSelectorRequest
+    | SelectorRequestWithCoordinates
+    | SelectorRequestWithPercentages
+    | TextSelectorRequest
+    | IdWithTextSelectorRequest
+)
 
 
 def tap(
     ctx: MobileUseContext,
     selector_request: SelectorRequest,
     dry_run: bool = False,
-    index: Optional[int] = None,
+    index: int | None = None,
 ):
     """
     Tap on a selector.
@@ -170,7 +171,7 @@ def long_press_on(
     ctx: MobileUseContext,
     selector_request: SelectorRequest,
     dry_run: bool = False,
-    index: Optional[int] = None,
+    index: int | None = None,
 ):
     long_press_on_body = selector_request.to_dict()
     if not long_press_on_body:
@@ -210,7 +211,7 @@ SwipeDirection = Annotated[
 class SwipeRequest(BaseModel):
     model_config = ConfigDict(extra="forbid")
     swipe_mode: SwipeStartEndCoordinatesRequest | SwipeStartEndPercentagesRequest | SwipeDirection
-    duration: Optional[int] = None  # in ms, default is 400ms
+    duration: int | None = None  # in ms, default is 400ms
 
     def to_dict(self):
         res = {}
@@ -256,7 +257,7 @@ def paste_text(ctx: MobileUseContext, dry_run: bool = False):
     return run_flow(ctx, ["pasteText"], dry_run=dry_run)
 
 
-def erase_text(ctx: MobileUseContext, nb_chars: Optional[int] = None, dry_run: bool = False):
+def erase_text(ctx: MobileUseContext, nb_chars: int | None = None, dry_run: bool = False):
     """
     Removes characters from the currently selected textfield (if any)
     Removes 50 characters if nb_chars is not specified.
@@ -274,7 +275,7 @@ def launch_app(ctx: MobileUseContext, package_name: str, dry_run: bool = False):
     return run_flow_with_wait_for_animation_to_end(ctx, flow_input, dry_run=dry_run)
 
 
-def stop_app(ctx: MobileUseContext, package_name: Optional[str] = None, dry_run: bool = False):
+def stop_app(ctx: MobileUseContext, package_name: str | None = None, dry_run: bool = False):
     if package_name is None:
         flow_input = ["stopApp"]
     else:
@@ -316,7 +317,7 @@ class WaitTimeout(Enum):
 
 
 def wait_for_animation_to_end(
-    ctx: MobileUseContext, timeout: Optional[WaitTimeout] = None, dry_run: bool = False
+    ctx: MobileUseContext, timeout: WaitTimeout | None = None, dry_run: bool = False
 ):
     if timeout is None:
         return run_flow(ctx, ["waitForAnimationToEnd"], dry_run=dry_run)
@@ -331,12 +332,10 @@ def run_flow_with_wait_for_animation_to_end(
 
 
 if __name__ == "__main__":
-    # long press, erase
-    # input_text(text="test")
-    # erase_text()
     ctx = MobileUseContext(
+        llm_config=initialize_llm_config(),
         device=DeviceContext(
-            host_platform="LINUX",
+            host_platform="WINDOWS",
             mobile_platform=DevicePlatform.ANDROID,
             device_id="emulator-5554",
             device_width=1080,
@@ -347,7 +346,6 @@ if __name__ == "__main__":
     )
     screen_data = get_screen_data(ctx.screen_api_client)
     from minitap.mobile_use.graph.state import State
-    from minitap.mobile_use.tools.mobile.erase_text import get_erase_text_tool
 
     dummy_state = State(
         latest_ui_hierarchy=screen_data.elements,
@@ -358,20 +356,33 @@ if __name__ == "__main__":
         focused_app_info=None,
         device_date="",
         structured_decisions=None,
-        executor_retrigger=False,
-        executor_failed=False,
+        complete_subgoals_by_ids=[],
         executor_messages=[],
         cortex_last_thought="",
         agents_thoughts=[],
     )
 
-    # invoke erase_text tool
-    input_resource_id = "com.google.android.settings.intelligence:id/open_search_view_edit_text"
-    command_output: Command = get_erase_text_tool(ctx=ctx).invoke(
+    # from minitap.mobile_use.tools.mobile.input_text import get_input_text_tool
+
+    # input_resource_id = "com.google.android.apps.nexuslauncher:id/search_container_hotseat"
+    # command_output: Command = get_input_text_tool(ctx=ctx).invoke(
+    #     {
+    #         "tool_call_id": uuid.uuid4().hex,
+    #         "agent_thought": "",
+    #         "text_input_resource_id": input_resource_id,
+    #         "text": "Hello World",
+    #         "state": dummy_state,
+    #         "executor_metadata": None,
+    #     }
+    # )
+    from minitap.mobile_use.tools.mobile.clear_text import get_clear_text_tool
+
+    input_resource_id = "com.google.android.apps.nexuslauncher:id/input"
+    command_output: Command = get_clear_text_tool(ctx=ctx).invoke(
         {
             "tool_call_id": uuid.uuid4().hex,
             "agent_thought": "",
-            "input_text_resource_id": input_resource_id,
+            "text_input_resource_id": input_resource_id,
             "state": dummy_state,
             "executor_metadata": None,
         }

minitap/mobile_use/controllers/platform_specific_commands_controller.py

@@ -1,6 +1,5 @@
 from datetime import date
 import json
-from typing import Optional
 
 from adbutils import AdbDevice
 from minitap.mobile_use.utils.logger import MobileUseLogger
@@ -20,8 +19,8 @@ def get_adb_device(ctx: MobileUseContext) -> AdbDevice:
 
 
 def get_first_device(
-    logger: Optional[MobileUseLogger] = None,
-) -> tuple[Optional[str], Optional[DevicePlatform]]:
+    logger: MobileUseLogger | None = None,
+) -> tuple[str | None, DevicePlatform | None]:
     """Gets the first available device."""
     try:
         android_output = run_shell_command_on_host("adb devices")
@@ -50,7 +49,7 @@ def get_first_device(
     return None, None
 
 
-def get_focused_app_info(ctx: MobileUseContext) -> Optional[str]:
+def get_focused_app_info(ctx: MobileUseContext) -> str | None:
     if ctx.device.mobile_platform == DevicePlatform.IOS:
         return None
     device = get_adb_device(ctx)

minitap/mobile_use/graph/graph.py

@@ -6,13 +6,11 @@ from langchain_core.messages import (
 from langgraph.constants import END, START
 from langgraph.graph import StateGraph
 from langgraph.graph.state import CompiledStateGraph
-from langgraph.prebuilt import ToolNode
+
 from minitap.mobile_use.agents.contextor.contextor import ContextorNode
 from minitap.mobile_use.agents.cortex.cortex import CortexNode
 from minitap.mobile_use.agents.executor.executor import ExecutorNode
-from minitap.mobile_use.agents.executor.executor_context_cleaner import (
-    executor_context_cleaner_node,
-)
+from minitap.mobile_use.agents.executor.tool_node import ExecutorToolNode
 from minitap.mobile_use.agents.orchestrator.orchestrator import OrchestratorNode
 from minitap.mobile_use.agents.planner.planner import PlannerNode
 from minitap.mobile_use.agents.planner.utils import (
@@ -21,6 +19,7 @@ from minitap.mobile_use.agents.planner.utils import (
     one_of_them_is_failure,
 )
 from minitap.mobile_use.agents.summarizer.summarizer import SummarizerNode
+from minitap.mobile_use.constants import EXECUTOR_MESSAGES_KEY
 from minitap.mobile_use.context import MobileUseContext
 from minitap.mobile_use.graph.state import State
 from minitap.mobile_use.tools.index import EXECUTOR_WRAPPERS_TOOLS, get_tools_from_wrappers
@@ -53,7 +52,7 @@ def post_cortex_gate(
     state: State,
 ) -> Literal["continue", "end_subgoal"]:
     logger.info("Starting post_cortex_gate")
-    if not state.structured_decisions:
+    if len(state.complete_subgoals_by_ids) > 0:
         return "end_subgoal"
     return "continue"
 
@@ -62,7 +61,7 @@ def post_executor_gate(
     state: State,
 ) -> Literal["invoke_tools", "skip"]:
     logger.info("Starting post_executor_gate")
-    messages = state.messages
+    messages = state.executor_messages
     if not messages:
         return "skip"
     last_message = messages[-1]
@@ -77,17 +76,6 @@ def post_executor_gate(
     return "skip"
 
 
-def post_executor_tools_gate(
-    state: State,
-) -> Literal["continue", "failed", "done"]:
-    logger.info("Starting post_executor_tools_gate")
-    if state.executor_failed:
-        return "failed"
-    if state.executor_retrigger:
-        return "continue"
-    return "done"
-
-
 async def get_graph(ctx: MobileUseContext) -> CompiledStateGraph:
     graph_builder = StateGraph(State)
 
@@ -100,12 +88,12 @@ async def get_graph(ctx: MobileUseContext) -> CompiledStateGraph:
     graph_builder.add_node("cortex", CortexNode(ctx))
 
     graph_builder.add_node("executor", ExecutorNode(ctx))
-    executor_tool_node = ToolNode(
-        get_tools_from_wrappers(ctx=ctx, wrappers=EXECUTOR_WRAPPERS_TOOLS)
+    executor_tool_node = ExecutorToolNode(
+        tools=get_tools_from_wrappers(ctx=ctx, wrappers=EXECUTOR_WRAPPERS_TOOLS),
+        messages_key=EXECUTOR_MESSAGES_KEY,
     )
     graph_builder.add_node("executor_tools", executor_tool_node)
 
-    graph_builder.add_node("executor_context_cleaner", executor_context_cleaner_node)
     graph_builder.add_node("summarizer", SummarizerNode(ctx))
 
     # Linking nodes
@@ -132,18 +120,9 @@ async def get_graph(ctx: MobileUseContext) -> CompiledStateGraph:
     graph_builder.add_conditional_edges(
         "executor",
         post_executor_gate,
-        {"invoke_tools": "executor_tools", "skip": "executor_context_cleaner"},
-    )
-    graph_builder.add_conditional_edges(
-        "executor_tools",
-        post_executor_tools_gate,
-        {
-            "continue": "executor",
-            "done": "executor_context_cleaner",
-            "failed": "executor_context_cleaner",
-        },
+        {"invoke_tools": "executor_tools", "skip": "summarizer"},
     )
-    graph_builder.add_edge("executor_context_cleaner", "summarizer")
+    graph_builder.add_edge("executor_tools", "summarizer")
     graph_builder.add_edge("summarizer", "contextor")
 
     return graph_builder.compile()