minitap-mobile-use 3.3.0__py3-none-any.whl

minitap/mobile_use/sdk/builders/__init__.py

@@ -0,0 +1,10 @@
+"""Builder classes for configuring mobile-use components."""
+
+from minitap.mobile_use.sdk.builders.agent_config_builder import AgentConfigBuilder
+from minitap.mobile_use.sdk.builders.task_request_builder import (
+    TaskRequestCommonBuilder,
+    TaskRequestBuilder,
+)
+from minitap.mobile_use.sdk.builders.index import Builders
+
+__all__ = ["AgentConfigBuilder", "TaskRequestCommonBuilder", "TaskRequestBuilder", "Builders"]

minitap/mobile_use/sdk/builders/agent_config_builder.py

@@ -0,0 +1,307 @@
+"""
+Builder for AgentConfig objects using a fluent interface.
+"""
+
+import copy
+
+from langchain_core.callbacks.base import Callbacks
+
+from minitap.mobile_use.clients.ios_client_config import BrowserStackClientConfig, IosClientConfig
+from minitap.mobile_use.config import get_default_llm_config, get_default_minitap_llm_config
+from minitap.mobile_use.context import DevicePlatform
+from minitap.mobile_use.sdk.constants import DEFAULT_PROFILE_NAME
+from minitap.mobile_use.sdk.types.agent import AgentConfig, AgentProfile, ServerConfig
+from minitap.mobile_use.sdk.types.task import TaskRequestCommon
+
+
+class AgentConfigBuilder:
+    """
+    Builder class providing a fluent interface for creating AgentConfig objects.
+
+    This builder allows for step-by-step construction of an AgentConfig with
+    clear methods that make the configuration process intuitive and type-safe.
+
+    Examples:
+        >>> builder = AgentConfigBuilder()
+        >>> config = (builder
+        ...     .add_profile(AgentProfile(name="HighReasoning", llm_config=LLMConfig(...)))
+        ...     .add_profile(AgentProfile(name="LowReasoning", llm_config=LLMConfig(...)))
+        ...     .for_device(DevicePlatform.ANDROID, "device123")
+        ...     .with_default_task_config(TaskRequestCommon(max_steps=30))
+        ...     .with_default_profile("HighReasoning")
+        ...     .build()
+        ... )
+    """
+
+    def __init__(self):
+        """Initialize an empty AgentConfigBuilder."""
+        self._agent_profiles: dict[str, AgentProfile] = {}
+        self._task_request_defaults: TaskRequestCommon | None = None
+        self._default_profile: str | AgentProfile | None = None
+        self._device_id: str | None = None
+        self._device_platform: DevicePlatform | None = None
+        self._servers: ServerConfig = get_default_servers()
+        self._graph_config_callbacks: Callbacks = None
+        self._cloud_mobile_id_or_ref: str | None = None
+        self._ios_client_config: IosClientConfig | None = None
+        self._browserstack_config: BrowserStackClientConfig | None = None
+        self._video_recording_enabled: bool = False
+
+    def add_profile(self, profile: AgentProfile, validate: bool = True) -> "AgentConfigBuilder":
+        """
+        Add an agent profile to the mobile-use agent.
+
+        Args:
+            profile: The agent profile to add
+        """
+        self._agent_profiles[profile.name] = profile
+        if validate:
+            profile.llm_config.validate_providers()
+        return self
+
+    def add_profiles(
+        self,
+        profiles: list[AgentProfile],
+        validate: bool = True,
+    ) -> "AgentConfigBuilder":
+        """
+        Add multiple agent profiles to the mobile-use agent.
+
+        Args:
+            profiles: List of agent profiles to add
+        """
+        for profile in profiles:
+            self.add_profile(profile=profile, validate=validate)
+        return self
+
+    def with_default_profile(self, profile: str | AgentProfile) -> "AgentConfigBuilder":
+        """
+        Set the default agent profile used for tasks.
+
+        Args:
+            profile: The name or instance of the default agent profile
+        """
+        self._default_profile = profile
+        return self
+
+    def for_device(
+        self,
+        platform: DevicePlatform,
+        device_id: str,
+    ) -> "AgentConfigBuilder":
+        """
+        Configure the mobile-use agent for a specific device.
+
+        Args:
+            platform: The device platform (ANDROID or IOS)
+            device_id: The unique identifier for the device
+        """
+        if self._cloud_mobile_id_or_ref is not None:
+            raise ValueError(
+                "Device ID cannot be set when a cloud mobile is already configured.\n"
+                "> for_device() and for_cloud_mobile() are mutually exclusive"
+            )
+        if self._browserstack_config is not None:
+            raise ValueError(
+                "Device ID cannot be set when BrowserStack is already configured.\n"
+                "> for_device() and for_browserstack() are mutually exclusive"
+            )
+        self._device_id = device_id
+        self._device_platform = platform
+        return self
+
+    def for_cloud_mobile(self, cloud_mobile_id_or_ref: str) -> "AgentConfigBuilder":
+        """
+        Configure the mobile-use agent to use a cloud mobile.
+
+        When using a cloud mobile, tasks are executed remotely via the Platform API,
+        and only PlatformTaskRequest can be used.
+
+        Args:
+            cloud_mobile_id_or_ref: The unique identifier or reference name for the cloud mobile.
+                Can be either a UUID (e.g., '550e8400-e29b-41d4-a716-446655440000')
+                or a reference name (e.g., 'my-test-device')
+        """
+        if self._device_id is not None:
+            raise ValueError(
+                "Cloud mobile device ID cannot be set when a device is already configured.\n"
+                "> for_device() and for_cloud_mobile() are mutually exclusive"
+            )
+        if self._browserstack_config is not None:
+            raise ValueError(
+                "Cloud mobile cannot be set when BrowserStack is already configured.\n"
+                "> for_cloud_mobile() and for_browserstack() are mutually exclusive"
+            )
+        self._cloud_mobile_id_or_ref = cloud_mobile_id_or_ref
+        return self
+
+    def for_browserstack(self, config: BrowserStackClientConfig) -> "AgentConfigBuilder":
+        """
+        Configure the mobile-use agent to use BrowserStack cloud devices.
+
+        When using BrowserStack, the agent connects to BrowserStack's cloud infrastructure
+        for iOS device automation. This is mutually exclusive with for_device() and
+        for_cloud_mobile().
+
+        Args:
+            config: BrowserStack configuration with credentials and device settings
+        """
+        if self._device_id is not None:
+            raise ValueError(
+                "BrowserStack cannot be set when a device is already configured.\n"
+                "> for_device() and for_browserstack() are mutually exclusive"
+            )
+        if self._cloud_mobile_id_or_ref is not None:
+            raise ValueError(
+                "BrowserStack cannot be set when a cloud mobile is already configured.\n"
+                "> for_cloud_mobile() and for_browserstack() are mutually exclusive"
+            )
+        self._browserstack_config = config
+        self._device_platform = DevicePlatform.IOS
+        return self
+
+    def with_default_task_config(self, config: TaskRequestCommon) -> "AgentConfigBuilder":
+        """
+        Set the default task configuration.
+
+        Args:
+            config: The task configuration to use as default
+        """
+        self._task_request_defaults = copy.deepcopy(config)
+        return self
+
+    def with_adb_server(self, host: str, port: int | None = None) -> "AgentConfigBuilder":
+        """
+        Set the ADB server host and port.
+
+        Args:
+            host: The ADB server host
+            port: The ADB server port
+        """
+        self._servers.adb_host = host
+        if port is not None:
+            self._servers.adb_port = port
+        return self
+
+    def with_servers(self, servers: ServerConfig) -> "AgentConfigBuilder":
+        """
+        Set the server settings.
+
+        Args:
+            servers: The server settings to use
+        """
+        self._servers = copy.deepcopy(servers)
+        return self
+
+    def with_graph_config_callbacks(self, callbacks: Callbacks) -> "AgentConfigBuilder":
+        """
+        Set the graph config callbacks.
+
+        Args:
+            callbacks: The graph config callbacks to use
+        """
+        self._graph_config_callbacks = callbacks
+        return self
+
+    def with_ios_client_config(self, config: IosClientConfig) -> "AgentConfigBuilder":
+        self._ios_client_config = copy.deepcopy(config)
+        return self
+
+    def with_video_recording_tools(self) -> "AgentConfigBuilder":
+        """
+        Enable video recording tools (start_video_recording, stop_video_recording).
+
+        When enabled, the agent will have access to tools for recording the device
+        screen and analyzing the video content using Gemini models.
+
+        IMPORTANT: This requires:
+        1. ffmpeg to be installed on the system (for video compression)
+        2. A video-capable model configured in utils.video_analyzer
+
+        Supported models for video_analyzer:
+        - gemini-3-flash-preview (recommended)
+        - gemini-3-pro-preview
+        - gemini-2.5-flash
+        - gemini-2.5-pro
+        - gemini-2.0-flash
+
+        Returns:
+            Self for method chaining
+
+        Raises:
+            FFmpegNotInstalledError: If ffmpeg is not installed
+            ValueError: When the agent is initialized if any profile lacks video_analyzer config
+        """
+        from minitap.mobile_use.utils.video import check_ffmpeg_available
+
+        check_ffmpeg_available()
+        self._video_recording_enabled = True
+        return self
+
+    def build(self, validate_profiles: bool = True) -> AgentConfig:
+        """
+        Build the mobile-use AgentConfig object.
+
+        Args:
+            default_profile: Name of the default agent profile to use
+
+        Returns:
+            A configured AgentConfig object
+
+        Raises:
+            ValueError: If default_profile is specified but not found in configured profiles
+        """
+        nb_profiles = len(self._agent_profiles)
+
+        if isinstance(self._default_profile, str):
+            profile_name = self._default_profile
+            default_profile = self._agent_profiles.get(profile_name, None)
+            if default_profile is None:
+                raise ValueError(f"Profile '{profile_name}' not found in configured agents")
+        elif isinstance(self._default_profile, AgentProfile):
+            default_profile = self._default_profile
+            if default_profile.name not in self._agent_profiles:
+                self.add_profile(default_profile, validate=validate_profiles)
+        elif nb_profiles <= 0:
+            llm_config = (
+                get_default_minitap_llm_config(validate=validate_profiles)
+                or get_default_llm_config()
+            )
+            default_profile = AgentProfile(
+                name=DEFAULT_PROFILE_NAME,
+                llm_config=llm_config,
+            )
+            self.add_profile(default_profile, validate=validate_profiles)
+        elif nb_profiles == 1:
+            # Select the only one available
+            default_profile = next(iter(self._agent_profiles.values()))
+        else:
+            available_profiles = ", ".join(self._agent_profiles.keys())
+            raise ValueError(
+                f"You must call with_default_profile() to select one among: {available_profiles}"
+            )
+
+        return AgentConfig(
+            agent_profiles=self._agent_profiles,
+            task_request_defaults=self._task_request_defaults or TaskRequestCommon(),
+            default_profile=default_profile,
+            device_id=self._device_id,
+            device_platform=self._device_platform,
+            servers=self._servers,
+            graph_config_callbacks=self._graph_config_callbacks,
+            cloud_mobile_id_or_ref=self._cloud_mobile_id_or_ref,
+            ios_client_config=self._ios_client_config,
+            browserstack_config=self._browserstack_config,
+            video_recording_enabled=self._video_recording_enabled,
+        )
+
+
+def get_default_agent_config():
+    return AgentConfigBuilder().build()
+
+
+def get_default_servers():
+    return ServerConfig(
+        adb_host="localhost",
+        adb_port=5037,
+    )

minitap/mobile_use/sdk/builders/index.py

@@ -0,0 +1,15 @@
+from minitap.mobile_use.sdk.builders.agent_config_builder import AgentConfigBuilder
+from minitap.mobile_use.sdk.builders.task_request_builder import TaskRequestCommonBuilder
+
+
+class BuildersWrapper:
+    @property
+    def AgentConfig(self) -> AgentConfigBuilder:
+        return AgentConfigBuilder()
+
+    @property
+    def TaskDefaults(self) -> TaskRequestCommonBuilder:
+        return TaskRequestCommonBuilder()
+
+
+Builders = BuildersWrapper()

minitap/mobile_use/sdk/builders/task_request_builder.py

@@ -0,0 +1,236 @@
+"""
+Builder for TaskRequest objects using a fluent interface.
+"""
+
+from pathlib import Path
+from typing import Self, TypeVar, cast
+
+from pydantic import BaseModel
+
+from minitap.mobile_use.constants import RECURSION_LIMIT
+from minitap.mobile_use.sdk.types.agent import AgentProfile
+from minitap.mobile_use.sdk.types.task import TaskRequest, TaskRequestCommon
+
+
+TIn = TypeVar("TIn", bound=BaseModel | None)
+TOut = TypeVar("TOut", bound=BaseModel)
+
+
+class TaskRequestCommonBuilder(BaseModel):
+    """
+    Builder class providing a fluent interface for creating TaskRequestCommon objects.
+    """
+
+    def __init__(self):
+        self._max_steps = RECURSION_LIMIT
+        self._record_trace = False
+        self._trace_path = Path("mobile-use-traces")
+        self._llm_output_path: Path | None = None
+        self._thoughts_output_path: Path | None = None
+        self._locked_app_package: str | None = None
+
+    def with_max_steps(self, max_steps: int) -> Self:
+        """
+        Set the maximum number of steps the task can take.
+
+        Args:
+            max_steps: Maximum number of steps
+        """
+        self._max_steps = max_steps
+        return self
+
+    def with_trace_recording(self, enabled: bool = True, path: str | None = None) -> Self:
+        """
+        Configure trace recording for the task.
+
+        Traces record screenshots and actions during execution.
+
+        Args:
+            enabled: Whether to enable trace recording
+            path: Directory path where traces should be saved
+        """
+        self._record_trace = enabled
+        if enabled and path:
+            self._trace_path = Path(path)
+        return self
+
+    def with_llm_output_saving(self, path: str) -> Self:
+        """
+        Configure LLM output saving for the task.
+
+        Args:
+            path: Path where to save the LLM output message
+        """
+        self._llm_output_path = Path(path)
+        return self
+
+    def with_thoughts_output_saving(self, path: str) -> Self:
+        """
+        Configure thoughts output saving for the task.
+
+        Args:
+            path: Path where to save the thoughts output message
+        """
+        self._thoughts_output_path = Path(path)
+        return self
+
+    def with_locked_app_package(self, package_name: str) -> Self:
+        """
+        Set the app package to lock execution to.
+
+        This ensures the specified app is launched and in the foreground before
+        the agentic loop starts.
+
+        Args:
+            package_name: Package name (Android, e.g., 'com.whatsapp') or
+                         bundle ID (iOS, e.g., 'com.apple.mobilesafari')
+        """
+        self._locked_app_package = package_name
+        return self
+
+    def build(self) -> TaskRequestCommon:
+        """
+        Build the TaskRequestCommon object.
+
+        Returns:
+            A configured TaskRequestCommon object
+
+        Raises:
+            ValueError: If required fields are missing
+        """
+        return TaskRequestCommon(
+            max_steps=self._max_steps,
+            record_trace=self._record_trace,
+            trace_path=self._trace_path,
+            llm_output_path=self._llm_output_path,
+            thoughts_output_path=self._thoughts_output_path,
+            locked_app_package=self._locked_app_package,
+        )
+
+
+class TaskRequestBuilder[TIn](TaskRequestCommonBuilder):
+    """
+    Builder class providing a fluent interface for creating TaskRequest objects.
+
+    This builder allows for step-by-step construction of a TaskRequest with
+    clear methods that make the configuration process intuitive and type-safe.
+
+    Examples:
+        >>> builder = TaskRequestBuilder[None](goal="Open Gmail and check unread emails")
+        >>> task_request = (
+        ...     builder
+        ...     .with_max_steps(30)
+        ...     .using_profile("LowReasoning")
+        ...     .with_output_description("A list of email subjects and senders")
+        ...     .build()
+        ... )
+    """
+
+    def __init__(self, goal: str):
+        """Initialize an empty TaskRequestBuilder."""
+        super().__init__()
+        self._goal = goal
+        self._profile: str | AgentProfile | None = None
+        self._name: str | None = None
+        self._output_description = None
+        self._output_format: type[TIn] | None = None
+
+    @classmethod
+    def from_common(cls, goal: str, common: TaskRequestCommon):
+        res = cls(goal=goal)
+        res._max_steps = common.max_steps
+        res._record_trace = common.record_trace
+        res._trace_path = common.trace_path
+        res._llm_output_path = common.llm_output_path
+        res._thoughts_output_path = common.thoughts_output_path
+        res._locked_app_package = common.locked_app_package
+        return res
+
+    def using_profile(self, profile: str | AgentProfile) -> "TaskRequestBuilder[TIn]":
+        """
+        Set the agent profile for executing the task.
+
+        Args:
+            profile: The agent profile to use
+        """
+        self._profile = profile
+        return self
+
+    def with_name(self, name: str) -> "TaskRequestBuilder[TIn]":
+        """
+        Set the name of the task - useful when recording traces.
+        Otherwise, a random name will be generated.
+
+        Args:
+            name: Name of the task
+        """
+        self._name = name
+        return self
+
+    def without_llm_output_saving(self) -> Self:
+        """
+        Disable LLM output saving for the task.
+        """
+        self._llm_output_path = None
+        return self
+
+    def without_thoughts_output_saving(self):
+        """
+        Disable thoughts output saving for the task.
+        """
+        self._thoughts_output_path = None
+        return self
+
+    def with_output_description(self, description: str) -> "TaskRequestBuilder[TIn]":
+        """
+        Set the description of the expected output format.
+        This is especially useful for data extraction tasks.
+
+        Args:
+            description: Description of the expected output format
+        """
+        self._output_description = description
+        return self
+
+    def with_output_format(self, output_format: type[TOut]) -> "TaskRequestBuilder[TOut]":
+        """
+        Set the pydantic model for the expected output format.
+
+        Args:
+            output_format: Pydantic model instance defining the output format
+        """
+        self._output_format = output_format  # type: ignore
+        return cast(TaskRequestBuilder[TOut], self)
+
+    def build(self) -> TaskRequest[TIn]:
+        """
+        Build the TaskRequest object.
+
+        Returns:
+            A configured TaskRequest object
+
+        Raises:
+            ValueError: If required fields are missing
+        """
+        if not self._goal:
+            raise ValueError("Task goal is required")
+
+        if self._output_format and self._output_description:
+            raise ValueError("Output format and description are mutually exclusive")
+
+        task_request = TaskRequest(
+            goal=self._goal,
+            profile=self._profile.name
+            if isinstance(self._profile, AgentProfile)
+            else self._profile,
+            task_name=self._name,
+            output_description=self._output_description,
+            output_format=self._output_format,
+            max_steps=self._max_steps,
+            record_trace=self._record_trace,
+            trace_path=self._trace_path,
+            llm_output_path=self._llm_output_path,
+            thoughts_output_path=self._thoughts_output_path,
+            locked_app_package=self._locked_app_package,
+        )
+        return task_request

minitap/mobile_use/sdk/constants.py

@@ -0,0 +1 @@
+DEFAULT_PROFILE_NAME = "default"

minitap/mobile_use/sdk/examples/README.md

@@ -0,0 +1,83 @@
+# mobile-use SDK Examples
+
+Location: `src/mobile_use/sdk/examples/`
+
+Run any example via:
+
+- `python src/mobile_use/sdk/examples/<filename>.py`
+
+## Practical Automation Examples
+
+These examples demonstrate two different ways to use the SDK, each applying an appropriate level of complexity for the task at hand:
+
+### platform_minimal_example.py - Painless integration with the Minitap platform
+
+This script shows the simplest way to run minitap :
+
+- Visit https://platform.minitap.ai to create a task and get your API key.
+- Initialize the agent with your API key: .init(api_key=...).
+- Ask the agent to run one of the tasks you’ve set up in the Minitap platform
+  (e.g., "like-instagram-post").
+- The task’s goal and settings live in the Minitap platform, you don’t need
+  to hardcode them here.
+- If you’ve created different profiles (LLM configurations) in the Minitap platform (like "fast-config"),
+  you can pick which one to use with the `profile` field.
+
+### simple_photo_organizer.py - Straightforward Approach
+
+Demonstrates the simplest way to use the SDK for quick automation tasks:
+
+- **Direct API calls** without builders or complex configuration
+- Creates a photo album and organizes photos from a specific date
+- Uses structured Pydantic output to capture results
+
+### smart_notification_assistant.py - Feature-Rich Approach
+
+Showcases more advanced SDK features while remaining practical:
+
+- Uses builder pattern for configuring the agent and overriding the default task configurations
+- Implements **multiple specialized agent profiles** for different reasoning tasks:
+  - Analyzer profile for detailed inspection of notifications
+  - Note taker profile for writing a summary of the notifications
+- Enables **tracing** for debugging and visualization
+- Includes **structured Pydantic models** with enums and nested relationships
+- Demonstrates proper **exception handling** for different error types
+- Shows how to set up task defaults for consistent configuration
+
+## Usage Notes
+
+- **Choosing an Approach**:
+
+  - Use the direct approach (like `platform_minimal_example.py`) for painless setup using the Minitap platform. You can configure any task, save, run, and monitor them with a few clicks.
+  - Use the simple approach (like `simple_photo_organizer.py`) for straightforward tasks, you configure settings yourself and every LLM call happens on your device.
+  - Use the builder approach (like `smart_notification_assistant.py`) when you need more customization.
+
+- **Device Detection**: The agent detects the first available device unless you specify one with `AgentConfigBuilder.for_device(...)`.
+
+- **Servers**: With default base URLs (`localhost:9998/9999`), the agent starts the servers automatically. When you override URLs, it assumes servers are already running.
+
+- **LLM API Keys**: Provide necessary keys (e.g., `OPENAI_API_KEY`) in a `.env` file at repo root; see `mobile_use/config.py`.
+
+- **Traces**: When enabled, traces are saved to a specified directory (defaulting to `./mobile-use-traces/`) and can be useful for debugging and visualization.
+
+- **Structured Output**: Pydantic models enable type safety when processing task outputs, making it easier to handle and chain results between tasks.
+
+## Locked App Execution
+
+You can restrict task execution to a specific app using the `with_locked_app_package()` method. This ensures the agent stays within the target application throughout the task execution.
+
+```python
+# Lock execution to WhatsApp
+result = await agent.run_task(
+    request=agent.new_task("Send message to Bob")
+        .with_locked_app_package("com.whatsapp")
+        .build()
+)
+```
+
+**When locked to an app:**
+
+- The system verifies the app is open before starting
+- If the app is accidentally closed or navigated away from, the Contextor agent will attempt to relaunch it
+- The Planner and Cortex agents will prioritize in-app actions
+

minitap/mobile_use/sdk/examples/__init__.py

@@ -0,0 +1 @@
+"""Example scripts for the mobile-use SDK."""