droidrun 0.3.0__py3-none-any.whl → 0.3.2__py3-none-any.whl

droidrun/cli/main.py

@@ -7,13 +7,21 @@ import click
 import os
 import logging
 import warnings
+from contextlib import nullcontext
 from rich.console import Console
 from droidrun.agent.droid import DroidAgent
 from droidrun.agent.utils.llm_picker import load_llm
 from droidrun.adb import DeviceManager
-from droidrun.tools import AdbTools, IOSTools, Tools
+from droidrun.tools import AdbTools, IOSTools
 from functools import wraps
 from droidrun.cli.logs import LogHandler
+from droidrun.telemetry import print_telemetry_message
+from droidrun.portal import (
+    download_portal_apk,
+    enable_portal_accessibility,
+    PORTAL_PACKAGE_NAME,
+    ping_portal,
+)
 
 # Suppress all warnings
 warnings.filterwarnings("ignore")
@@ -36,7 +44,6 @@ def configure_logging(goal: str, debug: bool):
     )
     logger.addHandler(handler)
 
-
     logger.setLevel(logging.DEBUG if debug else logging.INFO)
     logger.propagate = False
 
@@ -59,6 +66,8 @@ async def run_command(
     model: str,
     steps: int,
     base_url: str,
+    api_base: str,
+    vision: bool,
     reasoning: bool,
     reflection: bool,
     tracing: bool,
@@ -76,6 +85,7 @@ async def run_command(
     with log_handler.render() as live:
         try:
             logger.info(f"🚀 Starting: {command}")
+            print_telemetry_message()
 
             if not kwargs.get("temperature"):
                 kwargs["temperature"] = 0
@@ -92,7 +102,9 @@ async def run_command(
                 device = devices[0].serial
                 logger.info(f"📱 Using device: {device}")
             elif device is None and ios:
-                raise ValueError("iOS device not specified. Please specify the device base url (http://device-ip:6643) via --device")
+                raise ValueError(
+                    "iOS device not specified. Please specify the device base url (http://device-ip:6643) via --device"
+                )
             else:
                 logger.info(f"📱 Using device: {device}")
 
@@ -101,7 +113,11 @@ async def run_command(
             # LLM setup
             log_handler.update_step("Initializing LLM...")
             llm = load_llm(
-                provider_name=provider, model=model, base_url=base_url, **kwargs
+                provider_name=provider,
+                model=model,
+                base_url=base_url,
+                api_base=api_base,
+                **kwargs,
             )
             logger.info(f"🧠 LLM ready: {provider}/{model}")
 
@@ -120,11 +136,12 @@ async def run_command(
                 tools=tools,
                 max_steps=steps,
                 timeout=1000,
+                vision=vision,
                 reasoning=reasoning,
                 reflection=reflection,
                 enable_tracing=tracing,
                 debug=debug,
-                save_trajectories=save_trajectory
+                save_trajectories=save_trajectory,
             )
 
             logger.info("▶️  Starting agent execution...")
@@ -176,14 +193,14 @@ class DroidRunCLI(click.Group):
 @click.option(
     "--provider",
     "-p",
-    help="LLM provider (OpenAI, Ollama, Anthropic, Gemini, DeepSeek)",
-    default="Gemini",
+    help="LLM provider (OpenAI, Ollama, Anthropic, GoogleGenAI, DeepSeek)",
+    default="GoogleGenAI",
 )
 @click.option(
     "--model",
     "-m",
     help="LLM model name",
-    default="models/gemini-2.5-pro",
+    default="models/gemini-2.5-flash",
 )
 @click.option("--temperature", type=float, help="Temperature for LLM", default=0.2)
 @click.option("--steps", type=int, help="Maximum number of steps", default=15)
@@ -194,10 +211,24 @@ class DroidRunCLI(click.Group):
     default=None,
 )
 @click.option(
-    "--reasoning", is_flag=True, help="Enable/disable planning with reasoning", default=False
+    "--api_base",
+    help="Base URL for API (e.g., OpenAI, OpenAI-Like)",
+    default=None,
 )
 @click.option(
-    "--reflection", is_flag=True, help="Enable reflection step for higher reasoning", default=False
+    "--vision",
+    is_flag=True,
+    help="Enable vision capabilites by using screenshots",
+    default=False,
+)
+@click.option(
+    "--reasoning", is_flag=True, help="Enable planning with reasoning", default=False
+)
+@click.option(
+    "--reflection",
+    is_flag=True,
+    help="Enable reflection step for higher reasoning",
+    default=False,
 )
 @click.option(
     "--tracing", is_flag=True, help="Enable Arize Phoenix tracing", default=False
@@ -218,7 +249,9 @@ def cli(
     model: str,
     steps: int,
     base_url: str,
+    api_base: str,
     temperature: float,
+    vision: bool,
     reasoning: bool,
     reflection: bool,
     tracing: bool,
@@ -235,14 +268,14 @@ def cli(
 @click.option(
     "--provider",
     "-p",
-    help="LLM provider (OpenAI, Ollama, Anthropic, Gemini, DeepSeek)",
-    default="Gemini",
+    help="LLM provider (OpenAI, Ollama, Anthropic, GoogleGenAI, DeepSeek)",
+    default="GoogleGenAI",
 )
 @click.option(
     "--model",
     "-m",
     help="LLM model name",
-    default="models/gemini-2.5-pro",
+    default="models/gemini-2.5-flash",
 )
 @click.option("--temperature", type=float, help="Temperature for LLM", default=0.2)
 @click.option("--steps", type=int, help="Maximum number of steps", default=15)
@@ -253,10 +286,24 @@ def cli(
     default=None,
 )
 @click.option(
-    "--reasoning", is_flag=True, help="Enable/disable planning with reasoning", default=False
+    "--api_base",
+    help="Base URL for API (e.g., OpenAI or OpenAI-Like)",
+    default=None,
+)
+@click.option(
+    "--vision",
+    is_flag=True,
+    help="Enable vision capabilites by using screenshots",
+    default=False,
 )
 @click.option(
-    "--reflection", is_flag=True, help="Enable reflection step for higher reasoning", default=False
+    "--reasoning", is_flag=True, help="Enable planning with reasoning", default=False
+)
+@click.option(
+    "--reflection",
+    is_flag=True,
+    help="Enable reflection step for higher reasoning",
+    default=False,
 )
 @click.option(
     "--tracing", is_flag=True, help="Enable Arize Phoenix tracing", default=False
@@ -270,9 +317,7 @@ def cli(
     help="Save agent trajectory to file",
     default=False,
 )
-@click.option(
-    "--ios", is_flag=True, help="Run on iOS device", default=False
-)
+@click.option("--ios", is_flag=True, help="Run on iOS device", default=False)
 def run(
     command: str,
     device: str | None,
@@ -280,7 +325,9 @@ def run(
     model: str,
     steps: int,
     base_url: str,
+    api_base: str,
     temperature: float,
+    vision: bool,
     reasoning: bool,
     reflection: bool,
     tracing: bool,
@@ -297,13 +344,15 @@ def run(
         model,
         steps,
         base_url,
+        api_base,
+        vision,
         reasoning,
         reflection,
         tracing,
         debug,
         temperature=temperature,
         save_trajectory=save_trajectory,
-        ios=ios
+        ios=ios,
     )
 
 
@@ -325,17 +374,17 @@ async def devices():
 
 
 @cli.command()
-@click.argument("ip_address")
+@click.argument("serial")
 @click.option("--port", "-p", default=5555, help="ADB port (default: 5555)")
 @coro
-async def connect(ip_address: str, port: int):
+async def connect(serial: str, port: int):
     """Connect to a device over TCP/IP."""
     try:
-        device = await device_manager.connect(ip_address, port)
+        device = await device_manager.connect(serial, port)
         if device:
-            console.print(f"[green]Successfully connected to {ip_address}:{port}[/]")
+            console.print(f"[green]Successfully connected to {serial}:{port}[/]")
         else:
-            console.print(f"[red]Failed to connect to {ip_address}:{port}[/]")
+            console.print(f"[red]Failed to connect to {serial}:{port}[/]")
     except Exception as e:
         console.print(f"[red]Error connecting to device: {e}[/]")
 
@@ -356,16 +405,15 @@ async def disconnect(serial: str):
 
 
 @cli.command()
-@click.option("--path", required=True, help="Path to the APK file to install")
 @click.option("--device", "-d", help="Device serial number or IP address", default=None)
+@click.option("--path", help="Path to the Droidrun Portal APK to install on the device. If not provided, the latest portal apk version will be downloaded and installed.", default=None)
+@click.option(
+    "--debug", is_flag=True, help="Enable verbose debug logging", default=False
+)
 @coro
-async def setup(path: str, device: str | None):
-    """Install an APK file and enable it as an accessibility service."""
+async def setup(path: str | None, device: str | None, debug: bool):
+    """Install and enable the DroidRun Portal on a device."""
     try:
-        if not os.path.exists(path):
-            console.print(f"[bold red]Error:[/] APK file not found at {path}")
-            return
-
         if not device:
             devices = await device_manager.list_devices()
             if not devices:
@@ -381,66 +429,99 @@ async def setup(path: str, device: str | None):
                 f"[bold red]Error:[/] Could not get device object for {device}"
             )
             return
-        tools = Tools(serial=device)
-        console.print(f"[bold blue]Step 1/2: Installing APK:[/] {path}")
-        result = await tools.install_app(path, False, True)
 
-        if "Error" in result:
-            console.print(f"[bold red]Installation failed:[/] {result}")
-            return
+        if not path:
+            console.print("[bold blue]Downloading DroidRun Portal APK...[/]")
+            apk_context = download_portal_apk(debug)
         else:
-            console.print(f"[bold green]Installation successful![/]")
+            console.print(f"[bold blue]Using provided APK:[/] {path}")
+            apk_context = nullcontext(path)
 
-        console.print(f"[bold blue]Step 2/2: Enabling accessibility service[/]")
+        with apk_context as apk_path:
+            if not os.path.exists(apk_path):
+                console.print(f"[bold red]Error:[/] APK file not found at {apk_path}")
+                return
 
-        package = "com.droidrun.portal"
+            console.print(f"[bold blue]Step 1/2: Installing APK:[/] {apk_path}")
+            result = await device_obj.install_app(apk_path, True, True)
 
-        try:
-            await device_obj._adb.shell(
-                device,
-                "settings put secure enabled_accessibility_services com.droidrun.portal/com.droidrun.portal.DroidrunPortalService",
-            )
+            if "Error" in result:
+                console.print(f"[bold red]Installation failed:[/] {result}")
+                return
+            else:
+                console.print(f"[bold green]Installation successful![/]")
 
-            await device_obj._adb.shell(
-                device, "settings put secure accessibility_enabled 1"
-            )
+            console.print(f"[bold blue]Step 2/2: Enabling accessibility service[/]")
 
-            console.print("[green]Accessibility service enabled successfully![/]")
-            console.print(
-                "\n[bold green]Setup complete![/] The DroidRun Portal is now installed and ready to use."
-            )
+            try:
+                await enable_portal_accessibility(device_obj)
 
-        except Exception as e:
-            console.print(
-                f"[yellow]Could not automatically enable accessibility service: {e}[/]"
-            )
-            console.print(
-                "[yellow]Opening accessibility settings for manual configuration...[/]"
-            )
+                console.print("[green]Accessibility service enabled successfully![/]")
+                console.print(
+                    "\n[bold green]Setup complete![/] The DroidRun Portal is now installed and ready to use."
+                )
 
-            await device_obj._adb.shell(
-                device, "am start -a android.settings.ACCESSIBILITY_SETTINGS"
-            )
+            except Exception as e:
+                console.print(
+                    f"[yellow]Could not automatically enable accessibility service: {e}[/]"
+                )
+                console.print(
+                    "[yellow]Opening accessibility settings for manual configuration...[/]"
+                )
+
+                await device_obj.shell(
+                    "am start -a android.settings.ACCESSIBILITY_SETTINGS"
+                )
+
+                console.print(
+                    "\n[yellow]Please complete the following steps on your device:[/]"
+                )
+                console.print(
+                    f"1. Find [bold]{PORTAL_PACKAGE_NAME}[/] in the accessibility services list"
+                )
+                console.print("2. Tap on the service name")
+                console.print(
+                    "3. Toggle the switch to [bold]ON[/] to enable the service"
+                )
+                console.print("4. Accept any permission dialogs that appear")
+
+                console.print(
+                    "\n[bold green]APK installation complete![/] Please manually enable the accessibility service using the steps above."
+                )
 
-            console.print(
-                "\n[yellow]Please complete the following steps on your device:[/]"
-            )
-            console.print(
-                f"1. Find [bold]{package}[/] in the accessibility services list"
-            )
-            console.print("2. Tap on the service name")
-            console.print("3. Toggle the switch to [bold]ON[/] to enable the service")
-            console.print("4. Accept any permission dialogs that appear")
+    except Exception as e:
+        console.print(f"[bold red]Error:[/] {e}")
 
-            console.print(
-                "\n[bold green]APK installation complete![/] Please manually enable the accessibility service using the steps above."
-            )
+        if debug:
+            import traceback
+
+            traceback.print_exc()
+
+
+@cli.command()
+@click.option("--device", "-d", help="Device serial number or IP address", default=None)
+@click.option(
+    "--debug", is_flag=True, help="Enable verbose debug logging", default=False
+)
+@coro
+async def ping(device: str | None, debug: bool):
+    """Ping a device to check if it is ready and accessible."""
+    try:
+        device_obj = await device_manager.get_device(device)
+        if not device_obj:
+            console.print(f"[bold red]Error:[/] Could not find device {device}")
+            return
 
+        await ping_portal(device_obj, debug)
+        console.print(
+            "[bold green]Portal is installed and accessible. You're good to go![/]"
+        )
     except Exception as e:
         console.print(f"[bold red]Error:[/] {e}")
-        import traceback
+        if debug:
+            import traceback
 
-        traceback.print_exc()
+            traceback.print_exc()
 
 
 if __name__ == "__main__":
@@ -449,13 +530,15 @@ if __name__ == "__main__":
     provider = "GoogleGenAI"
     model = "models/gemini-2.5-flash"
     temperature = 0
-    api_key = os.getenv("GEMINI_API_KEY")
+    api_key = os.getenv("GOOGLE_API_KEY")
     steps = 15
+    vision = True
     reasoning = True
     reflection = False
     tracing = True
     debug = True
     base_url = None
+    api_base = None
     ios = False
     run_command(
         command=command,
@@ -464,11 +547,13 @@ if __name__ == "__main__":
         model=model,
         steps=steps,
         temperature=temperature,
+        vision=vision,
         reasoning=reasoning,
         reflection=reflection,
         tracing=tracing,
         debug=debug,
         base_url=base_url,
+        api_base=api_base,
         api_key=api_key,
-        ios=ios
+        ios=ios,
     )

droidrun/portal.py

@@ -0,0 +1,139 @@
+import requests
+import tempfile
+import os
+import contextlib
+from droidrun.adb import Device, DeviceManager
+import asyncio
+
+REPO = "droidrun/droidrun-portal"
+ASSET_NAME = "droidrun-portal"
+GITHUB_API_HOSTS = ["https://api.github.com", "https://ungh.cc"]
+
+PORTAL_PACKAGE_NAME = "com.droidrun.portal"
+A11Y_SERVICE_NAME = (
+    f"{PORTAL_PACKAGE_NAME}/com.droidrun.portal.DroidrunAccessibilityService"
+)
+
+
+def get_latest_release_assets(debug: bool = False):
+    for host in GITHUB_API_HOSTS:
+        url = f"{host}/repos/{REPO}/releases/latest"
+        response = requests.get(url)
+        if response.status_code == 200:
+            if debug:
+                print(f"Using GitHub release on {host}")
+            break
+
+    response.raise_for_status()
+    latest_release = response.json()
+
+    if "release" in latest_release:
+        assets = latest_release["release"]["assets"]
+    else:
+        assets = latest_release.get("assets", [])
+
+    return assets
+
+
+@contextlib.contextmanager
+def download_portal_apk(debug: bool = False):
+    assets = get_latest_release_assets(debug)
+
+    asset_url = None
+    for asset in assets:
+        if (
+            "browser_download_url" in asset
+            and "name" in asset
+            and asset["name"].startswith(ASSET_NAME)
+        ):
+            asset_url = asset["browser_download_url"]
+            break
+        elif "downloadUrl" in asset and os.path.basename(
+            asset["downloadUrl"]
+        ).startswith(ASSET_NAME):
+            asset_url = asset["downloadUrl"]
+            break
+        else:
+            if debug:
+                print(asset)
+
+    if not asset_url:
+        raise Exception(f"Asset named '{ASSET_NAME}' not found in the latest release.")
+
+    tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".apk")
+    try:
+        r = requests.get(asset_url, stream=True)
+        r.raise_for_status()
+        for chunk in r.iter_content(chunk_size=8192):
+            if chunk:
+                tmp.write(chunk)
+        tmp.close()
+        yield tmp.name
+    finally:
+        if os.path.exists(tmp.name):
+            os.unlink(tmp.name)
+
+
+async def enable_portal_accessibility(device: Device):
+    await device.shell(
+        f"settings put secure enabled_accessibility_services {A11Y_SERVICE_NAME}"
+    )
+    await device.shell("settings put secure accessibility_enabled 1")
+
+
+async def check_portal_accessibility(device: Device, debug: bool = False) -> bool:
+    a11y_services = await device.shell(
+        "settings get secure enabled_accessibility_services"
+    )
+    if not A11Y_SERVICE_NAME in a11y_services:
+        if debug:
+            print(a11y_services)
+        return False
+
+    a11y_enabled = await device.shell("settings get secure accessibility_enabled")
+    if a11y_enabled != "1":
+        if debug:
+            print(a11y_enabled)
+        return False
+
+    return True
+
+
+async def ping_portal(device: Device, debug: bool = False):
+    """
+    Ping the Droidrun Portal to check if it is installed and accessible.
+    """
+    try:
+        packages = await device.list_packages()
+    except Exception as e:
+        raise Exception(f"Failed to list packages: {e}")
+
+    if not PORTAL_PACKAGE_NAME in packages:
+        if debug:
+            print(packages)
+        raise Exception("Portal is not installed on the device")
+
+    if not await check_portal_accessibility(device, debug):
+        await device.shell("am start -a android.settings.ACCESSIBILITY_SETTINGS")
+        raise Exception(
+            "Droidrun Portal is not enabled as an accessibility service on the device"
+        )
+
+    try:
+        state = await device.shell(
+            "content query --uri content://com.droidrun.portal/state"
+        )
+        if not "Row: 0 result=" in state:
+            raise Exception("Failed to get state from Droidrun Portal")
+
+    except Exception as e:
+        raise Exception(f"Droidrun Portal is not reachable: {e}")
+
+
+async def test():
+    device = await DeviceManager().get_device()
+    await ping_portal(device, debug=False)
+
+
+if __name__ == "__main__":
+    asyncio.run(test())

droidrun/telemetry/__init__.py

@@ -0,0 +1,4 @@
+from .tracker import capture, flush, print_telemetry_message
+from .events import DroidAgentInitEvent, DroidAgentFinalizeEvent
+
+__all__ = ["capture", "flush", "DroidAgentInitEvent", "DroidAgentFinalizeEvent", "print_telemetry_message"]

droidrun/telemetry/events.py

@@ -0,0 +1,27 @@
+from typing import List
+from droidrun.agent.context import Task
+from pydantic import BaseModel
+
+class TelemetryEvent(BaseModel):
+    pass
+
+class DroidAgentInitEvent(TelemetryEvent):
+    goal: str
+    llm: str
+    tools: str
+    personas: str
+    max_steps: int
+    timeout: int
+    vision: bool
+    reasoning: bool
+    reflection: bool
+    enable_tracing: bool
+    debug: bool
+    save_trajectories: bool
+
+
+class DroidAgentFinalizeEvent(TelemetryEvent):
+    tasks: str
+    success: bool
+    output: str
+    steps: int