sentienceapi 0.90.16__py3-none-any.whl → 0.92.2__py3-none-any.whl

sentience/base_agent.py

@@ -1,3 +1,5 @@
+from typing import Optional
+
 """
 BaseAgent: Abstract base class for all Sentience agents
 Defines the interface that all agent implementations must follow
@@ -99,3 +101,96 @@ class BaseAgent(ABC):
             >>> # filtered now contains only relevant elements
         """
         return snapshot.elements
+
+
+class BaseAgentAsync(ABC):
+    """
+    Abstract base class for all async Sentience agents.
+
+    Provides a standard interface for:
+    - Executing natural language goals (act)
+    - Tracking execution history
+    - Monitoring token usage
+    - Filtering elements based on goals
+
+    Subclasses must implement:
+    - act(): Execute a natural language goal (async)
+    - get_history(): Return execution history
+    - get_token_stats(): Return token usage statistics
+    - clear_history(): Reset history and token counters
+
+    Subclasses can override:
+    - filter_elements(): Customize element filtering logic
+    """
+
+    @abstractmethod
+    async def act(self, goal: str, **kwargs) -> AgentActionResult:
+        """
+        Execute a natural language goal using the agent (async).
+
+        Args:
+            goal: Natural language instruction (e.g., "Click the login button")
+            **kwargs: Additional parameters (implementation-specific)
+
+        Returns:
+            AgentActionResult with execution details
+
+        Raises:
+            RuntimeError: If execution fails after retries
+        """
+        pass
+
+    @abstractmethod
+    def get_history(self) -> list[ActionHistory]:
+        """
+        Get the execution history of all actions taken.
+
+        Returns:
+            List of ActionHistory entries
+        """
+        pass
+
+    @abstractmethod
+    def get_token_stats(self) -> TokenStats:
+        """
+        Get token usage statistics for the agent session.
+
+        Returns:
+            TokenStats with cumulative token counts
+        """
+        pass
+
+    @abstractmethod
+    def clear_history(self) -> None:
+        """
+        Clear execution history and reset token counters.
+
+        This resets the agent to a clean state.
+        """
+        pass
+
+    def filter_elements(self, snapshot: Snapshot, goal: str | None = None) -> list[Element]:
+        """
+        Filter elements from a snapshot based on goal context.
+
+        Default implementation returns all elements unchanged.
+        Subclasses can override to implement custom filtering logic
+        such as:
+        - Removing irrelevant elements based on goal keywords
+        - Boosting importance of matching elements
+        - Filtering by role, size, or visual properties
+
+        Args:
+            snapshot: Current page snapshot
+            goal: User's goal (can inform filtering strategy)
+
+        Returns:
+            Filtered list of elements (default: all elements)
+
+        Example:
+            >>> agent = SentienceAgentAsync(browser, llm)
+            >>> snap = await snapshot_async(browser)
+            >>> filtered = agent.filter_elements(snap, goal="Click login")
+            >>> # filtered now contains only relevant elements
+        """
+        return snapshot.elements

sentience/browser.py

@@ -2,13 +2,19 @@
 Playwright browser harness with extension loading
 """
 
+import asyncio
 import os
 import shutil
 import tempfile
 import time
 from pathlib import Path
+from typing import Optional, Union
 from urllib.parse import urlparse
 
+from playwright.async_api import BrowserContext as AsyncBrowserContext
+from playwright.async_api import Page as AsyncPage
+from playwright.async_api import Playwright as AsyncPlaywright
+from playwright.async_api import async_playwright
 from playwright.sync_api import BrowserContext, Page, Playwright, sync_playwright
 
 from sentience._extension_loader import find_extension_path
@@ -16,7 +22,7 @@ from sentience.models import ProxyConfig, StorageState, Viewport
 
 # Import stealth for bot evasion (optional - graceful fallback if not available)
 try:
-    from playwright_stealth import stealth_sync
+    from playwright_stealth import stealth_async, stealth_sync
 
     STEALTH_AVAILABLE = True
 except ImportError:
@@ -37,6 +43,7 @@ class SentienceBrowser:
         record_video_dir: str | Path | None = None,
         record_video_size: dict[str, int] | None = None,
         viewport: Viewport | dict[str, int] | None = None,
+        device_scale_factor: float | None = None,
     ):
         """
         Initialize Sentience browser
@@ -109,6 +116,9 @@ class SentienceBrowser:
         else:
             self.viewport = viewport
 
+        # Device scale factor for high-DPI emulation
+        self.device_scale_factor = device_scale_factor
+
         self.playwright: Playwright | None = None
         self.context: BrowserContext | None = None
         self.page: Page | None = None
@@ -211,6 +221,10 @@ class SentienceBrowser:
             "user_agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36",
         }
 
+        # Add device scale factor if configured
+        if self.device_scale_factor is not None:
+            launch_params["device_scale_factor"] = self.device_scale_factor
+
         # Add proxy if configured
         if proxy_config:
             launch_params["proxy"] = proxy_config.to_playwright_dict()
@@ -574,3 +588,472 @@ class SentienceBrowser:
     def __exit__(self, exc_type, exc_val, exc_tb):
         """Context manager exit"""
         self.close()
+
+
+class AsyncSentienceBrowser:
+    """Async version of SentienceBrowser for use in asyncio contexts."""
+
+    def __init__(
+        self,
+        api_key: str | None = None,
+        api_url: str | None = None,
+        headless: bool | None = None,
+        proxy: str | None = None,
+        user_data_dir: str | Path | None = None,
+        storage_state: str | Path | StorageState | dict | None = None,
+        record_video_dir: str | Path | None = None,
+        record_video_size: dict[str, int] | None = None,
+        viewport: Viewport | dict[str, int] | None = None,
+        device_scale_factor: float | None = None,
+    ):
+        """
+        Initialize Async Sentience browser
+
+        Args:
+            api_key: Optional API key for server-side processing (Pro/Enterprise tiers)
+                    If None, uses free tier (local extension only)
+            api_url: Server URL for API calls (defaults to https://api.sentienceapi.com if api_key provided)
+            headless: Whether to run in headless mode. If None, defaults to True in CI, False otherwise
+            proxy: Optional proxy server URL (e.g., 'http://user:pass@proxy.example.com:8080')
+            user_data_dir: Optional path to user data directory for persistent sessions
+            storage_state: Optional storage state to inject (cookies + localStorage)
+            record_video_dir: Optional directory path to save video recordings
+            record_video_size: Optional video resolution as dict with 'width' and 'height' keys
+            viewport: Optional viewport size as Viewport object or dict with 'width' and 'height' keys.
+                     Examples: Viewport(width=1280, height=800) (default)
+                              Viewport(width=1920, height=1080) (Full HD)
+                              {"width": 1280, "height": 800} (dict also supported)
+                     If None, defaults to Viewport(width=1280, height=800).
+            device_scale_factor: Optional device scale factor to emulate high-DPI (Retina) screens.
+                               Examples: 1.0 (default, standard DPI)
+                                        2.0 (Retina/high-DPI, like MacBook Pro)
+                                        3.0 (very high DPI)
+                               If None, defaults to 1.0 (standard DPI).
+        """
+        self.api_key = api_key
+        # Only set api_url if api_key is provided, otherwise None (free tier)
+        if self.api_key and not api_url:
+            self.api_url = "https://api.sentienceapi.com"
+        else:
+            self.api_url = api_url
+
+        # Determine headless mode
+        if headless is None:
+            # Default to False for local dev, True for CI
+            self.headless = os.environ.get("CI", "").lower() == "true"
+        else:
+            self.headless = headless
+
+        # Support proxy from argument or environment variable
+        self.proxy = proxy or os.environ.get("SENTIENCE_PROXY")
+
+        # Auth injection support
+        self.user_data_dir = user_data_dir
+        self.storage_state = storage_state
+
+        # Video recording support
+        self.record_video_dir = record_video_dir
+        self.record_video_size = record_video_size or {"width": 1280, "height": 800}
+
+        # Viewport configuration - convert dict to Viewport if needed
+        if viewport is None:
+            self.viewport = Viewport(width=1280, height=800)
+        elif isinstance(viewport, dict):
+            self.viewport = Viewport(width=viewport["width"], height=viewport["height"])
+        else:
+            self.viewport = viewport
+
+        # Device scale factor for high-DPI emulation
+        self.device_scale_factor = device_scale_factor
+
+        self.playwright: AsyncPlaywright | None = None
+        self.context: AsyncBrowserContext | None = None
+        self.page: AsyncPage | None = None
+        self._extension_path: str | None = None
+
+    def _parse_proxy(self, proxy_string: str) -> ProxyConfig | None:
+        """
+        Parse proxy connection string into ProxyConfig.
+
+        Args:
+            proxy_string: Proxy URL (e.g., 'http://user:pass@proxy.example.com:8080')
+
+        Returns:
+            ProxyConfig object or None if invalid
+        """
+        if not proxy_string:
+            return None
+
+        try:
+            parsed = urlparse(proxy_string)
+
+            # Validate scheme
+            if parsed.scheme not in ("http", "https", "socks5"):
+                print(f"⚠️  [Sentience] Unsupported proxy scheme: {parsed.scheme}")
+                print("   Supported: http, https, socks5")
+                return None
+
+            # Validate host and port
+            if not parsed.hostname or not parsed.port:
+                print("⚠️  [Sentience] Proxy URL must include hostname and port")
+                print("   Expected format: http://username:password@host:port")
+                return None
+
+            # Build server URL
+            server = f"{parsed.scheme}://{parsed.hostname}:{parsed.port}"
+
+            # Create ProxyConfig with optional credentials
+            return ProxyConfig(
+                server=server,
+                username=parsed.username if parsed.username else None,
+                password=parsed.password if parsed.password else None,
+            )
+
+        except Exception as e:
+            print(f"⚠️  [Sentience] Invalid proxy configuration: {e}")
+            print("   Expected format: http://username:password@host:port")
+            return None
+
+    async def start(self) -> None:
+        """Launch browser with extension loaded (async)"""
+        # Get extension source path using shared utility
+        extension_source = find_extension_path()
+
+        # Create temporary extension bundle
+        self._extension_path = tempfile.mkdtemp(prefix="sentience-ext-")
+        shutil.copytree(extension_source, self._extension_path, dirs_exist_ok=True)
+
+        self.playwright = await async_playwright().start()
+
+        # Build launch arguments
+        args = [
+            f"--disable-extensions-except={self._extension_path}",
+            f"--load-extension={self._extension_path}",
+            "--disable-blink-features=AutomationControlled",
+            "--no-sandbox",
+            "--disable-infobars",
+            "--disable-features=WebRtcHideLocalIpsWithMdns",
+            "--force-webrtc-ip-handling-policy=disable_non_proxied_udp",
+        ]
+
+        if self.headless:
+            args.append("--headless=new")
+
+        # Parse proxy configuration if provided
+        proxy_config = self._parse_proxy(self.proxy) if self.proxy else None
+
+        # Handle User Data Directory
+        if self.user_data_dir:
+            user_data_dir = str(self.user_data_dir)
+            Path(user_data_dir).mkdir(parents=True, exist_ok=True)
+        else:
+            user_data_dir = ""
+
+        # Build launch_persistent_context parameters
+        launch_params = {
+            "user_data_dir": user_data_dir,
+            "headless": False,
+            "args": args,
+            "viewport": {"width": self.viewport.width, "height": self.viewport.height},
+            "user_agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36",
+        }
+
+        # Add device scale factor if configured
+        if self.device_scale_factor is not None:
+            launch_params["device_scale_factor"] = self.device_scale_factor
+
+        # Add proxy if configured
+        if proxy_config:
+            launch_params["proxy"] = proxy_config.to_playwright_dict()
+            launch_params["ignore_https_errors"] = True
+            print(f"🌐 [Sentience] Using proxy: {proxy_config.server}")
+
+        # Add video recording if configured
+        if self.record_video_dir:
+            video_dir = Path(self.record_video_dir)
+            video_dir.mkdir(parents=True, exist_ok=True)
+            launch_params["record_video_dir"] = str(video_dir)
+            launch_params["record_video_size"] = self.record_video_size
+            print(f"🎥 [Sentience] Recording video to: {video_dir}")
+            print(
+                f"   Resolution: {self.record_video_size['width']}x{self.record_video_size['height']}"
+            )
+
+        # Launch persistent context
+        self.context = await self.playwright.chromium.launch_persistent_context(**launch_params)
+
+        self.page = self.context.pages[0] if self.context.pages else await self.context.new_page()
+
+        # Inject storage state if provided
+        if self.storage_state:
+            await self._inject_storage_state(self.storage_state)
+
+        # Apply stealth if available
+        if STEALTH_AVAILABLE:
+            await stealth_async(self.page)
+
+        # Wait a moment for extension to initialize
+        await asyncio.sleep(0.5)
+
+    async def goto(self, url: str) -> None:
+        """Navigate to a URL and ensure extension is ready (async)"""
+        if not self.page:
+            raise RuntimeError("Browser not started. Call await start() first.")
+
+        await self.page.goto(url, wait_until="domcontentloaded")
+
+        # Wait for extension to be ready
+        if not await self._wait_for_extension():
+            try:
+                diag = await self.page.evaluate(
+                    """() => ({
+                    sentience_defined: typeof window.sentience !== 'undefined',
+                    registry_defined: typeof window.sentience_registry !== 'undefined',
+                    snapshot_defined: window.sentience && typeof window.sentience.snapshot === 'function',
+                    extension_id: document.documentElement.dataset.sentienceExtensionId || 'not set',
+                    url: window.location.href
+                })"""
+                )
+            except Exception as e:
+                diag = f"Failed to get diagnostics: {str(e)}"
+
+            raise RuntimeError(
+                "Extension failed to load after navigation. Make sure:\n"
+                "1. Extension is built (cd sentience-chrome && ./build.sh)\n"
+                "2. All files are present (manifest.json, content.js, injected_api.js, pkg/)\n"
+                "3. Check browser console for errors (run with headless=False to see console)\n"
+                f"4. Extension path: {self._extension_path}\n"
+                f"5. Diagnostic info: {diag}"
+            )
+
+    async def _inject_storage_state(self, storage_state: str | Path | StorageState | dict) -> None:
+        """Inject storage state (cookies + localStorage) into browser context (async)"""
+        import json
+
+        # Load storage state
+        if isinstance(storage_state, (str, Path)):
+            with open(storage_state, encoding="utf-8") as f:
+                state_dict = json.load(f)
+            state = StorageState.from_dict(state_dict)
+        elif isinstance(storage_state, StorageState):
+            state = storage_state
+        elif isinstance(storage_state, dict):
+            state = StorageState.from_dict(storage_state)
+        else:
+            raise ValueError(
+                f"Invalid storage_state type: {type(storage_state)}. "
+                "Expected str, Path, StorageState, or dict."
+            )
+
+        # Inject cookies
+        if state.cookies:
+            playwright_cookies = []
+            for cookie in state.cookies:
+                cookie_dict = cookie.model_dump()
+                playwright_cookie = {
+                    "name": cookie_dict["name"],
+                    "value": cookie_dict["value"],
+                    "domain": cookie_dict["domain"],
+                    "path": cookie_dict["path"],
+                }
+                if cookie_dict.get("expires"):
+                    playwright_cookie["expires"] = cookie_dict["expires"]
+                if cookie_dict.get("httpOnly"):
+                    playwright_cookie["httpOnly"] = cookie_dict["httpOnly"]
+                if cookie_dict.get("secure"):
+                    playwright_cookie["secure"] = cookie_dict["secure"]
+                if cookie_dict.get("sameSite"):
+                    playwright_cookie["sameSite"] = cookie_dict["sameSite"]
+                playwright_cookies.append(playwright_cookie)
+
+            await self.context.add_cookies(playwright_cookies)
+            print(f"✅ [Sentience] Injected {len(state.cookies)} cookie(s)")
+
+        # Inject LocalStorage
+        if state.origins:
+            for origin_data in state.origins:
+                origin = origin_data.origin
+                if not origin:
+                    continue
+
+                try:
+                    await self.page.goto(origin, wait_until="domcontentloaded", timeout=10000)
+
+                    if origin_data.localStorage:
+                        localStorage_dict = {
+                            item.name: item.value for item in origin_data.localStorage
+                        }
+                        await self.page.evaluate(
+                            """(localStorage_data) => {
+                                for (const [key, value] of Object.entries(localStorage_data)) {
+                                    localStorage.setItem(key, value);
+                                }
+                            }""",
+                            localStorage_dict,
+                        )
+                        print(
+                            f"✅ [Sentience] Injected {len(origin_data.localStorage)} localStorage item(s) for {origin}"
+                        )
+                except Exception as e:
+                    print(f"⚠️  [Sentience] Failed to inject localStorage for {origin}: {e}")
+
+    async def _wait_for_extension(self, timeout_sec: float = 5.0) -> bool:
+        """Poll for window.sentience to be available (async)"""
+        start_time = time.time()
+        last_error = None
+
+        while time.time() - start_time < timeout_sec:
+            try:
+                result = await self.page.evaluate(
+                    """() => {
+                        if (typeof window.sentience === 'undefined') {
+                            return { ready: false, reason: 'window.sentience undefined' };
+                        }
+                        if (window.sentience._wasmModule === null) {
+                             return { ready: false, reason: 'WASM module not fully loaded' };
+                        }
+                        return { ready: true };
+                    }
+                """
+                )
+
+                if isinstance(result, dict):
+                    if result.get("ready"):
+                        return True
+                    last_error = result.get("reason", "Unknown error")
+            except Exception as e:
+                last_error = f"Evaluation error: {str(e)}"
+
+            await asyncio.sleep(0.3)
+
+        if last_error:
+            import warnings
+
+            warnings.warn(f"Extension wait timeout. Last status: {last_error}")
+
+        return False
+
+    async def close(self, output_path: str | Path | None = None) -> str | None:
+        """
+        Close browser and cleanup (async)
+
+        Args:
+            output_path: Optional path to rename the video file to
+
+        Returns:
+            Path to video file if recording was enabled, None otherwise
+        """
+        temp_video_path = None
+
+        if self.record_video_dir:
+            try:
+                if self.page and self.page.video:
+                    temp_video_path = await self.page.video.path()
+                elif self.context:
+                    for page in self.context.pages:
+                        if page.video:
+                            temp_video_path = await page.video.path()
+                            break
+            except Exception:
+                pass
+
+        if self.context:
+            await self.context.close()
+            self.context = None
+
+        if self.playwright:
+            await self.playwright.stop()
+            self.playwright = None
+
+        if self._extension_path and os.path.exists(self._extension_path):
+            shutil.rmtree(self._extension_path)
+
+        # Clear page reference after closing context
+        self.page = None
+
+        final_path = temp_video_path
+        if temp_video_path and output_path and os.path.exists(temp_video_path):
+            try:
+                output_path = str(output_path)
+                Path(output_path).parent.mkdir(parents=True, exist_ok=True)
+                shutil.move(temp_video_path, output_path)
+                final_path = output_path
+            except Exception as e:
+                import warnings
+
+                warnings.warn(f"Failed to rename video file: {e}")
+                final_path = temp_video_path
+
+        return final_path
+
+    async def __aenter__(self):
+        """Async context manager entry"""
+        await self.start()
+        return self
+
+    async def __aexit__(self, exc_type, exc_val, exc_tb):
+        """Async context manager exit"""
+        await self.close()
+
+    @classmethod
+    async def from_existing(
+        cls,
+        context: AsyncBrowserContext,
+        api_key: str | None = None,
+        api_url: str | None = None,
+    ) -> "AsyncSentienceBrowser":
+        """
+        Create AsyncSentienceBrowser from an existing Playwright BrowserContext.
+
+        Args:
+            context: Existing Playwright BrowserContext
+            api_key: Optional API key for server-side processing
+            api_url: Optional API URL
+
+        Returns:
+            AsyncSentienceBrowser instance configured to use the existing context
+        """
+        instance = cls(api_key=api_key, api_url=api_url)
+        instance.context = context
+        pages = context.pages
+        instance.page = pages[0] if pages else await context.new_page()
+
+        # Apply stealth if available
+        if STEALTH_AVAILABLE:
+            await stealth_async(instance.page)
+
+        # Wait for extension to be ready
+        await asyncio.sleep(0.5)
+
+        return instance
+
+    @classmethod
+    async def from_page(
+        cls,
+        page: AsyncPage,
+        api_key: str | None = None,
+        api_url: str | None = None,
+    ) -> "AsyncSentienceBrowser":
+        """
+        Create AsyncSentienceBrowser from an existing Playwright Page.
+
+        Args:
+            page: Existing Playwright Page
+            api_key: Optional API key for server-side processing
+            api_url: Optional API URL
+
+        Returns:
+            AsyncSentienceBrowser instance configured to use the existing page
+        """
+        instance = cls(api_key=api_key, api_url=api_url)
+        instance.page = page
+        instance.context = page.context
+
+        # Apply stealth if available
+        if STEALTH_AVAILABLE:
+            await stealth_async(instance.page)
+
+        # Wait for extension to be ready
+        await asyncio.sleep(0.5)
+
+        return instance