|
| 1 | +import os |
| 2 | +from typing import Tuple |
| 3 | + |
| 4 | +from dotenv import load_dotenv |
| 5 | +from playwright.sync_api import Browser |
| 6 | +from playwright.sync_api import Error as PlaywrightError |
| 7 | +from playwright.sync_api import Page |
| 8 | +import requests |
| 9 | + |
| 10 | +from ..shared.base_playwright import BasePlaywrightComputer |
| 11 | + |
| 12 | + |
| 13 | +load_dotenv() |
| 14 | + |
| 15 | + |
| 16 | +class AnchorBrowser(BasePlaywrightComputer): |
| 17 | + """ |
| 18 | + Computer implementation for Anchor browser (https://anchorbrowser.io) |
| 19 | + Requires an API key in the .env file as ANCHOR_API_KEY |
| 20 | +
|
| 21 | + IMPORTANT: The `goto` and navigation tools are already implemented and recommended |
| 22 | + when using the Anchor computer to help the agent navigate more effectively. |
| 23 | + """ |
| 24 | + |
| 25 | + def __init__( |
| 26 | + self, |
| 27 | + width: int = 1024, |
| 28 | + height: int = 900, |
| 29 | + proxy_active: bool = True, |
| 30 | + adblock_active: bool = True, |
| 31 | + popup_blocking_active: bool = True, |
| 32 | + captcha_active: bool = True, |
| 33 | + timeout: int = 15, |
| 34 | + idle_timeout: int = 2, |
| 35 | + debug: bool = False, |
| 36 | + ): |
| 37 | + """Initialize the Anchor browser session""" |
| 38 | + super().__init__() |
| 39 | + self.api_key = os.getenv("ANCHOR_API_KEY") |
| 40 | + if not self.api_key: |
| 41 | + raise ValueError("ANCHOR_API_KEY not found in .env file") |
| 42 | + |
| 43 | + self.debug = debug |
| 44 | + self.base_url = "https://api.anchorbrowser.io/api" |
| 45 | + self.base_ws_url = "wss://connect.anchorbrowser.io" |
| 46 | + self.session_id = None |
| 47 | + self.dimensions = (width, height) |
| 48 | + self.proxy_config = {"active": proxy_active} |
| 49 | + self.adblock_config = { |
| 50 | + "active": adblock_active, |
| 51 | + "popup_blocking_active": popup_blocking_active, |
| 52 | + } |
| 53 | + self.captcha_config = {"active": captcha_active} |
| 54 | + self.timeout = timeout |
| 55 | + self.idle_timeout = idle_timeout |
| 56 | + self._browser = None |
| 57 | + self._page = None |
| 58 | + |
| 59 | + if self.debug: |
| 60 | + print(f"Anchor browser initialized with viewport {width}x{height}") |
| 61 | + |
| 62 | + def _get_browser_and_page(self) -> Tuple[Browser, Page]: |
| 63 | + """ |
| 64 | + Get browser and page objects. |
| 65 | + For Anchor, we don't have direct browser/page objects, but we simulate them |
| 66 | + for compatibility with the BasePlaywrightComputer interface. |
| 67 | + """ |
| 68 | + # Create a session with Anchor API |
| 69 | + width, height = self.dimensions |
| 70 | + response = requests.post( |
| 71 | + f"{self.base_url}/sessions", |
| 72 | + headers={"anchor-api-key": f"{self.api_key}"}, |
| 73 | + json={ |
| 74 | + "width": width, |
| 75 | + "height": height, |
| 76 | + "useProxy": self.proxy_config["active"], |
| 77 | + "solveCaptcha": self.captcha_config["active"], |
| 78 | + "sessionTimeout": self.timeout, |
| 79 | + "sessionIdleTimeout": self.idle_timeout, |
| 80 | + "adBlocker": self.adblock_config["active"], |
| 81 | + "popupBlockingActive": self.adblock_config["popup_blocking_active"], |
| 82 | + "headless": False, |
| 83 | + }, |
| 84 | + ) |
| 85 | + response.raise_for_status() |
| 86 | + self.session_id = response.json().get("id") |
| 87 | + if not self.session_id: |
| 88 | + raise ValueError("Failed to create Anchor browser session") |
| 89 | + |
| 90 | + browser = self._playwright.chromium.connect_over_cdp( |
| 91 | + f"{self.base_ws_url}/?sessionId={self.session_id}" |
| 92 | + ) |
| 93 | + context = browser.contexts[0] |
| 94 | + context.on("page", self._handle_new_page) |
| 95 | + page = context.pages[0] |
| 96 | + page.goto("https://bing.com") |
| 97 | + return browser, page |
| 98 | + |
| 99 | + def _handle_new_page(self, page: Page): |
| 100 | + """Handle the creation of a new page.""" |
| 101 | + print("New page created") |
| 102 | + self._page = page |
| 103 | + page.on("close", self._handle_page_close) |
| 104 | + |
| 105 | + def _handle_page_close(self, page: Page): |
| 106 | + """Handle the closure of a page.""" |
| 107 | + print("Page closed") |
| 108 | + if self._page == page: |
| 109 | + if self._browser.contexts[0].pages: |
| 110 | + self._page = self._browser.contexts[0].pages[-1] |
| 111 | + else: |
| 112 | + print("Warning: All pages have been closed.") |
| 113 | + self._page = None |
| 114 | + |
| 115 | + def __exit__(self, exc_type, exc_val, exc_tb): |
| 116 | + """Clean up resources when exiting""" |
| 117 | + if self.session_id: |
| 118 | + requests.delete( |
| 119 | + f"{self.base_url}/sessions/{self.session_id}", |
| 120 | + headers={"anchor-api-key": f"{self.api_key}"}, |
| 121 | + ) |
| 122 | + if self.debug: |
| 123 | + print(f"Ended Anchor session: {self.session_id}") |
| 124 | + self.session_id = None |
| 125 | + |
| 126 | + def screenshot(self) -> str: |
| 127 | + """ |
| 128 | + Capture a screenshot of the current viewport using CDP. |
| 129 | +
|
| 130 | + Returns: |
| 131 | + str: A base64 encoded string of the screenshot. |
| 132 | + """ |
| 133 | + try: |
| 134 | + # Get CDP session from the page |
| 135 | + cdp_session = self._page.context.new_cdp_session(self._page) |
| 136 | + |
| 137 | + # Capture screenshot using CDP |
| 138 | + result = cdp_session.send( |
| 139 | + "Page.captureScreenshot", {"format": "png", "fromSurface": True} |
| 140 | + ) |
| 141 | + |
| 142 | + return result["data"] |
| 143 | + except PlaywrightError as error: |
| 144 | + print( |
| 145 | + f"CDP screenshot failed, falling back to standard screenshot: {error}" |
| 146 | + ) |
| 147 | + return super().screenshot() |
0 commit comments