chore: browser_open (#3122)

Co-authored-by: Wendong-Fan <w3ndong.fan@gmail.com>

Author: nitpicker55555

camel/toolkits/hybrid_browser_toolkit/hybrid_browser_toolkit.py

@@ -25,6 +25,66 @@ class HybridBrowserToolkit(BaseToolkit):
     This wrapper allows users to choose between:
     - 'typescript': WebSocket-based implementation using TypeScript/Node.js
     - 'python': Pure Python implementation using Playwright directly
+
+    Args:
+        mode (Literal["typescript", "python"]): Implementation mode. -
+            'typescript': Uses WebSocket-based TypeScript implementation -
+            'python': Uses pure Python Playwright implementation. Defaults to
+            "typescript".
+        headless (bool): Whether to run browser in headless mode.
+            Defaults to True.
+        user_data_dir (Optional[str]): Directory for user data
+            persistence. Defaults to None.
+        stealth (bool): Whether to enable stealth mode. Defaults to
+            False.
+        web_agent_model (Optional[BaseModelBackend]): Model for web
+            agent operations. Defaults to None.
+        cache_dir (str): Directory for caching. Defaults to "tmp/".
+        enabled_tools (Optional[List[str]]): List of enabled tools.
+            Defaults to None.
+        browser_log_to_file (bool): Whether to log browser actions to
+            file. Defaults to False.
+        log_dir (Optional[str]): Custom directory path for log files.
+            If None, defaults to "browser_log". Defaults to None.
+        session_id (Optional[str]): Session identifier. Defaults to None.
+        default_start_url (str): Default URL to start with. Defaults
+            to "https://google.com/".
+        default_timeout (Optional[int]): Default timeout in
+            milliseconds. Defaults to None.
+        short_timeout (Optional[int]): Short timeout in milliseconds.
+            Defaults to None.
+        navigation_timeout (Optional[int]): Navigation timeout in
+            milliseconds. Defaults to None.
+        network_idle_timeout (Optional[int]): Network idle timeout in
+            milliseconds. Defaults to None.
+        screenshot_timeout (Optional[int]): Screenshot timeout in
+            milliseconds. Defaults to None.
+        page_stability_timeout (Optional[int]): Page stability timeout
+            in milliseconds. Defaults to None.
+        dom_content_loaded_timeout (Optional[int]): DOM content loaded
+            timeout in milliseconds. Defaults to None.
+        viewport_limit (bool): Whether to filter page snapshot
+            elements to only those visible in the current viewport.
+            Defaults to False.
+        connect_over_cdp (bool): Whether to connect to an existing
+            browser via Chrome DevTools Protocol. Defaults to False.
+            (Only supported in TypeScript mode)
+        cdp_url (Optional[str]): WebSocket endpoint URL for CDP
+            connection. Required when connect_over_cdp is True.
+            Defaults to None. (Only supported in TypeScript mode)
+        cdp_keep_current_page (bool): When True and using CDP mode,
+        won't create new pages but use the existing one. Defaults to False.
+            (Only supported in TypeScript mode)
+        full_visual_mode (bool): When True, browser actions like click,
+            browser_open, visit_page, etc. will return 'full visual mode'
+            as snapshot instead of actual page content. The
+            browser_get_page_snapshot method will still return the actual
+            snapshot. Defaults to False.
+        **kwargs: Additional keyword arguments passed to the
+            implementation.
+
+    Returns:
+        HybridBrowserToolkit instance of the specified implementation.
     """
 
     def __new__(

camel/toolkits/hybrid_browser_toolkit/ws_wrapper.py

@@ -109,6 +109,7 @@ def __init__(self, config: Optional[Dict[str, Any]] = None):
         self._send_lock = asyncio.Lock()
         self._receive_task = None
         self._pending_responses: Dict[str, asyncio.Future[Dict[str, Any]]] = {}
+        self._browser_opened = False
         self._server_ready_future = None
 
         self.browser_log_to_file = (config or {}).get(
@@ -391,6 +392,9 @@ async def start(self):
 
         await self._send_command('init', self.config)
 
+        if self.config.get('cdpUrl'):
+            self._browser_opened = True
+
     async def stop(self):
         """Stop the WebSocket connection and server."""
         if self.websocket:
@@ -400,11 +404,12 @@ async def stop(self):
                     timeout=2.0,
                 )
 
-            # Close websocket connection
             with contextlib.suppress(Exception):
                 await self.websocket.close()
             self.websocket = None
 
+        self._browser_opened = False
+
         # Gracefully stop the Node process before cancelling the log reader
         if self.process:
             try:
@@ -448,12 +453,13 @@ async def disconnect_only(self):
 
         This is useful for CDP mode where the browser should remain open.
         """
-        # Close websocket connection
         if self.websocket:
             with contextlib.suppress(Exception):
                 await self.websocket.close()
             self.websocket = None
 
+        self._browser_opened = False
+
         # Stop the Node process
         if self.process:
             try:
@@ -709,17 +715,31 @@ async def open_browser(
         response = await self._send_command(
             'open_browser', {'startUrl': start_url}
         )
+        self._browser_opened = True
         return response
 
     @action_logger
     async def close_browser(self) -> str:
         """Close browser."""
         response = await self._send_command('close_browser', {})
+        self._browser_opened = False
         return response['message']
 
     @action_logger
     async def visit_page(self, url: str) -> Dict[str, Any]:
-        """Visit a page."""
+        """Visit a page.
+
+        In non-CDP mode, automatically opens browser if not already open.
+        """
+        if not self._browser_opened:
+            is_cdp_mode = bool(self.config.get('cdpUrl'))
+
+            if not is_cdp_mode:
+                logger.info(
+                    "Browser not open, automatically opening browser..."
+                )
+                await self.open_browser()
+
         response = await self._send_command('visit_page', {'url': url})
         return response
 

examples/workforce/eigent.py

@@ -324,7 +324,6 @@ def search_agent_factory(
         "browser_get_som_screenshot",
     ]
     web_toolkit_custom = HybridBrowserToolkit(
-        mode="python",
         headless=False,
         enabled_tools=custom_tools,
         browser_log_to_file=True,