enhance: browser_som_screenshot (#3059)

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

Author: nitpicker55555

camel/toolkits/hybrid_browser_toolkit/config_loader.py

@@ -44,6 +44,9 @@ class BrowserConfig:
     connect_over_cdp: bool = False
     cdp_url: Optional[str] = None
 
+    # Full visual mode configuration
+    full_visual_mode: bool = False
+
 
 @dataclass
 class ToolkitConfig:
@@ -116,6 +119,8 @@ def from_kwargs(cls, **kwargs) -> 'ConfigLoader':
                 toolkit_kwargs["session_id"] = value
             elif key == "enabledTools":
                 toolkit_kwargs["enabled_tools"] = value
+            elif key == "fullVisualMode":
+                browser_kwargs["full_visual_mode"] = value
 
         browser_config = BrowserConfig(**browser_kwargs)
         toolkit_config = ToolkitConfig(**toolkit_kwargs)
@@ -146,6 +151,7 @@ def to_ws_config(self) -> Dict[str, Any]:
             "viewport_limit": self.browser_config.viewport_limit,
             "connectOverCdp": self.browser_config.connect_over_cdp,
             "cdpUrl": self.browser_config.cdp_url,
+            "fullVisualMode": self.browser_config.full_visual_mode,
         }
 
     def get_timeout_config(self) -> Dict[str, Optional[int]]:

camel/toolkits/hybrid_browser_toolkit/hybrid_browser_toolkit.py

@@ -50,6 +50,7 @@ def __new__(
         viewport_limit: bool = False,
         connect_over_cdp: bool = False,
         cdp_url: Optional[str] = None,
+        full_visual_mode: bool = False,
         **kwargs: Any,
     ) -> Any:
         r"""Create a HybridBrowserToolkit instance with the specified mode.
@@ -98,6 +99,11 @@ def __new__(
             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)
+            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.
 
@@ -129,6 +135,7 @@ def __new__(
                 viewport_limit=viewport_limit,
                 connect_over_cdp=connect_over_cdp,
                 cdp_url=cdp_url,
+                full_visual_mode=full_visual_mode,
                 **kwargs,
             )
         elif mode == "python":

camel/toolkits/hybrid_browser_toolkit/hybrid_browser_toolkit_ts.py

@@ -98,6 +98,7 @@ def __init__(
         viewport_limit: bool = False,
         connect_over_cdp: bool = False,
         cdp_url: Optional[str] = None,
+        full_visual_mode: bool = False,
     ) -> None:
         r"""Initialize the HybridBrowserToolkit.
 
@@ -143,6 +144,9 @@ def __init__(
             cdp_url (Optional[str]): WebSocket endpoint URL for CDP
             connection (e.g., 'ws://localhost:9222/devtools/browser/...').
             Required when connect_over_cdp is True. Defaults to None.
+            full_visual_mode (bool): When True, browser actions like click,
+            browser_open, visit_page, etc. will not return snapshots.
+            Defaults to False.
         """
         super().__init__()
         RegisteredAgentToolkit.__init__(self)
@@ -167,6 +171,7 @@ def __init__(
             enabled_tools=enabled_tools,
             connect_over_cdp=connect_over_cdp,
             cdp_url=cdp_url,
+            full_visual_mode=full_visual_mode,
         )
 
         # Legacy attribute access for backward compatibility
@@ -182,6 +187,7 @@ def __init__(
         self._default_start_url = browser_config.default_start_url
         self._session_id = toolkit_config.session_id or "default"
         self._viewport_limit = browser_config.viewport_limit
+        self._full_visual_mode = browser_config.full_visual_mode
 
         # Store timeout configuration for backward compatibility
         self._default_timeout = browser_config.default_timeout
@@ -648,22 +654,29 @@ async def browser_click(self, *, ref: str) -> Dict[str, Any]:
 
             # Add tab information
             tab_info = await ws_wrapper.get_tab_info()
-            result.update(
-                {
-                    "tabs": tab_info,
-                    "current_tab": next(
-                        (
-                            i
-                            for i, tab in enumerate(tab_info)
-                            if tab.get("is_current")
-                        ),
-                        0,
+
+            response = {
+                "result": result.get("result", ""),
+                "snapshot": result.get("snapshot", ""),
+                "tabs": tab_info,
+                "current_tab": next(
+                    (
+                        i
+                        for i, tab in enumerate(tab_info)
+                        if tab.get("is_current")
                     ),
-                    "total_tabs": len(tab_info),
-                }
-            )
+                    0,
+                ),
+                "total_tabs": len(tab_info),
+            }
 
-            return result
+            if "newTabId" in result:
+                response["newTabId"] = result["newTabId"]
+
+            if "timing" in result:
+                response["timing"] = result["timing"]
+
+            return response
         except Exception as e:
             logger.error(f"Failed to click element: {e}")
             return {
@@ -1377,6 +1390,8 @@ def clone_for_new_session(
             screenshot_timeout=self._screenshot_timeout,
             page_stability_timeout=self._page_stability_timeout,
             dom_content_loaded_timeout=self._dom_content_loaded_timeout,
+            viewport_limit=self._viewport_limit,
+            full_visual_mode=self._full_visual_mode,
         )
 
     def get_tools(self) -> List[FunctionTool]: