Merge branch 'main' into add-ad-use-demo

Author: MagMueller

browser_use/agent/service.py

@@ -1312,7 +1312,7 @@ async def take_step(self, step_info: AgentStepInfo | None = None) -> tuple[bool,
 		Returns:
 		        Tuple[bool, bool]: (is_done, is_valid)
 		"""
-		if len(self.history.history) == 0:
+		if step_info is not None and step_info.step_number == 0:
 			# First step
 			self._log_first_step_startup()
 			await self._execute_initial_actions()

browser_use/browser/session.py

@@ -2255,3 +2255,125 @@ async def cdp_client_for_node(self, node: EnhancedDOMTreeNode) -> CDPSession:
 				self.logger.debug(f'Failed to get CDP client for target {node.target_id}: {e}, using main session')
 
 		return await self.get_or_create_cdp_session()
+
+	async def take_screenshot(
+		self,
+		path: str | None = None,
+		full_page: bool = False,
+		format: str = 'png',
+		quality: int | None = None,
+		clip: dict | None = None,
+	) -> bytes:
+		"""Take a screenshot using CDP.
+
+		Args:
+			path: Optional file path to save screenshot
+			full_page: Capture entire scrollable page beyond viewport
+			format: Image format ('png', 'jpeg', 'webp')
+			quality: Quality 0-100 for JPEG format
+			clip: Region to capture {'x': int, 'y': int, 'width': int, 'height': int}
+
+		Returns:
+			Screenshot data as bytes
+		"""
+		import base64
+
+		from cdp_use.cdp.page import CaptureScreenshotParameters
+
+		cdp_session = await self.get_or_create_cdp_session()
+
+		# Build parameters dict explicitly to satisfy TypedDict expectations
+		params: CaptureScreenshotParameters = {
+			'format': format,
+			'captureBeyondViewport': full_page,
+		}
+
+		if quality is not None and format == 'jpeg':
+			params['quality'] = quality
+
+		if clip:
+			params['clip'] = {
+				'x': clip['x'],
+				'y': clip['y'],
+				'width': clip['width'],
+				'height': clip['height'],
+				'scale': 1,
+			}
+
+		params = CaptureScreenshotParameters(**params)
+
+		result = await cdp_session.cdp_client.send.Page.captureScreenshot(params=params, session_id=cdp_session.session_id)
+
+		if not result or 'data' not in result:
+			raise Exception('Screenshot failed - no data returned')
+
+		screenshot_data = base64.b64decode(result['data'])
+
+		if path:
+			Path(path).write_bytes(screenshot_data)
+
+		return screenshot_data
+
+	async def screenshot_element(
+		self,
+		selector: str,
+		path: str | None = None,
+		format: str = 'png',
+		quality: int | None = None,
+	) -> bytes:
+		"""Take a screenshot of a specific element.
+
+		Args:
+			selector: CSS selector for the element
+			path: Optional file path to save screenshot
+			format: Image format ('png', 'jpeg', 'webp')
+			quality: Quality 0-100 for JPEG format
+
+		Returns:
+			Screenshot data as bytes
+		"""
+
+		bounds = await self._get_element_bounds(selector)
+		if not bounds:
+			raise ValueError(f"Element '{selector}' not found or has no bounds")
+
+		return await self.take_screenshot(
+			path=path,
+			format=format,
+			quality=quality,
+			clip=bounds,
+		)
+
+	async def _get_element_bounds(self, selector: str) -> dict | None:
+		"""Get element bounding box using CDP."""
+
+		cdp_session = await self.get_or_create_cdp_session()
+
+		# Get document
+		doc = await cdp_session.cdp_client.send.DOM.getDocument(params={'depth': 1}, session_id=cdp_session.session_id)
+
+		# Query selector
+		node_result = await cdp_session.cdp_client.send.DOM.querySelector(
+			params={'nodeId': doc['root']['nodeId'], 'selector': selector}, session_id=cdp_session.session_id
+		)
+
+		node_id = node_result.get('nodeId')
+		if not node_id:
+			return None
+
+		# Get bounding box
+		box_result = await cdp_session.cdp_client.send.DOM.getBoxModel(
+			params={'nodeId': node_id}, session_id=cdp_session.session_id
+		)
+
+		box_model = box_result.get('model')
+		if not box_model:
+			return None
+
+		content = box_model['content']
+		return {
+			'x': min(content[0], content[2], content[4], content[6]),
+			'y': min(content[1], content[3], content[5], content[7]),
+			'width': max(content[0], content[2], content[4], content[6]) - min(content[0], content[2], content[4], content[6]),
+			'height': max(content[1], content[3], content[5], content[7]) - min(content[1], content[3], content[5], content[7]),
+		}

browser_use/browser/watchdogs/downloads_watchdog.py

@@ -16,6 +16,7 @@
 
 from browser_use.browser.events import (
 	BrowserLaunchEvent,
+	BrowserStateRequestEvent,
 	BrowserStoppedEvent,
 	FileDownloadedEvent,
 	NavigationCompleteEvent,
@@ -34,6 +35,7 @@ class DownloadsWatchdog(BaseWatchdog):
 	# Events this watchdog listens to (for documentation)
 	LISTENS_TO: ClassVar[list[type[BaseEvent[Any]]]] = [
 		BrowserLaunchEvent,
+		BrowserStateRequestEvent,
 		BrowserStoppedEvent,
 		TabCreatedEvent,
 		TabClosedEvent,
@@ -81,6 +83,26 @@ async def on_TabClosedEvent(self, event: TabClosedEvent) -> None:
 		"""Stop monitoring closed tabs."""
 		pass  # No cleanup needed, browser context handles target lifecycle
 
+	async def on_BrowserStateRequestEvent(self, event: BrowserStateRequestEvent) -> None:
+		"""Handle browser state request events."""
+		cdp_session = self.browser_session.agent_focus
+		if not cdp_session:
+			return
+
+		url = await self.browser_session.get_current_page_url()
+		if not url:
+			return
+
+		target_id = cdp_session.target_id
+		self.event_bus.dispatch(
+			NavigationCompleteEvent(
+				event_type='NavigationCompleteEvent',
+				url=url,
+				target_id=target_id,
+				event_parent_id=event.event_id,
+			)
+		)
+
 	async def on_BrowserStoppedEvent(self, event: BrowserStoppedEvent) -> None:
 		"""Clean up when browser stops."""
 		# Cancel all CDP event handler tasks

browser_use/dom/serializer/serializer.py

@@ -127,10 +127,6 @@ def _is_interactive_cached(self, node: EnhancedDOMTreeNode) -> bool:
 	def _create_simplified_tree(self, node: EnhancedDOMTreeNode, depth: int = 0) -> SimplifiedNode | None:
 		"""Step 1: Create a simplified tree with enhanced element detection."""
 
-		# Prevent infinite recursion by limiting depth to 30 levels
-		if depth > 30:
-			return None
-
 		if node.node_type == NodeType.DOCUMENT_NODE:
 			# for all cldren including shadow roots
 			for child in node.children_and_shadow_roots:

docs/customize/examples/prompting-guide.mdx renamed to docs/customize/agent/prompting-guide.mdx

@@ -74,6 +74,42 @@
     {
       "source": "/customize/examples/chain-agents",
       "destination": "/customize/examples/follow-up-tasks"
+    },
+    {
+      "source": "/customize/examples/fast-agent",
+      "destination": "/examples/templates/fast-agent"
+    },
+    {
+      "source": "/customize/examples/follow-up-tasks",
+      "destination": "/examples/templates/follow-up-tasks"
+    },
+    {
+      "source": "/customize/examples/parallel-browser",
+      "destination": "/examples/templates/parallel-browser"
+    },
+    {
+      "source": "/customize/examples/playwright-integration",
+      "destination": "/examples/templates/playwright-integration"
+    },
+    {
+      "source": "/customize/examples/sensitive-data",
+      "destination": "/examples/templates/sensitive-data"
+    },
+    {
+      "source": "/customize/examples/secure",
+      "destination": "/examples/templates/secure"
+    },
+    {
+      "source": "/customize/examples/more-examples",
+      "destination": "/examples/templates/more-examples"
+    },
+    {
+      "source": "/customize/examples/ad-use",
+      "destination": "/examples/apps/ad-use"
+    },
+    {
+      "source": "/customize/examples/vibetest-use",
+      "destination": "/examples/apps/vibetest-use"
     }
   ],
   "navigation": {
@@ -99,6 +135,7 @@
                 "pages": [
                   "customize/agent/basics",
                   "customize/agent/supported-models",
+                  "customize/agent/prompting-guide",
                   "customize/agent/output-format",
                   "customize/agent/all-parameters"
                 ]
@@ -126,20 +163,6 @@
                   "customize/tools/response"
                 ]
               },
-              {
-                "group": "Examples",
-                "icon": "folder-open",
-                "pages": [
-                  "customize/examples/fast-agent",
-                  "customize/examples/follow-up-tasks",
-                  "customize/examples/parallel-browser",
-                  "customize/examples/playwright-integration",
-                  "customize/examples/sensitive-data",
-                  "customize/examples/secure",
-                  "customize/examples/more-examples",
-                  "customize/examples/prompting-guide"
-                ]
-              },
               {
                 "group": "Integration",
                 "icon": "plug",
@@ -150,6 +173,32 @@
               }
             ]
           },
+          {
+            "group": "Examples",
+            "pages": [
+              {
+                "group": "Templates",
+                "icon": "folder",
+                "pages": [
+                  "examples/templates/fast-agent",
+                  "examples/templates/follow-up-tasks",
+                  "examples/templates/parallel-browser",
+                  "examples/templates/playwright-integration",
+                  "examples/templates/sensitive-data",
+                  "examples/templates/secure",
+                  "examples/templates/more-examples"
+                ]
+              },
+              {
+                "group": "Apps",
+                "icon": "box-open",
+                "pages": [
+                  "examples/apps/ad-use",
+                  "examples/apps/vibetest-use"
+                ]
+              }
+            ]
+          },
           {
             "group": "Development",
             "pages": [

docs/docs.json

@@ -0,0 +1,71 @@
+---
+title: "Ad-Use (Ad Generator)"
+description: "Generate Instagram ads from landing pages using browser agents and Google's Nano Banana 🍌."
+icon: "image"
+mode: "wide"
+---
+
+<Note>
+This demo requires browser-use v0.7.6+.
+</Note>
+
+<video
+  controls
+  className="w-full aspect-video rounded-xl"
+  src="https://github.com/user-attachments/assets/7fab54a9-b36b-4fba-ab98-a438f2b86b7e">
+</video>
+
+## Features
+
+1. Agent visits your target website
+2. Captures brand name, tagline, and key selling points
+3. Takes a clean screenshot for design reference
+4. Creates a scroll-stopping Instagram ad with 🍌
+
+## Setup
+
+Make sure the newest version of browser-use is installed (with screenshot functionality):
+```bash
+pip install -U browser-use
+```
+
+Export your Gemini API key, get it from: [Google AI Studio](https://makersuite.google.com/app/apikey) 
+```
+export GOOGLE_API_KEY='your-google-api-key-here'
+```
+
+## Normal Usage
+
+```bash
+# Basic - Generate ad from any website
+python ad_generator.py https://www.apple.com/iphone-16-pro/
+
+# Debug Mode - See the browser in action
+python ad_generator.py https://www.apple.com/iphone-16-pro/ --debug
+```
+
+## Programmatic Usage
+```python
+import asyncio
+from ad_generator import create_ad_from_landing_page
+
+async def main():
+    results = await create_ad_from_landing_page(
+        url="https://your-landing-page.com",
+        debug=False
+    )
+    print(f"Generated ads: {results}")
+
+asyncio.run(main())
+```
+
+## Output
+
+Generated ads are saved in the `output/` directory with:
+- **PNG image files** (ad_style_timestamp.png) - Actual generated ads from Gemini 2.5 Flash Image
+- **Prompt files** (ad_style_timestamp_prompt.txt) - The prompts used for generation  
+- **Landing page screenshots** for reference
+
+## Source Code
+
+Full implementation: [https://github.com/browser-use/browser-use/tree/main/examples/apps/ad-use](https://github.com/browser-use/browser-use/tree/main/examples/apps/ad-use)