Skip to content

Open-source Python library for AI browser agents, plus Browser Use Cloud: stealth browsers and a hosted agent API.

Fix extract content

Magnus Müller39ff164008f0
diff --git a/browser_use/controller/service.py b/browser_use/controller/service.py
index 0eea6301a..cefe839d4 100644
--- a/browser_use/controller/service.py
+++ b/browser_use/controller/service.py
@@ -364,9 +364,16 @@ class Controller(Generic[Context]):
import markdownify
if extract_links:
- content = markdownify.markdownify(page_html, heading_style='ATX', bullets='-')
+ content = markdownify.markdownify(page_html, heading_style='ATX', bullets='-', strip=['style', 'script'])
else:
- content = markdownify.markdownify(page_html, heading_style='ATX', bullets='-', strip=['a'])
+ content = markdownify.markdownify(page_html, heading_style='ATX', bullets='-', strip=['a', 'style', 'script'])
+
+ # Remove weird positioning artifacts
+ import re
+
+ content = re.sub(r'❓\s*\[\d+\]\s*\w+.*?Position:.*?Size:.*?\n?', '', content, flags=re.MULTILINE | re.DOTALL)
+ content = re.sub(r'Primary: UNKNOWN\n\nNo specific evidence found', '', content, flags=re.MULTILINE | re.DOTALL)
+ content = re.sub(r'UNKNOWN CONFIDENCE', '', content, flags=re.MULTILINE | re.DOTALL)
except Exception as e:
raise RuntimeError(f'Could not convert html to markdown: {type(e).__name__}')