Automate browser based workflows with AI
Python aiapiautomationbrowserbrowser-automationcomputergptllmplaywrightpowerautomatepuppeteerpythonrpaseleniumvisionworkflow
fix: block Liferay favicon redirect that strands login flow (#6083)
Aaron Perez0f9fece51e07
diff --git a/skyvern/forge/agent_functions.py b/skyvern/forge/agent_functions.pyindex 44a33fc33..7f98ea5be 100644--- a/skyvern/forge/agent_functions.py+++ b/skyvern/forge/agent_functions.py@@ -535,6 +535,9 @@ class AgentFunction:"""Return an org-scoped API key; returns None in the base implementation."""return None+ async def setup_browser_context_extensions(self, browser_context: Any, **kwargs: Any) -> None:+ """Attach cloud-only listeners/route handlers to a fresh BrowserContext. OSS no-op."""+async def validate_step_execution(self,task: Task,diff --git a/skyvern/webeye/browser_factory.py b/skyvern/webeye/browser_factory.pyindex 277f3e0e1..ac07bd687 100644--- a/skyvern/webeye/browser_factory.py+++ b/skyvern/webeye/browser_factory.py@@ -389,6 +389,7 @@ class BrowserContextFactory:set_popup_video_listener(browser_context=browser_context, browser_artifacts=browser_artifacts)set_download_file_listener(browser_context=browser_context, **kwargs)set_dialog_handler(browser_context=browser_context)+ await app.AGENT_FUNCTION.setup_browser_context_extensions(browser_context=browser_context, **kwargs)proxy_location: ProxyLocation | None = kwargs.get("proxy_location")if proxy_location is not None:diff --git a/skyvern/webeye/scraper/scraper.py b/skyvern/webeye/scraper/scraper.pyindex f68ef9900..0049975ca 100644--- a/skyvern/webeye/scraper/scraper.py+++ b/skyvern/webeye/scraper/scraper.py@@ -37,6 +37,33 @@ from skyvern.webeye.scraper.scraped_page import (from skyvern.webeye.utils.page import SkyvernFrameLOG = structlog.get_logger()+++async def build_scraping_failed_reason(browser_state: BrowserState, requested_url: str) -> str:+ """Build the user-facing ScrapingFailed reason with the requested URL plus the landed URL when they differ.++ Query strings are stripped because OAuth/SSO URLs commonly carry secrets in query params.+ """+ safe_requested = strip_query_params(requested_url)+ safe_landed: str | None = None+ try:+ page = await browser_state.get_working_page()+ if page is not None and page.url:+ safe_landed = strip_query_params(page.url)+ except asyncio.CancelledError:+ raise+ except Exception:+ LOG.debug("Could not resolve landed URL for ScrapingFailed reason", exc_info=True)++ base = (+ "Skyvern failed to load the website. "+ "The page may have navigated unexpectedly or become unresponsive during analysis."+ )+ if safe_landed and safe_landed != safe_requested and safe_landed not in {"about:blank", ""}:+ return f"{base} Requested URL: {safe_requested}. Current URL: {safe_landed}."+ return f"{base} URL: {safe_requested}."++RESERVED_ATTRIBUTES = {"accept", # for input file"alt",@@ -204,7 +231,7 @@ async def scrape_website(if isinstance(e, FailedToTakeScreenshot):raise eelse:- raise ScrapingFailed() from e+ raise ScrapingFailed(reason=await build_scraping_failed_reason(browser_state, url)) from eLOG.info("Scraping failed, will retry", max_retries=max_retries, num_retry=num_retry, url=url, wait_seconds=0.5)await asyncio.sleep(0.5)return await scrape_website(