vodorod/Dorod-Sky
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Add SCROLL action to extract-action prompt (#SKY-7924) (#4743)

Browse Source
This commit is contained in:
pedrohsdb
2026-02-13 09:42:49 -08:00
committed by GitHub
parent 0243ae1375
commit 53a89c3e44
7 changed files with 109 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
skyvern/webeye/actions/handler.py
View File

@@ -2185,9 +2185,40 @@ async def handle_scroll_action(
task: Task,
step: Step,
) -> list[ActionResult]:
if action.x and action.y:
if action.element_id:
# Element-based scrolling from extract-action prompt. Uses
# scrollNearestScrollableContainer() from domUtils.js which walks the DOM to find
# the nearest scrollable ancestor or sibling container relative to the element.
scroll_direction = "down" if action.scroll_y >= 0 else "up"
scrolled = False
dom = DomUtil(scraped_page=scraped_page, page=page)
skyvern_element = await dom.safe_get_skyvern_element_by_id(action.element_id)
if skyvern_element:
try:
scrolled = await skyvern_element.locator.evaluate(
"(el, direction) => scrollNearestScrollableContainer(el, direction)",
scroll_direction,
)
except Exception:
LOG.warning(
"JavaScript scroll evaluation failed, falling back to mouse wheel",
element_id=action.element_id,
exc_info=True,
)
else:
LOG.warning("Could not resolve element for scroll action", element_id=action.element_id)
if not scrolled:
LOG.warning(
"Could not find scrollable container near element, falling back to mouse wheel",
element_id=action.element_id,
)
await page.mouse.wheel(action.scroll_x, action.scroll_y)
elif action.x and action.y:
# Coordinate-based scrolling from CUA/UI-TARS agents
await page.mouse.move(action.x, action.y)
await page.mouse.wheel(action.scroll_x, action.scroll_y)
await page.mouse.wheel(action.scroll_x, action.scroll_y)
else:
await page.mouse.wheel(action.scroll_x, action.scroll_y)
return [ActionSuccess()]