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

Make SkyvernPage extend Playwright (#3934)

Browse Source
This commit is contained in:
Stanislav Novosad
2025-11-10 09:02:37 -07:00
committed by GitHub
parent 2e3879bc37
commit 23ebd72d4a
4 changed files with 88 additions and 234 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
skyvern/core/script_generations/script_skyvern_page.py
View File

@@ -123,8 +123,6 @@ class ScriptSkyvernPage(SkyvernPage):
action: ActionType,
*args: Any,
prompt: str = "",
data: str | dict[str, Any] = "",
intention: str = "",
**kwargs: Any,
) -> Any:
"""
@@ -150,9 +148,11 @@ class ScriptSkyvernPage(SkyvernPage):
}
# Backward compatibility: use intention if provided and prompt is empty
intention = kwargs.get("intention", None)
if intention and not prompt:
prompt = intention
data = kwargs.get("data", None)
meta = ActionMetadata(prompt, data)
call = ActionCall(action, args, kwargs, meta)
@@ -199,7 +199,6 @@ class ScriptSkyvernPage(SkyvernPage):
action_type=action,
intention=prompt,
status=action_status,
data=data,
kwargs=kwargs,
call_result=call.result,
)
@@ -262,7 +261,6 @@ class ScriptSkyvernPage(SkyvernPage):
action_type: ActionType,
intention: str = "",
status: ActionStatus = ActionStatus.pending,
data: str | dict[str, Any] = "",
kwargs: dict[str, Any] | None = None,
call_result: Any | None = None,
) -> Action | None:
@@ -396,7 +394,7 @@ class ScriptSkyvernPage(SkyvernPage):
# If screenshot creation fails, don't block execution
pass
async def goto(self, url: str, timeout: float = settings.BROWSER_LOADING_TIMEOUT_MS) -> None:
async def goto(self, url: str, **kwargs: Any) -> None:
url = render_template(url)
url = prepend_scheme_and_validate_url(url)
@@ -405,10 +403,8 @@ class ScriptSkyvernPage(SkyvernPage):
if context and context.script_mode:
print(f"🌐 Navigating to: {url}")
await self.page.goto(
url,
timeout=timeout,
)
timeout = kwargs.pop("timeout", settings.BROWSER_ACTION_TIMEOUT_MS)
await self.page.goto(url, timeout=timeout, **kwargs)
if context and context.script_mode:
print(" ✓ Page loaded")

172
skyvern/core/script_generations/skyvern_page.py
View File

@@ -3,7 +3,6 @@ from __future__ import annotations
import asyncio
import copy
from dataclasses import dataclass
from enum import StrEnum
from typing import Any, Callable, Literal, overload
import structlog
@@ -19,10 +18,6 @@ from skyvern.webeye.actions.action_types import ActionType
LOG = structlog.get_logger()
class Driver(StrEnum):
PLAYWRIGHT = "playwright"
@dataclass
class ActionMetadata:
prompt: str = ""
@@ -41,7 +36,7 @@ class ActionCall:
error: Exception | None = None # populated if failed
class SkyvernPage:
class SkyvernPage(Page):
"""
A lightweight adapter for the selected driver that:
1. Executes actual browser commands
@@ -54,21 +49,32 @@ class SkyvernPage:
page: Page,
ai: SkyvernPageAi,
) -> None:
super().__init__(page)
self.page = page
self.current_label: str | None = None
self._ai = ai
def __getattribute__(self, name: str) -> Any:
page = object.__getattribute__(self, "page")
if hasattr(page, name):
for cls in type(self).__mro__:
if cls is Page:
break
if name in cls.__dict__:
return object.__getattribute__(self, name)
return getattr(page, name)
return object.__getattribute__(self, name)
async def _decorate_call(
self,
fn: Callable,
action: ActionType,
*args: Any,
prompt: str = "",
data: str | dict[str, Any] = "",
intention: str = "", # backward compatibility
**kwargs: Any,
) -> Any:
return await fn(self, *args, prompt=prompt, data=data, intention=intention, **kwargs)
return await fn(self, *args, prompt=prompt, **kwargs)
@staticmethod
def action_wrap(
@@ -79,20 +85,17 @@ class SkyvernPage:
skyvern_page: SkyvernPage,
*args: Any,
prompt: str = "",
data: str | dict[str, Any] = "",
intention: str = "", # backward compatibility
**kwargs: Any,
) -> Any:
return await skyvern_page._decorate_call(
fn, action, *args, prompt=prompt, data=data, intention=intention, **kwargs
)
return await skyvern_page._decorate_call(fn, action, *args, prompt=prompt, **kwargs)
return wrapper
return decorator
async def goto(self, url: str, timeout: float = settings.BROWSER_LOADING_TIMEOUT_MS) -> None:
await self.page.goto(url, timeout=timeout)
async def goto(self, url: str, **kwargs: Any) -> None:
timeout = kwargs.pop("timeout", settings.BROWSER_ACTION_TIMEOUT_MS)
await self.page.goto(url, timeout=timeout, **kwargs)
######### Public Interfaces #########
@@ -103,9 +106,6 @@ class SkyvernPage:
*,
prompt: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str | None: ...
@@ -115,9 +115,6 @@ class SkyvernPage:
*,
prompt: str,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str | None: ...
@@ -128,9 +125,6 @@ class SkyvernPage:
*,
prompt: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str | None:
"""Click an element using a CSS selector, AI-powered prompt matching, or both.
@@ -144,8 +138,7 @@ class SkyvernPage:
selector: CSS selector for the target element.
prompt: Natural language description of which element to click.
ai: AI behavior mode. Defaults to "fallback" which tries selector first, then AI.
data: Additional context data for AI processing.
timeout: Maximum time to wait for the click action in milliseconds.
**kwargs: All Playwright click parameters (timeout, force, modifiers, etc.)
Returns:
The selector string that was successfully used to click the element, or None.
@@ -163,12 +156,16 @@ class SkyvernPage:
```
"""
# Backward compatibility
intention = kwargs.pop("intention", None)
if intention is not None and prompt is None:
prompt = intention
if not selector and not prompt:
raise ValueError("Missing input: pass a selector and/or a prompt.")
timeout = kwargs.pop("timeout", settings.BROWSER_ACTION_TIMEOUT_MS)
data = kwargs.pop("data", None)
context = skyvern_context.current()
if context and context.ai_mode_override:
ai = context.ai_mode_override
@@ -206,8 +203,8 @@ class SkyvernPage:
)
if selector:
locator = self.page.locator(selector, **kwargs)
await locator.click(timeout=timeout)
locator = self.page.locator(selector)
await locator.click(timeout=timeout, **kwargs)
return selector
@@ -219,11 +216,9 @@ class SkyvernPage:
*,
prompt: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
totp_identifier: str | None = None,
totp_url: str | None = None,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str: ...
@overload
@@ -234,11 +229,9 @@ class SkyvernPage:
value: str | None = None,
selector: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
totp_identifier: str | None = None,
totp_url: str | None = None,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str: ...
@action_wrap(ActionType.INPUT_TEXT)
@@ -249,11 +242,9 @@ class SkyvernPage:
*,
prompt: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
totp_identifier: str | None = None,
totp_url: str | None = None,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str:
"""Fill an input field using a CSS selector, AI-powered prompt matching, or both.
@@ -267,8 +258,6 @@ class SkyvernPage:
value: The text value to input into the field.
prompt: Natural language description of which field to fill and what value.
ai: AI behavior mode. Defaults to "fallback" which tries selector first, then AI.
data: Additional context data for AI processing.
timeout: Maximum time to wait for the fill action in milliseconds.
totp_identifier: TOTP identifier for time-based one-time password fields.
totp_url: URL to fetch TOTP codes from for authentication.
@@ -293,12 +282,16 @@ class SkyvernPage:
"""
# Backward compatibility
intention = kwargs.pop("intention", None)
if intention is not None and prompt is None:
prompt = intention
if not selector and not prompt:
raise ValueError("Missing input: pass a selector and/or a prompt.")
timeout = kwargs.pop("timeout", settings.BROWSER_ACTION_TIMEOUT_MS)
data = kwargs.pop("data", None)
return await self._input_text(
selector=selector,
value=value or "",
@@ -317,19 +310,21 @@ class SkyvernPage:
value: str,
ai: str | None = "fallback",
prompt: str | None = None,
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
totp_identifier: str | None = None,
totp_url: str | None = None,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str:
# Backward compatibility
intention = kwargs.pop("intention", None)
if intention is not None and prompt is None:
prompt = intention
if not selector and not prompt:
raise ValueError("Missing input: pass a selector and/or a prompt.")
timeout = kwargs.pop("timeout", settings.BROWSER_ACTION_TIMEOUT_MS)
data = kwargs.pop("data", None)
return await self._input_text(
selector=selector,
value=value,
@@ -419,8 +414,7 @@ class SkyvernPage:
*,
prompt: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
**kwargs: Any,
) -> str: ...
@overload
@@ -431,8 +425,7 @@ class SkyvernPage:
files: str | None = None,
selector: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
**kwargs: Any,
) -> str: ...
@action_wrap(ActionType.UPLOAD_FILE)
@@ -443,17 +436,19 @@ class SkyvernPage:
*,
prompt: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str:
# Backward compatibility
intention = kwargs.pop("intention", None)
if intention is not None and prompt is None:
prompt = intention
if not selector and not prompt:
raise ValueError("Missing input: pass a selector and/or a prompt.")
timeout = kwargs.pop("timeout", settings.BROWSER_ACTION_TIMEOUT_MS)
data = kwargs.pop("data", None)
context = skyvern_context.current()
if context and context.ai_mode_override:
ai = context.ai_mode_override
@@ -466,7 +461,7 @@ class SkyvernPage:
try:
file_path = await download_file(files)
locator = self.page.locator(selector)
await locator.set_input_files(file_path)
await locator.set_input_files(file_path, **kwargs)
except Exception as e:
error_to_raise = e
selector = None
@@ -501,7 +496,7 @@ class SkyvernPage:
file_path = await download_file(files)
locator = self.page.locator(selector)
await locator.set_input_files(file_path, timeout=timeout)
await locator.set_input_files(file_path, timeout=timeout, **kwargs)
return files
@overload
@@ -512,9 +507,6 @@ class SkyvernPage:
*,
prompt: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str | None: ...
@@ -526,9 +518,6 @@ class SkyvernPage:
value: str | None = None,
selector: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str | None: ...
@@ -540,9 +529,6 @@ class SkyvernPage:
*,
prompt: str | None = None,
ai: str | None = "fallback",
data: str | dict[str, Any] | None = None,
timeout: float = settings.BROWSER_ACTION_TIMEOUT_MS,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> str | None:
"""Select an option from a dropdown using a CSS selector, AI-powered prompt matching, or both.
@@ -557,8 +543,6 @@ class SkyvernPage:
value: The option value to select.
prompt: Natural language description of which option to select.
ai: AI behavior mode. Defaults to "fallback" which tries selector first, then AI.
data: Additional context data for AI processing.
timeout: Maximum time to wait for the select action in milliseconds.
Returns:
The value that was successfully selected.
@@ -581,12 +565,16 @@ class SkyvernPage:
"""
# Backward compatibility
intention = kwargs.pop("intention", None)
if intention is not None and prompt is None:
prompt = intention
if not selector and not prompt:
raise ValueError("Missing input: pass a selector and/or a prompt.")
timeout = kwargs.pop("timeout", settings.BROWSER_ACTION_TIMEOUT_MS)
data = kwargs.pop("data", None)
context = skyvern_context.current()
if context and context.ai_mode_override:
ai = context.ai_mode_override
@@ -631,46 +619,30 @@ class SkyvernPage:
async def wait(
self,
seconds: float,
prompt: str | None = None,
data: str | dict[str, Any] | None = None,
intention: str | None = None,
**kwargs: Any,
) -> None:
await asyncio.sleep(seconds)
@action_wrap(ActionType.NULL_ACTION)
async def null_action(
self, prompt: str | None = None, data: str | dict[str, Any] | None = None, intention: str | None = None
) -> None:
async def null_action(self, **kwargs: Any) -> None:
return
@action_wrap(ActionType.SOLVE_CAPTCHA)
async def solve_captcha(
self, prompt: str | None = None, data: str | dict[str, Any] | None = None, intention: str | None = None
) -> None:
async def solve_captcha(self, prompt: str | None = None) -> None:
raise NotImplementedError("Solve captcha is not supported outside server context")
@action_wrap(ActionType.TERMINATE)
async def terminate(
self,
errors: list[str],
prompt: str | None = None,
data: str | dict[str, Any] | None = None,
intention: str | None = None,
) -> None:
async def terminate(self, errors: list[str], **kwargs: Any) -> None:
# TODO: update the workflow run status to terminated
return
@action_wrap(ActionType.COMPLETE)
async def complete(
self, prompt: str | None = None, data: str | dict[str, Any] | None = None, intention: str | None = None
) -> None:
async def complete(self, prompt: str | None = None) -> None:
"""Stub for complete. Override in subclasses for specific behavior."""
@action_wrap(ActionType.RELOAD_PAGE)
async def reload_page(
self, prompt: str | None = None, data: str | dict[str, Any] | None = None, intention: str | None = None
) -> None:
await self.page.reload()
async def reload_page(self, **kwargs: Any) -> None:
await self.page.reload(**kwargs)
return
@action_wrap(ActionType.EXTRACT)
@@ -680,7 +652,7 @@ class SkyvernPage:
schema: dict[str, Any] | list | str | None = None,
error_code_mapping: dict[str, str] | None = None,
intention: str | None = None,
data: str | dict[str, Any] | None = None,
**kwargs: Any,
) -> dict[str, Any] | list | str | None:
"""Extract structured data from the page using AI.
@@ -689,7 +661,6 @@ class SkyvernPage:
schema: JSON Schema defining the structure of data to extract.
error_code_mapping: Mapping of error codes to custom error messages.
intention: Additional context about the extraction intent.
data: Additional context data for AI processing.
Returns:
Extracted data matching the provided schema, or None if extraction fails.
@@ -711,12 +682,11 @@ class SkyvernPage:
# Returns: {"name": "...", "price": 29.99}
```
"""
data = kwargs.pop("data", None)
return await self._ai.ai_extract(prompt, schema, error_code_mapping, intention, data)
@action_wrap(ActionType.VERIFICATION_CODE)
async def verification_code(
self, prompt: str | None = None, data: str | dict[str, Any] | None = None, intention: str | None = None
) -> None:
async def verification_code(self, prompt: str | None = None) -> None:
return
@action_wrap(ActionType.SCROLL)
@@ -724,9 +694,7 @@ class SkyvernPage:
self,
scroll_x: int,
scroll_y: int,
prompt: str | None = None,
data: str | dict[str, Any] | None = None,
intention: str | None = None,
**kwargs: Any,
) -> None:
await self.page.evaluate(f"window.scrollBy({scroll_x}, {scroll_y})")
@@ -736,9 +704,7 @@ class SkyvernPage:
keys: list[str],
hold: bool = False,
duration: float = 0,
prompt: str | None = None,
data: str | dict[str, Any] | None = None,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> None:
await handler_utils.keypress(self.page, keys, hold=hold, duration=duration)
@@ -747,9 +713,7 @@ class SkyvernPage:
self,
x: int,
y: int,
prompt: str | None = None,
data: str | dict[str, Any] | None = None,
intention: str | None = None,
**kwargs: Any,
) -> None:
await self.page.mouse.move(x, y)
@@ -759,9 +723,7 @@ class SkyvernPage:
start_x: int,
start_y: int,
path: list[tuple[int, int]],
prompt: str | None = None,
data: str | dict[str, Any] | None = None,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> None:
await handler_utils.drag(self.page, start_x, start_y, path)
@@ -771,9 +733,7 @@ class SkyvernPage:
x: int,
y: int,
direction: Literal["down", "up"],
prompt: str | None = None,
data: str | dict[str, Any] | None = None,
intention: str | None = None, # backward compatibility
**kwargs: Any,
) -> None:
await handler_utils.left_mouse(self.page, x, y, direction)