website·Indexed 31 days ago
https://ascript.cnDocumentation for AScript on Android, featuring screen image capture, OCR, and AI visual recognition via the ascript_ai module. Includes guides on object detection, value extraction using ai.ask_value, action simulation, and integrating YOLO cloud-trained models.
node module in AScript for Android allows developers to retrieve information about screen elements or perform actions on specific UI controls using a control selector.from ascript.android import nodeAScript provides a suite of cloud-based services to support developers, including:
hint parameter to provide additional context about what to look for without changing the output format.ai.find("确定按钮", hint="在底部弹窗里，不是顶部导航栏那个")
ai.ask_value("图中算式的结果", value_type=int, hint="只算红框里那一道")ai.exists() for a boolean check of an element's presence. Use ai.ask_value(..., value_type=bool) for semantic questions about the page state. Use ai.ask() for open-ended descriptions (suitable for logging, not for conditional logic).# Check if element exists
if ai.exists("登录按钮"):
ai.click("登录按钮")
# Semantic page check
if ai.ask_value("当前是不是支付成功页", value_type=bool):
print("Order complete")
# General description
print(ai.ask("当前是什么页面，用户在做什么"))hint= parameter in AI methods to provide additional context about what to look for without changing the output format. This helps the model distinguish between similar elements.ai.find("Confirm Button", hint="In the bottom popup, not the top navigation bar")
ai.ask_value("Result of equation", value_type=int, hint="Only calculate the one in the red box")For devices requiring higher privileges than standard accessibility:
adb shell commands without a USB cable).To initialize and bind an HID device:
wait()) accept an image= parameter. To save overhead, capture the screen once and reuse the image object for multiple queries. Supported image types include None (auto-capture), PIL.Image, ndarray (OpenCV BGR), or a string file path.from ascript.ios import screen
from ascript.ios.vlm import ascript_ai as ai
# Capture once, reuse multiple times to save cost/time
shot = screen.capture()
price = ai.ask_value("Total Price", value_type=float, image=shot)
count = ai.ask_value("Item Count", value_type=int, image=shot)AScript provides a visual tool (YOLO Tools) to train YOLOv11, v8, and v5 models without needing a manual Python environment. The process involves four main stages:
Ocr class provides high-level methods that handle automatic screen capturing and use the current default engine. All these methods support an optional image parameter (Bitmap or cv2 ndarray) to avoid repeated screen captures, and a rect parameter [x, y, x1, y1] to limit the search area.from ascript.android.screen import Ocr
# 1. Find the first occurrence of text
# Returns dict with text, rect, center_x, center_y, confidence or None
result = Ocr.find("Login")
# 2. Find all occurrences (pass text=None to get all text on screen)
# Returns list[dict] or None
all_text = Ocr.find_all()
# 3. Check if text exists
# Returns True/False
if Ocr.exists("Login"):
print("Button exists")
# 4. Find and click text
# timeout: 0 = try once, >0 = retry until timeout
Ocr.click("Confirm", timeout=5)
# 5. Wait for text to appear
# Returns dict or None
result = Ocr.wait("Loading Complete", timeout=15)
# 6. Wait for text to appear and then click it
Ocr.wait_click("Start", timeout=10)libusb driver via Homebrew.brew install libusb