Skip to main content

​
Basic Usage

from askui import VisionAgent

with VisionAgent() as agent:
    agent.act("""
        Open Firefox
        Navigate to amazon.com
        Search for "mechanical keyboard"
        Click first result
        Add to cart
    """)

​
Writing Effective Instructions

# Good - visual description
agent.act('Click the orange "Add to Cart" button')

# Good - positional context
agent.act('Click the search icon in the top right')

# Bad - technical terms agent can't see
agent.act('Click the element with id="add-to-cart"')

​
Patterns

​
Form Filling

agent.act("""
    Fill registration form:
    - First Name: Jane
    - Email: jane@example.com
    - Password: SecurePass123!
    Check "I agree to terms"
    Click "Create Account"
""")

​
Data Extraction

result = agent.get("""
    From the order confirmation, extract:
    - Order number
    - Total amount
    - Delivery date
""")

​
Conditional

agent.act("""
    If there's a notification about "pending approval":
        Click it and approve
    Otherwise:
        Close the panel
""")