genpark-screen-context

GenPark Screen Context is a zero-dependency autonomous UI agent perception engine implementing Accessibility Tree (AXTree) coordinate bounding box normalization, viewport occlusion and clipping polygon analysis, and real-time omnichannel e-commerce intent trigger detection.

flowchart TD
    A[Raw Accessibility Tree Nodes] --> B[AXTree Normalizer]
    B --> C[Normalized Center & Interactive Bounding Boxes]
    D[Candidate Elements & Viewport Bounds] --> E[Occlusion & Clipping Polygon Solver]
    E --> F[Visibility Ratio & Visual Prominence Score]
    G[DOM Screen Text Tokens] --> H[Omnichannel Intent Pattern Engine]
    H --> I[Checkout, Price Comparison & Cancellation Alerts]

Features

• AXTree Coordinate Normalizer: Maps arbitrary screen resolutions to unit $[0, 1]$ bounding box vectors and interactive element classifications.
• Viewport Occlusion Solver: Computes geometric intersection, offscreen detection, and center-weighted visual prominence scores.
• Omnichannel Intent Matcher: Detects purchase, price comparison, and churn risk triggers from DOM token streams.

Quickstart

from client import GenParkScreenContext

sc = GenParkScreenContext()
res = sc.match_omnichannel_intent_trigger(["checkout", "order", "total"])
print("Intent:", res["primary_intent"])

License

MIT