AI-powered desktop automation that reads messages and interacts with desktop applications using LLM without direct web APIs.
Challenge 1
Hardcoded coordinate dependencies for UI interaction causing brittleness during resolution changes.
Solution
Transitioning to adaptive image recognition (OpenCV/PyAutoGUI) to identify UI elements dynamically, improving reliability across environments.
Challenge 2
Maintaining human-like interaction speed to avoid detection as a bot.
Solution
Implemented randomized execution delays and human-like typing simulation to vary response cadence.
Challenge 3
Data extraction from non-API desktop applications.
Solution
Utilized OCR (Tesseract) and clipboard parsing techniques to read text directly from the application interface.
Desktop automation, LLM prompt engineering, advanced GUI scripting