Add a new `computer` tool that enables agents to control desktop GUIs: - Take screenshots - Click, double-click, right-click at coordinates - Type text, press keys, hotkey combinations - Scroll in any direction - Move cursor, drag operations - Get screen size and cursor position Integrates with cua-computer-server which can run in: - Sandbox mode (cua-xfce, cua-ubuntu Docker images) - Node mode (Linux with supervisor setup) Includes setup documentation and example screenshot.
81 KiB
1024x768px
81 KiB
1024x768px