The robot scan UI#

glitched ui

opens a page in your browser. Add robot registers one: its name, address and why you're allowed to test it, then plain-language questions (the same ones as glitched onboard) and, optionally, an SBOM; Robot details changes the answers later. The page shows each robot drawn as a robot, every open finding pinned to the part it's in, and a panel for each finding: what's wrong, the evidence, the recommended fix, and Recheck, which re-runs just that check. Scan now runs the passive scan; the first time for each robot it first shows exactly what will be sent - the address, each check's ports and requests, and what's never sent - and waits for you to start it (What's sent shows this any time; glitched device plan <name> in a terminal). While it runs, the panel goes through each check live - every port and request it tries, with the result - and ends with a summary. Robots added with glitched device add show up too.

Pictures. Each robot starts with a 3D render from the built-in library that fits its profile (wheeled robot, robot arm, quadruped, humanoid, drone or controller box), on a transparent background with its parts already placed (tools/render_library.py makes them in Blender). Change picture picks another, or uploads your own photo (PNG, JPEG or WebP, up to 8 MB). On a photo, Edit parts lets you click where each part is - network, cloud link, main controller, debug port, USB/serial - and rename them; findings for parts that aren't on the picture stay in the list. Photos are stored with your other Glitched data (~/.local/share/glitched/robot-images) and never leave your computer; location and other metadata is removed from JPEGs.

The UI runs on your computer only: it listens on 127.0.0.1, needs the one-time key in the link glitched ui prints (don't share it), and refuses requests from other websites. It uses the same database, checks and safety rules as the command line; nothing is sent anywhere else. --no-browser prints the link instead of opening it; --port picks the port.