dox docs

Keyboard

dox is built for the keyboard first. The canvas is a 2D space, so most navigation is a camera move rather than a scroll.

Canvas

Input Action
Wheel Zoom, one step at a time, anchored at the cursor. Steps run from 50% to 200%.
Middle-drag Pan the canvas. Works over windows and inside app iframes.
Right-click on empty space App menu — the list of everything installed.
Ctrl / Cmd + . Reset the camera to the origin.

Windows

Input Action
Left-drag the 4px edge Move the window.
Left-drag a corner Resize.
Right-click the top edge Window menu: pin, focus, close.
Shift + click Add a window to the selection.
+ or = Focus the selection, one by one.
p Pin the selection — keep it on top and out of the way.
Backspace Close the selection.
Esc Close any open menu.

Clipboard and layout

These operate on the window selection, not on text.

Input Action
Ctrl / Cmd + C Copy the selected windows, including the image inside an image window.
Ctrl / Cmd + V Paste them, offset slightly.
Ctrl / Cmd + A Select every window.
Ctrl / Cmd + B Box-pack: arrange every window into a tidy shelf layout around the view centre.
Ctrl / Cmd + S Save the current layout as a .dox scene file.
Ctrl / Cmd + O Load a .dox scene.

A scene file is the camera position plus one entry per window: app, position, size, stacking order and the parameters it was opened with. It is a plain file, so a layout can be saved, shared and opened like any other document.

Inside apps

Where Input Action
Explorer Double-click Open a file, a folder, or a session.
Explorer Ctrl / Cmd + C, X, V Copy, cut, paste files. Copy also writes the paths to the system clipboard.
Explorer Drag a row onto the canvas Images become image windows, .sss files open the agent, folders open a new explorer.
Explorer Drag from empty space Marquee-select.
Editor — Autosaves on a short delay; there is no save button.
Agent Enter Send.
Agent Ctrl / Cmd + C Interrupt the running turn, including a bash call in progress.
Agent /stop Same interrupt, for when there is no keyboard (mobile).
Agent / <command> Run a shell command directly, without asking the model.

The bottom bar

Element What it does
Right side Window count, front-end memory, back-end memory, camera position, zoom.