Clicking a node · Semantix

Clicking is the main verb in the 3D view. One click flies the camera in, opens a card of information about the thing you clicked, and lights the sky to show what that thing is connected to.

What it touches#

When a node is focused, it lights up together with its wires and its direct neighbours, and the rest of the scene sinks into the dark. One hop, one light: the things it imports, the things that import it, and the folder it lives in, all at once.

That's the "what does this touch" instrument. Click a file you're about to refactor and read what stayed bright.

The lighting recomputes as the scene changes, so expanding a folder while a node is focused brings its new neighbours up with it.

The card#

A glass card attaches itself beside the node, carrying what Semantix knows about it.

The header is the file's icon and name, with what it is as a caption underneath — its language, directory for a folder, or the extension for a file whose language Semantix doesn't recognise: toml, lock. Below that, three sections:

symbols — how many functions, classes and types are in it.

actions — the three most recent things an agent did to this file: wrote · 12s ago, edited · 3m ago, read · 1h ago. If nothing has touched it, it reads idle.

history — a longer list, up to eight. It only appears once the file has actually been written or edited.

Note

These two sections are an agent's log, not the file's. They record what agents did during this session — your own edits in the code editor, changes made outside Semantix, and git history never appear here. A file you rewrote by hand five minutes ago still reads idle.

Tip

The history rows are clickable. Click one and the chat jumps to that exact change — the message and the diff where it happened.

That's the shortest path from "why does this file look like this" to the conversation that made it that way. You're standing in the project, you see a file that surprises you, and two clicks later you're reading the reasoning.

While an agent is working, the card shows only the live code section — what's being written or read — and the three sections come back when the agent is idle. The card's text can be selected and copied.

The three buttons#

The card header carries a small pill of icon buttons.

Expand in place (a +, folders only) — unfolds the folder's contents into the scene you're already in, and flies to them. Once open, the same button reads − (Collapse).

Move to (folders only) — opens the folder as a graph root of its own, in a new tab, or switches to the tab already showing it.

Drill (a </> glyph) — goes into the thing:

You drill	You get
A folder	That folder as its own 3D tab
A file	That file open in the code editor, in a new tab

How the card goes away#

It isn't a window — there's no title bar, and you can't drag or resize it; the × in its corner is the one control. It's attached to the node. Four things dismiss it:

Click empty space.
Press Escape. (If a strike was still running, the first Escape cancels that and the second closes the card.)
Fly far enough away. The card fades as you retreat and lets go entirely once you're well clear of it. Moving the camera nearby doesn't dismiss it.
Turn away — it hides when the node it's anchored to leaves the screen.

Only one at a time#

Focusing is single-selection: one card, on the node you last clicked. The exception is an agent at work, which opens its own cards on the files it's touching — up to three at once. That's Watching the agent work.