The Canvas

The Canvas is the project board: an infinite surface where each open piece of work becomes a card. It is the first screen you see when you open a new project, and it already comes with a conversation card ready — starting a project in Elyra is a conversation, not configuration.

A card is not a disposable window. Its content (a conversation, a terminal, a note, a file) lives in the app’s main process. Changing the layout, minimizing or maximizing a card does not interrupt what was running.

What exists on the board

Card	What it is
Conversation	A conversation with an assistant. The project’s main conversation card is the maestro.
Terminal	A full terminal inside a card, with a plain shell or already with an assistant.
Note	Markdown text stored in a .md file, used as memory and as a message between you and the agents. See Notes on the Canvas.
Note group	A binder of notes with pages, so dozens of cards do not spread across the board.
File	A real project file opened as a card, in the same editor as the Files screen.
Browser	A page opened inside the board. It is not a tab of the Browser screen.
Phone	A phone card (emulator). It only appears when the emulator is available in the build; otherwise the item says it is not available in this version yet.

The File card has no button of its own in the bar: it is born when you open a real file (from the tree, from search, or from a link). You create the other kinds — see Creating cards.

Every workspace has its own Canvas

The Canvas is per workspace, not per project. Two workspaces of the same project keep separate boards: what you arrange in one does not appear in, or move, the other. That is what lets you keep a parallel work board without mixing the two.

As a practical consequence, every command-line command that touches the Canvas needs to know which workspace it is aiming at. Running from inside a card, the target is that card’s workspace; outside one, you point at the workspace explicitly.

The empty board

With no card on the board, the Canvas shows the invitation “Nothing on the board yet” and explains the two paths: the bar at the top to open a conversation, a terminal, or a note, and right-clicking the background to create a card at the point you indicate.

That invitation deliberately has no button of its own — the bar is what creates cards, and a second path there would duplicate the assistant-detection logic. The background is still a work area: right-clicking and dragging to pan work normally over it.

The background

The Canvas background is a static star field, painted once, with no parallax: the stars stay pinned to the screen and are not recalculated when you drag the board. It is surface decoration — it represents nothing about your work and does not change behavior with the board’s content.

Seeing and finding cards

To know what exists on the board without scanning the screen at a zoomed-out level, use the Layers panel: it lists the workspace’s cards with text search and a type filter, and moves the camera to the chosen card.

You can also list the cards from the command line — useful for an agent that needs to know what exists on the board before acting.

Limits worth knowing now

• The app must be open. Several Canvas operations from the command line are refused when there is no application window open; the target is not guessed.
• The workspace’s Canvas must have been opened on screen at least once. Without that, command-line commands fail instead of inventing a board that does not exist.
• The separation between screens is not complete yet. The product decision is that Canvas, Conversation, Terminal and Browser are independent places, but the central area has not been fully separated — do not count on “each screen showing only its own content” as guaranteed behavior.