Visual Canvas

The Canvas is the visual authoring lens of the Model Workspace's Build tab. It renders your model as a graph — one node per variable and per constraint, connected to the objective — so you can see the structure of the optimization problem at a glance and edit it by direct manipulation.

The canvas is a view of the same canonical model as the other lenses: drop a variable on the canvas and it appears in the JSON editor; ask the AI assistant to add a constraint and a node shows up on the canvas. Every change autosaves to the draft.

Working on the canvas

• Variables — add decision variables and set their type (continuous, integer, binary) and bounds.
• Constraints — add constraints and connect the variables that participate in them; edit the expression and its sense (<=, >=, =).
• Objective — set the sense (minimize/maximize) and the expression to optimize.

The model-at-a-glance rail (variables by type, constraints, density, problem class, health score) updates live as you edit, and the Analyze tab always reflects what is on the canvas.

Undo / redo

Canvas edits support undo/redo, and your working state is autosaved continuously — the explicit milestones are commits.

Large models

The canvas renders a node per model element, which stops being useful (and fast) at scale. Above 800 elements (variables + constraints combined), the workspace skips the visual canvas automatically:

• Build → Canvas shows a "model too large" notice instead of the graph
• The model remains fully editable through the JSON editor and JModel lenses
• Analyze and Solve work exactly the same

This guard exists so that importing, say, a 40,000-variable assignment model never locks your browser. For models of that scale, a declarative JModel formulation with indexed families is a much better representation than 40,000 nodes anyway.

When to use which lens