pixelplace-web.vercel.app

describe_canvas

Get a coarse TEXT MAP of what is currently drawn — a small character grid where each character is the dominant color of a region and "." is transparent, plus a legend, the painted bounding box, and coverage. Tool results cannot contain images, so this is how you check that a shape landed where you meant, that the subject is centred, or that you did not accidentally fill the background. It shows layout and silhouette, not detail — for anything about how the art LOOKS, ask the person.

Page: /

Input schema

{
  "type": "object",
  "properties": {
    "frame": {
      "type": "integer",
      "minimum": 0,
      "description": "Frame to describe. Defaults to the visible one."
    },
    "detail": {
      "type": "integer",
      "minimum": 8,
      "maximum": 48,
      "description": "Maximum grid size on the long edge. The map preserves aspect ratio and reports its actual mapWidth/mapHeight. Default 24. Higher costs more tokens."
    }
  }
}