Skip to main content

​
TL;DR

Get precise chunks for your retrieval and embedded workflows. Example of bounding boxes returned by the API Example: Document with detected bounding boxes highlighting different content regions

​
Endpoint

POST https://api.trycardinal.ai/split
Content-Type: multipart/form-data
Auth: X-API-KEY: <API_KEY>
You may provide either file or fileUrl.

​
Example Response

{
  "pages": [
    {
      "page": 1,
      "width": 612,
      "height": 792,
      "bounding_boxes": [
        {
          "bounding_box": {
            "min_x": 1.2558,
            "min_y": 0.9282,
            "max_x": 2.4513,
            "max_y": 1.1343
          },
          "content": "Test"
        },
        {
          "bounding_box": {
            "min_x": 2.6974,
            "min_y": 1.2862,
            "max_x": 4.9843,
            "max_y": 1.4512
          },
          "content": "Text!"
        }
      ]
    }
  ]
}
Fields
  • width, height — page size in points (1 inch = 72 points)
  • bounding_boxes[].bounding_box — box in inches (min_x, min_y, max_x, max_y)
  • bounding_boxes[].content — text captured for that region

​
Units

Bounding boxes are returned in inches, relative to the original PDF coordinate system.
  • Origin: top-left of the page
  • Format: (min_x, min_y, max_x, max_y) → inches
You’ll also see width and height (in points) for each page in the response.
💡 Recommended workflow:
  1. Convert inches → points (multiply by 72).
  2. Normalize by the PDF’s page size in points (width and height from the response).
  3. Convert to percentages for rendering highlights.
This ensures bounding boxes scale correctly across different PDF sizes (e.g., 600–800 point wide pages).

​
Conversion Example

Here’s how you can implement the recommended workflow on the frontend: 
const convertExtractedDataToHighlightAreas = (): HighlightArea[] => {
  const boundingBoxData = Array.isArray(extractedData)
    ? extractedData
    : extractedData?.pages?.[currentPage - 1]?.bounding_boxes;

  if (!boundingBoxData || boundingBoxData.length === 0) {
    console.log('No bounding box data found');
    return [];
  }

  const INCHES_TO_POINTS = 72;

  return boundingBoxData.map((box: any, index: number) => {
    // Convert from inches to points
    const xInPoints = box.bounding_box.min_x * INCHES_TO_POINTS;
    const yInPoints = box.bounding_box.min_y * INCHES_TO_POINTS;
    const widthInPoints = (box.bounding_box.max_x - box.bounding_box.min_x) * INCHES_TO_POINTS;
    const heightInPoints = (box.bounding_box.max_y - box.bounding_box.min_y) * INCHES_TO_POINTS;

    // Convert to percentages relative to PDF dimensions
    const leftPercent = (xInPoints / pdfWidth) * 100;
    const topPercent = (yInPoints / pdfHeight) * 100;
    const widthPercent = (widthInPoints / pdfWidth) * 100;
    const heightPercent = (heightInPoints / pdfHeight) * 100;

    return {
      pageIndex: currentPage - 1,
      left: Math.max(0, Math.min(100, leftPercent)),
      top: Math.max(0, Math.min(100, topPercent)),
      width: Math.max(0, Math.min(100 - leftPercent, widthPercent)),
      height: Math.max(0, Math.min(100 - topPercent, heightPercent)),
    };
  });
};

​
API Reference

RAG API

​
Coming Soon

  • Variable chunking — flexible bounding box grouping (not just paragraphs or words).