test/document-parsing.test.js

import test from 'node:test';
import assert from 'node:assert/strict';
import { parseDocumentRegions } from '../src/document-parsing.js';

test('parses document regions across multiple page images', () => {
  const text = `image_index=0 title [58, 78, 677, 100]
Stellar engines and Dyson bubbles can be stable

image_index=0 text [58, 116, 235, 131]
Colin R. McInnes*

image_index=1 equation [100, 200, 800, 300]
E = mc^2`;
  assert.deepEqual(parseDocumentRegions(text, 2), [
    { imageId: 0, label: 'title', type: 'box', coordinates: [58, 78, 677, 100], content: 'Stellar engines and Dyson bubbles can be stable' },
    { imageId: 0, label: 'text', type: 'box', coordinates: [58, 116, 235, 131], content: 'Colin R. McInnes*' },
    { imageId: 1, label: 'equation', type: 'box', coordinates: [100, 200, 800, 300], content: 'E = mc^2' },
  ]);
});

test('preserves multiline region content', () => {
  const [region] = parseDocumentRegions('image_index=0 code [0, 0, 500, 500]\nconst x = 1;\nreturn x;', 1);
  assert.equal(region.content, 'const x = 1;\nreturn x;');
});

test('accepts harmless header whitespace', () => {
  const parsed = parseDocumentRegions(`image_index = 0   title [ 58,78, 677 , 100 ]
Liquid Foundation Model`, 1);
  assert.equal(parsed.length, 1);
  assert.deepEqual(parsed[0].coordinates, [58, 78, 677, 100]);
});

test('accepts model-chosen labels and zero-to-one coordinates', () => {
  const parsed = parseDocumentRegions(`image_index=0 Quoted label / callout [0.318, 0.045, 0.707, 0.067]
The second image shows the cows in their natural habitat.

image_index=0 image [202, 78, 797, 486]
Diagram of Liquid Foundation Model architecture.`, 1);
  assert.deepEqual(parsed, [
    { imageId: 0, label: 'Quoted label / callout', type: 'box', coordinates: [318, 45, 707, 67], content: 'The second image shows the cows in their natural habitat.' },
    { imageId: 0, label: 'image', type: 'box', coordinates: [202, 78, 797, 486], content: 'Diagram of Liquid Foundation Model architecture.' },
  ]);
});

test('keeps complete regions when the final generated region is truncated', () => {
  const parsed = parseDocumentRegions(`image_index=0 title [58, 78, 677, 100]
Liquid Foundation Model

image_index=0 text [58, 116, 530, 155]
Model architecture description

image_index=0 table [`, 1);
  assert.equal(parsed.length, 2);
  assert.equal(parsed[1].content, 'Model architecture description');
});

test('rejects malformed document layouts', () => {
  assert.equal(parseDocumentRegions('ordinary answer', 1), null);
  assert.equal(parseDocumentRegions('image_index=1 text [0, 0, 1, 1]\ncontent', 1), null);
  assert.equal(parseDocumentRegions('image_index=0 text [10, 10, 5, 20]\ncontent', 1), null);
  assert.equal(parseDocumentRegions('image_index=0 text [0, nope, 1, 1]\ncontent', 1), null);
  assert.equal(parseDocumentRegions('image_index=0 text [0, 0, 1, 1]', 1), null);
});