@link-assistant/web-capture 1.10.7 → 1.10.8

package/CHANGELOG.md

@@ -1,5 +1,14 @@
 # @link-assistant/web-capture
 
+## 1.10.8
+
+### Patch Changes
+
+- 8cee5e4: Keep JS/Rust kreuzberg parity for the html-to-markdown 3.6 image dimensions
+  change (issue #137). Export `normalizeStructuredKeys` and add regression tests
+  asserting inline image `dimensions` are exposed as `{ width, height }`, mirroring
+  the Rust `inline_image_to_json` fix.
+
 ## 1.10.7
 
 ### Patch Changes

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@link-assistant/web-capture",
-  "version": "1.10.7",
+  "version": "1.10.8",
   "description": "CLI and microservice to render web pages as HTML, Markdown, or PNG",
   "main": "src/index.js",
   "bin": {

package/src/kreuzberg.js

@@ -39,7 +39,19 @@ function toSnakeCase(key) {
   return key.replace(/[A-Z]/g, (match) => `_${match.toLowerCase()}`);
 }
 
-function normalizeStructuredKeys(value) {
+/**
+ * Recursively normalize object keys to snake_case.
+ *
+ * Used to give the structured `metadata`, `tables`, and `images` payloads a
+ * stable shape regardless of the casing the native binding emits. In
+ * particular, inline image `dimensions` are exposed as `{ width, height }` —
+ * mirroring the Rust `inline_image_to_json` mapping after the
+ * html-to-markdown 3.6 `ImageDimensions` change (see issue #137).
+ *
+ * @param {*} value - Value to normalize
+ * @returns {*} The value with all object keys converted to snake_case
+ */
+export function normalizeStructuredKeys(value) {
   if (Array.isArray(value)) {
     return value.map(normalizeStructuredKeys);
   }

package/tests/unit/kreuzberg.test.js

@@ -1,6 +1,7 @@
 import {
   convertWithKreuzberg,
   isKreuzbergAvailable,
+  normalizeStructuredKeys,
 } from '../../src/kreuzberg.js';
 
 let available;
@@ -207,3 +208,42 @@ describe('kreuzberg html-to-markdown integration', () => {
     expect(result.content).toContain('Item 3');
   });
 });
+
+describe('kreuzberg structured key normalization (issue #137 parity)', () => {
+  // Mirrors the Rust `inline_image_to_json` regression tests: after the
+  // html-to-markdown 3.6 change, inline image dimensions are a structured
+  // `{ width, height }` value rather than a tuple. The normalization pipeline
+  // that backs `result.images` must preserve those `width`/`height` keys.
+  it('preserves width/height keys on inline image dimensions', () => {
+    const images = normalizeStructuredKeys([
+      {
+        format: 'png',
+        filename: 'pixel.png',
+        description: 'one pixel',
+        dimensions: { width: 800, height: 600 },
+        source: 'ImgDataUri',
+      },
+    ]);
+
+    expect(images).toHaveLength(1);
+    expect(images[0].dimensions).toEqual({ width: 800, height: 600 });
+    expect(images[0].format).toBe('png');
+    expect(images[0].filename).toBe('pixel.png');
+  });
+
+  it('leaves images without dimensions untouched', () => {
+    const images = normalizeStructuredKeys([
+      { format: 'svg', source: 'SvgElement', dimensions: null },
+    ]);
+
+    expect(images[0].dimensions).toBeNull();
+  });
+
+  it('snake_cases nested structured keys', () => {
+    const normalized = normalizeStructuredKeys({
+      openGraph: { siteName: 'Example' },
+    });
+
+    expect(normalized).toEqual({ open_graph: { site_name: 'Example' } });
+  });
+});