Python
language boundary
JavaScript
Before
Python backend content.find('music') → 10 code points
10
Stored row tracked_changes .start_offset = 10 bare int — units undocumented
10
Transform → 10 copies 10 verbatim
10 — unconverted
JS frontend content.slice(10, 15) →  musi ✗ wrong
After
Python backend content.find('music') → 10 code points
10
Stored row tracked_changes .start_offset = 10 bare int — units undocumented
10
Transform → 10 copies 10 verbatim
10
NEW Conversion codePointToCodeUnitOffset( content, 10) → 11 code units
11
JS frontend content.slice(11, 16) → music ✓ correct
the backend never changes
the conversion lives at the boundary
Hover any stage to see what it does.
The fix is one box, inserted at the language boundary. Before, the bare offset 10 crossed from Python to JavaScript unconverted and sliced one character early; after, the frontend converts 10 → 11 before slicing. Everything upstream is untouched.