- Extract filepath from ConversionResult.input.file instead of indexing safe_file_paths
- Add content_paths field to track source filepath for each converted document
- Ensures correct filepath metadata even when some files fail conversion
- Add comprehensive test for filepath metadata with conversion failures
Addresses Cursor Bugbot comment on PR #3813
Co-Authored-By: João <joao@crewai.com>
5dc40bff9e