LlamaParse strange document segmentation

This is more of a LlamaParse quirk than a problem with the central learning points in the Event-Driven Agentic Document course, but if you look closely into the slicing LlamaParse does of the resumé you see this:


Notice the two lines in the red boxes: LlamaParse moved them to be bullets of the next section (when they were really the last bullets of the previous one), presumably just because of a page break.

This probably means that the vector indexes produced will not be ideal, right?

Is this a bug in LlamaParse? Or a limitation of the layout of the specific resumé PDF being used? Any way to work around it?

1 Like