LlamaIndex
Log in

Find answers from the community

Updated last month

Instruct LLamaParse to ignore top header as headline

SSaltuk
·
Is it possible to instruct LLamaParse to not scan the top header as headline? I want to use the MarkdownParser, but it horribly fails, because of these headers in my document
Attachment
image.png
L
S
24 comments
LLogan M
the parsing_instructions param may help with that actually
SSaltuk
I tried it, doesn't seem to be consistent
SSaltuk
Probably have to do it manually then ^^ - found a tool online that whitens the headers out. Trying that out now
SSaltuk
whitening out also doesnt seem to work rip
LLogan M
I tried it, doesn't seem to be consistent -- tbh it might just be some iterating on the specific instructions to get it to work well
SSaltuk
I just found out - there is also a bounding_box parameter. Probably more useful for my use-case 🙂
SSaltuk
Thanks though!
SSaltuk
I think bounding_box is not working xD
SSaltuk
Atleast in the Web-UI it doesn't work
SSaltuk
Need to check if it works via API
LLogan M
hmmm seems like it works for premium mode, weird
LLogan M
(In the UI)
LLogan M
flagging this to the team though
SSaltuk
it only works on premium mode?
SSaltuk
not in normal mode?
SSaltuk
did you try it?
LLogan M
It might be a bug, but yea, I tried it with your screenshot
LLogan M
0.1,0,0,0 for bounding_box excluded the header only when I set premium mode 🤔
SSaltuk
It doesn't work via API either.
SSaltuk
Then i will use premium_mode for now. I'm really in need to make this work ASAP xD
SSaltuk
I can't afford premium mode
SSaltuk
xD
LLogan M
Hopefully another price drop soon! 🙏
LLogan M
Straightening out some infra costs
Add a reply
Sign up and join the conversation on Discord