Find answers from the community

Updated last year

Metadata

ttarpus

I then leverage llama-index, and get bogus page number

Attachment

20 comments

LLogan M

That's very strange.

I would start with a fresh/new pinecone db

You can confirm the metadata before inserting by checking
document.metadata

I imagine something like this

Plain Text

pinecone_index = <create new vector index>

...
 
documents = SimpleDirectoryReader("./data").load_data()

for doc in documents: 
    print(doc.metadata)

index = VectorStoreIndex.from_documents(documents, ....)

nodes = index.as_retriever().retrieve("query")

for n in nodes:
    print(n.node.metadata)

ttarpus

Hi!!!!

ttarpus

I will try that. You guys are AWESOME

ttarpus

here is the metadata before going to the db

ttarpus

so it seem the metadata are ok before going in, once inside, and upon retrieval

ttarpus

but when I run the llama_index query the metadata are off

ttarpus

woah, oddly. I re-built/re-indexed pinecone and the issue is gone now

ttarpus

very wierd!

ttarpus

page 3. great

Attachment

ttarpus

it seems to work now!

ttarpus

ok, another question 🙂

I need citations for my knowledge graph / graph db.

Do you guys support this?

LLogan M

you can check the response object of a query engine to see details of the nodes used to create the response

response.source_nodes

ttarpus

so it is possible to use llama index for knowledge graph citations too?

this is great news.

i think so, yes! 🙏

wow

aweosme

I will try it and let you know!

TTheresa

just wanted to pop in here and say THANK YOU SO MUCH for your help so far Logan!!! we'd been trying to troubleshoot the weird citation issue for a while and never thought to start w a fresh new pinecone db so TY!!

LLogan M

Happy to help! :dotsHARDSTYLE:

Add a reply