Vectorstore with a kor library (on top of langchain)

[ahmed-bhs]

I am testing a kor library and I find it really helpful when extracting pieces of information from a low-quality text (like parsed online ads). The problem is a kor prompt usually gets long and when combined with an original text, it easily exceeds the token limit of OpenAI.

When working with pure LangChain, I use vectorstores to grab the most relevant chunks of text.

Do you know how to use vectorstores with kor? That would really help! Thanks.

[eyurtsev]

Hi @ahmed-bhs,

There's no pre-built solution. You can iterate over all documents chunks as shown here: https://eyurtsev.github.io/kor/document_extraction.html

If you want to do something more efficient (but potentially more error prone), here's a discussion that might be useful: #181