Hi guys
When I am trying to upsert document with pinecone, I see this error “ErrorWithoutStackTrace: PineconeClient: Error calling upsert: ErrorWithoutStackTrace: metadata size is 210710 bytes, which exceeds the limit of 40960 bytes per vector”
How can I resolve this issue?
Thank you
Max
You’ll need to upload less metadata per record. A common reason for this error is uploading large documents. Often times the solution is to split large documents into chunks before embedding them and upsetting to pinecone. Typically a large document is not well represented by a single embedding.
Thank you, I was wondering if I can do the chunk spliting on pinecone? How do I do it?
Right now, I am using langchain to upload text into an index.
I am not even sending any metadata to the index, but I still get the error. Why could that be the case?
Hey im facing the same strange issue. I load my data, split them as usual and try put in inside the vs but immfacing every time the same issue: problem.py (github.com)
So metadata contains every time just a “source” field which will be rendered directly before try pass it to poinecone with the url in it but im still getting the error