When adding documents to the knowledge base, I’m struggling with many documents having an error status.
As a workaround, I tried adding a process in the pipeline to extract only text from documents and perform chunking, but the error rate (subjectively) hasn’t changed.
Are there any good solutions for this?