[Question]: Create structured output in RAG Flow

[Firos333]

Hi All,

I have been working on the platform for the last two weeks and am impressed with its features. I successfully connected the locally hosted DeepSeek R1 model using OLLMA. I would like to know how to achieve the following functionalities with a locally hosted LLM:

When uploading multiple PDFs, such as invoice documents, I must extract the invoice number and date from each PDF. However, the current upload to the knowledge base combines outputs from all PDFs. How can I modify the process to extract and manage data from each PDF individually? The output is sometimes vague. It would be more helpful to provide it in a structured format like Excel or JSON. How can this be implemented?

[KevinHuSh]

Could you tell us your story, elaberate on your scenario?