unstructured-pdf-generation

To generate synthetic PDFs for RAG evaluation, provide a content description to the generation tool, which creates the PDF documents alongside companion JSON files containing evaluation questions and uploads them to Unity Catalog Volumes.

Synthetic data generation for unstructured PDFs uses an LLM to create realistic document content based on your detailed descriptions, producing both the PDF files and JSON evaluation guidelines for testing unstructured data pipelines.

Yes, you can store generated PDF documents in Unity Catalog Volumes by specifying your target catalog and schema during generation, which automatically uploads the synthetic PDFs and JSON evaluation files to the specified location.

The best way to create test PDFs for a RAG pipeline is to generate synthetic documents with companion JSON evaluation files, allowing you to test retrieval accuracy using pre-defined questions and guidelines stored directly in Unity Catalog.

You do not need to manually configure document size as a fixed value, because the synthetic PDF generation process supports configurable document sizes based on your provided descriptions and use case requirements.