unstructured-pdf-generation

You generate synthetic PDFs for RAG testing by using the generate_pdf_documents tool with a content description, desired count, and catalog/schema configuration. The Skill outputs realistic PDF documents and accompanying JSON metadata to your specified location.

Synthetic PDF generation is used to create realistic test datasets and demo materials for retrieval-augmented generation pipelines without needing real documents. It enables robust evaluation workflows for document-backed QA tasks by providing controlled, customizable content.

Yes, you can upload generated PDF documents directly to Unity Catalog. The Skill supports automatic uploads to Unity Catalog volumes, allowing you to specify catalog, schema, volume, and folder parameters for organized storage of your synthetic data.

Yes, the PDF generation tool creates accompanying JSON metadata files for each generated document. These JSON files contain questions and evaluation guidelines specifically designed for testing retrieval systems and evaluating RAG pipeline performance.

The best way to create test datasets for unstructured data use cases is using LLM-generated PDFs with customizable content. This approach provides realistic documents paired with JSON metadata, enabling comprehensive testing of document processing and retrieval workflows.

You customize the storage location for generated PDF documents by specifying the volume and folder parameters in the generate_pdf_documents tool. You can also use the overwrite_folder parameter to control whether existing files in the target directory are replaced.