Skip to content

Guides

Task-focused guides for working with SEED once it is installed.

Sections

  • Plan: turn free text, example data, a formal schema, real documents, or an ERD into one canonical InferredSchema that drives either output modality.
  • Structured Data: generate tabular and relational data as CSV, Parquet, Excel, or JSON from a schema, and evaluate it.
  • Creating a Document Type: define a new document type with a JSON schema and optional steering docs, then generate your first document from it.
  • Schema from Documents: reverse-engineer a schema from real example documents, or split a concatenated packet into one schema per document type.
  • Single Document: generate one document — a rendered PDF plus its paired ground-truth JSON label.
  • Batch Generation: run the pipeline many times in parallel from a single diversity brief.
  • Packets: generate coordinated sets of inter-related documents that share context and merge into a single multi-document PDF.
  • Generation Choices: control per-document variation with x-probability optional fields and table presentation variations.