---

name: document-outputs description: "Export DoclingDocument content to Markdown, JSON, YAML, HTML, text, DocTags, WebVTT, table files, images, and RAG chunks." disable-model-invocation: true metadata: disco-role: operating license: MIT

Document Outputs

Use this sub-skill after a ConversionResult or DoclingDocument already exists and the task is about exporting, serializing, extracting tables/pictures, or chunking content for downstream use.

Do not use this sub-skill to obtain a conversion result, tune extraction pipelines, or call remote service chunking APIs; route those to the conversion, pipeline configuration, advanced pipeline, or remote service sub-skills.

Fast Routes

• For document file/string exports, use references/export-and-serialization.md.
• For table and picture post-processing, use references/export-and-serialization.md and the bundled scripts/export_tables_from_doc.py helper.
• For RAG chunking with HybridChunker, HierarchicalChunker, or line-based token chunking, use references/chunking.md.
• For JSON validation, referenced images, empty tables/pictures, tokenizer extras, oversized chunks, and repeated table headers, use references/troubleshooting.md.

Output Checklist

1. Start from result.document or a loaded DoclingDocument; conversion setup is out of scope here.
2. Choose the output contract first: human-readable Markdown/HTML/text, machine-readable JSON/YAML/dict, model-oriented DocTags/DocLang, caption-oriented WebVTT, or chunk objects for RAG.
3. Preserve images only when conversion generated images and the export ImageRefMode matches the delivery target.
4. For tables, prefer table.export_to_dataframe(doc=doc) when Python post-processing is available; use the bundled helper when only Docling JSON is available.
5. For RAG, tune HybridChunker around the embedding tokenizer and table-header behavior before changing the upstream conversion pipeline.