Supported Formats

Complete reference for input and output formats supported by Duckling.

Input Formats

Documents

Images

Technical Documents

Output Formats

Text Formats

Markdown (.md)

Best for documentation and content that needs formatting.

# Document Title

## Section 1

This is a paragraph with **bold** and *italic* text.

- List item 1
- List item 2

| Column 1 | Column 2 |
|----------|----------|
| Data 1   | Data 2   |

HTML (.html)

Web-ready format with styling preserved.

<h1>Document Title</h1>
<h2>Section 1</h2>
<p>This is a paragraph with <strong>bold</strong> and <em>italic</em> text.</p>

Plain Text (.txt)

Simple text without any formatting.

Document Title

Section 1

This is a paragraph with bold and italic text.

Structured Formats

JSON (.json)

Full document structure in JSON format. Lossless representation.

{
  "title": "Document Title",
  "sections": [
    {
      "heading": "Section 1",
      "level": 2,
      "content": [
        {
          "type": "paragraph",
          "text": "This is a paragraph..."
        }
      ]
    }
  ]
}

DocTags (.doctags)

Tagged document format for semantic analysis.

<document>
  <title>Document Title</title>
  <section level="2">
    <heading>Section 1</heading>
    <paragraph>This is a paragraph...</paragraph>
  </section>
</document>

YAML (.yaml)

Structured export using Docling's YAML serializer (or PyYAML fallback). Useful for config-driven pipelines.

WebVTT (.vtt)

Timed text export for transcripts and captions (especially from ASR pipeline output).

DCLX (.dclx)

Docling exchange format for round-tripping parsed documents between Docling tools.

HTML split by page (.split.html)

HTML export with page boundaries preserved (html_split_page format) for paginated web views.

DocLang (.dclg.xml)

AI-native XML document format from the DocLang open standard. Duckling exports DocLang via Docling's export_to_doclang() API. The format preserves document structure, semantic roles, layout, and bounding-box geometry in a single machine-readable representation suited for LLM and agent pipelines.

<doclang>
  <heading level="1">
    <location value="48"/><location value="40"/>
    <location value="420"/><location value="72"/>
    Document Title
  </heading>
  <text>This is a paragraph...</text>
  <table>
    <ched/>Column 1<ched/>Column 2<nl/>
    <fcel/>Data 1<fcel/>Data 2<nl/>
  </table>
</doclang>

Requires Docling 2.118.0+ and docling-core 2.90.0+ (<3) with DocLang spec 0.7 support via export_to_doclang() (see backend/requirements.txt). If DocLang is missing after conversion, upgrade dependencies and convert again — existing jobs converted before the upgrade will not have a DocLang file on disk. Duckling writes .dclg.xml (same as the Docling CLI); the DocLang spec also accepts the shorter .dclg extension.

Document Tokens (.tokens.json)

Token-level representation for NLP applications.

{
  "tokens": [
    {"text": "Document", "type": "word", "position": 0},
    {"text": "Title", "type": "word", "position": 1}
  ]
}

RAG Formats

RAG Chunks (.chunks.json)

Document chunks optimized for retrieval-augmented generation.

{
  "chunks": [
    {
      "id": 1,
      "text": "This is the first chunk of text...",
      "meta": {
        "headings": ["Section 1"],
        "page": 1,
        "token_count": 128
      }
    }
  ]
}

Format Selection Guide

API Format Parameter

When using the API, specify the format in the export endpoint:

# Download as Markdown
curl http://localhost:5001/api/export/{job_id}/markdown

# Download as JSON
curl http://localhost:5001/api/export/{job_id}/json

# Download as HTML
curl http://localhost:5001/api/export/{job_id}/html

# Download as DocLang
curl http://localhost:5001/api/export/{job_id}/doclang

MIME Types