Docling Document Intelligence

Docling Document Intelligence

by docling-project (IBM)
MIT
Repository
Featured markdown Document Processing
document-processingragpdfocrchunking

Summary

Parse, convert, chunk, and structurally analyze PDFs, DOCX, PPTX, HTML, and images with IBM's open-source Docling toolkit.

Features

  • Parses PDF/DOCX/PPTX/HTML/images to Markdown or structured JSON
  • Hybrid heading+token chunking purpose-built for RAG ingestion
  • Standard vs VLM pipeline selection guidance
  • Structural analysis: headings, tables, figures, reading order

Install This Skill

Add this skill to your favorite AI agent in a few steps.

Any AI agent

This skill is plain instructions — it works with any assistant that accepts custom instructions or system prompts.

  1. Copy the skill content with the button below.
  2. Paste it into your agent's instruction file or system prompt (for example AGENTS.md, .cursorrules, or a custom instructions field).
  3. Ask the agent to apply the skill whenever the task matches.

Skill Content

Markdown Content

Copy this content and use it with your preferred AI agent

# Docling Document Intelligence Skill

Use this skill to parse, convert, chunk, and analyze documents with Docling. It handles local file paths and URLs, outputting Markdown or structured JSON (DoclingDocument).

## Scope
| Task | Covered |
|---|---|
| Parse PDF/DOCX/PPTX/HTML/image | Yes |
| Convert to Markdown | Yes |
| Export as DoclingDocument JSON | Yes |
| Chunk for RAG (hybrid heading + token) | Yes (Python API) |
| Analyze structure (headings, tables, figures) | Yes (Python API) |
| OCR for scanned PDFs | Yes (auto-enabled) |

## Choose a pipeline
| Pipeline | Best for | Key tradeoff |
|---|---|---|
| Standard (default) | Born-digital PDFs, speed | No GPU needed |
| VLM | Complex layouts, handwriting, formulas | Needs GPU, slower |

```bash
docling report.pdf --output /tmp/
docling report.pdf --to json --output /tmp/
docling report.pdf --pipeline vlm --output /tmp/
```

Description

Docling Document Intelligence wraps IBM's open-source Docling library, a widely used document-conversion toolkit, as an agent skill for parsing, converting, chunking, and structurally analyzing documents. It handles local file paths and URLs across PDF, DOCX, PPTX, HTML, and image inputs, outputting either clean Markdown or a fully structured JSON representation that preserves headings, tables, figures, and reading order.

Conversion runs primarily through the docling CLI, with the Python API reserved for features the CLI does not expose, including hybrid heading-plus-token chunking built for RAG ingestion, VLM remote-API endpoint configuration, and a hybrid text-extraction mode. The skill documents a genuine decision point most document tools gloss over: a standard pipeline that is fast and needs no GPU versus a VLM pipeline that is GPU-dependent and slower but far better on complex layouts, handwriting, and formulas, with guidance on picking correctly per document. Because Docling is mature and already used broadly in RAG and document-AI pipelines, this skill is a strong default whenever a document needs to be converted, chunked for retrieval, or structurally analyzed rather than just OCR'd.

Related Skills

GLM-OCR Table Recognition

GLM-OCR Table Recognition

by Z.ai (Zhipu AI)

Extract complex tables, including merged cells and multi-row headers, from images and PDFs into clean Markdown.

Document Processing
ocrtable-extractiondocument-processing
PaddleOCR Document Parsing

PaddleOCR Document Parsing

by PaddlePaddle

Turn PDFs and scanned images into structured Markdown/JSON: cell-level tables, LaTeX formulas, correct multi-column order.

Document Processing
ocrdocument-processingpdf
Browse all skills →