DS4SD/docling
Docling
Docling is a multimodal content converter and document parser designed to transform PDFs, Office files, and HTML into structured Markdown or JSON for generative AI applications. It functions as an OCR document processor and a PDF layout analyzer that extracts tables, charts, and hierarchical structures while preserving the original page layout.
The system operates as a local-first inference engine, allowing for the processing of sensitive data in air-gapped environments without external network connectivity. It can also be deployed as an API or a Model Context Protocol server to provide parsing capabilities to external AI agents and language model frameworks.
Its capabilities include converting diverse file formats into a unified intermediate representation, utilizing optical character recognition for scanned documents, and transforming graphical charts into tables or text. The tool supports multi-format exports and includes speech-to-text capabilities for audio transcription.
A command line interface is provided for parsing documents from local paths or URLs into structured files.
Features
- AI-Ready Structured Conversions - Transforms complex PDFs and diverse document types into structured formats suitable for generative AI applications.
- LLM-Ready Markdown Converters - Transforms PDFs and office documents into structured Markdown optimized for LLM and RAG consumption.
- Local Inference Engines - Executes all parsing and transcription models on local hardware to allow operation in air-gapped environments.
- Document Layout Analysis - Uses specialized models to identify structural elements like headers, tables, and lists to maintain document hierarchy.
- PDF Document Analyzers - Analyzes complex PDF layouts to extract tables, charts, and hierarchical structures.
- RAG Data Pipelines - Converts diverse document formats into machine-readable text to power RAG pipelines.
- Optical Character Recognitions - Combines native text parsing with optical character recognition to process both digital and scanned documents.
- OCR Document Processors - Ships a processing engine that extracts text from scanned documents and images via OCR.
- Multimodal Content Converters - Transforms diverse file formats including audio and images into machine-readable text and structured data.
- Multi-Format Document Parsing - Converts PDF, Office documents, HTML, EPUB, and emails into a unified machine-readable representation.
- Optical Character Recognition - Extracts text from scanned PDFs and images to process content that is not natively selectable.
- Local Data Processing - Parses sensitive files in air-gapped environments to ensure data privacy without relying on external cloud APIs.
- Structured PDF Conversions - Analyzes page layout, table structures, and formulas within PDF files to preserve the original document hierarchy.
- MCP Servers - Runs as an API or MCP server to provide parsing capabilities to external agents and applications.
- Document Digitization Tools - Converts Office files, emails, and PDFs into unified formats like Markdown or JSON for organizational archival.
- Table Structure Reconstructions - Analyzes visual cell boundaries and alignments to transform complex PDF tables into structured machine-readable data.
- Chart-to-Table Parsers - Utilizes vision models to interpret graphical chart elements and convert them into descriptive text or tables.
- Air-Gapped Execution - Processes sensitive data in air-gapped environments without requiring any external network connectivity.
- Data Extraction And Generation - Fast document parsing and format conversion tool.
- Data Ingestion and Parsing - Tool for parsing and exporting documents into structured formats.
- Data Processing - Document preparation toolkit for generative AI workflows.
- Data Processing Tools - Library to prepare diverse document formats for generative AI.
