Docling API

Document processing API powered by Docling.

Features

• Multi-format support: PDF, DOCX, PPTX, XLSX, HTML, images, and more
• Multiple output formats: Markdown, plain text, JSON
• Streaming support: Regular streaming, per-page streaming, and SSE
• Bulk processing: Process multiple documents concurrently
• Image extraction: Extract embedded images and page renders

Installation

uv sync

Usage

Start the server:

uv run uvicorn main:app --reload

API documentation available at: http://localhost:8000/docs

API Endpoints

Documents

Endpoint	Method	Description
/documents/process	POST	Convert a document to specified format
/documents/process/stream	POST	Stream converted content
/documents/process/stream/pages	POST	Stream content page by page (NDJSON)
/documents/process/sse	POST	Stream with Server-Sent Events
/documents/process/bulk	POST	Process multiple documents

Images

Endpoint	Method	Description
/images/extract	POST	Extract images from a document

Health

Endpoint	Method	Description
/health/live	GET	Liveness probe
/health/ready	GET	Readiness probe

Project Structure

api-docs-1/
├── main.py                    # FastAPI app entry point
├── api/
│   └── routes/
│       ├── documents.py       # Document processing endpoints
│       ├── images.py          # Image extraction endpoint
│       └── health.py          # Health check endpoints
├── core/
│   ├── config.py              # Configuration settings
│   ├── schemas.py             # Pydantic response models
│   └── error_handlers.py      # Centralized error handling
└── services/
    ├── file_utils.py          # File handling utilities
    ├── docling_converter.py   # Document conversion logic
    ├── docling_streaming.py   # Streaming utilities
    ├── docling_service.py     # High-level orchestration
    └── image_service.py       # Image extraction service

Configuration

Environment variables:

Variable	Default	Description
MAX_FILE_SIZE_MB	50	Maximum upload file size in MB
DOCLING_OCR_ENABLED	true	Enable/disable OCR
DOCLING_OCR_LANGS	en	OCR languages (comma-separated)
OMP_NUM_THREADS	4	CPU threads for processing

OCR Notes

Docling auto-selects the best available OCR engine:

• RapidOCR (default): Works out of the box with PyTorch
• EasyOCR: Install with uv add easyocr
• Tesseract: Requires system installation

The "RapidOCR returned empty result" warnings are normal for pages without scanned text.

Frontend

A React test client is included in the frontend/ directory.

Setup

cd frontend
npm install
npm run dev

Open http://localhost:5173 to test all API endpoints.

Features

• Single file processing (all formats)
• Streaming responses
• Per-page streaming
• Server-Sent Events (SSE)
• Bulk file processing
• Image extraction
• Health checks

License

MIT