Document Processor API

A FastAPI application that extracts structured content from PDF documents. This API processes PDF files and extracts content in a structured format including text, tables, and images with their positions and hierarchy.

Features

• Extract content from PDF documents
• OCR capability with multiple pipeline options (fast, full, hybrid)
• Page and slice screenshots with configurable format and quality
• Content hierarchy extraction with parent-child relationships
• Table data extraction
• Customizable processing options (page range, image format, quality, scale)

Models

Enums

• OcrPipeline: Options for OCR processing

  • FAST: Optimized for speed
  • FULL: Optimized for accuracy
  • HYBRID: Balance between speed and accuracy

• ImageFormat: Supported image formats for screenshots

  • JPEG
  • PNG
  • WEBP

Data Models

• Image: Represents an image

  • data: Base64 encoded image data
  • content_type: MIME type of the image
  • width: Width of the image in pixels
  • height: Height of the image in pixels

• Position: Represents the position of a slice in a document

  • page_num: The page number of the slice
  • top: The top position of the slice
  • right: The right position of the slice
  • bottom: The bottom position of the slice
  • left: The left position of the slice
  • coord_origin: The coordinate origin of the slice

• Slice: Represents a slice of content extracted from a document

  • slice_num: The number of the slice in the document
  • level: The level of the slice in the document hierarchy
  • ref: The reference ID of the slice
  • parent_ref: The reference ID of the parent slice
  • label: The label of the slice
  • content_text: The content text of the slice
  • caption_text: The caption of the slice (for tables and pictures)
  • table_data: The table data associated with the slice (for tables)
  • positions: The positions of the slice in the document
  • screenshot: The screenshot of the slice

• Page: Represents a page in a document

  • page_num: The page number of the document
  • width: The width of the page
  • height: The height of the page
  • screenshot: The screenshot of the page
  • slices: The slices on the page

• ProcessRequest: Request model for document processing

  • ocr_pipeline: The OCR pipeline to use (default: HYBRID)
  • first_page: The first page number to process (default: 1)
  • last_page: The last page number to process (default: None = all pages)
  • image_format: The image format for screenshots (default: WEBP)
  • image_quality: The quality of the image, 0-100 (default: 80)
  • image_scale: The scale factor for the images (default: 2.0)

• ProcessResponse: Response model for document processing

  • document: The document name
  • content_type: The MIME type of the document
  • size: The size of the document in bytes
  • pages: The pages of the document

• HealthResponse: Response model for health check

  • version: The version of the service
  • build_id: The build ID of the service
  • commit_sha: The commit SHA of the service
  • up_since: The date and time when the service started

API Endpoints

Health Check

GET /health

Returns the health status of the service.

Response:

{
  "version": "1.0.0",
  "build_id": "12345",
  "commit_sha": "abcdef",
  "up_since": "2025-04-30 12:00:00"
}

Process Document

POST /process

Process a PDF document and extract structured content.

Request Form Parameters:

Parameter	Type	Default	Description
file	File	-	The PDF document to process
ocr_pipeline	OcrPipeline	HYBRID	The OCR pipeline to use
first_page	int	1	The first page number to process
last_page	int or null	null	The last page number to process
extract_pages_screenshot	bool	true	Whether to extract page screenshots
extract_slices_screenshot	bool	true	Whether to extract slice screenshots
image_format	ImageFormat	WEBP	The image format for screenshots
image_quality	int	80	The quality of images (0-100)
image_scale	float	2.0	The scale factor for images

Response: A ProcessResponse object containing the structured content of the document.

Installation

Prerequisites

• Python 3.10+
• FastAPI
• PDF processing libraries
• OCR capabilities

Setup

1. Clone the repository

   git clone https://github.com/yourusername/document-processor-api.git
   cd document-processor-api

2. Install dependencies

   pip install -r requirements.txt

3. Run the application

   uvicorn app.api:api --reload

Environment Variables

The application uses the following environment variables:

• ENV: Environment (development, production)
• VERSION: Application version
• BUILD_ID: Build identifier
• COMMIT_SHA: Git commit SHA

License

This project is licensed under the MIT License - see the LICENSE file for details.

Author

• Joan Fabrégat j@fabreg.at