Docspell

Docspell is a self-hosted document management system designed to organize files from scanners, emails, and other digital sources. It functions as a personal document organizer for households, families, and small companies to unify and categorize their paperwork. The system helps users manage the volume of files generated by digitizing physical documents, providing a centralized location for storage and retrieval.

The software is deployed as a server with a mobile-friendly single-page application as the primary user interface. Users can interact with the system through a web browser, a command line interface, or an Android application specifically designed for uploading files from mobile devices. It can be installed via Docker, Debian packages, the Nix package manager, or Helm charts for Kubernetes environments.

Key features

• Automated metadata and tag suggestions using machine learning
• Optical Character Recognition (OCR) for document text extraction
• Full-text search across stored documents
• Integration with email for document ingestion
• Support for custom metadata and correspondent association
• NLP-based date and entity discovery
• REST/HTTP API for external integration

The backend is written in Scala using a functional style and relies on the Stanford Core NLP library for its machine learning capabilities. This allows the system to learn from existing documents to suggest tags and identify correspondents without manual entry. For file processing, it orchestrates several external tools, including Tesseract for OCR, unoconv for file conversion, and ocrmypdf for PDF processing. The frontend is built as a single-page application using Elm and Tailwind CSS, ensuring a responsive experience across different devices.

Docspell is an AGPLv3 licensed tool for users who require a private, self-managed system to automate the indexing and retrieval of digital documents.