Paperless-ngx is the most widely used self-hosted document management system in the private and small-business space: paper is scanned, put through automatic text recognition, tagged, and ends up in a searchable archive. The project is the actively maintained community successor to the earlier Paperless variants.
Articles on Paperless-ngx (1)
The consume pipeline
The basic principle is an assembly line: documents enter the system through the consume folder, by mail retrieval, or by upload. There they pass through OCR (text recognition via OCRmyPDF and Tesseract, configurable for multiple languages), are stored as a searchable PDF/A archive copy, and are classified automatically. A learning mechanism proposes correspondent, document type, and tags based on the recognized text, and it becomes more accurate with every manual correction. The original is preserved untouched, while the archive version serves display and search.
Organization without a folder tree
Instead of nested folders, Paperless-ngx organizes by metadata: correspondents (who), document types (what), tags (arbitrary facets), and date fields. Full-text search combines these dimensions with the OCR text; saved views represent recurring queries. Physical storage in the file system is handled by configurable storage paths, which turn metadata into a comprehensible structure, so that the archive remains readable outside the application as well.
Architecture and operation
The reference installation runs as a Docker Compose stack: web application, database (PostgreSQL or SQLite/MariaDB), Redis as the task broker, and the volumes for originals, archive, and index. As with any containerized service, the data lives in the volumes, and the separation of originals, archive, and database determines the backup strategy; the built-in export additionally produces a portable full backup. Because archives grow over the years while local storage is finite, storage architecture is part of the picture, for example moving the document holdings to larger or remote storage targets.
Assessment
The strength of the system lies in the combination of automation and data sovereignty: once set up, the assembly line processes incoming paper almost unattended, and the entire collection stays in the owner’s hands as open files plus metadata, with no subscription and no cloud dependency.
