Paperless-ngx, the open-source tool that turns scanned paper into a searchable, self-hosted document archive, released version 3.1.0 on August 27, 2026. Unlike several recent updates from other self-hosted projects, this one isn’t a security patch. There’s no CVE anywhere in the changelog. It’s a straightforward feature and quality release for people already running Paperless-ngx as their day-to-day archive.
That’s worth flagging because it hasn’t been the norm lately. Recent weeks alone have brought active RCE exploitation in Gitea, a nine-bug security bulletin from n8n, and a critical proxy bypass in Portainer. Paperless-ngx 3.1.0 breaks that pattern: it’s shipping new capabilities, not putting out a fire.
For anyone who hasn’t used it, Paperless-ngx runs OCR on scanned documents and photographed paperwork, then indexes the extracted text so the whole archive becomes searchable. It suggests tags, document types, and correspondents once it has learned from your documents, and it has become a common self-hosted alternative to scanning things straight into Google Drive or Dropbox. The project’s own comparison list names Mayan EDMS and Docspell as its closest open-source peers.
The biggest change in 3.1.0 is a new workflow action that applies AI suggestions automatically. Paperless-ngx has offered machine-learning suggestions for tags, document types, and correspondents for some time, but a person always had to review and accept them first. Now that step can be built into a workflow rule, so routine documents get classified without anyone opening them, while anything unusual can still go through manual review. A recurring invoice from the same vendor, for instance, can be tagged and filed the moment it lands, with nobody needing to open it first.
A handful of other changes round out the release:
- Selective remote OCR. Each document can now use a remote OCR engine or stay local, chosen per document instead of through one sitewide setting.
- OIDC group sync. Identity provider groups can map directly to Paperless-ngx’s superuser and staff roles, which helps if permissions are already centralized through Authentik, Keycloak, or another OIDC provider.
- Document versioning. A new option merges documents as versions of the same file, with a version dropdown, a counter, and related indexing fixes.
- Export changes. ZIP exports now let you choose the compression type and level, and the exporter uses less memory and I/O on large archives.
- Smaller fixes. Customizable icons for saved views, a websocket heartbeat, jitter added to IMAP polling, a Docker image rebuilt on Python 3.14, and interface fixes covering Android navigation, a Firefox printing regression, and the mobile search dropdown.
The versioning change fixes something that has bothered Paperless-ngx users for years. Re-scanning or replacing a document used to mean either overwriting the original or keeping near-duplicates around with no link between them. Merging documents as versions keeps that history in one place, with the version count visible right from the document view. That’s useful for anything that gets amended over time: contracts, insurance paperwork, recurring official forms. The changelog also lists index corrections tied to the new version history, so search results should reflect the right revision rather than an orphaned duplicate.
Of everything in this release, the workflow-driven AI suggestions and the per-document OCR toggle are probably worth setting up first. Both swap an all-or-nothing setting for something you can control per document, which matters if your archive mixes routine mail with anything sensitive. The OIDC group sync matters most if an identity provider already handles logins elsewhere in your homelab: Paperless-ngx permissions no longer have to be managed separately by hand.
Full setup details, the Docker Compose stack, and a comparison with alternatives like Nextcloud are on our Paperless-ngx application page. For the complete changelog and the maintainer’s own release notes, see the official Paperless-ngx 3.1.0 release on GitHub.