Document Workflow brings lightweight document management directly into OpenCloud, inspired by apps like Paperless-ngx and Papra. Organize your documents with searchable text, tags and custom fields, with optional AI assistance to suggest filenames, folders, and document details. Review suggestions before filing or automate the process from your chosen inbox without leaving OpenCloud.
0.20.3 Beta. Tested with OpenCloud 9 and Web SDK 8.2.1. It is an independent community app, not an official OpenCloud product.
- Watch folder to manually or automatically check for incoming documents.
- Extracts PDF text and reads scanned pages locally using Tesseract OCR with optional convert to PDF for unsupported filetypes. (can still tag and add fields manually to unsupported file types)
- Uses OPTIONAL AI with local Ollama or compatible cloud AI endpoint for suggestions and automations. With AI off the OCR and field/tag matching rules can still work.
- Files documents using configurable categories, naming patterns, custom fields and tag rules. Optional AI tag review reuses related existing tags and can propose new tags.
- Preserves original PDF bytes and OpenCloud permissions. Duplicate detection with exact byte matches and similarity matches.
- Provides optional unattended processing with OpenCloud App Tokens
- Customizable backups stored in OpenCloud.
- Includes English, Japanese, Spanish, French and German UI, plus light, dark and system appearance. OCR languages and interpretation rules can be extended separately.
All screenshots use fictional documents. Select an image to view it at full size.
| Inbox | Document review |
|---|---|
 |
 |
| Document library | Processing settings |
|---|---|
 |
 |
| Activity | Light mode |
|---|---|
 |
 |
Download the matching installation kit from Releases. Extract it, verify SHA256SUMS, and run ./install.sh install as an administrator on the Docker host. See Quick install for prerequisites, options and rollback.
The kit includes the Web app and the companion image build context, with Node, SQLite, Poppler, Tesseract and initial English/Japanese OCR data inside the container. A separate isolated conversion service supplies LibreOffice, MuPDF and image-to-PDF conversion; the installer manages it automatically. It has no OpenCloud credentials, storage mounts, public ports or internet route. Docker/Compose, Bash and sha256sum must already be available. Building the image downloads packages from the official Node/Debian sources. AI models are optional and configured separately.
The OpenCloud store ZIP installs the frontend only. Processing requires the companion service. Traefik routing can be configured by the installer; Caddy and other proxies use the external-proxy option and an administrator-configured same-origin route. See Installation boundaries.
After installation, choose Inbox and destination folders in Settings, configure models and OCR, and start with Review first. Enable automatic processing only after checking proposals against representative documents.
See Source builds, Architecture, Contributing, Security and Release/submission notes. Source tests require Poppler (pdfinfo, pdftotext, pdftoppm) and Tesseract with English/Japanese data on the development host; the end-user installation kit supplies these inside its container. Test fixtures are synthetic. Personal filing taxonomies and live credentials are not shipped as defaults.
New application code is AGPL-3.0-only. The upstream OpenCloud skeleton and dependencies retain their licenses; see Third-party notices, LICENSES/ and the dependency inventory.
I use AI tools to help with development, including analysis, code, tests and documentation. I care about security, privacy and protecting people's data, and I try to reflect that in how I build these projects.
Validation records distinguish checks actually run from checks still required. See validation and contribution guidance.