Who This Is For
Intermediate.
Think of it like a filing cabinet that reads its own labels for you, instead of flipping through folders trying to remember which one has last year's insurance renewal.
What You'll Build
A searchable document library with at least three document types and an automated intake workflow.
Prerequisites
- Docker host
- A flatbed or document-feeder scanner (or a mobile scanning app)
Key Directories
consume/, drop new documents here for automatic processingmedia/, where processed documents are storedexport/, for exportsdata/, application database and index
Deploy with Docker Compose
Paperless-ngx publishes an official Compose file with Redis and the app tied together, start from that rather than a custom setup.
OCR Basics
Every document gets OCR'd on intake so its text becomes searchable, this is the slowest step, especially on low-power hardware like a Raspberry Pi.
Tags, Correspondents, and Document Types
Set these up before bulk-importing anything, retrofitting organization onto hundreds of documents later is painful.
Example Workflow
- Scan a bill
- Drop the PDF into the consume folder
- OCR runs automatically
- Metadata gets applied via matching rules
- The document becomes searchable
Mail Ingestion
Paperless-ngx can also pull attachments directly from an email inbox on a schedule, useful for digital-only bills and statements.
Automatic Matching Rules
Define rules so, e.g., anything from a specific correspondent automatically gets tagged and filed correctly.
Security & Backup Notes
- These documents are often sensitive, back up both the media folder and the database, and keep the whole stack off any public-facing port
Troubleshooting
- OCR is slow, expected on low-power hardware; consider a more capable host if throughput matters
- Wrong language support, Tesseract language packs need to be configured explicitly for non-English documents
- Documents tagged incorrectly, refine matching rules; they're pattern-based and need iteration
- Permission problems with the consume folder, check the mounted volume's ownership matches the container user
- Backups miss the media folder or database, both are required; the database alone is useless without the actual document files
Lab Finish Line
A searchable document library with at least three document types and an automated intake workflow.