📖 Chapter 04 — Documents & Paperless

The pile of paper in the kitchen drawer, but searchable. Forever.

Learning Objectives

• Set up Paperless-ngx with OCR and auto-tagging
• Configure the consume folder so dropped scans are processed automatically
• Establish a family workflow for incoming paper
• Search years of scanned documents with the same precision as Google

Why Paperless

Every family has a pile of paper. Tax returns from 2014. The dishwasher warranty. The kid's vaccination records. The deed to the house. Insurance policies. The receipt for the laptop you bought in 2019. Most of it is rarely needed; when you do need it, you can't find it.

Paperless-ngx is a self-hosted document management system. You feed it scanned PDFs or photos; it runs OCR to extract the text, then auto-tags and stores the documents. When you need to find something, you search by content, date, tag, or correspondent.

What it replaces

Paperless-ngx replaces:

• The kitchen drawer full of "important documents"
• The accordion folder of tax returns
• The shoebox of receipts
• The "I know I saved that email somewhere" feeling

It does not replace:

• Original physical documents you legally need (passports, original deeds, car titles) — keep those in a fireproof safe
• Active working files (a contract you're negotiating, a project plan you're editing) — those go in their normal folder

Setting up Paperless-ngx

From Volume 1, Chapter 08: Paperless-ngx is already installed. Now to configure it for the family.

The consume folder

Paperless-ngx's killer feature is the consume folder. You drop a scanned PDF or photo into the folder; Paperless picks it up, processes it (OCR, auto-tag, store), and removes it from the folder.

The consume folder is at /mnt/tank/Apps/paperless/consume/ by default. Set up a SMB share for it (call it Scan) so the family can drop scans from any device.

From any laptop, drop a scanned PDF into \\NAS\Scan. Within a minute, it appears in the Paperless library, OCR'd, tagged, and searchable.

How the family scans

Three workflows, in order of how often you'll use them:

1. The flatbed scanner (most common for important docs)

You have a physical document. You want a high-quality scan.

• Use a flatbed scanner (Epson, Brother, Fujitsu — all work with SANE on Linux/TrueNAS, or with vendor software on Windows/Mac).
• Scan to PDF at 300 DPI, color (for color documents) or grayscale (for text).
• Save the PDF directly to \\NAS\Scan.
• Paperless picks it up, processes it, you forget about it.

2. The phone camera (most common for receipts and ad-hoc)

You're out, you get a receipt, you want it archived.

• Snap a photo with your phone.
• Use the Paperless mobile app (or any app that can save to a WebDAV folder) to upload.
• Or: email the photo to a dedicated address, and have Paperless's email-import feature pull it in.

3. Email forwarding (for statements, bills, anything digital)

You get a digital statement (utility bill, bank statement, insurance document).

• Forward the email to your Paperless email-import address.
• Paperless saves the email body and any attachments as documents.

OCR and auto-tagging

OCR (optical character recognition) is the part that turns a scanned image into searchable text. Paperless uses Tesseract, which is good but not perfect.

For a clean scan (300 DPI, straight orientation, decent lighting), Tesseract gets 99%+ accuracy. For a phone photo of a receipt on a crumpled table, maybe 80%. The 80% is still searchable; you just might miss a word or two.

What Paperless auto-detects

Paperless looks for patterns in the OCR'd text to suggest tags:

• Correspondent: the entity that sent/issued the document. Detected from letterheads, "From:" lines, and known patterns. Examples: "Comcast", "Toyota Financial Services", "State Farm".
• Document type: what kind of document. Detected from keywords and structure. Examples: "Invoice", "Receipt", "Contract", "Warranty", "Tax Return".
• Date: when the document was issued. Detected from common date formats.
• Tags: manual or rule-based. You define the rules.

Writing your own rules

After Paperless has been running for a month, you'll see the patterns. Write rules:

• If the document contains the word "W-2" or "1099" → tag as "Taxes" and set document type to "Tax Document"
• If the correspondent is "Comcast" → tag as "Utilities"
• If the date is in January and the document contains "Statement" → tag as "Year-End Statement"

Paperless has a Rules UI for this. The conversation's recommendation: start simple, add rules as you see patterns. Don't over-engineer upfront.

The "inbox" workflow

The most useful single workflow: process new paper as it arrives, in batches.

1. Designate a physical inbox (a tray, a drawer, a folder on the counter).
2. Incoming paper goes in the inbox. Don't file it; just put it in the inbox.
3. Once a week, the household admin (TK) processes the inbox: scan everything, drop the scans into \\NAS\Scan, shred the paper (or keep the original in a fireproof safe if it's legally important).
4. Paperless auto-processes. The inbox is empty. The family can find anything.

This is the workflow that actually works. The hard part is the weekly processing — it's a 15-30 minute job that has to be on the calendar.

Searching

The whole point of Paperless is being able to find things. The search syntax:

• Full-text: any word in any OCR'd document
• By tag: tag:warranty, tag:2024, tag:taxes
• By correspondent: correspondent:comcast
• By date: after:2024-01-01 before:2024-12-31
• By document type: type:invoice, type:receipt

Examples that come up in real life:

• "show me the warranty for the dishwasher" — search "dishwasher", filter to tag:warranty
• "what was the deductible on the homeowner's insurance" — search "deductible", filter to correspondent:state-farm
• "TK's 2022 W-2" — search "W-2", filter to year:2022

Backup

Paperless data is in /mnt/tank/Apps/paperless/. This is covered by the Apps dataset's snapshot policy. But snapshots are not backups (Volume 6 will cover this properly). For now, an additional safeguard: export the Paperless database monthly and store the export somewhere off the NAS. Paperless has a built-in "Export" feature in the admin UI.

When NOT to use Paperless

Paperless is for paper. It's not for:

• Active working files (a contract you're editing). Use your normal documents folder.
• Photos. Use Immich.
• Music. Use Jellyfin or a music player.
• Videos. Same.

One app per job. Don't make Paperless do things it isn't good at.

The single biggest mistake

The single biggest mistake people make with Paperless is not processing the inbox regularly. The system works beautifully when you scan everything weekly. It becomes a graveyard of unscanned paper when you "get to it later."

The whole point of the system is that paper becomes searchable. Paper that's still in the physical inbox isn't searchable. Process the inbox every week. Non-negotiable.

Engineering Note

Summary

Paperless-ngx is the family document archive. Scan incoming paper, drop in the consume folder, Paperless OCR's and auto-tags. A weekly inbox-processing session keeps the system current. Search by content, tag, correspondent, or date. Export the database monthly as an off-NAS backup. Don't use Paperless for active files, photos, music, or video.

Checklist

• ⬜ Set up the \\NAS\Scan SMB share pointing at the Paperless consume folder
• ⬜ Verify: drop a test PDF into the share, watch it appear in Paperless within a minute
• ⬜ Configure the email-import feature (if you want to forward statements)
• ⬜ Install the Paperless mobile app (if the family scans from phones)
• ⬜ Set up a recurring weekly scan session on the family calendar
• ⬜ After a month, write your first 3 rules based on patterns you've seen

Looking Ahead

Chapter 05 is mobile uploads — the patterns and tradeoffs of getting photos and files from phones onto the NAS. Most of this is covered by Immich (Chapter 02), but the chapter also covers non-photo uploads: documents, videos, files you want to share from the phone.

Ch 04 · v0.1 · drafted from the original ChatGPT conversation, July 2026