Importing a paperless-ngx export

paperless-ngx ships a document exporter that writes every original file plus a manifest.json describing it. Thaw reads that manifest, so nothing you set up is lost; it just becomes hints.

1. Run the exporter

From your paperless host:

docker compose exec webserver document_exporter ../export

That writes the originals (and archived OCR copies) into the export folder alongside manifest.json. Zip the folder:

cd export && zip -r ../paperless-export.zip .

2. Get it to Thaw

If the ZIP is under the browser upload limit, drop it on the Import page. Most real exports are larger, so during early access the path is the command line:

aws s3 cp paperless-export.zip s3://<your inbound bucket>/imports/
php artisan thaw:import:archive <family> s3://<your inbound bucket>/imports/paperless-export.zip

We run this for early-access members; send us a note and we will do it with you. A browser path for large archives is coming with the cloud folder connectors.

3. What Thaw does with it

The archive is expanded in slices so a thousand-file export never times out. For each document in the manifest, Thaw imports the original file (not the archived OCR copy), and passes along as hints:

• the paperless title
• the correspondent’s name
• the document type’s name
• the tags
• the created date and the archive serial number

The reader is told these came from a previous system and may be wrong. It confirms from the page. In practice they help a lot: a correspondent of “GEICO” and a type of “Insurance” make the reader’s job easy, and a wrong one is simply overridden.

4. Watch it fill in

The import page shows people and things appearing as they are found, a histogram by year, categories as they are sorted, and the first checkup items. Sorting a thousand documents takes well under an hour; the full read continues in the background, current documents first.

5. Turn the Mac mini off

Exact duplicates are recognised by content and linked, so if you exported twice, or the same bill was scanned twice in 2019, you get one document and a note. When the import shows done, the checkup will already have told you at least one thing you did not know.