Skip to content

OCR Queue

Requires Mistral API key

The OCR Queue uses Mistral AI's OCR service. You need a Mistral API key to use this feature.

Some documents scan poorly – handwritten notes, faxes, low-resolution scans. When Paperless-ngx can't extract readable text, the AI has nothing to work with and the document gets skipped or poorly tagged.

The OCR Queue is the rescue system for these documents.

How it works

When Paperless-AI encounters a document with very little or unreadable text, it automatically adds it to the OCR Queue instead of skipping it silently.

From /ocr you can then:

  1. See all queued documents
  2. Press Process Queue to send them through Mistral's OCR API
  3. Mistral reads the document image and extracts proper text
  4. The improved text is written back to Paperless-ngx
  5. Optionally, the AI tagging pipeline runs automatically on the freshly extracted text

Progress is shown in real time while the queue processes.

Enabling it

In /settings, enable Mistral OCR Queue and enter your Mistral API key. That's it – documents with poor OCR will now be queued automatically.

Which documents end up in the queue?

Documents whose extracted text falls below a minimum quality threshold are added automatically. You can also add documents manually from the History page if you notice a specific one was poorly classified.

Cost

The OCR feature uses Mistral's paid API. Check Mistral's pricing for current rates. For most home users, the cost is negligible.