This guide gets the finished app running with local OCR and suggestions. If you want to build it lesson by lesson instead, start with the tutorials. Check requirements first: Node 22.17 or newer, about 4 GB of free disk, and network access for the first run.

​
Get the code

npm ci writes about 2.8 GB to node_modules; it takes a few minutes on a first install and is much faster when npm’s cache is warm.

​
Start the local service

In one terminal:
Wait for Redact QVAC service listening on http://127.0.0.1:4317. Models are not loaded yet; they load on the first analysis. If port 4317 is taken, see How to change the port.

​
Start the app

In a second terminal, from the same redact directory:
Open http://localhost:3101.

​
Run the first analysis

  1. Choose a file, for example public/test-files/redact/chat-private.png.
  2. Pick Private or Confidential.
  3. Optionally fill Additional direction, for example Redact direct identifiers, account numbers, and reference codes. What you write here changes which fields are suggested.
  4. Select Prepare for review.
The first analysis downloads about 550 MB of models into .qvac/models and takes a few minutes, depending on your connection. The service terminal prints [QVACRegistryClient] [INFO] download lines meanwhile. Later analyses reuse the cache; OCR then takes roughly 20–30 seconds per page. If analysis fails or the service isn’t running, the review screen still opens and you can mark the document by hand.

​
Next