Label Verification

Compliance review tools

Faster review of alcohol beverage labels.

Upload a label image and the system reads the visible text, classifies the beverage, and runs the routine TTB checks an agent would normally do by hand. Use the single- label workflow to review one application, or the batch workflow to process many labels at once.

How to use this prototype

  1. Step 1

    Open a workflow.

    Choose single-label review or batch.

  2. Step 2

    Upload the label image.

    PNG, JPEG, GIF, or WEBP. Application fields are optional.

  3. Step 3

    Review the results.

    Open any row for the expected and extracted text side by side.

Review a single label

Upload one label image. The system extracts the visible fields, classifies the beverage type, and runs the standard TTB compliance checks. Provide optional COLA application values to also see a field-by-field comparison.

Open workflow

Review a batch of labels

Upload many label images at once. Add an expected-fields CSV to compare each label against its application values, or skip the CSV to run a compliance-only audit. Results stream into a sortable table you can export.

Open workflow

Security and limits

Prototype only: not for production COLA decisions. Uploads are processed in memory, discarded after the response, and capped at 50 images per batch,4.0 MB per image, and 512 KB per CSV.

AI data flow

Images are sent from the server to Vercel AI Gateway and the configured model provider for text extraction. Gateway credentials stay server-side; no model API key is exposed to the browser.

Performance target

Sarah's 5-second target is the product goal. Current prototype latency depends on provider response time; the app uses low-detail extraction and streaming batch rows to reduce perceived wait.