What's new in OleeVision.

Each extractor can now hold a list of allowed IP addresses, in the "Allowed IP addresses" box on its page. Leave it empty and nothing changes. Add your server’s address and every other address is refused before a read is spent, so a copied API key is worthless to anyone else. Use it only where the calling server has a fixed IP, most serverless hosts change address and would lock themselves out.

If an extractor reads far more documents in a day than it usually does, its owner gets an email with today’s count and the usual figure. Nothing is blocked or slowed, and it is one email per extractor per day. This is what a copied API key looks like from the outside. New extractors are exempt until they have a normal to be compared against.

Send your key as "Authorization: Bearer YOUR_KEY" and post to /api/v1/extract. The old /api/v1/extract/YOUR_KEY form returns 410. Your key still works, just move it into the header.

The per-extractor CORS switch has been withdrawn. Using it meant publishing your API key in page JavaScript, where any visitor could read it and spend your reads. The capture widget does the same job safely: your server trades the key for a token worth one scan, on one website, for fifteen minutes. If an extractor had CORS on, its page now explains what to do.

A built-in assistant now lives in the workspace. Ask product questions in plain language and Don answers and links you straight to the right screen.

The inbox is now "Reads". Browse every submitted document with a thumbnail, its extracted data, OCR text and webhook status, filtered by extractor and date.

A cleaner three-column workspace, a unified type system, and a new marketing home page that shows a live document-to-JSON extraction.

Extractors can now require captured consent before processing, with provenance recorded on every read for auditability.

Live charts for scans over time, success rate, failures and latency, plus alerts before you reach your monthly limit.

Push extraction results to your systems with signed webhooks, and manage cross-origin access per extractor.

Extraction latency and field accuracy improved across invoices, IDs and receipts.