A PDF extraction API for Sanad
The OCR beta accepts PDF uploads and returns extracted text, with page selection, language hints, and explicit processing limits.
Turn a scanned document into usable text
Sanad now has a PDF OCR beta. A client submits a PDF through an authenticated multipart request and receives the extracted text. Page selection and language hints give the caller control over the material being processed.
A scanned document contains more than words in reading order. Orientation, page layout, image resolution, and tables all affect what can be recovered. The API makes the extraction step available as its own operation so an application can inspect the text before relying on it elsewhere.
Bound the work of each request
The service renders selected pages and processes them sequentially. Upload, page-count, image-size, and request-capacity limits keep one document from consuming unbounded resources. Per-key admission controls limit concurrent work and return explicit retry behavior when capacity is unavailable.
The public response focuses on the extraction result and request status. It does not require callers to choose an underlying inference provider or understand the infrastructure used to process each page.
Use the beta as an extraction step
The beta is intended for controlled integration and inspection of results. The extraction path keeps document bytes and recognized text out of application persistence and logs; clients should still review the service’s terms and privacy notice before submitting a document.
Recognition errors can change a number, omit a row, or alter reading order. Applications should preserve a route back to the original page when the extracted text informs an important decision. This release provides text extraction; answering questions with source verification is a separate problem.