Internal document workspace

OCR settings

Configure the OCR engine and check worker health. Monitor and retry documents in Processing jobs.

Active engine—Loading configuration…

Local worker health—Checking the worker…

Document timeout—Worker limit for each complete document

Saved changes apply to the next queued document. A document already being processed continues with its current configuration. Azure credentials are managed from the Azure OCR setup page and are never displayed here.

Processing configuration

OCR engine settings

Choose how new uploads and manual retries are processed.

OCR providerSelect a local or configured cloud engine.PaddleOCR executionAuto uses the worker's deployment default, with CPU and GPU available as explicit choices.PaddleOCR model profileSmall is recommended for CPU processing. Medium prioritizes accuracy but can take substantially longer; Tiny reduces compute and memory use further.

Recognition language setMultilingual recognizes mixed Chinese, English, Japanese, and many other languages with one PP-OCRv6 model.PDF render DPI (reserved)

DPI

Reserved for providers that render PDF pages locally.

Worker request timeout

sec

Applies to the next attempt, including all pages. A running attempt keeps its original limit; health checks use a shorter timeout.PaddleOCR worker URLUse the Docker service URL in production, or the published host URL when the API runs outside Docker. Authentication is managed by the deployment secret.

CPU inference threads

threads

Applies when the worker runs on CPU. Usually 2–4 threads gives the best balance; leave capacity for PaperMind and Windows.Maximum image dimension

Bounds OCR memory use for large scans and photos.