OCR Providers

OCR Providers

An OCR provider is the engine that turns a document image into text before an Extraction Profile pulls fields out of it. You can register a self-hosted engine (a RapidOCR / PaddleOCR sidecar) or a cloud document-AI service (AWS Textract, Google Document AI, Azure Document Intelligence). DataMagik tries your providers in order and falls back to AI vision — so extraction works even before you add one.

Adding a provider

1. Click New provider and give it a Name.
2. Pick a Type (self-hosted sidecar, AWS Textract, Google Document AI, or Azure Document Intelligence).
3. For a self-hosted sidecar, set the Endpoint URL. For a cloud service, paste its Credentials (JSON) — for example AWS keys as {"access_key_id":"…","secret_access_key":"…","region":"us-east-1"}.
4. Set the Order (providers are tried lowest-first) and make sure Enabled is on. Save.

Use the Test button to confirm a provider responds before relying on it.

How extraction picks an engine

ocr.extract tries your enabled providers in order; if none succeed (or none are registered) it falls back to AI vision. That means Extraction Profiles work out of the box, and adding a dedicated provider is an optimization for cost, speed, or accuracy.