OCR Then Extract
A 6-step recipe with Mindee. We can help you get access.
Recipe · 6 steps
OCR Then Extract
with Mindee
OCR a document for raw text, then extract structured fields from it.
- Send the document to the asynchronous OCR queue to capture its full text.
- Poll the shared jobs endpoint until the OCR job reports Processed or Failed.
- Read the recognized per-page text from the completed OCR inference.
- Send the same file to the extraction queue against the chosen extraction model to read its structured fields.
- Poll the shared jobs endpoint until the extraction job reports Processed or Failed.
- Retrieve the completed extraction inference and read the structured fields parsed from the document.
What you'll need to have ready
- authorization: Mindee API key sent in the Authorization header.
- ocrModelId: UUID of the OCR utility model.
- extractionModelId: UUID of the extraction model to apply after OCR.
- file: The document file to upload as binary form data.
- filename: Optional filename to associate with the uploaded document.
For your developer
Arazzo workflow mindee:ocr-then-extract, 6 operations: Enqueue_OCR_Product_Inference_v2_products_ocr_enqueue_post → Get_Job_Status_v2_jobs__job_id__get → Get_OCR_Product_Result_v2_products_ocr_results__inference_id__get → Enqueue_Extraction_Product_Inference_v2_products_extraction_enqueue_post → Get_Job_Status_v2_jobs__job_id__get → Get_Extraction_Product_Result_v2_products_extraction_results__inference_id__get. Specs on APIs.io.