The OCR process in the inference_schema and document_data_capture steps is currently too slow.
Instead of relying on a vLLM as is done now, it might be better to use an external OCR service to improve performance, since the current approach is causing noticeable delays.
The OCR process in the
inference_schemaanddocument_data_capturesteps is currently too slow.Instead of relying on a vLLM as is done now, it might be better to use an external OCR service to improve performance, since the current approach is causing noticeable delays.