Skip to main content

Usage

Basic Batch Processing​

To run the standard OCR application which processes all PDFs in a directory:

python ocr_app.py \
    --input_dir input/tourism \
    --output_dir output/tourism \
    --prompt_file input/tourism/prompt.txt

[!IMPORTANT] This application defaults to VLLM_USE_V1=0 to ensure compatibility with DeepSeek-OCR's custom logits processors.

Multi-Agent Workflow​

For complex documents requiring structured JSON output, use the workflow script:

python ocr_workflow_v2.py \
    --input_dir input/orgchart \
    --output_dir output/orgchart_workflow_v2 \
    --prompt_file input/orgchart/prompt.txt