feat: add gpt-4o OCR for documents and RAG Q&A

[jeonghoonkang]

Summary

• use gpt-4o to transcribe uploaded receipt images or PDF documents
• embed extracted text and answer questions through a simple RAG pipeline
• Base64-encode receipt images before sending them to the OCR model
• show upload progress in the Streamlit interface
• show images one at a time with arrow navigation while OCR text remains hidden
• provide a chat-style Q&A box for asking about the recognized text
• cache processed receipts and display the time taken for each Q&A answer
• allow jumping to a specific receipt by filename and store OCR results in a merged JSON file

Testing

• python -m py_compile apps/receipt_ocr/receipt_ocr_app.py && echo "py_compile success"

---

https://chatgpt.com/codex/tasks/task_e_688a94ac3f808331bd8ab4da9e07125f