All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
You should also add project tags for each release in Github, see Managing releases in a repository.
- Merged the MSF-OCR-Streamlit repository into this repository
- User authenticates with DHIS2 password rather than hard coded passkey
- Table headers are corrected based on key-value pairs from DHIS2
- Payload with key-value pairs for DHIS2 is displayed to the user before uploading
- Images are resized before being sent to OpenAI
- Added all UI features such as DHIS2 authentication, multiple tally sheet processing, left bar on streamlit, buttons present in LLM version of app into docTR version.
- User must confirm each page and key-value pairs before they're allowed to upload, so upload buttons is not initially selectable
- Bug where a user's changes aren't saved every other time or when confirming in the Streamlit app
- Requests to OpenAI are multithreaded to speed up time to get results for multiple images
- More comprehensive querying in DHIS2 for organization and dataset names
- Added OpenAI backed OCR functionality
- Module names changed to meet PEP 8 naming conventions
- Server settings and API keys set via environment variables instead of settings file
- Notebook for downloading sample test data
- msfocr.data.data_upload_DHIS2 created for sending key/value pairs to a DHIS2 server
- msfocr.docTR created to implement extracting tables from images using image2table and docTR
- Initial package structure created