How tamil-phy/nhs-healthsign works
File structure
• Top-level scripts:
• pdf2md.py – PDF → Markdown conversion using Docling.
• main.py – Core analysis engine (disease screening, vitals analysis, timeline monitoring).
• streamlit-ui.py – Streamlit dashboard for interactive exploration.
• Config directories/files:
• INTERESTED_READCODES.json, red_flag_read_codes_subset.json – clinical codes for disease screening.
• INTERESTED_VITALS.json – vital sign definitions and units.
• secrets.json – API keys / sensitive settings.
• Output directory:
• output_trial/ – JSON artifacts (code_matches.json, timeline.json, vital_matches.json) from a sample run.
• Supporting files:
• INTERESTED_*.json – domain data driving the analysis.
• main.py and pdf2md.py reference these configs.
Extract README insights
• The pipeline has three phases: ingestion (PDF → Markdown), processing (screen codes, vitals, timelines), and presentation (Streamlit).
• Dependencies include Python, Docling, fuzzywuzzy, OpenAI (if used), and Streamlit.