Hi Streamlit Community! 
I’m İsa, a registered nurse turned software developer. Throughout my clinical career, I realized that accurate documentation and risk assessment (like calculating NEWS2 scores) are critical for patient safety but often consume valuable time that could be spent on patient care.
So, I decided to bridge the gap between healthcare and tech. I built NurseFlow, a Clinical Decision Support System, entirely in Streamlit.
Live Demo: https://nurseflow.streamlit.app 
GitHub Repo: GitHub - ruin7-cloud/nurseflow
What does it do?
NurseFlow is designed to automate nursing handovers. Instead of typing everything manually, a nurse can:
-
Snap a photo of a patient monitor (Streamlit file uploader + Gemini Vision). -
Record a voice note (using streamlit-mic-recorder). -
Paste unstructured notes.
The app extracts vital signs into structured JSON and generates a professional ISBAR Handover Report in PDF format.
The “Hybrid Logic” Architecture
The most challenging part was ensuring safety. We know LLMs can hallucinate numbers. To solve this, I used a Hybrid Approach:
-
Google Gemini 2.5 Flash-Lite: Used only for extraction (OCR/Transcribing) and formatting the report text.
-
Python Logic Engine: The critical NEWS2 Risk Score calculation is handled by a deterministic Python function (Hard-coded logic), NOT the AI. This ensures 100% mathematical accuracy for the risk score.
Tech Stack
-
Frontend: Streamlit
(of course!) -
AI Model: Google Gemini 1.5 Flash
-
Audio:
streamlit-mic-recorder -
PDF:
fpdflibrary -
Vision:
Pillow
Screenshots
I built this project to demonstrate how Streamlit allows domain experts (like nurses) to prototype powerful tools rapidly.
I’d love to hear your feedback or suggestions on how to improve the UX for clinical settings!
Thanks! 
Best regards,
Isa Sahin, RN