CareFlow

A hackathon project: early warning for care-coordination delays in hospitals. It watches operational workflow events (not clinical data), rebuilds each case's timeline and flags steps that are about to miss, or have missed, their deadline. One example is a completed lab result that nobody has reviewed after 2 hours.

• Delay detection is rule-based: the SLA rules are in backend/app/sla_config.py. An LLM (gpt-4o-mini) only explains why a delay matters and what to check first.
• Events come from the public HAPI FHIR R4 server (backend/app/fhir_ingest.py). Only workflow metadata is used: case IDs, event types, timestamps and departments. No diagnoses, lab values or clinical notes.
• The backend is FastAPI and the dashboard is React (Vite).

The first version (Streamlit + Gemini) is archived at CareFlowAI, with a demo video.

Run locally

pip install -r requirements.txt
export OPENAI_API_KEY=...
cd backend && uvicorn app.main:app --reload     # API on http://127.0.0.1:8000

In a second terminal:

cd frontend && npm install && npm run dev