Offline multilingual cardiovascular health AI assistant for low-resource African settings, built for the ADTC 2026 Laptop LLM Track.
# AfriPulse CDSS — Version 1.0.0
### Offline Cardiovascular Clinical Decision Support System (CDSS) with On-Device Large Language Model (LLM) Explainability
Submitted as a competition-ready entry for the **ADTC Healthcare & Medical Challenge**.
---
## 🌟 Executive Overview & Purpose
AfriPulse is a clinical-grade, **completely offline**, desktop-first and responsive mobile-ready web hybrid application designed for healthcare practitioners operating in resource-limited, offline, or high-privacy clinical environments.
Cardiovascular disease (CVD) remains a primary global health concern, and early detection of Stage 1/Stage 2 Hypertension and elevated cardiovascular risk is critical for protecting target organs (heart, brain, kidneys). AfriPulse implements the latest deterministic medical protocols in clinical decision trees to assist assessors, nurses, and community assessors in identifying risk early, validating vital signs, tracking physiological trends, compiling health reports, and utilizing a **local, fully offline GGUF Llama 3.2 3B LLM model** to explain diagnostic conclusions without ever requiring internet connectivity.
---
## 🏛️ System Architecture Overview
The system is designed around a strictly private, local-first sandbox model. By utilizing in-browser data stores mirroring SQLite behavior and localized state-caches, AfriPulse guarantees that **no user data or clinical metadata ever leaves the physical host device**.
```text
[ User Interface ]
(React 19, Tailwind CSS, Motion Animations)
│
▼
[ System Context & State ]
(Global Theme, Connection Monitor, Logs)
│
▼
[ Repository Abstraction ]
(Typed Queries, Filters, Pagination, Transactions)
│
▼
[ Local Database Engine ]
(InMemory Sandbox, LocalStorage Mirror)
│
▼
[ Clinical Decision Support Engine ]
(WHO Guidelines, CVD Risk Estimators)
│
┌────────────────┴────────────────┐
▼ ▼
[ Ask AfriPulse AI ] [ Dynamic Reports ]
(Llama 3.2 3B Offline LLM) (CSV …