SehatSathi: A Hybrid Edge–AI Healthcare Assistant for Offline-First Community-Level Medical Support

More than 60 percent of the population of Pakistan lives in rural or under-resourced areas, where timely access to effective healthcare is a significant challenge. This paper presents SehatSathi, a hybrid edge-AI integrated healthcare assistant designed with an offline-first architecture, supporting basic symptom triage, healthcare advice, and facilitating local doctor-patient appointment requests. The proposed system conducts symptom processing and disease prediction locally using multiple machine learning models for continuous operations in a low-connectivity environment. All the user interactions, including appointment requests, are persisted locally on the device and automatically synchronized when network access becomes available. The system supports Urdu, Sindhi, and English. A conversational chatbot powered by LLaMA can also be integrated through an API. Component-based modularity ensures that the system architecture preserves privacy, explainability, and graceful degradation when network connectivity is unavailable. The empirical assessment focuses on offline predictive performance, system responsiveness, and operational availability under low-connectivity conditions. Results show stable predictive performance and reliable usability even without internet connectivity.