lil-shan/Koode

# Koode + Parayu AI-Powered Hospital Intake System (Edge AI for Public Healthcare, Kerala) Source document: koode_parayu.docx --- ## Overview Koode + Parayu is a fully offline, edge AI-based hospital intake and navigation system designed for high-volume government hospitals. - Parayu → AI-powered symptom intake kiosk (voice-first, Malayalam supported) - Koode Bot → ESP32-based robot for patient navigation The system replaces manual reception workflows with structured AI-driven triage. --- ## Problem Statement Government hospitals face severe intake bottlenecks: - Overloaded reception staff handling hundreds of patients - No structured triage → incorrect department routing - No symptom history passed to doctors - Elderly patients struggle with navigation - No epidemic pattern detection This is primarily a workflow inefficiency problem. --- ## Solution Architecture ### Components | Component | Function | |-----------------|----------------------------------| | Parayu Kiosk | AI-based symptom intake | | Koode Bot | Physical patient navigation | | Doctor Dashboard| Pre-consult clinical summaries | | SQLite DB | Patient records + analytics | | MQTT Broker | Communication layer | --- ## Key Features - Fully offline operation (no cloud dependency) - Malayalam, Hindi, English voice intake - Structured AI-driven clinical interview - Automatic department assignment - Real-time doctor dashboard - ESP32 robot navigation via MQTT - Symptom aggregation for epidemic detection - Privacy-first architecture --- ## AI Stack ### Models | Model | Role | |--------------------------------|------------------------------| | Gemma 4 E2B (Q4_K_M) | Clinical reasoning + reports | | Whisper (faster-whisper small) | Speech-to-text | | RAG System | Context injection | ### Why Gemma 4 - N …