AI Hospital Management is full-stack, production-ready software. The backend is built with Python FastAPI and the frontend with React 18 and TypeScript. The data layer runs on PostgreSQL 16, Redis 7, and ChromaDB, and the entire stack ships with a Docker Compose configuration and a one-command setup script so you can get a working environment running quickly. It is designed for developers and technical teams who need a substantial, working codebase for a hospital or clinical management system without building one from scratch.
The software covers ten AI and ML capabilities: an agentic triage system that produces ESI 1–5 acuity scores with routing recommendations; a differential diagnosis engine with ICD-10 codes, confidence scores, and chain-of-thought reasoning augmented by RAG over clinical guidelines; a clinical SOAP note generator that converts text or voice transcriptions into structured notes; a drug interaction checker covering 50+ common interaction pairs with severity classification using a hybrid rule-based and LLM approach; a 30-day readmission risk predictor using Gradient Boosted Trees; an AI appointment optimizer using constraint satisfaction and LLM reasoning; a real-time vital signs anomaly detector using Isolation Forest with WebSocket alerts; patient similarity embeddings stored in ChromaDB for clinical decision support; a medical RAG knowledge base over ICD-10 codes, a drug database, and clinical guidelines; and a predictive billing module that estimates insurance approval probability from CPT codes. The LLM layer uses GPT-4o via the OpenAI API, which requires a valid OpenAI API key to operate AI features.
Beyond the AI layer, the software includes patient registration and management, appointment scheduling, bed management, pharmacy and drug inventory, laboratory order tracking, prescriptions, billing and invoicing, medical records, and doctor management. Three role-based access levels are supported out of the box: Admin, Doctor, and Nurse, all secured with JWT authentication and refresh tokens. A pytest test suite covering AI endpoints, appointments, and patient flows is included. Demo seed data covers 8 doctors, 50 patients, 100 appointments, 60 medical records, 40 prescriptions, 30 lab orders, 20 invoices, and a complete drug inventory.
Training and reference data used in this software is synthetically generated. The readmission risk model is trained on synthetic MIMIC-III-style data, and the triage and diagnosis training datasets are synthetically produced. Synthetic data is used deliberately to improve class coverage and to ensure no real patient information is embedded in or distributed with the codebase.
Requirements: Docker and Docker Compose v24+ are recommended for deployment. For manual setup, Python 3.11+, Node.js 20+, PostgreSQL 16, Redis 7, and ChromaDB are required. An OpenAI API key is required for all LLM-powered AI features.
