Ahmessharf/Nabd-AI

Nabd AI is a secure medical RAG system for Egyptian healthcare that translates colloquial Arabic symptoms into clinical queries, retrieves evidence-based medical knowledge, and uses AI agents for diagnosis support, appointment booking, and hospital workflow automation with full privacy (zero-egress). # Nabd-AI: Medical-RAG Smart Hospital System Nabd-AI is a secure, production-ready **Medical Retrieval-Augmented Generation (RAG)** system tailored for the Egyptian healthcare sector. It translates colloquial Arabic symptoms into precise clinical queries, retrieves evidence-based medical knowledge, and utilizes intelligent AI agents for diagnosis support, appointment booking, and hospital workflow automation. Built with **Clean Architecture** principles, Nabd-AI enforces zero-egress data privacy to secure sensitive patient health information. --- ## 🚀 Key Features - **Colloquial Arabic Translation**: Translates Egyptian dialect symptom descriptions into formal medical terminology and structured clinical queries. - **Multimodal Document Ingestion**: Employs **Docling (VLM)**, **EasyOCR**, and **Pillow** to dynamically parse structured JSON/CSV data and unstructured PDF medical reports. - **Clean Architecture & Decoupled Design**: A completely modular design separating API routes, business controllers, vector database adapters, and LLM providers. - **Hybrid Storage & Vector Search**: Utilizes **FAISS** vector database with **Sentence Transformers (`all-MiniLM-L6-v2`)** for ultra-fast and semantic context retrieval. - **Production-Ready FastAPI Server**: Decoupled REST endpoints allowing external frontends (e.g., React, Flutter) to integrate seamlessly. - **Automated RAG Evaluation**: Incorporates the **RAGAS framework** to continuously measure RAG quality using metrics like *Faithfulness*, *Context Recall*, *Context Precision*, and *Answer Correctness*. --- ## 📂 Project Architecture ``` d:/NABD/ ├── Data/ # Ingested datasets & source documents ├── src/ # Core application code │ ├── assets/ # Static assets and archived files │ ├── controllers/ # Ingest, process, and chat orchestration logic │ │ ├── ChatController.py │ │ ├── DataController.py │ │ └── ProcessController.py │ ├ …