chouaibai/patient-assistance-chatbot-api-v1

Arabic/Darija Patient Chatbot API # Arabic/Darija Patient Assistance Chatbot API A local FastAPI MVP that structures medication-related requests written in Arabic, Moroccan Darija, French, or mixed language. The project is intentionally simple: it does not train a machine learning model, does not use a database, and does not require a paid API. It uses clear rule-based NLP plus RapidFuzz medicine matching so the behavior is easy to explain during a Master IT demo. ## Problem Context In Morocco, patients often describe pharmacy needs using a mix of Moroccan Darija, Arabic, and French. A useful assistant should understand messages such as `bghit 2 boites doliprane`, `avez-vous amoxicilline 500mg ?`, or `بغيت دواء للسخانة` and convert them into structured information. This project focuses only on request understanding. It does not diagnose, prescribe, recommend dosage, or replace a healthcare professional. ## Objectives - Accept Arabic, Darija, French, and mixed-language text. - Detect the request intent. - Extract medicine name, quantity, unit, dosage, medicine form, and simple symptom when available. - Match medicine names using local aliases and fuzzy matching. - Return clean JSON for a future chatbot, pharmacy interface, or mobile app. - Keep the solution local, explainable, testable, and easy to present. ## Features - FastAPI backend with automatic Swagger documentation. - Rule-based normalization, language detection, intent detection, and entity extraction. - RapidFuzz medicine matching from `data/medicines.json`. - Single-text and batch analysis endpoints. - Optional local voice endpoint using Whisper when voice dependencies are installed. - Streamlit demo interface for a quick visual presentation. - Pytest coverage for API endpoints and NLP examples. - Documentation for architecture, API examples, and evaluation. ## Architecture ```mermaid flowchart LR A["User text"] --> B["Normalize text"] B --> C["Detect language"] B --> D["Detect intent"] B --> E["Extract entities"] B --> F["Match …