what is -rag- fr?

This project is a local sales management system built around a single real estate development called Aurelia Skyline, using a technique called retrieval augmented generation, or RAG, to let an AI assistant answer questions using the development's actual property documents rather than guessing. The frontend is built with React and Vite, the backend uses FastAPI, and data is stored locally in SQLite. The core RAG logic, which handles parsing documents, splitting them into chunks, turning those chunks into searchable vectors, and generating AI answers, lives in its own module. The system is designed around three types of users. Administrators manage user accounts, upload property information, and maintain the knowledge base the AI draws from. Sales staff manage customer leads, log follow up notes, match customers to suitable units, and get AI assistance while working. Customers can browse available units, ask an AI advisor questions about the development, and submit a purchase interest form. According to the README, login with role based routing, the admin dashboard, knowledge base uploads, user management, property listing uploads, lead management, follow up tracking, the customer facing browsing and AI chat experience, and the core backend API are already working. Vector search currently uses Alibaba Cloud's DashScope embedding service. Several things remain unfinished: fully editable structured property records after upload, a review and feedback loop for AI generated answers, and proper production grade login sessions, since authentication currently relies mainly on a simple user ID header rather than secure tokens. Demo accounts for each role are provided so the system can be tried immediately. Setup requires running the FastAPI backend and the Vite frontend separately, and configuring API keys for DeepSeek and DashScope in an environment file, though the embedding step can fall back to a local, offline method if no DashScope key is available. The README also warns that once DashScope embedding is enabled, uploaded documents are sent to Alibaba Cloud for processing, so only non sensitive files should be uploaded. Written in TypeScript, this project would mainly interest developers building AI assisted sales tools, particularly for real estate.