abhilashreddychitiki/memory-concierge

Memory Concierge is an AI-powered hotel concierge prototype built around the idea that returning guests should never have to repeat their preferences. It was created as a demonstration for a fictional hotel called The Aurelian Vale. Staff use a dashboard to select an arriving guest, and the app automatically generates a personalized welcome, room preparation details, a tomorrow morning brief, and a guest-facing message, all informed by stored guest history. The app also simulates real-time events such as flight delays, dietary changes, or quiet-room requests. When an event is selected, the system generates an updated stay plan and a message for the guest. A voice button converts the welcome message to audio using ElevenLabs text-to-speech. Under the hood, the app is built with Next.js 14 using the App Router, written in TypeScript, and styled with Tailwind CSS and Framer Motion for animations. The AI responses come from NVIDIA NIM running a large language model called meta/llama-3.1-70b-instruct. Rather than asking the model for a free-form paragraph, the app requests specific structured JSON fields, welcome message, room status, dinner note, tomorrow brief, and concierge note, which makes the output easier to display reliably in the dashboard. Guest profiles are stored in a local JSON file. The project requires API keys for NVIDIA NIM and ElevenLabs to run. It is licensed under MIT.