RobotRanger

Muse Concierge is a voice agent that answers hotel guest calls, handles requests from start to finish, and sends a delivery robot to the door, so front desk staff can stay with the guests in front of them. Hotel delivery robots already save the walk, but every request still goes through one person on the phone. With Muse Concierge, a guest simply calls and speaks naturally, and the call ends in one of three ways: - Delivery: for towels or an Uber Eats order, the agent places one order and gives an ETA, then a robot brings it to the room. - Answers: questions about opening hours, Wi-Fi, breakfast, parking or late check-out fees are answered from a seeded hotel knowledge base, with times and prices quoted from it rather than from the model's memory. - Escalation: when a request needs a person, such as arranging a late check-out, a complaint or a billing issue, the agent raises a ticket for the front desk, which staff track and resolve in the console. It promises a follow-up, not the outcome. It uses four Meta models. Muse Voice Transcribe hears the guest, with a median of 538 ms after they stop speaking. Muse Spark 1.1 holds the conversation and acts through tool calls: placing orders, checking delivery status, looking up hotel information and raising escalations. Muse Wonder 1.0 runs the fleet: it assigns each order and re-plans when a robot fails. In our demo we inject a battery fault mid-delivery, and Wonder chooses how to recover from options the engine has calculated, so it can't invent a route or an ETA. If a model is slow or wrong, a fallback rule keeps the robots moving. At the door, SAM 3.1 and Spark confirm the room number, and SAM confirms the cabin is empty before a delivery counts as done. Staff get a live console with calls, fleet status, escalation tickets, inventory and a stock ledger. The robots, hotel floor and guest are simulated in MuJoCo in the browser. Live Demo Link: https://muse-concierge.vercel.app/