Conversational Robot Barista, Introducing Kubi

Second Space, a 24-hour co-working space in Kaohsiung, Taiwan, has introduced Kubi, a conversational robot barista and reception desk, for automated operations. Kubi leverages ElevenLabs' Conversational AI to provide users with friendly and memorable interactions.

Kubi adopts a sophisticated multi-sensor architecture for human-like interaction. The system is based on a microservices architecture that communicates through real-time event streams, with each service independently managed and scaled.

The key service composition is as follows:

• Environment/Tablet Camera Service: Real-time face and object recognition using PyTorch.
• Payment Service: Payment and receipt reporting processing based on Kotlin JVM.
• Milk & Bean Dispensers/Water Jet: Low-latency precision control using Arduino.
• Receipt Printing: Receipt printing via Node + TypeScript.

The core engine that orchestrates all microservices is a central event-driven core called BigBoy. BigBoy acts as a non-blocking event processor that receives events from all services, selects the appropriate scenario, and schedules actions.

Scenario acts as a non-blocking compiler for the robot's motion events. It automatically generates context-appropriate motion or GIFs through an LLM, and manages the scheduler to ensure voice, facial expressions, and movement are perfectly synchronized.