Show HN: Hellocafe.ai – open-source voice and chat enabled AI ordering system (www.hellocafe.ai)

Narra Labs published Hellocafe.ai, an open‑source, voice- and chat-enabled café ordering system that runs entirely on self‑hosted AI models. The demo implements a full ordering flow (cart, subtotal, conversational UI) and is deployed on just two Hetzner servers — one CPU node and one CPU+GPU node — showing the stack can operate outside big‑cloud providers. The project is presented as an experiment; it doesn’t actually prepare drinks and Narra Labs invites contact for similar production work. For the AI/ML community this is a compact, reproducible example of conversational commerce built with self‑hosted components. Technically it demonstrates the feasibility of combining speech (voice input/output) and text chat with stateful order management on modest infrastructure, implying practical tradeoffs between model size, latency, and cost (GPU for inference, CPU for orchestration). The open‑source release makes it a useful baseline for privacy‑sensitive deployments, edge/onsite solutions for small businesses, and experimentation with different STT/TTS and LLM choices. It’s a compelling reference for teams wanting to evaluate self‑hosting versus cloud APIs and to prototype integrated voice + chat experiences for point‑of‑sale scenarios.