Arc-AI is a clean, general-purpose AI chat assistant at ai.rapidcloud.eu. You sign in with your RapidCloud (Entra ID) account and start a conversation immediately — text or voice — with no setup. It runs as an installable progressive web app on desktop and mobile, with a canvas view for working visually alongside the chat. It is the lightweight counterpart to RapidCloud Chat, built for quick questions and everyday assistance rather than heavy multi-agent retrieval.
Ask anything. Stay in the EU.
Arc-AI is a clean, fast AI assistant that lives at one address: ai.rapidcloud.eu. Open it, sign in with your RapidCloud account, and start a conversation — by typing or by talking. No model picker, no API keys, no setup.
It installs as an app on your phone and laptop, switches between chat and a visual canvas view, and answers quickly enough to keep up with a real workday. Voice chat means you can think out loud; canvas means you can think on a surface.
Because it runs on RapidCloud’s EU-hosted platform, your conversations stay on infrastructure you control — not a public AI cloud. It’s the lightweight sibling to RapidCloud Chat: built for speed and everyday questions.
Under the hood▾
How Arc-AI is built
Arc-AI is a React/Vite single-page app, shipped as an installable PWA with offline-aware caching, a canvas view, and Web Speech voice input. The frontend talks to arc-ai-server, an Express + Socket.IO backend that streams responses over WebSockets for low-latency token delivery.
Model calls don’t go straight to a provider. The backend routes every request through RapidCloud’s LiteLLM gateway, which presents a single OpenAI-compatible endpoint over a pool of models and enforces a per-app virtual key. That means models can be swapped, rate-limited, or cost-tracked centrally without touching the app.
Authentication is delegated to the RapidCloud auth portal (Entra ID, restricted to @rapidcloud.eu), and the public hostname is fronted by a Cloudflare Tunnel — no inbound ports exposed. The whole stack runs on RapidCloud’s EU-hosted Proxmox cluster on an isolated VLAN, so prompts, completions, and session data remain on infrastructure you operate, not a third-party AI cloud.
