Back to Wave Wiki

Cloud Workstation

From Wave Wiki — the open reference for the Wave OS ecosystem.

1. About

Cloud Workstation is a full desktop environment running on Theta EdgeCloud. It provides an Arch Linux + Wayland desktop accessible directly from the Wave OS browser via WebRTC. Users get a real Linux desktop with Ollama for local AI, Sunshine for game streaming, and full development tools — all running on Theta's decentralized GPU infrastructure. Cloud Workstation bridges the gap between cloud compute and local productivity, enabling users to run desktop-class applications without local hardware.

2. Tech Stack

OS: Arch Linux with the Wayland display server. GPU Compute: Theta EdgeCloud, with multiple GPU types available. Game Streaming: Sunshine (NVIDIA encoder) for low-latency video.

AI: Ollama for local LLM inference (Llama 3, Mistral, and others). Display Protocol: WebRTC for browser-based desktop access, via an HTML5 WebRTC client embedded in a Wave OS iframe.

File Transfer: Harbor integration provides bidirectional file sync. Session Management: the ComputeSession entity tracks the SSH command and access URL per session.

3. Install

Cloud Workstation is a built-in Wave OS application. No installation required — accessible from the Wave OS app grid. Launching a workstation requires sufficient credits for the selected GPU type. Sessions can be started, stopped, and resumed from the Wave OS desktop.

4. Pricing & Credits

Pricing is per-hour based on GPU type, with tfuel_cost tracked per session. Credits are deducted in real-time during active sessions. Idle sessions are automatically stopped after a configurable timeout (default: 15 minutes). Persistent storage is optional and charged per GB per month. Ollama models are free to run (open-weight, no per-token cost). Free tier includes 30 minutes per day with an entry-level GPU.

5. Known Limitations

Cold start takes 2-5 minutes for new workstation sessions. WebRTC requires a stable broadband connection (minimum 10 Mbps for 1080p). There is no GPU passthrough for multi-GPU workloads — one GPU per session. Maximum session duration is 24 hours (auto-restart available). Persistent storage is capped at 100GB per workstation. Arch Linux is a rolling release, so occasional breaking updates require admin intervention. Ollama model sizes are limited by available GPU VRAM.

6. Privacy & Data

Workstation sessions are isolated with no cross-user visibility. ComputeSession metadata (GPU type, duration, cost) is stored in the Wave OS database with row-level security. File contents within the workstation are stored on Theta EdgeCloud, not in the Wave OS database. The WebRTC connection is encrypted end-to-end (DTLS-SRTP). Ollama conversations are stored locally in the workstation and deleted on session termination, unless persistent storage is configured. SSH access uses key-based authentication only — no password access. Session logs (start, stop, duration) are stored in the SessionLog entity for billing audit.

7. Published

Published: Yes, by xBuildy.