I built an open‑source tool to turn your random home GPUs & Ollama into an OpenAI‑compatible cloud (no port forwarding)

Description and use

Viento turns local GPU‑based LLMs into remote OpenAI‑compatible endpoints via an outbound WebSocket tunnel, eliminating port forwarding and cloud GPU costs.

Problem

Developers with home GPUs running Ollama or vLLM struggled to access them remotely from laptops or cloud‑hosted projects due to complex port forwarding, unreliable tunnels, or costly cloud GPU rentals.

Outcome

Viento enables developers to expose local GPU‑based LLMs as OpenAI‑compatible endpoints without port forwarding, reducing setup complexity, improving security, and eliminating cloud GPU costs.

Workflow

• Install Viento via pip: pip install viento
• Run the Viento node daemon on the local machine
• Configure your application to use the Viento endpoint (e.g., https://viento.onrender.com/v1/chat/completions)
• The Viento gateway forwards requests over an outbound WebSocket to the local engine
• Tokens are streamed back via SSE, with automatic backpressure and queueing handled by the gateway

Evidence

Viento runs as a tiny background node daemon on your machine.

Source

It opens an outbound WebSocket tunnel (wss://) to a lightweight cloud gateway mesh.

Source

Whenever your app makes a standard OpenAI-format API call, the gateway routes the job down through the websocket to your local engine.

Source