Connect a host installation

Run rho directly on a workstation or server and connect it to an existing Nexus — the stack's Nexus on another machine, or a standalone deployment.

Install rho on the host

Clone the public repository and run the host installer. It installs a portable Ruby and the packaged gems into a user-owned prefix and prints launcher and PATH instructions. Linux with glibc and macOS are supported on x64 and arm64.

git clone https://github.com/jasl/cybros.ai.git
cd cybros.ai
bash install/install.sh

The default full profile includes the browser page. The dev profile adds Playwright and Chromium for the model's optional browser tools. The installer prints prerequisites it does not install, such as a compiler and download tools.

Pair and configure

In an interactive terminal, replacing the example address:

rho setup --nexus-url https://nexus.example

The wizard lets you create an account in Nexus if needed, approve the device code, configure Telegram, optionally configure a provider through the same human OAuth login, and choose a saved default model. rho saves this terminal login for later setup commands. rho setup is available in full and agent modes; a runner-only installation uses rho connect instead and hosts no conversation UI.

Full mode pairs a bundled Runner reserved for that rho instance, including its eligible Bots and Subagent work. A runner-only installation can instead serve several Agents; its manager approves the sharing scope during pairing.

Run it

Start the foreground server in one terminal and use another under the same rho home. A running service applies saved settings immediately.

# Terminal 1
rho server

# Terminal 2
rho status
rho models --workload text_generation
rho console --open

rho console --open opens rho's public URL. Choose Connect to Nexus or Use a device code to establish the browser session, then create a conversation with an available model and runner. The runner uses its default working directory.

For a quick noninteractive text check with a saved default model:

rho run "Say hello"

This is a normal, potentially billed model request.

Keep it running

rho server stays in the foreground. For a persistent service, let your deployment's service manager own start, stop and restart. Never start a second daemon on the same rho home.

Full details — modes, binding to other addresses, transport assertions, environment variables and the separate runner container — are in Deployment layouts.