Rho stands for R-centered Human–AI Orchestration: an agent-native desktop workbench for R. It combines a persistent R workspace, project-aware code editing, scientific outputs, and an AI collaborator in one application. The user remains in control: editor, Console, and approved Agent actions all work with the same live Workspace R session.
- Project-aware R editing with a Monaco editor, multiple documents, a real file tree, source execution, and project/session restoration.
- One persistent Workspace R powered by Ark, shared by manual Console work, editor execution, and approved Agent actions.
- Scientific output surfaces for Console output, Environment objects, plots, Problems, and durable run history with provenance.
- Ask, Plan, and Act modes for explanation, planning, and reviewed actions against the current project and R session.
- Provider-first model settings with model discovery, visible capability evidence, explicit capability routing, optional Base URL overrides, and API keys kept in the operating system credential store.
- Reviewable file changes so Agent-proposed project edits can be inspected before they are applied.
- Resizable, persistent workspace layout for Files, editor, Agent, Environment, Console, Plots, and Problems.
- Local-first runtime with no Python, Jupyter Server, JupyterLab, or Electron dependency.
Rho currently has development builds for Windows x64 and Apple Silicon macOS. It requires:
- Windows 10/11 with Microsoft Edge WebView2 Runtime, or Apple Silicon macOS 14 or later;
- R 4.4 or later;
aisdk1.5.0 or later and a configured model only for Agent features; the pinnedaisdk.providerspackage is additionally required when using DeepSeek, Moonshot, Kimi Code, Stepfun, Volcengine, AiHubMix, xAI, OpenRouter, Bailian, or NVIDIA.
Unsigned local builds are for development review only. See the Windows prototype guide and the macOS support specification for platform-specific status and prerequisites.
- Launch Rho and open an R project directory.
- Open or create an
.Rfile, then run a selection, the current line, or the complete file in Workspace R. - Inspect results in Console, Environment, Plots, Problems, and Runs.
- Open Model settings, create a Provider connection, import or add a model, then assign that model to the routes you intend to use.
- Use Ask or Plan for read-only help, or Act for actions that require review and approval.
Workspace R is authoritative for project execution and scientific objects. Agent R handles LLM orchestration, while the Rust broker owns transport, approvals, revisions, persistence, and process lifecycle. See the architecture documentation for details, or use the documentation index to browse design, implementation, project, bug-fix, and release documents.
Rho is under active development. Windows packaging, Apple Silicon macOS packaging, and the core project workflow are implemented; installed-candidate acceptance, release signing/publication, macOS x64, and Linux packaging remain in progress.