Installation

Up and running
in minutes.

Install the latest release with a single command, or build from source. You'll need a local LLM server and a model.

★

Quick install (one-liner)

The fastest way to get the latest release binary — no build step required.

Linux / macOS:

curl -fsSL https://mnemosyne-systems.github.io/orangu/install.sh | sh

Windows (Command Prompt):

curl -fsSL https://mnemosyne-systems.github.io/orangu/install.cmd -o install.cmd && install.cmd

Windows (PowerShell):

Invoke-WebRequest -Uri https://mnemosyne-systems.github.io/orangu/install.cmd -OutFile install.cmd; .\install.cmd

Installs the whole stack — orangu, orangu-coordinator, orangu-server, the benchmarking tool orangu-bench, and the model builder orangu-gguf — to ~/.local/bin (Linux/macOS) or %USERPROFILE%\.local\bin (Windows), and warns if it is not on your PATH. Override the location with INSTALL_DIR. The four-page cheat sheet lands in ~/.orangu (override with DOC_DIR), and the complete manual is published as a PDF with every release. Prefer to build from source? Follow the steps below.

01

Install Rust and dependencies

orangu requires Rust and Cargo. On Fedora / RHEL:

dnf install -y git rust cargo

On other distributions, on macOS, and on Windows, install Rust via rustup.rs. Windows builds natively against the MSVC toolchain — no WSL required.

02

Clone and build

git clone https://github.com/mnemosyne-systems/orangu.git
cd orangu
cargo build --release

The release binary will be at target/release/orangu. To install system-wide:

sudo install -Dm755 target/release/orangu /usr/local/bin/orangu
03

Start a local LLM server

The installer ships orangu-server, the project's own pure-Rust GGUF engine. Configure it first, then download a model and serve it on the default API port 8100.

# configure, then fetch a model and serve it
orangu-server -i
orangu-server download unsloth/gemma-4-E2B-it-GGUF
orangu-server --all unsloth/gemma-4-E2B-it-GGUF

The order matters: downloading and serving both read the models directory from the configuration -i writes. That directory defaults to the Hugging Face hub cache (~/.cache/huggingface/hub); orangu-server download fetches models straight into it, and anything already there is found without moving it. orangu also talks to any other OpenAI-compatible server — see Serving models per role for a starting model per role.

04

Configure and launch

Run the interactive setup wizard. It auto-detects your model and walks through every option:

orangu --init

Or copy the sample config and edit it directly:

cp doc/etc/orangu.conf ~/.config/orangu.conf
orangu --config ~/.config/orangu.conf
05

First commands to try

/help # list all commands
/tools # list model-facing tools
/status # git status
/list_files # browse workspace
/model # switch models
/session # list sessions
Optional tools

Enhance your experience.

delta

Syntax-highlighting pager for /diff. Set it as your Git pager and orangu pipes through it automatically.

bat

A cat replacement with syntax highlighting, used automatically by /show_file when installed.

gh / glab

Required for /pull_request, /comment, and /close on GitHub and GitLab.