Install#
ipx is delivered as a self-contained native executable, compiled with Nuitka. Whichever
package manager you use, the only thing you need installed is that package manager - there is no Python, Node or
Rust runtime to manage.
You have |
Install command |
What you get |
|---|---|---|
pip (Python 3.9+) |
|
The pure-Python package; |
npm (Node 14+) |
|
A tiny launcher plus the native binary for your OS and CPU, installed as a platform package. |
cargo (Rust 1.70+) |
|
A Rust launcher with the native binary embedded. |
macOS / Linux shell |
|
The release binary in |
Windows PowerShell |
|
The release binary in |
Nothing |
Download from the GitHub releases |
A single file per platform plus a |
Supported platforms for the native binary: Windows (x64, arm64), macOS (Intel, Apple silicon) and Linux (x64, arm64).
Check the install#
$ ipx --version
ipx 0.1.0
$ ipx doctor
ipx doctor verifies connectivity, your sign-in, the organization context, whether the local Ollama runtime is
running and whether the host-server port is free.
Updating#
Use the same tool you installed with: pip install -U inferenceport-cli, npm update -g @inferenceport/ipx,
cargo install ipx-cli --force, or re-run the install script.
Shell completion#
$ ipx completion bash >> ~/.bashrc
$ ipx completion zsh > ~/.zfunc/_ipx
$ ipx completion fish > ~/.config/fish/completions/ipx.fish
First-run notes#
The first launch of the native binary unpacks itself once into a per-version cache folder (a second or two); later launches start immediately.
Local-model features need the Ollama runtime. If you already use the InferencePort AI desktop app it is provided for you; otherwise install it from https://ollama.com. Cloud features need nothing extra.