Skip to main content

Installation

This page gets the pie CLI onto your machine. After this, see Initial setup to create ~/.pie/config.toml and run a prompt.

Supported platforms​

PlatformBuilds
Linux x86_64portable, cuda12.6, cuda12.8, cuda13.0, and portable-cuda*
Linux aarch64portable
macOS Apple Siliconportable

Windows native and Intel Macs are not supported. On Windows, use WSL and the Linux build.

portable includes the ggml-backed embedded driver and is the fallback everywhere. The CUDA flavors are Linux x86_64 only and include the embedded cuda_native driver.

Install the CLI​

curl -fsSL https://pie-project.org/install.sh | bash

The installer downloads a release tarball, installs pie into ~/.local/bin by default, and auto-picks a flavor:

  • NVIDIA driver >= 580: cuda13.0
  • NVIDIA driver >= 525: cuda12.8
  • otherwise: portable

Override the flavor or install directory with environment variables:

curl -fsSL https://pie-project.org/install.sh | PIE_FLAVOR=cuda12.8 bash
curl -fsSL https://pie-project.org/install.sh | PIE_FLAVOR=portable bash
curl -fsSL https://pie-project.org/install.sh | PIE_INSTALL_DIR=/usr/local/bin bash

Valid PIE_FLAVOR values are portable, cuda12.6, cuda12.8, cuda13.0, portable-cuda12.6, portable-cuda12.8, and portable-cuda13.0.

To install a specific release tag:

curl -fsSL https://pie-project.org/install.sh | PIE_VERSION=v0.1.2 bash

If pie is not on your PATH, add the install directory:

echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc

Verify​

pie --version
pie doctor
pie driver list

pie doctor reports platform readiness, compiled-in embedded drivers, GPU visibility, config status, and subprocess-driver venv resolution.

Optional Python drivers​

The pie binary can run embedded drivers without Python. Python is only needed for subprocess drivers: dev, vllm, and sglang.

Install uv, then let pie driver print or run the matching venv recipe:

pie driver dev install ~/.pie/venvs/dev --run
pie driver dev set venv ~/.pie/venvs/dev
pie driver dev doctor

Use the same flow for vLLM or SGLang:

pie driver vllm install ~/.pie/venvs/vllm --run
pie driver vllm set venv ~/.pie/venvs/vllm

pie driver sglang install ~/.pie/venvs/sglang --run
pie driver sglang set venv ~/.pie/venvs/sglang

The CLI creates Python 3.12 virtual environments and installs the matching wheels (pie-driver-dev[cu128], pie-driver-vllm, or pie-driver-sglang). A model can also override the interpreter with venv or python under [model.driver.options].

Install from source​

For development:

git clone https://github.com/pie-project/pie.git
cd pie

# Default features: embedded portable driver plus always-linked dummy.
cargo install --path server

# Include the embedded CUDA driver too.
CUDACXX=/usr/local/cuda/bin/nvcc \
  cargo install --path server --features driver-portable,driver-cuda

The installed CLI is still pie. Re-run pie doctor after building to confirm which embedded drivers are compiled in.

Next​