Install

System Requirements

Archi is deployed using a Python-based CLI onto containers. It requires:

docker version 24+ or podman version 5.4.0+ (for containers)
python 3.10.0+ (for the CLI)

Note: We support either running open-source models locally or connecting to existing APIs. If you plan to run open-source models on your machine's GPUs, see the Advanced Setup & Deployment section.

Installation

Clone the Archi repository:

git clone https://github.com/archi-physics/archi.git

Check out the latest stable tag (recommended for users; stay on main only if you're actively developing):

cd archi
git checkout $(git describe --tags $(git rev-list --tags --max-count=1))

Install Archi (from inside the repository):

pip install -e .

This installs Archi's dependencies and the CLI tool. Verify the installation with:

which archi

The command prints the path to the archi executable.

Helm deployments

Archi also supports rendering and installing a Helm chart through the archi install command. This is useful when you want to deploy archi into a Kubernetes cluster instead of running the compose-based workflow locally. The standard command uses the local helm installation to deploy archi using the created chart. With the dry-run flag the CLI command only creates the chart and required templates.

For a full walkthrough, see the Helm deployment guide.

Show Full Installation Script

# Clone the repository
git clone https://github.com/archi-physics/archi.git
cd archi
export ARCHI_DIR=$(pwd)

# (Optional) Checkout the latest stable tag (recommended for users)
# Skip this if you're developing and want the tip of main.
git checkout $(git describe --tags $(git rev-list --tags --max-count=1))

# (Optional) Create and activate a virtual environment
python3 -m venv archi_venv
source archi_venv/bin/activate

# Install dependencies
cd "$ARCHI_DIR"
pip install -e .

# Verify installation
which archi