Installing

To use the VCP CLI tool, you will need:

  • Python version 3.10 or greater

  • Access to a command-line terminal or shell on a MacOS or Linux systems.

  • For the benchmarks commands, you will need to be running on an Intel/AMD64 architecture CPU with NVIDIA GPU, running Linux with NVIDIA drivers.

  • For other commands (e.g. data) you will need a Virtual Cells Platform account (register here)

From PyPi

The Virtual Cells Platform (VCP) CLI is published to PyPi.

We recommend installing the tool into a fresh virtual environment, for example using venv, to create an activate a virtual environment:

python -m venv vcp-cli
source vcp-cli/bin/activate

Core Installation (Minimal)

Install only the core CLI functionality (auth, config, version commands):

pip install vcp-cli

This installs a minimal footprint without heavy dependencies like MLflow or benchmarking tools.

Full Installation

Install all features (model, data, and benchmarks commands):

pip install 'vcp-cli[all]'

Selective Installation

Install only the features you need:

# For model operations only
pip install 'vcp-cli[model]'

# For data operations only
pip install 'vcp-cli[data]'

# For benchmarks only
pip install 'vcp-cli[benchmarks]'

# Combine multiple features
pip install 'vcp-cli[model,data]'

Optional Feature Groups:

  • model - Model operations (init, stage, submit) - requires MLflow, numpy, copier, GitPython

  • data - Data operations (search, download, describe, preview, summary) - requires boto3 type stubs for better IDE support

  • benchmarks - Benchmark operations (run, list, get) - requires cz-benchmarks, anndata

Update Package

To update the package to the latest version, run:

# Core installation
pip install --upgrade vcp-cli

# Full installation with all features
pip install --upgrade 'vcp-cli[all]'