pim-cli 0.2.1

A pip-inspired CLI tool for managing ML models via a Pimfile

pim-cli

pim is a CLI tool to declaratively install and manage machine learning models from a Pimfile.

THIS IS VERY MUCH IN DEVELOPMENT AT THE MOMENT WARNING

🚀 Why Pim?

Modern AI workflows require more than just installing Python packages — they depend on large pretrained models, often fetched from different frameworks (like Hugging Face, TorchVision, or Scikit-learn) with inconsistent interfaces and storage behaviors.

Pim is a lightweight, framework-agnostic CLI tool that solves this by treating models as first-class dependencies — just like packages — with simple, reproducible install workflows.

✅ What Pim Does

• Installs pretrained models from Hugging Face, TorchVision, and Scikit-learn
• Uses a Pimfile (like Pipfile or requirements.txt) to declare which models your project depends on
• Centralizes model downloads in a shared cache (~/.pim) or a user-defined directory
• Supports authentication for gated models (e.g., Hugging Face token access)
• Lets you run pim install from anywhere — no manual Python scripting required

🧠 Why It’s Different

Feature	pip / conda	pim
Installs Python packages	✅	❌
Declarative install file (Pimfile)	✅	✅
Installs pretrained ML models	❌	✅
Unified CLI for multiple frameworks	❌	✅
Model caching & directory control	❌	✅
Handles gated models & auth	❌	✅

Pim doesn’t replace tools like pip or conda — it complements them by handling the growing complexity of working with large-scale pretrained model artifacts in a clean, declarative, and scriptable way.

Installation

pip install pim-cli

Usage

pim install

License

MIT