pypi

A private PyPI repository hosted on GitHub Pages using dumb-pypi.

Quick Start

1. Enable GitHub Pages: See SETUP.md for detailed setup instructions
2. Add packages: Copy your .whl or .tar.gz files to the appropriate container directory under packages/ (e.g., packages/nvcr-io-nvidia-l4t-jetpack-r36-4-0/)
3. Push to main: The CI/CD workflow will automatically build and deploy the indexes
4. Install packages: Use pip install --index-url https://pypi.juno-labs.com/<container-name>/ your-package

Documentation

• SETUP.md - Complete setup guide for first-time configuration
• USAGE.md - Detailed usage instructions and examples
• packages/README.md - Quick reference for adding packages

Overview

This repository automatically generates and publishes PyPI-compatible package indexes for different Docker container builds. Each container directory under packages/ gets its own PyPI index, allowing you to install packages built specifically for your container environment.

How It Works

1. Upload packages: Add your .whl, .tar.gz, or .zip files to a container-specific directory under packages/ (e.g., packages/nvcr-io-nvidia-l4t-jetpack-r36-4-0/)
2. Automatic indexing: GitHub Actions automatically generates separate PyPI indexes for each container using dumb-pypi
3. Publishing: The indexes are deployed to GitHub Pages with a root navigation page
4. Installation: Use pip to install packages from the appropriate container-specific index

Adding Packages

# Copy your package to the appropriate container directory
cp dist/mypackage-1.0.0-py3-none-any.whl packages/nvcr-io-nvidia-l4t-jetpack-r36-4-0/

# Commit and push
git add packages/nvcr-io-nvidia-l4t-jetpack-r36-4-0/mypackage-1.0.0-py3-none-any.whl
git commit -m "Add mypackage 1.0.0 for nvcr-io-nvidia-l4t-jetpack-r36-4-0"
git push

The CI/CD workflow will automatically:

• Scan each container directory under packages/ for package files
• Generate separate indexes for each container
• Create a root navigation page to browse available containers
• Deploy to GitHub Pages

Installing Packages

Once the indexes are published, you can install packages using pip with the container-specific index:

# Install a specific package from a container-specific index
pip install --index-url https://pypi.juno-labs.com/nvcr-io-nvidia-l4t-jetpack-r36-4-0/ mypackage

# Use as an extra index alongside PyPI
pip install --extra-index-url https://pypi.juno-labs.com/nvcr-io-nvidia-l4t-jetpack-r36-4-0/ mypackage

Or add to your requirements.txt:

--extra-index-url https://pypi.juno-labs.com/nvcr-io-nvidia-l4t-jetpack-r36-4-0/
mypackage==1.0.0

Or configure in pip.conf or ~/.pip/pip.conf:

[global]
extra-index-url = https://pypi.juno-labs.com/nvcr-io-nvidia-l4t-jetpack-r36-4-0/

Configuration

The PyPI indexes are built using the following configuration:

• Package lists: Automatically generated from files in each container directory
• Package URLs: Raw GitHub content URLs for package files in each container directory
• Output structure: index/ with subdirectories for each container (deployed to GitHub Pages)
• Container metadata: Read from info.yaml in each container directory

GitHub Pages Setup

To enable GitHub Pages for this repository:

1. Go to repository Settings → Pages
2. Set Source to "GitHub Actions"
3. The workflow will automatically deploy the index

Workflow

The CI/CD workflow (.github/workflows/build-index.yml) triggers on:

• Pushes to the main branch that modify files in packages/
• Manual workflow dispatch

Supported Package Formats

• .whl - Python wheel files
• .tar.gz - Gzipped tar source distributions
• .zip - Zip source distributions

Example

See the example in the problem statement:

$ dumb-pypi \
    --package-list pkg.txt \
    --packages-url https://my-pypi-packages.s3.amazonaws.com/ \
    --output-dir my-built-index

This repository automates this process using GitHub Actions and GitHub Pages.