Support This Project

All projects made with passion 💙

portapkg

Portable, offline-friendly Python package bundler and installer.

Bundle a package + all its dependencies on an online machine, then install on an air-gapped machine with nothing but Python and pip.

Quick start

Online machine — bundle a package

# Install portapkg
pip install portapkg

# Bundle a package for all 6 platforms + 5 Python versions
portapkg bundle instrumation

# Bundle for specific platforms only
portapkg bundle instrumation --platforms win_amd64,macosx_13_0_arm64

# Bundle for specific platforms and python versions (faster!)
portapkg bundle instrumation --platforms win_amd64 --python-versions 313

# Bundle a snapshot of the current environment (single platform, most reliable)
portapkg bundle instrumation --snapshot

# List all bundles
portapkg list
portapkg list --json    # machine-readable, for scripts / AI agents

# Show bundle details
portapkg info instrumation
portapkg info instrumation --json

# Update / re-fetch a bundle
portapkg update instrumation

# Verify a bundle covers a target platform/Python before shipping it
portapkg verify instrumation --platforms win_amd64 --python-versions 313

Output bundle structure:

bundles/
└── instrumation/
    ├── manifest.json
    └── wheels/
        ├── instrumation-0.3.0-py3-none-any.whl
        ├── pyserial-3.5-py2.py3-none-any.whl
        └── ...

Offline machine — install from USB

Copy the bundles/ folder and portapkg.py to a USB drive, then on the offline machine:

# List available bundles
python portapkg.py list

# Install a bundled package (works without internet)
python portapkg.py install instrumation

# Install to a custom path
python portapkg.py install instrumation --target ./myenv/Lib/site-packages

The portapkg.py file is fully self-contained — stdlib only, no dependencies.

How it works

1. Dependency resolution: pip download <package> resolves the full transitive dependency tree using pip's own resolver. For multi-platform bundles, the resolver also scans each wheel's METADATA for platform-conditional deps (e.g. pefile; sys_platform == "win32") and recursively resolves any missing deps found.
2. Multi-platform fetch: For each dependency, wheels are downloaded across multiple platforms × Python versions using pip download --platform --python-version --only-binary=:all:.
3. Layered fallback strategy:
   • Layer 1: Exact binary wheel for the target platform/Python.
   • Layer 2: Platform-constrained source distribution.
   • Layer 3: Unconstrained source distribution — catches sdist-only packages like Eel.
   • Layer 4: Version bump — queries pip index versions for newer releases with binary wheel coverage (e.g. msgpack gaining cp313 wheels in a later patch).
4. Offline install: pip install --no-index --find-links <wheels_dir> installs from local files only.

Development

Setup

# Clone the repo
git clone https://github.com/your-org/portapkg-py && cd portapkg-py

# Install in editable mode with dev dependencies
pip install -e ".[dev]"

Running tests

# Run all tests
pytest

# Run tests with coverage
pytest --cov=portapkg --cov-report=term-missing

# Run a specific test file
pytest tests/test_platform.py

# Run a specific test
pytest tests/test_platform.py::TestParseWheelFilename

# Run linter
pip install ruff
ruff check portapkg/ tests/ portapkg.py

Project structure

portapkg/
├── portapkg.py                ← standalone offline installer (stdlib only)
├── portapkg/
│   ├── cli.py                 ← entrypoint: bundle/list/info/update
│   ├── bundler/
│   │   ├── fetch.py           ← multi-platform wheel download
│   │   ├── resolver.py        ← dependency resolution (pip download)
│   │   └── manifest.py        ← manifest.json read/write
│   └── installer/
│       ├── install.py         ← offline install logic
│       └── platform.py        ← OS/arch/python detection
├── tests/                     ← pytest test suite
├── pyproject.toml             ← packaging & tool config
└── .github/workflows/         ← CI / release workflows

CI / CD

• CI (.github/workflows/test.yml): runs on push/PR to main — tests across 3 OS × 5 Python versions, plus linting with ruff.
• Release (.github/workflows/publish.yml): manual workflow_dispatch — specify a version tag, and it builds wheels + sdist and creates a GitHub Release with the artifacts attached.

Platform matrix

Python versions: 3.9, 3.10, 3.11, 3.12, 3.13

Error handling

• Missing binary wheel: warns and falls back through layered strategy (source dist → unconstrained sdist → version bump)
• Platform-conditional deps: automatically resolved by scanning wheel METADATA for Requires-Dist lines with sys_platform / platform_system / implementation_name markers, evaluated with packaging.markers.
• No compatible wheel on offline machine: clear human-readable error with suggested fix (--snapshot)
• Missing pip: detected before any install attempt
• Missing bundle: tells you which directory to copy

Configuration

PORTAPKG_BUNDLES_DIR

By default, bundles are stored in ./bundles/ relative to the current working directory. You can override this with the PORTAPKG_BUNDLES_DIR environment variable:

export PORTAPKG_BUNDLES_DIR=/path/to/bundles
portapkg bundle instrumation   # stores in /path/to/bundles/instrumation/

Both the CLI (portapkg) and the standalone script (portapkg.py) respect this environment variable.

Requirements

• Online machine: Python 3.9+, pip (any version)
• Offline machine: Python 3.9+, pip (any version)
• No third-party packages are used by portapkg itself — stdlib only.