Python and pip caching in GitHub Actions

Python projects can have significant dependency install times, especially with packages such as numpy, pandas or tensorflow that require compilation. The cache brings that time down from several minutes to a few seconds.

What you will learn

Enable the pip cache built into setup-python
Cache Poetry and Pipenv depending on your dependency manager
Configure actions/cache for fine-grained control of the paths
Cache a complete virtual environment to skip the install altogether
Avoid the traps: a cache invalidated by a Python upgrade, a double cache

The built-in cache of setup-python

The simplest method uses the cache built into actions/setup-python:

- uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
  with:
    python-version: '3.11'
    cache: 'pip'  # Enables the pip cache

- run: pip install -r requirements.txt

With Poetry

If you manage your dependencies with Poetry, tell setup-python about it: the cache will then key off poetry.lock.

- uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
  with:
    python-version: '3.11'
    cache: 'poetry'

- run: |
    pip install poetry
    poetry install

With Pipenv

The same principle for Pipenv: the cache then follows Pipfile.lock.

- uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
  with:
    python-version: '3.11'
    cache: 'pipenv'

- run: |
    pip install pipenv
    pipenv install --dev

Manual caching with actions/cache

For more control, use actions/cache directly:

- uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
  with:
    python-version: '3.11'

- uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
  with:
    path: ~/.cache/pip
    key: pip-${{ runner.os }}-${{ hashFiles('**/requirements*.txt') }}
    restore-keys: |
      pip-${{ runner.os }}-

- run: pip install -r requirements.txt

The cache paths per manager

If you configure actions/cache by hand, the path: depends on the manager and on the operating system. This table gives the locations to cache.

Manager	Linux path	macOS path	Windows path
pip	`~/.cache/pip`	`~/Library/Caches/pip`	`~\AppData\Local\pip\Cache`
Poetry	`~/.cache/pypoetry`	`~/Library/Caches/pypoetry`	`%APPDATA%\pypoetry\Cache`
Pipenv	`~/.cache/pipenv`	`~/Library/Caches/pipenv`	`%USERPROFILE%\.pipenv`

Caching the virtual environment

For maximum gains, cache the complete virtual environment:

- uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
  with:
    python-version: '3.11'

- name: Cache virtualenv
  uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
  id: cache-venv
  with:
    path: .venv
    key: venv-${{ runner.os }}-${{ hashFiles('**/requirements*.txt') }}

- name: Create venv and install deps
  if: steps.cache-venv.outputs.cache-hit != 'true'
  run: |
    python -m venv .venv
    source .venv/bin/activate
    pip install -r requirements.txt

- name: Run tests
  run: |
    source .venv/bin/activate
    pytest

Gain: the install is skipped entirely when the cache exists.

A complete Python workflow

name: Python CI

on: [push, pull_request]

# No rights by default: the job asks for the minimum
permissions: {}

jobs:
  test:
    runs-on: ubuntu-24.04
    permissions:
      contents: read
    strategy:
      matrix:
        python-version: ['3.10', '3.11', '3.12']

    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false

      - name: Setup Python ${{ matrix.python-version }}
        uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
        with:
          python-version: ${{ matrix.python-version }}
          cache: 'pip'
          cache-dependency-path: |
            requirements.txt
            requirements-dev.txt

      - name: Install dependencies
        run: |
          pip install --upgrade pip
          pip install -r requirements.txt
          pip install -r requirements-dev.txt

      - name: Run linters
        run: |
          ruff check .
          mypy .

      - name: Run tests
        run: pytest --cov=src --cov-report=xml

      - name: Upload coverage
        uses: codecov/codecov-action@fb8b3582c8e4def4969c97caa2f19720cb33a72f # v7.0.0
        with:
          files: ./coverage.xml

Caching with Poetry (advanced)

name: Python CI with Poetry

on: [push, pull_request]

permissions: {}

jobs:
  test:
    runs-on: ubuntu-24.04
    permissions:
      contents: read
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false

      - name: Setup Python
        uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
        with:
          python-version: '3.11'

      - name: Install Poetry
        uses: snok/install-poetry@a783c322200f0519c7926aa6faa857c4e23e9263 # v1.4.2
        with:
          version: '1.8.2'
          virtualenvs-create: true
          virtualenvs-in-project: true

      - name: Cache Poetry virtualenv
        uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
        with:
          path: .venv
          key: poetry-${{ runner.os }}-${{ hashFiles('**/poetry.lock') }}
          restore-keys: |
            poetry-${{ runner.os }}-

      - name: Install dependencies
        run: poetry install --no-interaction

      - name: Run tests
        run: poetry run pytest

The specific optimisations

Packages with native dependencies

For numpy, scipy and pandas, which carry compiled dependencies:

- name: Cache pip wheels
  uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
  with:
    path: |
      ~/.cache/pip
      ~/.local/lib/python*/site-packages
    key: pip-wheels-${{ runner.os }}-${{ hashFiles('**/requirements*.txt') }}

Pre-commit hooks

Pre-commit environments are slow to rebuild. Caching them makes the hook almost instant from the second run onwards.

- name: Cache pre-commit
  uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
  with:
    path: ~/.cache/pre-commit
    key: pre-commit-${{ runner.os }}-${{ hashFiles('.pre-commit-config.yaml') }}

- run: pre-commit run --all-files

The common mistakes

A cache invalidated by a Python upgrade

A pip cache depends on the Python version. If the key does not include it, an upgrade restores packages compiled for the previous one, and everything breaks.

# ❌ The key does not change when Python changes
key: pip-${{ hashFiles('requirements.txt') }}

# ✅ Include the Python version
key: pip-${{ runner.os }}-py${{ matrix.python-version }}-${{ hashFiles('requirements.txt') }}

A conflict with the setup-python cache

Stacking the cache: of setup-python and an actions/cache on the same path produces two managers stepping on each other. Keep only one.

# ❌ A double cache, with unpredictable behaviour
- uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
  with:
    cache: 'pip'
- uses: actions/cache@5a3ec84eff668545956fd18022155c47e93e2684 # v4.2.3
  with:
    path: ~/.cache/pip

# ✅ Use one or the other
- uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
  with:
    cache: 'pip'

Key points

The built-in cache of setup-python (cache: 'pip') is enough for most projects, in one line.
For Poetry or Pipenv, name the right manager: cache: 'poetry', cache: 'pipenv'.
actions/cache gives you control over the paths: useful for compiled wheels or pre-commit hooks.
Caching the complete virtualenv skips the install outright, the maximum gain.
Include the Python version in the cache key, otherwise a Python upgrade restores an incompatible cache.
Never stack the cache: of setup-python and an actions/cache on the same path: pick one.

Next steps

Java cache: the same reasoning about keys and paths applied to Maven and Gradle, useful as soon as a repository mixes several languages.
Concurrency: the complementary lever, which avoids paying for several Python installs on commits that are already obsolete.
Introduction to runners: what explains why the machine starts blank on every run, and therefore why the pip cache exists.