Contributing Guide#

Dev Installation#

To set up for local development (requires poetry):

$ git clone
$ cd aiohttp-client-cache
$ poetry install -E all -E docs

Pre-commit hooks#

CI jobs will run code style checks, type checks, linting, etc. If you would like to run these same checks locally, you can use pre-commit. This is optional but recommended.

To install pre-commit hooks:

pre-commit install

To manually run checks on all files:

pre-commit run --all-files
# Alternative alias with nox:
nox -e lint

To disable pre-commit hooks:

pre-commit uninstall


Test Layout#

Tests are divided into unit and integration tests:

  • Unit tests can be run without any additional setup, and don’t depend on any external services

  • Integration tests depend on additional services, which are easiest to run using Docker (see Integration Tests section below)

Running Tests#

  • Run pytest to run all tests

  • Run pytest test/unit to run only unit tests

  • Run pytest test/integration to run only integration tests

For CI jobs (including PRs), these tests will be run for each supported python version. You can use nox to do this locally, if needed:

nox -e test

Or to run tests for a specific python version:

nox -e test-3.10

To generate a coverage report:

nox -e cov

See nox --list for a ful list of available commands.

Integration Test Containers#

A live web server and backend databases are required to run integration tests, and docker-compose config is included to make this easier. First, install docker and install docker-compose.

Then, run:

docker-compose up -d
pytest test/integration

To test DragonflyDB you need to stop a Redis container (if running) and run docker compose -f dragonflydb.yaml up. No other changes are required, you can run related tests with e.g. pytest test -k redis.


Sphinx is used to generate documentation.

To build the docs locally:

$ nox -e docs

To preview:

# MacOS:
$ open docs/_build/index.html
# Linux:
$ xdg-open docs/_build/html/index.html


Documentation is automatically built and published by Readthedocs whenever code is merged into the main branch.

Sometimes, there are differences in the Readthedocs build environment that can cause builds to succeed locally but fail remotely. To help debug this, you can use the readthedocs/build container to build the docs. A configured build container is included in docker-compose.yml to simplify this.

Run with:

docker compose up -d --build
docker exec readthedocs make all

Pull Requests#

Here are some general guidelines for submitting a pull request:

  • If the changes are trivial, just briefly explain the changes in the PR description.

  • Otherwise, please submit an issue describing the proposed change prior to submitting a PR.

  • Add unit test coverage for your changes

  • If your changes add or modify user-facing behavior, add documentation describing those changes

  • Submit the PR to be merged into the main branch.

Notes for Maintainers#


  • Releases are built and published to PyPI based on git tags.

  • Milestones will be used to track progress on major and minor releases.

  • GitHub Actions will build and deploy packages to PyPI on tagged commits on the main branch.

Release steps:

  • Update the version in both pyproject.toml and aiohttp_client_cache/

  • Make sure the release notes in are up to date

  • Push a new tag, e.g.: git tag v0.1.0 && git push origin v0.1.0

  • This will trigger a deployment. Verify that this completes successfully and that the new version can be installed from pypi with pip install

  • A readthedocs build will be triggered by the new tag. Verify that this completes successfully.

Downstream builds:

  • We also maintain a Conda package, which is automatically built and published by conda-forge whenever a new release is published to PyPI. The feedstock repo only needs to be updated manually if there are changes to dependencies.

  • For reference: repology lists additional downstream packages maintained by other developers.

Code Layout#

Here is a brief overview of the main classes and modules. See API Reference for more complete documentation.

  • session.CacheMixin, session.CachedSession: A mixin and wrapper class, respectively, for aiohttp.ClientSession. There is little logic here except wrapping ClientSession._request() with caching behavior.

  • response.CachedResponse: A wrapper class built from an aiohttp.ClientResponse, with additional cache-related info. This is what is serialized and persisted to the cache.

  • backends.base.CacheBackend: Most of the caching logic lives here, including saving and retrieving responses. It contains two BaseCache objects for storing responses and redirects, respectively.

  • backends.base.BaseCache: Base class for lower-level storage operations, overridden by individual backends.

  • Other modules under backends.*: Backend implementations that subclass CacheBackend + BaseCache

  • cache_control: Utilities for determining cache expiration and other cache actions

  • cache_keys: Utilities for creating cache keys