All the metadata in the conda-build recipe is specified in the meta.yaml file. It contains info such as the package name, version, source URL, build dependencies, runtime requirements, tests and licensing. It supports Jinja and platform selectors.

Conda build can produce variants of the same package (e.g. different Python versions). This file controls the variables injected to meta.yaml. This is the main mechanism behind ‘conda-forge-pinning’.

Distutils/setuptools script that provides a function to install a Python package with or without compiled extensions.

Accompanying metadata file for setup.py that defines some configuration parameters.

Static file that provides all the information necessary to create an installable Python package. This alleviates the main problem of setup.py: its contents are dynamic and need to be executed to get retrieved.

A file specifying conda packages and what channels to get them from. Users can also add pip packages here. It can be ingested by conda/mamba env and micromamba create. The tool will solve the environment and install the required packages.

Contains the definitions of an already solved conda environment. It can be generated by conda-lock, and consumed by conda install -f.

Similar to conda-lock, Poetry can freeze an existing virtual environment, writing the exact package definitions to this file.

A better way of specifying requirements for a project. Instead of keeping several requirements.txt files around, you can specify different categories in the same file.

When Pipenv solves a Pipfile (finds the dependencies of the specified requirements in Pipfile), the resulting list is “frozen” to this lock file, including exact versions and hashes.

A file containing a list of packages (and some command line options) to be installed by pip.

Conda channels store package metadata for each architecture (linux-64, win-64, etc) in these JSON files. Each package reports here which package they depend on, among other things.

Stripped down repodata.json which only contains the last build of each package, plus their dependencies, recursively. It simplifies the solver stage in some cases.

A channel-wide metadata file that contains an index of all the conda packages present across all architectures in the channel.

Conda packages are distributed as bzip2-compressed tarballs, including the package files and some metadata.

The .conda file format was introduced in conda 4.7 as a more compact, and thus faster, alternative to a tarball. The .conda file format consists of an outer, uncompressed ZIP-format container, with 2 inner compressed .tar files.

A source code distribution available in PyPI, packaged as .tar.gz. Pip will prefer wheels by default.

Platform-specific build of a PyPI package that can contain compiled libraries. Pip will prefer this even if a sdist is present. Extension is .whl.