Prokki 🦖

Prokki is a lightweight, blazing-fast reverse proxy cache for Python package indexes (like PyPI). It acts as an middleman between your Python package installer (e.g., pip, uv or poetry) and the public index, reducing bandwidth usage, improving install speeds, and enhancing reliability in CI/CD pipelines.

Usage

• Create a configuration file, config.toml for example:

host = "0.0.0.0"
port = 8080

cache = "index-cache"

log.severity = "Info"

[[index]]
name = "pypi"
url = "https://pypi.org/simple"

[[index]]
name = "torch-cu118"
url = "https://download.pytorch.org/whl/cu118"

• Create a volume for packages (you can also mount a folder instead):

docker volume create prokki-packages-cache

• Start Prokki:

docker run \
    --detach \
    --name prokki \
    --publish 5000:8080 \
    --volume prokki-packages-cache:/index-cache \
    --volume $(pwd)/config.toml:/config.toml \
    ghcr.io/senysenyseny16/prokki

By default it:

• listens on port 8080 (in this example, it's mapped to port 5000)

You can view available options using the following command:

docker run -it --rm ghcr.io/senysenyseny16/prokki --help

• Specify it as the index for your package manager; in this example, uv is used:

uv pip install torch --index http://<host>:<port>/<index>