1. Overview

The piwheels project is designed to automate building of wheels from packages on PyPI for a set of pre-configured ABIs. As the name suggests, it was originally built for Raspberry Pis but there’s nothing particular in the codebase that should limit it to that platform. The system relies on the following components:

Component	Description
piw-master	Coordinates the various build slaves, using the database to store all relevant information, and keeps the web site up to date.
piw-slave	Builds packages on behalf of the piwheels master. Is intended to run on separate machines to the master, partly for performance and partly for security.
piw-monitor	Provides a friendly curses-based UI for interacting with the piwheels master.
piw-sense	Provides a friendly Sense HAT-based UI for interacting with the piwheels master.
piw-initdb	A simple maintenance script for initializing or upgrading the database to the current version.
piw-import	A tool for importing wheels manually into the piwheels database and file-system.
piw-add	A tool for manually adding packages or versions to the database (and therefore, the build queue).
piw-remove	A tool for manually removing builds from the database and file-system.
piw-rebuild	A tool for regenerating certain elements of the piwheels web-site.
piw-logger	A tool for transferring download statistics into the piwheels database.
database server	Currently only PostgreSQL is supported (and frankly that’s all we’re ever likely to support). This provides the master’s data store.
web server	Anything that can serve from a static directory is fine here. We use Apache in production.

Note

At present the master is a monolithic application, but the internal architecture is such that it could, in future, be split into three parts: one that deals exclusively with the database server, one that deals exclusively with the file-system served by the web server, and one that talks to the piwheels slave and monitor processes.