How should I interpret PyPI downloads compared with GitHub stars?

Downloads and stars measure very different things. Installing a package can happen automatically because it's a dependency, part of a requirements file, or run by a CI pipeline. Starring a project requires a person to make an intentional click, so 11,000 downloads versus 36 stars isn't unusual and doesn't tell you much about user satisfaction.

A weekday-heavy pattern with lower weekend activity can suggest business or CI usage, but the sample is still too small and noisy to draw strong conclusions. A baseline around 350 downloads per week is encouraging, but the more useful signals are repeat users, issue reports, pull requests, dependency relationships, and people mentioning that they use the package. I'd track the trend over several months rather than interpreting any single peak.

Sharp spikes often come from infrastructure rather than a sudden rush of human users. A new release can cause mirrors to resynchronize every file, and caches may expire at roughly the same time. Scheduled CI jobs or a dependent project rebuilding can have the same effect. If the spikes line up with your releases, mirror synchronization is a strong possibility.

Make sure you're looking at the download count with CI downloads excluded. The link includes CI downloads, so many of those requests may come from build runners reinstalling the package repeatedly. Even after filtering, the result will still include mirrors, caches, and automated tooling, so treat it as evidence that machines are touching the package—not as a count of individual users.