Hugging Face automates weekly releases of huggingface_hub with AI-assisted changelogs and human oversight

Hugging Face has transitioned the huggingface_hub Python client to weekly releases using a fully automated CI pipeline, replacing a manual process that previously shipped every four to six weeks. The new workflow is triggered manually from GitHub Actions and runs a sequence of deterministic and AI-assisted steps in a single file, .github/workflows/release.yml.

The pipeline automates version bumps, tagging, PyPI publishing, downstream test branch creation, and internal announcements. An open-weights model—currently GLM-5.2 from Z.ai—drafts release notes and a Slack announcement, which are then reviewed and edited by a human before publication. The model’s output is cross-checked against a deterministic manifest of PRs included in the release to prevent omissions or fabrications.

Before the model drafts notes, a Python script extracts PR numbers from squash-merge commits and stores them as ground truth. The model’s draft is then validated against this manifest; any discrepancies trigger an automated fix cycle. Only after the changelog matches the manifest exactly does the workflow allow human review and final publication.

The entire stack uses only open-source tools and open-weights models, with no proprietary APIs or vendor-locked components. The workflow is designed to be reusable by other maintainers, aligning with Hugging Face’s goal of providing a transparent, community-friendly process.