How huggingface/blog works
Project type and purpose
• This repository is a content-as-code static-site project for the Hugging Face Blog.
• It holds Markdown posts, asset folders (thumbnails), configuration files (_blog.yml, _tags.yml, _redirects.yml), CI workflows under .github/workflows, and helper scripts.
• On each merge, the content is automatically published to the Hugging Face blog platform.
File structure
• Top-level:
– .github/workflows/validate-yaml.yml invokes CI validations.
– scripts/validate-yaml.ts enforces front-matter and config schema.
– assets/… folders contain thumbnails for each post.
– Root Markdown files (*.md) are the blog posts.
– YAML files (_blog.yml, _tags.yml, _redirects.yml) drive the navigation and metadata.
– notebooks/ holds Jupyter notebooks but is primarily content, not code.
• No application code or frontend/backend services live here—everything is content for a static generator.
Key insights from README
• Contributors create a new Markdown file + asset folder, update _blog.yml, then open a PR.
• On merge, the Hub’s static-site pipeline renders and deploys the post.
• Images should live in a separate “documentation-images” dataset to keep the repo lean.
Essential details
• Validation script under CI.
• Config files (_blog.yml, _tags.yml).
• That notebooks and raw Markdown are inputs to the build, not runtime services.
• Make it clear that this is a static pipeline—no live servers except for the CDN.