For data safety, offline access, and integration with a local knowledge base, I needed to bulk-export Yuque notes to local Markdown. I ended up using yuque-dl to pull hundreds of documents plus images in one pass, and along the way contributed a "download all knowledge bases at once" feature back to the tool via a PR.

The blog was originally hosted on GitHub Pages, accessed through Cloudflare CDN. This migration moves build output directly to Cloudflare Pages, eliminating the GitHub Pages middle layer for faster edge delivery and better cache control. A blue-green deployment strategy ensures zero-downtime migration.

This round I did three things to the blog search box: switched from the default search: true to the slimsearch plugin, reined in the number of search results, and fixed an accidental navigation on Enter with a Pinyin IME. Below is each issue, its root cause, the fix, and the pitfalls I hit.

The blog domain is proxied through Cloudflare with GitHub Pages as the origin. To improve static asset delivery and reduce redundant origin fetches, a Cache Rule was configured in the Cloudflare Dashboard to cache build artifacts, images, icons, and fonts at the edge.

This blog mainly serves two types of content: one is technical tinkering and troubleshooting records, and the other is life, photography, and personal interests. Many articles are not standard tutorials but post-mortems I put together after encountering problems in real environments.