ahgraber/AIMLbling-about:用 Hugo 搭建的 LLM 与生成式 AI 博客
ahgraber/AIMLbling-about
ahgraber/AIMLbling-about 是一个用 Hugo 和 Hextra 主题搭建的博客,记录作者对 LLM、生成式 AI 及 arxiv.org 论文的思考。
AI/MLbling About
In which I organize my thoughts about LLMs, Generative AI, cool papers from arxiv.org, and other marginalia on topics in data science and machine learning.
Quickstart
This blog is built with Hugo using the Hextra theme.
hugo and its prerequisites git, go and dart-sass are required.
First-time setup
Hugo fetches private modules (e.g. ai-treadmill) via HTTPS.
GitHub credentials must be available to git for this to work.
Authenticate with the GitHub CLI and store credentials in the system keychain:
gh auth login --git-protocol https --web
I also use just to define standard tasks (see ./justfile and ./hugo/justfile),
such as automatically creating a new branch per post for drafting (just hugo new "<title>"),
and publishing posts when ready (just hugo publish "<relative path to index.md>").
Drafting new posts
Drafts are managed in their own blog/<title> branch.
Content is written using "standard" markdown (Goldmark, specifically). The Hextra theme provides additional features, like shortcodes for callouts, cards, hide/show details, etc. Footnotes are supported, as is LaTeX, Mermaid diagrams, and syntax highlighting.
Once a draft is finished, it can be published (task hugo:publish path="<relative path to index.md>") on its branch,
then PR'd into main.
Deployment
Once the pull request is complete, a Github Action will build a new multi-arch container image and push to ghcr.io/ahgraber/aimlbling-about.
Versioning is datetime-based (YYYY.MM.DDHHMM) since semver doesn't really make sense (no sense of major.minor.patch for a blog).
Hosting
It is hosted on my kubernetes (k3s) homelab. A Renovate workflow in my homelab repo watches for new image versions. Once detected, it will automatically create a PR to deploy the new blog image to prod.
Integrations
[ ] giscus - a comments system powered by GitHub Discussions
[x] umami analytics - a simple, fast, privacy-focused alternative to Google Analytics
来源:GitHub · github.com