The Motivation

After years of using fragmented note-taking tools, I decided to build a dedicated space for my technical reflections. My requirements were simple: speed, minimalism, Markdown native, and zero maintenance overhead.

I landed on Hugo paired with the PaperMod theme. Hugo’s Go-based engine provides blazing-fast build times, while PaperMod offers a clean, distraction-free aesthetic.

Architecture Overview

The site is hosted on GitHub and utilizes GitHub Actions for automated building and deployment.

  • Static Site Generator: Hugo (Extended version)
  • Theme: PaperMod (Home-info mode)
  • Deployment: GitHub Pages
  • Automation: GitHub Actions
  • Configuration: YAML (hugo.yaml)

Key Technical Setup

1. Configuration (hugo.yaml)

While many Hugo sites use TOML, I preferred YAML for its readability. I configured PaperMod in home-info mode to serve as a minimalist landing page rather than a heavy profile view.

params:
  homeInfoParams:
    Title: "👋 Welcome to Njx'Log"
    Content: >
        Hi, this is Jiaxing Ni. I share my AI/ML learning notes here.

2. Special Pages & Search

PaperMod supports local search out-of-the-box using Fuse.js. To enable this, I ensured the build output includes JSON:

outputs:
  home:
    - HTML
    - RSS
    - JSON # Required for search

I also set up dedicated layouts for archives and search by creating specific Markdown files in the content/ directory with layout parameters.

3. CI/CD Workflow

The deployment is fully automated. Every push to the main branch triggers a workflow in .github/workflows/hugo.yaml. It sets up the Hugo environment, builds the static assets, and uploads them to GitHub Pages.

- name: Setup Hugo
  uses: peaceiris/actions-hugo@v3
  with:
    hugo-version: '0.153.0'
    extended: true # Necessary for SCSS compilation

My Content Workflow

To maintain consistency, I follow these self-imposed rules:

  1. Creation: Use hugo new posts/your-post.md to scaffold new content.
  2. Timezone: Always use +08:00 in frontmatter.
  3. Assets: Store images in static/ and link them via relative paths.

Conclusion

“The faintest ink is better than the best memory.”

This blog is more than just a repository of notes; it’s a laboratory for my engineering thoughts. Moving forward, I’ll be sharing more deep dives into LLM fine-tuning, RAG pipelines, and MLOps.

Stay tuned for more updates.