On how this site was built

Table of Contents


I decided on building this website with a key principle in mind: simplicity. I wanted a platform that was easy to manage and expand upon in the future. My initial requirements focused on:

And specifically for the resume I wanted:


As a broke university student I don’t like paying for hosting, or paying for electricity to host on my own. Therefore, I opted for the best completely free service which absolutely is Oracle Cloud Infrastructure and their always free tier deal where you get 2x AMD, 4x ARM Ampere A1 and 24GB RAM to allocate to VMs as you see fit. I already had an account made in the UK Region but as it seems I couldn’t manage to spawn any instance as there is a huge shortage at the moment, most likely all the people eager to host Minecraft servers.

So, to cut the story short I had to make a script which I ran for a couple of days to spawn a single core AMD instance which I then used to run the same script this time trying to spawn a 4x ARM VM. After a whole two weeks I had no luck and I started looking into alternatives. Turns out if you add in a credit card and upgrade to their pay as you go plan you have priority to these free resources, and, most importantly, they are still free. After a few more problems with them declining all my credit and debit cards from multiple different banks I had a long chain of emails with their support, finally got my account upgraded to pay as you go, and managed to get an ARM instance.

Static Site Generator

Looked around a bit and I decided to go with Zola to generate static pages from markdown as it seems to be very well liked by the community and as a bonus it’s also written in Rust which resonated with my preferences. For the themes I really liked Duckqill and I just changed a few settings around.

After configuring things and making slight changes to how it looks I now can just run zola build --output-dir /var/www/licu.dev/public --force to have my site’s new version built.


My biggest problem with regularly updating my resume is that I always have to shift things around to have it fit on a single page, and when I decide to try a new look I have to make changes in multiple places. Another problem is up until now I was not using any sort of version control. I know I could do all these things with LaTeX but honestly I was never a big fan of it and instead I took this opportunity to use Typst. I wanted to use a standard for the data layer and jsonresume seems to be the only somewhat used one. I really tried to use it but unfortunately it does not provide all the sections that I consider important and neither the exact format I saw to make logical sense for me, so I just drew inspiration from it. For extra readability I also chose to keep my data in YAML format.

I included the command to compile the .typ files to PDF (so I can display them on my site as an iframe) to a script alongside the zola build command typst compile --font-path resume/ resume/main.typ static/resume/Resume.pdf

Web server

While I had experience with Nginx, I opted for Caddy for this project due to its ease of use and automated certificate management. Nginx, while powerful, requires separate configuration for features like Let’s Encrypt integration, which I find cumbersome. Caddy streamlines this process by handling it automatically, saving me time and effort. I also had a weird preference where I wanted specific paths of the website to be represented by a personal subdomain such as blog.licu.dev instead of licu.dev/blog. Turns out this was not that hard to configure. In my Caddyfile, I’ve defined separate sections for licu.dev, *.licu.dev (wildcard for all subdomains), and individual subdomains like blog.licu.dev and resume.licu.dev. At the moment this is the whole Caddyfile which is incredibly short and easy to read and modify.

    email [email protected]
    acme_dns  cloudflare {env.CLOUDFLARE_API_TOKEN}

(default) {
  header X-Clacks-Overhead "GNU Terry Pratchett"
  encode zstd gzip

www.licu.dev {
    redir https://licu.dev{uri} permanent

*.licu.dev, licu.dev {
    import default

    root * /var/www/licu.dev/public

    handle_path /blog* {
        redir * https://blog.licu.dev{uri} permanent

    handle_path /resume* {
        redir * https://resume.licu.dev{uri} permanent


blog.licu.dev {
    root * /var/www/licu.dev/public/blog


resume.licu.dev {
    root * /var/www/licu.dev/public/resume



While I considered using Docker Swarm or Kubernetes for better orchestration, I opted for a simpler approach using Docker Compose. I created Dockerfiles for the services and hosted them on GitHub Container Registry (GHCR). I then used a Watchtower container to periodically check and apply updates. For services functioning as “jobs” (like building the new site version), I made them sleep indefinitely to prevent Watchtower from attempting to update or start non-running containers.

Here’s the GitHub Actions CI/CD workflow I use for the personal-site:

name: Create and publish a Docker image

on: push

  REGISTRY: ghcr.io
  IMAGE_NAME: ${{ github.repository }}
  DOCKER_TARGET_PLATFORM: linux/arm64/v8

    runs-on: ubuntu-latest
      contents: read
      packages: write

      - name: Checkout repository
        uses: actions/checkout@v3

      - name: Log in to the Container registry
        uses: docker/login-action@v3
          registry: ${{ env.REGISTRY }}
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}

      - name: Set up QEMU
        uses: docker/setup-qemu-action@v3

      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@v3

      - name: Extract metadata (tags, labels) for Docker
        id: meta
        uses: docker/metadata-action@v5
          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
          tags: |

      - name: Build and push Docker image
        uses: docker/build-push-action@v5
          context: .
          platforms: linux/arm64
          push: true
          tags: ${{ steps.meta.outputs.tags }}
          labels: ${{ steps.meta.outputs.labels }}
          provenance: false

And the docker-compose file that ties everything together:

version: "3"
    image: containrrr/watchtower
      - /var/run/docker.sock:/var/run/docker.sock
      - /root/.docker/config.json:/config.json
    command: --interval 60 --cleanup

    image: iarekylew00t/caddy-cloudflare:latest
    restart: unless-stopped
      - 80:80
      - 443:443
      - ./data/caddy/Caddyfile:/etc/caddy/Caddyfile
      # - ./data/caddy/caddy_data:/data
      # - ./data/caddy/caddy_config:/config
      - ./data/personal-site/var/www:/var/www/licu.dev:ro

    image: mongo:latest
    restart: unless-stopped
      - ./data/mongo/db:/data/db

    image: ghcr.io/lmihaig/personal-site:latest
      - ./data/personal-site/var/www:/var/www/licu.dev

All the code for the infrastructure can be found here and the code for the blog + resume part of the site is here

Read Previous

On what this blog is

Go to top