Suggested by Kim via support, who also provided the relevant docker-compose config (some non-PDF-relevant parts removed):

 services:
  bookstack-app:
    image: lscr.io/linuxserver/bookstack:latest
    container_name: bookstack-app
    environment:
      ### PDF Export
      - EXPORT_PDF_COMMAND="curl --request POST http://pdf-proxy:3070/forms/chromium/convert/html --form 'files=@{input_html_path};filename=index.html;type=text/html' --form paperWidth=8.27 --form paperHeight=11.7 --form marginTop=0.5 --form marginBottom=0.5 --form marginLeft=0.5 --form marginRight=0.5 -o '{output_pdf_path}'"
      - EXPORT_PDF_COMMAND_TIMEOUT=45
    restart: unless-stopped
    depends_on:
      - pdf-proxy
    networks:
      - bookstack-network


  pdf-app:
    image: gotenberg/gotenberg:latest
    container_name: pdf-app
    shm_size: 1gb
    networks:
      - pdf-network
    restart: always

  pdf-proxy:
    image: caddy:2.9-alpine
    container_name: pdf-proxy
    command: caddy reverse-proxy --from :3070 --to http://pdf-app:3000
    networks: 
      - pdf-network
      - bookstack-network
    restart: always

networks:
  bookstack-network:
    driver: bridge
  
  pdf-network:
    driver: ipvlan
    internal: true

Support ticket 206 for reference

• Update head.html files to be served as plain text downloads instead of HTML. - Already active

Todo

• Update module builder to be a pre-deploy script to convert all hacks into static ZIPs.
  • Will be a lot more efficient, and show issues on build/deploy instead of to the user on use.
  • Make local dev easier as all will be part of the site.
  • Should cache on the modification time for the relevant index.md.
• Update module builder to actually insert head.html into the head.
  • Currently, it's inserted at body-start, which is a bit janky with <link/> elements, and goes against expectations.
• Doc Updates
• Blogpost