Dotfiles

My personal Linux desktop configuration files, managed with GNU Stow.

Components overview

• Compositors: Hyprland (primary) and Niri
• Desktop Shell & UI: DankMaterialShell, Waybar, Rofi, Quickshell
• Terminals: Kitty, Ghostty
• Theming: Dynamic wallpaper-based material theming using Matugen and Wallust, with support for GTK applications
• Other utilities: Zsh, VS Code, Swaylock

How to replicate

1. Install dependencies: The required packages are listed in pkglist.txt (Arch official repos) and aurlist.txt (AUR). Install them using pacman and an AUR helper (e.g., yay or paru):

   sudo pacman -S --needed - < pkglist.txt
   yay -S --needed - < aurlist.txt

2. Clone the repository:

   git clone https://github.com/thalesraymond/dotfiles.git ~/dotfiles
   cd ~/dotfiles

3. Deploy configurations: Use GNU Stow to create symbolic links in your home directory. To deploy all configurations:

   stow *

   (Note: To stow individual components, specify the folder name, e.g., stow hypr)

Media server (Docker Compose)

A self-hosted media automation stack lives in docker/media-server/. It is not part of the Stow model (excluded via docker/.stow-local-ignore) — it is a standalone Docker Compose project running against the secondary HDD at /mnt/media_storage.

Service	Role	Port
qBittorrent	Torrent client (WebUI)	8080 (TCP + UDP 6881)
FlareSolverr	Cloudflare challenge solver	internal (http://flaresolverr:8191)
Prowlarr	Indexer manager	9696
Sonarr	TV show automation	8989
Radarr	Movie automation	7878
Bazarr	Subtitle automation	6767
Maintainerr	Auto-cleanup rules (Jellyfin + *arrs)	6246
Jellyfin	Media server (transcoding)	8096

Storage layout

/mnt/media_storage
├── downloads/            # qBittorrent downloads (shared with Sonarr/Radarr)
└── media/
    ├── movies/           # Radarr-managed movies  → Jellyfin /data/movies
    └── tv/               # Sonarr-managed shows   → Jellyfin /data/tvshows

Initialize it (all containers run as PUID=1000/PGID=1000, so the HDD must be owned by UID/GID 1000):

sudo mkdir -p /mnt/media_storage/{downloads,media/movies,media/tv}
sudo chown -R 1000:1000 /mnt/media_storage

Run the stack

cd docker/media-server
docker compose up -d        # stop with: docker compose down

• All services use TZ=America/Sao_Paulo and restart: unless-stopped.
• Downloads and media are on the same filesystem, so Sonarr/Radarr use hardlinks/atomic moves (no copy + delete). Point them at http://qbittorrent:8080 on the compose network.
• Bazarr (subtitles): WebUI at 6767 → Settings → Sonarr/Radarr, add both apps. It mounts the same /tv//movies paths, so no path mapping is needed.
• FlareSolverr (Cloudflare-protected indexers): in Prowlarr → Settings → Indexers, set the FlareSolverr API URL to http://flaresolverr:8191, then enable Use FlareSolverr on the affected indexers.
• Maintainerr (auto-cleanup): in its Settings, connect Jellyfin (http://jellyfin:8096 + API key from Jellyfin → Dashboard → API Keys) and Radarr/Sonarr (API keys from their Settings → General). Media is mounted at the same /tv//movies paths the *arrs report, so rules can delete files directly.
• Per-service config lives in docker/media-server/config/ (gitignored) — back it up separately.

Start on boot (optional)

sudo systemctl link ~/dotfiles/docker/media-server/media-server.service
sudo systemctl enable --now media-server.service

Edit WorkingDirectory= in the unit first if this repo is not at ~/dotfiles. Containers also self-restart via their restart: policy should the Docker daemon restart.

Jellyfin hardware acceleration (NVIDIA NVENC)

The compose file reserves the GPU for the jellyfin service via the NVIDIA container runtime — install the toolkit once on the host:

sudo pacman -S nvidia-container-toolkit     # Arch official repos
sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker

Then in Jellyfin → Dashboard → Playback → Transcoding, select NVIDIA NVENC and enable the codecs you want.

Run the stack on the native engine. docker compose must use the default context (system Docker Engine), not Docker Desktop: Desktop runs its engine inside a VM that cannot see the host's /dev/dri or /dev/nvidia*, which fails with error gathering device information while adding custom device "/dev/dri". The NVIDIA runtime is registered in /etc/docker/daemon.json for the native daemon only.

docker context use default

On an Intel/AMD-only host, delete the deploy: GPU reservation and the NVIDIA_* environment variables from the jellyfin service, and add devices: - /dev/dri:/dev/dri for VAAPI/QSV.