An alternative to the *arr stack

What’s in a typical *arr stack

Many setups add more: Overseerr or Jellyseerr for requests, Bazarr for subtitles, Recyclarr to sync quality profiles, Unpackerr to extract downloads, a VPN container and a reverse proxy to reach it all from outside.

Where the time goes

• Wiring. Each app needs the others’ addresses and API keys. Change one and something downstream stops working.
• Paths. The download client and Sonarr or Radarr have to see the same files at the same paths, or imports fail. Hence remote path mappings and the long guides on hardlinks.
• Updates. Five release cycles, five changelogs and five things that can break after a pull.
• Access. Five web interfaces on five ports, plus a reverse proxy, VPN or open ports to use them away from home.
• People. One login per app, shared by everyone in the house.

The same job in one container

Here is a trimmed-down *arr stack in Docker Compose:

services:
  sonarr:
    image: lscr.io/linuxserver/sonarr
    ports: ["8989:8989"]
    volumes: ["./sonarr:/config", "/srv/data:/data"]
  radarr:
    image: lscr.io/linuxserver/radarr
    ports: ["7878:7878"]
    volumes: ["./radarr:/config", "/srv/data:/data"]
  prowlarr:
    image: lscr.io/linuxserver/prowlarr
    ports: ["9696:9696"]
    volumes: ["./prowlarr:/config"]
  sabnzbd:
    image: lscr.io/linuxserver/sabnzbd
    ports: ["8080:8080"]
    volumes: ["./sabnzbd:/config", "/srv/data:/data"]
  qbittorrent:
    image: lscr.io/linuxserver/qbittorrent
    environment: ["WEBUI_PORT=8081"]
    ports: ["8081:8081", "6881:6881", "6881:6881/udp"]
    volumes: ["./qbittorrent:/config", "/srv/data:/data"]
# …then PUID/PGID and TZ for each, API keys copied
# between the apps, and a reverse proxy in front.

And here is everything Nabbed runs at home:

services:
  agent:
    image: ghcr.io/getnabbed/agent:0.2
    restart: unless-stopped
    environment:
      NABBED_API_URL: https://api.nabbed.media
      NABBED_LIBRARY_ROOTS: /media
    volumes:
      - data:/data
      - ./downloads:/downloads
      - /srv/media:/media
    ports: ["6881:6881", "6881:6881/udp"]  # optional

volumes:
  data:

The agent searches, downloads and imports in one process, so there are no API keys between apps and no path mappings: the folder it downloads to and the library it imports into are both its own. The Docker walkthrough explains each line, and the install guide adds the hardening options.

What Nabbed covers

You keep Plex or Jellyfin: the agent triggers a library scan after each import. You also keep your indexers and Usenet provider, and any tools for things outside TV and films, such as music.

Why one agent is easier to live with

• Imports are all or nothing. The agent verifies, extracts, renames and moves a release, and writes nothing to your library unless every step succeeds.
• It keeps going offline. If the web app is unreachable, downloads carry on and events are sent once it’s back.
• It only writes where you say. Library folders are opt-in on the agent’s machine; by default it can’t write anywhere.
• It’s auditable. The agent is open source under the MIT licence, so you can read every byte it sends.

Comparing Nabbed with Sonarr and Radarr specifically, including when they’re the better choice? See the Sonarr and Radarr comparison.

Questions