Repository navigation
Proxmox
Streamline is one static binary with release archives for linux/amd64 and arm64, which suits an LXC container well: no Docker inside it, no runtime, and ffmpeg from Debian.
- 1. Create the container
- 2. Install Streamline
- 3. Mount your media
- 4. Share ownership in an unprivileged container
- 5. Hardware transcoding (VAAPI)
- Updating
- What the script installs
- Proxmox VE Helper-Scripts
In the Proxmox UI: Create CT with a Debian 13 (or 12, or Ubuntu 24.04) template, unprivileged, with Nesting enabled (the default for new unprivileged containers; the service's systemd sandboxing needs it). Two cores, 2 GiB of memory and 8 GiB of disk are plenty — your media does not live on that disk.
Or from the host shell:
pct create 120 local:vztmpl/debian-13-standard_13.1-2_amd64.tar.zst \
--hostname streamline --unprivileged 1 --features nesting=1 \
--cores 2 --memory 2048 --rootfs local-lvm:8 \
--net0 name=eth0,bridge=vmbr0,ip=dhcp --onboot 1
pct start 120(pveam available --section system lists the templates you can download; use the name you actually have.)
Inside the container, as root (pct enter 120):
apt-get update && apt-get install -y curl
curl -fsSL https://raw.githubusercontent.com/DataHearth/streamline/main/deploy/lxc/streamline.sh | bashIt downloads the latest release for the container's architecture, checks it against the release's checksums.txt (and that file's cosign signature when cosign is installed), installs ffmpeg, creates a streamline user, writes /etc/streamline/config.yaml with data_dir: /var/lib/streamline, and starts a hardened systemd service on port 8080.
The service will not stay up until the library folders point at the mount you add in step 3. The installer writes the default paths (/media/movies, /media/series), which the streamline user cannot create, so the journal shows create library path /media/movies: … permission denied and systemd keeps restarting the unit. That is expected: do step 3 now and restart.
The first start that succeeds creates the admin account admin@streamline.local with a generated password, printed once to the service's output, which the journal keeps:
journalctl -u streamline | grep 'default admin'It is never written to the config file. Lost it before logging in? Set a new one: systemctl stop streamline && runuser -u streamline -- streamline -c /etc/streamline/config.yaml auth set-password admin@streamline.local --generate && systemctl start streamline (as the streamline user, so the database files keep their owner).
Then continue with First-Run Setup. The library folders go in the config file (step 3 there), and must be the paths of the mount you add next.
| Variable | Effect |
|---|---|
STREAMLINE_VERSION=3.2.0 |
Install that release instead of the latest |
STREAMLINE_VERIFY_SIGNATURE=yes |
Refuse to install unless cosign is present and the signature checks out (no skips it; the default, auto, checks when cosign is installed) |
Pass them to bash: curl -fsSL … | STREAMLINE_VERSION=3.2.0 bash.
Downloads and media must be one mount, for the reason in the folder rule. Bind-mount the host directory holding both into the container as a mount point, from the host shell:
pct set 120 -mp0 /tank/data,mp=/mnt/data
pct reboot 120Then in /etc/streamline/config.yaml:
library:
movie_path: /mnt/data/media/movies
series_path: /mnt/data/media/series
music_path: /mnt/data/media/music # if you use music
ebook_path: /mnt/data/media/books # if you use books
audiobook_path: /mnt/data/media/audiobooks
download_path: /mnt/data/downloadsand systemctl restart streamline. Your torrent client — in another container, or Streamline's built-in engine — must see the same folders at the same paths.
Note
The service is sandboxed with ProtectSystem=full and ProtectHome=true: it can write anywhere except /usr, /boot, /etc (but for /etc/streamline) and /home. Mount your media under /mnt, /srv or /media, not /home.
An unprivileged container shifts every uid and gid by 100000: the streamline user (say uid 999 inside) is uid 100999 on the host, and a share owned by a host user shows up inside as nobody:nogroup, unwritable. Pick one:
Give the share to the shifted ids. Simplest when only this container writes there. On the host:
pct exec 120 -- id streamline # e.g. uid=999(streamline) gid=999(streamline)
chown -R 100999:100999 /tank/data # 100000 + the ids insideShare a group with other containers. When Jellyfin, a torrent client and Streamline all write the same tree, give them one group and map it straight through. With a host group media of gid 1001, and media (gid 1001) created inside each container, add to /etc/pve/lxc/120.conf:
lxc.idmap: u 0 100000 65536
lxc.idmap: g 0 100000 1001
lxc.idmap: g 1001 1001 1
lxc.idmap: g 1002 101002 64534
allow it once in /etc/subgid on the host (root:1001:1), then inside the container usermod -aG media streamline and make the tree group-writable on the host (chgrp -R media /tank/data && chmod -R g+rwX /tank/data, plus chmod g+s on directories so new files keep the group). The service runs with UMask=0002, so what Streamline creates stays writable by the group.
The transcoder can encode on an Intel or AMD GPU. Pass the render node through, from the host:
pct set 120 -dev0 /dev/dri/renderD128,gid=$(pct exec 120 -- getent group render | cut -d: -f3)
pct reboot 120(dev0 is Proxmox VE 8.2 and later; it maps the node to the container's render group, which the script already added the streamline user to.) Then install the VA driver inside the container — intel-media-va-driver for Intel (Broadwell and later; if encoding fails, intel-media-va-driver-non-free from Debian's non-free component), i965-va-driver for older Intel chips, mesa-va-drivers for AMD — and restart the service. Settings → Transcoding reports whether the device answered.
Debian's ffmpeg has VAAPI but no libvmaf, so transcoding.verify.min_vmaf parses and is never enforced on this install.
streamline-updateIt fetches and verifies the latest release before stopping anything, so a failed download leaves the running version alone; then it stops the service, snapshots the database with the new binary's own streamline backup (a manual backup in /var/lib/streamline/backups/, kept under backup.retention.manual, 10 by default; if the snapshot fails, the update is abandoned and the running version restarted), swaps the binary, refreshes the unit and starts it again. The new version migrates the database on boot, and migrations only go forward: the backup is how you go back. Streamline also takes its own pre-migration backup into the same folder before migrating (Backups); streamline backup list shows both and streamline restore puts either back. streamline-update is the same script as the install with update, so curl -fsSL …/streamline.sh | bash -s -- update does the same thing.
Take a Proxmox backup of the container now and then too: the backup above covers a bad migration, not a lost disk.
| Path | What |
|---|---|
/usr/local/bin/streamline |
The binary |
/etc/streamline/config.yaml |
Config, mode 0600, owned by streamline — the web UI writes settings back to it |
/var/lib/streamline |
data_dir: database, posters, built-in torrent state |
/etc/systemd/system/streamline.service |
The unit (journalctl -u streamline -f for logs) |
/usr/local/sbin/streamline-update |
The updater |
The same script works on any Debian or Ubuntu machine with systemd, a VM or bare metal included.
deploy/lxc/community-scripts/ holds the same install in the Proxmox VE Helper-Scripts layout (ct/streamline.sh creates the container, install/streamline-install.sh installs into it, and running update in the container updates it). It will be submitted there once the project meets their listing requirements; until then the container scripts can only run from their repository, so use the steps above.
🎬 Operating Streamline
- Installation
- First-Run Setup
- Adding Movies and TV
- Importing an Existing Library
- Activity and Calendar
- Requests and Users
- Music and Books
- Uploading files
- Extensions
- Send to Kindle
- NixOS and Nix
- Proxmox VE (LXC)
- Troubleshooting
- Roadmap
⚙️ Advanced
- Configuration Reference
- Authentication and SSO
- Quality Profiles and Naming
- Quality Profiles and Custom Formats
- Scheduled Jobs
- REST API
- Writing an extension
- Observability and Logging
- GitOps and Kubernetes