Installation

Docker

Run the Fliks server and its database with Docker Compose, the recommended way to self-host it.

In short

Download the example Compose file, set a database password and your media folder, then run docker compose up -d and open http://<host>:4848.

Run Fliks#

You need Docker with the Compose plugin and a folder with your videos.

bash
curl -LO https://raw.githubusercontent.com/fliks-app/fliks/main/docker-compose.example.yml
mv docker-compose.example.yml docker-compose.yml

Edit docker-compose.yml:

  1. Replace changeme with the same password in POSTGRES_PASSWORD and DB_PASSWORD.
  2. Set your media folder: /path/to/your/media:/medias. You'll type /medias when adding a library.
bash
docker compose up -d

Open http://<host>:4848 and follow the Quick start from step 2.

Port 4848 is taken, or pin a version

Change the left side of '4848:4848' under ports: (for example '8080:4848'); leave PORT alone. To pin a release instead of :latest, edit the image tag (tags are on the package page), then run docker compose up -d again.

Reference#

You don't need any of this for a first install.

Volumes#

What each volume holds
Volume Contents
/medias Your library root(s). Read-only is enough to browse and play; downloads, deleting media and saving subtitles next to a video need write access.
/downloads Only needed if you install the download plugin; use the same path here and in your download client.
/app/conf The JWT signing key and the engine token, both generated at first start. Lose the key and every session is invalidated.
/app/data Artwork, uploaded avatars, database backups, plugin data (plugins/), and a regenerable cache/ subfolder. Not disposable: avatars can't be re-fetched.
/app/transcode The HLS segment cache. Regenerable, safe to wipe. The image sets FLIKS_TRANSCODE_DIR to this path.

Running as a specific user#

Run as a specific user (NAS, Kubernetes)

Some platforms pin the container to a non-root uid: the TrueNAS catalog uses 568, Unraid uses 99:100, Kubernetes' runAsNonRoot and OpenShift's arbitrary uid policy pick their own. The image's /app/conf, /app/data and /app/transcode are group-writable by root (g=u), so any uid works as long as group 0 is in its group list:

yaml
services:
  fliks:
    user: '568:568'
    group_add: ['0']

Chromecast on a bridge network#

Casting from a bridge network

The default bridge network works for casting. The device you cast from (a browser, the phone app, the desktop app) finds the Chromecast itself; the server takes no part in discovery. What the Chromecast does need is to reach the server directly, at the address Fliks hands it: set Public address under Settings > General if that address isn't the one the Chromecast can reach (see Reverse proxy).

Limit CPU and memory on a small box

On a shared NAS or single-board computer, cap the container so transcodes don't starve the rest of the box. A 4K HDR transcode takes about 700 MB of RAM. The engine sizes itself on the CPUs of the limit: software transcodes share them (each gets at least 2 threads), so more simultaneous transcodes run slower, and none is refused or queued. Background jobs (scans, thumbnails, marker detection) run on the CPUs minus one (at least one) and wait or pause while a transcode needs the CPU. Hardware encodes don't use that budget. The Concurrent background jobs setting overrides it (see Streaming and transcoding).

yaml
services:
  fliks:
    mem_limit: 4g
    cpus: '4'

Hardware transcoding#

See Hardware acceleration for the devices and environment variables each vendor needs (Intel QSV/VAAPI, NVIDIA NVENC, AMD through VAAPI). The Intel and NVIDIA stacks are amd64-only; on arm64 (a Raspberry Pi 5, an ARM NAS) transcodes fall back to CPU.

What is in the image#

The published image is a single container: the Rust server (fliks-server) behind fliks-engine, the streaming engine that owns port 4848 and forwards the API to the server, the built Angular web client served as static assets, and a self-contained FFmpeg build with hardware-acceleration support baked in. A separate PostgreSQL container holds the database. It runs on linux/amd64 and linux/arm64.

Edit this page on GitHub

Esc