Introduction

Jellything is a free server software to serve media like movies, videos and music through a website.

Source Code

Getting Started

Installation

From the AUR

This is the recommended option. It will also install a systemd service and system user for that service.

git clone https://aur.archlinux.org/jellything-git.git
cd jellything-git
makepkg -si

From source

Requirements:

  • rustup
  • esbuild
  • nasm
  • meson
  • ninja
  • cmake
  • dav1d
  • ffmpeg (only if you use transcoding)

Jellything was only tested for x86_64-unknown-linux-gnu and aarch64-unknown-linux-gnu targetss. Others probably work too.

git clone --recursive https://codeberg.org/metamuffin/jellything.git
cd jellything
cargo build --release
# Global installation
cp target/release/{jellything,jellytool} /usr/local/bin
# User installation
cp target/release/{jellything,jellytool} ~/.local/bin

Setup

1. Writing configurations

First write your configuration files whereever you want. The AUR package uses /etc/jellything.yaml.

# This hostname must be identical to how other instances reach you.
hostname: example.org
brand: "Jellything"
slogan: ""
admin_username: admin

# All of these paths can be customized. See "Paths"
media_path: "/srv/media"
asset_path: "/var/lib/jellything/assets"
database_path: "/var/lib/jellything/db"
temp_path: "/tmp/jellything"
cache_path: "/var/cache/jellything"
secrets_path: "/etc/jellysecrets.yaml" # points to the file below
# jellysecrets.yaml; filled with placeholders
admin_password: "xxxxxx"

# Both these keys should be initialized randomly.
# Use `head -c 32 /dev/random | base64`
cookie_key: "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx="
session_key: "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx="

# Credentials for remote instances. Keep this empty if you are just starting.
federation:
  "example.org": { username: "examplefed", password: "xxxxxxx" }

api:
  fanart_tv: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
  tmdb: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
  omdb: xxxxxxxx
  # This is the Trakt Application `client_id`
  trakt: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
  tvdb: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

2. Creating required directories

Next create directories in place of cache_path, temp_path, media_path. if jellything is not permitted to do so itself. Also obtain the default assets from the jellything-assets repo.

mkdir -p $cache_path $temp_path $media_path
git clone https://codeberg.org/metamuffin/jellything-assets.git $asset_path

3. Preparing your first import

In library_path, create files like below. This file provides an entry point to your library. Its exact meaning is described in The Import Guide.

Next place your favorite movies in media_path and use the import helper to quickly generate import instructions for it. This requires at least Trakt and TMDB keys to work.

# Jellytool will show an interactive wizard to select the correct metadata source.
# The files will be renamed to include the Trakt id in the name.
jellytool add big-buck-bunny.mkv
jellytool add agent-327-operation-barbershop.mkv
jellytool add spring.mkv

Launch your Platform

tip

If you used the AUR package, this should be as simple as systemctl start jellything.

For other installations, run the jellything binary with the configuration file's path as its first argument or set it in JELLYTHING_CONFIG. Jellything will start and serve your library at https://127.0.0.1:8080/

jellything /etc/jellything.yaml
# Or
JELLYTHING_CONFIG=/etc/jellything.yaml jellything

It is also advised to use jellything with a reverse proxy. Configure the network interface with the environment variables BIND_ADDR and PORT. Logging to stderr can be configured with the LOG variable and env_logger syntax.

Initial import

Next log in with the admin accound and visit /admin/dashboard to run an initial import.

Migrating the Database

caution

The current architecture for saving to the database will be completely rewritten soon.

With some updates the database serialization for changes. This requires a migration process. You need to follow this procedure either with every update.

1. Export the database

Use the jellytool of the previous version to export the database to JSON.

mv /path/to/db /path/to/db.old # Rename the DB to avoid conflict later
jellytool.old migrate export /path/to/db.old export /tmp/jdb

2. Run migrations on the JSON dump

This is not implemented yet. It usually just works without anyway.

3. Import the database

Now import your library back to where it usually lives using the jellytool of the current version.

jellytool migrate export /path/to/db import /tmp/jdb

4. Delete old Databases

Delete the old database and the JSON dump, they are not required anymore.

caution

Confirm that everything still works and no data is lost before!

rm /path/to/db.old
rm -r /tmp/jdb

Paths

  • media_path: All your media (videos, movies, etc.) and import flags live here. Jellything will only require read access to this directory.
  • asset_path: Static assets for the page. This includes fallback images, fonts and the front page. Used read-only.
    • error.avif: Used when images cant be displayed because of an error. This image is not required to be AVIF despite the name extension.
    • fallback-<kind>.avif: Fallback image when there is no asset available. AVIF also not required. <kind> is a node kind or "Person".
    • front.htm: Contents of the front page. The typical jellything page scaffold is placed around it.
    • logo.svg: Logo of the platform to replace the text in the top left of every page.
    • fonts/material-icons.woff2 Material Icons Font. Get this from online.
  • database_path: The database stores all imported nodes, user accounts and such. Write access required.
  • temp_path: Where to place short-lived files. Write access required. Can be in volatile memory (e.g. /tmp).
  • cache_path: In here jellything will place dozens of transcoded imagery, saved api responses and media metadata. This folder will likely be the biggest. Write access required.
  • secrets_path: Path to the secrets YAML file. Read access required. Should ideally not be readable by anything else than jellything.

Jellything's Import System

In normal operation, jellything serves all metadata from only the database.

Jellything Rust API

For making your own applications that implement client functionality, use the jellyclient crate. The jellycommon crate exposes commonly used structs like those used in the library and for jhls.

Generated Documentation

Jellything HTTP API

Most endpoints require the Accept header to be present and set to application/json and image/avif respectively. Any endpoint returning JSON, will report errors with an object containing error string in the error key. Routes marked with * require authentification.

The jellyclient crate already implements most API functionality. The jellycommon crate provides useful structs for deserializing data (also reexported in jellyclient).

# Cargo.toml
[depedencies]
jellyclient = { git = "https://codeberg.org/metamuffin/jellything.git" }

General

GET /api/version

Returns API version number.

POST /api/create_session

Request body contains JSON with keys username, password, expire (in seconds) and drop_permissions (a list of permissions, that this session cannot use). The Response contains the session cookie as a string in JSON.

GET* /n/<id>?<parents>&<children>&<filter..>

Request a library node with userdata attached. If parents or children flag is set, all parents/children will be returned too, otherwise those lists are empty. Returns ApiNodeResponse.

GET* /n/<id>/userdata

Returns only NodeUserData for that node.

GET* /search?<query>&<page>

Returns ApiSearchResponse.

GET* /items?<filter..>&<page>

Returns ApiItemsResponse.

GET* /home

Returns ApiHomeResponse.

Assets

Endpoints like .../poster redirect to the asset that you need and automatically choose fallbacks. Alternatively you can directly request assets with /asset/<token>, but there you need to implement the fallback behaviour yourself. Returned images are coded with AVIF. The width parameter is the width of the resolution you want to image to be.

[!WARNING] The actual returned resolution must not be exactly what you requested. Currently it is rounded up to the next power of two.

GET* /asset/<token>?<width>

This is the final asset endpoint where images are returned from. Other endpoints redirect here. The token part is an opaque string that you obtain from somewhere else like a redirect or from within the Node struct.

GET* /n/<id>/poster?<width>

GET* /n/<id>/backdrop?<width>

GET* /n/<id>/thumbnail?<t>&<width>

Returns a single frame from some track video of the media at a given time t.

GET* /n/<id>/person/<index>/asset?<group>&<width>

Returns headshot of a person from that node.

Stream

GET* /n/<id>/stream?<params..>

Responds with the stream directly or a redirect to the actual source in case of federation.

  • ?whep&<track...>&<seek>
    • WHEP Endpoint for streaming that set of tracks. The format used is decided by the server.
  • ?whepcontrol&<token>
    • WebSocket endpoint for controlling WHEP playback. TODO schema
  • ?remux&<track...>&<container>
  • ?hlssupermultivariant&<container>
    • Returns m3u8/HLS playlist of all known multi-variant playlists, one for each segment. The plylist is updated for live media.
  • ?hlsmultivariant&<segment>&<container>
    • Returns m3u8/HLS playlist of all track formats' variant playlists.
  • ?hlsvariant&<segment>&<track>&<container>&<format>
    • Returns m3u8/HLS playlist of all known fragments of this track format. The playlist is updated for live media.
  • ?info
    • Returns JSON StreamInfo.
  • ?fragmentindex&<segment>&<track>
    • Returns time ranges for every fragment of this track.
  • ?fragment&<segment>&<track>&<index>&<container>&<format>
export type FragmentIndex = TimeRange[];
export interface TimeRange {
  start: number;
  end: number;
}
export interface SubtitleCue extends TimeRange {
  content: string;
}
export interface StreamInfo {
  name?: string;
  segments: SegmentInfo[];
}
export interface SegmentInfo {
  name?: string;
  duration: number;
  tracks: TrackInfo[];
}
export type TrackKind = "video" | "audio" | "subtitles";
export interface TrackInfo {
  name?: string;
  language?: string;
  kind: TrackKind;
  formats: FormatInfo[];
}
export type StreamContainer = "webm" | "matroska" | "mpeg4" | "jvtt" | "webvtt";
export interface FormatInfo {
  codec: string;
  bitrate: number;
  remux: boolean;
  containers: StreamContainer[];

  width?: number;
  height?: number;
  channels?: number;
  samplerate?: number;
  bit_depth?: number;
}