Discord4J is a fast, powerful, unopinionated, reactive library to enable quick and easy development of Discord bots for Java, Kotlin, and other JVM languages using the official Discord Bot API.
In this example for v3.1, whenever a user sends a !ping
message the bot will immediately respond with Pong!
.
public final class ExampleBot {
public static void main(final String[] args) {
final String token = args[0];
final DiscordClient client = DiscordClient.create(token);
final GatewayDiscordClient gateway = client.login().block();
gateway.on(MessageCreateEvent.class).subscribe(event -> {
final Message message = event.getMessage();
if ("!ping".equals(message.getContent())) {
final MessageChannel channel = message.getChannel().block();
channel.createMessage("Pong!").block();
}
});
gateway.onDisconnect().block();
}
}
-
๐ Reactive - Discord4J follows the reactive-streams protocol to ensure Discord bots run smoothly and efficiently regardless of size.
-
๐ Official - Automatic rate limiting, automatic reconnection strategies, and consistent naming conventions are among the many features Discord4J offers to ensure your Discord bots run up to Discord's specifications and to provide the least amount of surprises when interacting with our library.
-
๐ ๏ธ Modular - Discord4J breaks itself into modules to allow advanced users to interact with our API at lower levels to build minimal and fast runtimes or even add their own abstractions.
-
โ๏ธ Powerful - Discord4J can be used to develop any bot, big or small. We offer many tools for developing large-scale bots from custom distribution frameworks, off-heap caching, and its interaction with Reactor allows complete integration with frameworks such as Spring and Micronaut.
-
๐ซ Community - We pride ourselves on our inclusive community and are willing to help whenever challenges arise; or if you just want to chat! We offer help ranging from Discord4J specific problems, to general programming and web development help, and even Reactor-specific questions. Be sure to visit us on our Discord server!
- Creating a new Gradle project with IntelliJ (recommended)
- Creating a new Maven project with IntelliJ
- Creating a new Gradle project with Eclipse
- Creating a new Maven project with Eclipse
repositories {
mavenCentral()
}
dependencies {
implementation 'com.discord4j:discord4j-core:3.1.3'
}
repositories {
mavenCentral()
}
dependencies {
implementation("com.discord4j:discord4j-core:3.1.3")
}
<dependencies>
<dependency>
<groupId>com.discord4j</groupId>
<artifactId>discord4j-core</artifactId>
<version>3.1.3</version>
</dependency>
</dependencies>
libraryDependencies ++= Seq(
"com.discord4j" % "discord4j-core" % "3.1.3"
)
Discord4J 3.1.x introduces performance and API enhancements, a plethora of new features, and dependency upgrades. A Migration Guide is provided to aide users and ensure a smooth and readily available transition.
Discord4J | Support | Gateway/API | Intents |
---|---|---|---|
v3.2.x | In development | v8 | Mandatory, non-privileged default |
v3.1.x | Current | v6 | Optional, no intent default |
v3.0.x | Maintenance only | v6 | No intents support |
See this wiki page for more details about compatibility.
We would like to give a special thanks to all of our sponsors for providing us the funding to continue developing and hosting repository resources as well as driving forward initiatives for community programs. In particular, we would like to give a special shoutout to these wonderful individuals:
Here are some real-world examples of large bots using Discord4J:
- Groovy - An easy to use music bot used by more than 4 million servers.
- ZeroTwo - An anime multi-purpose bot used in more than 250K servers.
- DisCal - Implements Google Calendar into Discord as seamlessly and comprehensively as possible; serving nearly 14K servers.
- Shadbot - A configurable multipurpose bot with music, gambling mini-games, video game stats, and more; serving almost 12K servers.
Do you own a large bot using Discord4J? Ask an admin in our Discord or submit a pull request to add your bot to the list!
Discord4J uses Project Reactor as the foundation for our asynchronous framework. Reactor provides a simple yet extremely powerful API that enables users to reduce resources and increase performance.
public final class ExampleBot {
public static void main(final String[] args) {
final String token = args[0];
final DiscordClient client = DiscordClient.create(token);
client.login().flatMapMany(gateway -> gateway.on(MessageCreateEvent.class))
.map(MessageCreateEvent::getMessage)
.filter(message -> "!ping".equals(message.getContent()))
.flatMap(Message::getChannel)
.flatMap(channel -> channel.createMessage("Pong!"))
.blockLast();
}
}
Discord4J also provides several methods to aide in better reactive chain compositions, such as GatewayDiscordClient#withGateway
and EventDispatcher#on
with an error handling overload.
final String token = args[0];
final DiscordClient client = DiscordClient.create(token);
client.withGateway(gateway -> {
final Publisher<?> pingPong = gateway.on(MessageCreateEvent.class, event ->
Mono.just(event.getMessage())
.filter(message -> "!ping".equals(message.getContent()))
.flatMap(Message::getChannel)
.flatMap(channel -> channel.createMessage("Pong!")));
final Publisher<?> onDisconnect = gateway.onDisconnect()
.doOnTerminate(() -> System.out.println("Disconnected!"));
return Mono.when(pingPong, onDisconnect);
}).block();
By utilizing Reactor, Discord4J has native integration with Kotlin coroutines when paired with the kotlinx-coroutines-reactor library.
val token = args[0]
val client = DiscordClient.create(token)
client.withGateway {
mono {
it.on(MessageCreateEvent::class.java)
.asFlow()
.collect {
val message = it.message
if (message.content == "!ping") {
val channel = message.channel.awaitSingle()
channel.createMessage("Pong!").awaitSingle()
}
}
}
}
.block()
// IMAGE_URL = https://cdn.betterttv.net/emote/55028cd2135896936880fdd7/3x
// ANY_URL = https://www.youtube.com/watch?v=5zwY50-necw
final MessageChannel channel = ...
channel.createEmbed(spec ->
spec.setColor(Color.RED)
.setAuthor("setAuthor", ANY_URL, IMAGE_URL)
.setImage(IMAGE_URL)
.setTitle("setTitle/setUrl")
.setUrl(ANY_URL)
.setDescription("setDescription\n" +
"big D: is setImage\n" +
"small D: is setThumbnail\n" +
"<-- setColor")
.addField("addField", "inline = true", true)
.addField("addFIeld", "inline = true", true)
.addField("addFile", "inline = false", false)
.setThumbnail(IMAGE_URL)
.setFooter("setFooter --> setTimestamp", IMAGE_URL)
.setTimestamp(Instant.now())
).block();
Users typically prefer working with names instead of IDs. This example will demonstrate how to search for all members that have a role with a specific name.
final Guild guild = ...
final Set<Member> roleMembers = new HashSet<>();
for (final Member member : guild.getMembers().toIterable()) {
for (final Role role : member.getRoles().toIterable()) {
if ("Developers".equalsIgnoreCase(role.getName())) {
roleMembers.add(member);
break;
}
}
}
return roleMembers;
Alternatively, using Reactor:
final Guild guild = ...
return guild.getMembers()
.filterWhen(member -> member.getRoles()
.map(Role::getName)
.any("Developers"::equalsIgnoreCase));
Discord4J provides full support for voice connections and the ability to send audio to other users connected to the same channel. Discord4J can accept any Opus audio source with LavaPlayer being the preferred solution for downloading and encoding audio from YouTube, SoundCloud, and other providers.
To get started, you will first need to instantiate and configure an, conventionally global, AudioPlayerManager
.
public static final AudioPlayerManager PLAYER_MANAGER;
static {
PLAYER_MANAGER = new DefaultAudioPlayerManager();
// This is an optimization strategy that Discord4J can utilize to minimize allocations
PLAYER_MANAGER.getConfiguration().setFrameBufferFactory(NonAllocatingAudioFrameBuffer::new);
AudioSourceManagers.registerRemoteSources(PLAYER_MANAGER);
AudioSourceManagers.registerLocalSource(PLAYER_MANAGER);
}
Next, we need to allow Discord4J to read from an AudioPlayer
to an AudioProvider
.
public final class LavaPlayerAudioProvider extends AudioProvider {
private final AudioPlayer player;
private final MutableAudioFrame frame;
public LavaPlayerAudioProvider(final AudioPlayer player) {
// Allocate a ByteBuffer for Discord4J's AudioProvider to hold audio data for Discord
super(ByteBuffer.allocate(StandardAudioDataFormats.DISCORD_OPUS.maximumChunkSize()));
// Set LavaPlayer's AudioFrame to use the same buffer as Discord4J's
frame = new MutableAudioFrame();
frame.setBuffer(getBuffer());
this.player = player;
}
@Override
public boolean provide() {
// AudioPlayer writes audio data to the AudioFrame
final boolean didProvide = player.provide(frame);
if (didProvide) {
getBuffer().flip();
}
return didProvide;
}
}
Typically, audio players will have queues or internal playlists for users to be able to automatically cycle through songs as they are finished or requested to be skipped over. We can manage this queue externally and pass it to other areas of our code to allow tracks to be viewed, queued, or skipped over by creating an AudioTrackScheduler
.
public final class AudioTrackScheduler extends AudioEventAdapter {
private final List<AudioTrack> queue;
private final AudioPlayer player;
public AudioTrackScheduler(final AudioPlayer player) {
// The queue may be modifed by different threads so guarantee memory safety
// This does not, however, remove several race conditions currently present
queue = Collections.synchronizedList(new LinkedList<>());
this.player = player;
}
public List<AudioTrack> getQueue() {
return queue;
}
public boolean play(final AudioTrack track) {
return play(track, false);
}
public boolean play(final AudioTrack track, final boolean force) {
final boolean playing = player.startTrack(track, !force);
if (!playing) {
queue.add(track);
}
return playing;
}
public boolean skip() {
return !queue.isEmpty() && play(queue.remove(0), true);
}
@Override
public void onTrackEnd(final AudioPlayer player, final AudioTrack track, final AudioTrackEndReason endReason) {
// Advance the player if the track completed naturally (FINISHED) or if the track cannot play (LOAD_FAILED)
if (endReason.mayStartNext) {
skip();
}
}
}
Currently, Discord only allows 1 voice connection per server. Working within this limitation, it is logical to think of the 3 components we have worked with thus far (AudioPlayer
, LavaPlayerAudioProvider
, and AudioTrackScheduler
) to be correlated to a specific Guild
, naturally unique by some Snowflake
. Logically, it makes sense to combine these objects into one, so that they can be put into a Map
for easier retrieval when connecting to a voice channel or when working with commands.
public final class GuildAudioManager {
private static final Map<Snowflake, GuildAudioManager> MANAGERS = new ConcurrentHashMap<>();
public static GuildAudioManager of(final Snowflake id) {
return MANAGERS.computeIfAbsent(id, ignored -> new GuildAudioManager());
}
private final AudioPlayer player;
private final AudioTrackScheduler scheduler;
private final LavaPlayerAudioProvider provider;
private GuildAudioManager() {
player = PLAYER_MANAGER.createPlayer();
scheduler = new AudioTrackScheduler(player);
provider = new LavaPlayerAudioProvider(player);
player.addListener(scheduler);
}
// getters
}
Finally, we need to connect to the voice channel. After connecting you are given a VoiceConnection
object where you can utilize it later to disconnect from the voice channel by calling VoiceConnection#disconnect
.
final VoiceChannel channel = ...
final GuildAudioManager manager = GuildAudioManager.of(channel.getGuildId());
final AudioProvider provider = manager.getProvider();
final VoiceConnection connection = channel.join(spec -> spec.setProvider(provider)).block();
// In the AudioLoadResultHandler, add AudioTrack instances to the AudioTrackScheduler (and send notifications to users)
PLAYER_MANAGER.loadItemOrdered(manager, "https://www.youtube.com/watch?v=dQw4w9WgXcQ", new AudioLoadResultHandler() { /* overrides */ })
Typically, after everyone has left a voice channel, the bot should disconnect automatically as users typically forget to disconnect the bot manually. This problem can be solved rather elegantly using a reactive approach over an imperative one as the example below demonstrates.
final VoiceChannel channel = ...
final Mono<Void> onDisconnect = channel.join(spec -> { /* TODO Initialize */ })
.flatMap(connection -> {
// The bot itself has a VoiceState; 1 VoiceState signals bot is alone
final Publisher<Boolean> voiceStateCounter = channel.getVoiceStates()
.count()
.map(count -> 1L == count);
// After 10 seconds, check if the bot is alone. This is useful if
// the bot joined alone, but no one else joined since connecting
final Mono<Void> onDelay = Mono.delay(Duration.ofSeconds(10L))
.filterWhen(ignored -> voiceStateCounter)
.switchIfEmpty(Mono.never())
.then();
// As people join and leave `channel`, check if the bot is alone.
// Note the first filter is not strictly necessary, but it does prevent many unnecessary cache calls
final Mono<Void> onEvent = channel.getClient().getEventDispatcher().on(VoiceStateUpdateEvent.class)
.filter(event -> event.getOld().flatMap(VoiceState::getChannelId).map(channel.getId()::equals).orElse(false))
.filterWhen(ignored -> voiceStateCounter)
.next()
.then();
// Disconnect the bot if either onDelay or onEvent are completed!
return Mono.first(onDelay, onEvent).then(connection.disconnect());
});