v0.3.0 - Valhalla: Voice Engine

Discord Voice Engine

Release v0.3.0 implements the full Discord Voice Gateway and audio pipeline for Dusk, developed strictly following the Asgard Protocol (10 Stage-Gates, Fail-First testing).

New Modules

Module	Responsibility
Dusk::Voice::Gateway	Voice WebSocket handshake (OP 0→2→3→4→5→6), heartbeat, SELECT_PROTOCOL
Dusk::Voice::UDP	74-byte IP Discovery, public IP/port parsing, RTP frame dispatch
Dusk::Voice::Audio	PCM→Opus pipeline via ffmpeg (Proc::Async), 20ms Supply of Buf frames
Dusk::Voice::Crypto	xsalsa20_poly1305 encryption via libsodium (NativeCall), correct RTP header
Dusk::Voice::Client	Full facade: connect(), play(Str), stop(), disconnect()

New Error Types

• Dusk::Error::LibsodiumNotFound — with apt install libsodium-dev hint
• Dusk::Error::FfmpegNotFound — with apt install ffmpeg hint
• Dusk::Error::VoiceUDPDiscoveryFailed — UDP timeout with clear message
• Dusk::Error::VoiceSessionTimeout — Voice WS session expired (codes 4006/4014)

Tests (Asgard Gate 05 — Fail-First)

• t/14-voice-gateway.t — 6 subtests (OP sequence, heartbeat, error handling)
• t/15-voice-udp.t — 5 subtests (IP Discovery packet structure, timeout)
• t/16-voice-audio.t — 5 subtests, ffmpeg-guarded (skips gracefully if absent)
• t/17-voice-crypto.t — 6 subtests, libsodium-guarded (RTP header, encryption)

System Requirements

# Debian/Ubuntu
apt install libsodium-dev ffmpeg

# macOS
brew install libsodium ffmpeg

Quick Start

use Dusk::Voice::Client;

my $voice = Dusk::Voice::Client.new(
    guild-id   => $guild-id,
    channel-id => $channel-id,
    user-id    => $bot-user-id,
    session-id => $state.session-id,
    token      => $server.token,
    endpoint   => $server.endpoint,
);

await $voice.connect();
await $voice.play('/path/to/audio.mp3');
$voice.disconnect();