A playback fleet, not a single script
Our music platform runs on eleven Go playback servers behind one bot. Each chat stays on its server for the length of a session, and new chats go to the least-loaded server among the healthy ones only.
Health comes before load: a broken server can never look idle and quietly swallow new chats, which is the most common way music bots fail at scale.
Every track verified before it plays
A song that cuts off halfway is worse than a song that fails. Every download is checked against the real length of the track, and if the first source delivers a partial file, a second source is tried before anything reaches the voice chat.
What your music bot gets
- Queues, playlists, seeking, skip and loop.
- Voice-chat transport over WebRTC with jitter buffering.
- Automatic failover when a data centre drops the call.
- Idle detection, so assistants leave empty chats on their own.
- Rolling updates that never stop playback for everyone at once.