Voice Transport API — WebSocket, LiveKit

transport

import "github.com/lookatitude/beluga-ai/v2/voice/transport"

Package transport provides the audio transport interface and registry for the Beluga AI voice pipeline. Transports handle bidirectional audio I/O between clients and the voice pipeline, abstracting the underlying protocol (WebSocket, LiveKit, Daily, etc.).

LiveKit is treated as a transport, not a framework dependency. LiveKit’s server provides WebRTC transport, while Beluga handles all STT/LLM/TTS processing through the frame-based pipeline.

Core Interface

The AudioTransport interface provides bidirectional audio I/O:

type AudioTransport interface {
    Recv(ctx context.Context) (<-chan voice.Frame, error)
    Send(ctx context.Context, frame voice.Frame) error
    AudioOut() io.Writer
    Close() error
}

Registry Pattern

Providers register via Register in their init() function and are created with New. Use List to discover available providers.

import _ "github.com/lookatitude/beluga-ai/v2/voice/transport"

t, err := transport.New("websocket", transport.Config{URL: "ws://..."})
frames, err := t.Recv(ctx)
for frame := range frames {
    // process incoming audio frame
}

Pipeline Integration

Use AsVoiceTransport to adapt an AudioTransport to the voice.Transport interface expected by the [voice.VoicePipeline].

Built-in Transport

The package includes a WebSocketTransport implementation registered as “websocket”. Configure it with NewWebSocketTransport and options WithWSSampleRate and WithWSChannels.

Configuration

The Config struct supports URL, authentication token, sample rate, channel count, and provider-specific extras. Use functional options WithURL, WithToken, WithSampleRate, and WithChannels.

Available Providers

• websocket — Built-in WebSocket transport (voice/transport)
• livekit — LiveKit WebRTC rooms (voice/transport/providers/livekit)
• daily — Daily.co rooms (voice/transport/providers/daily)
• pipecat — Pipecat server (voice/transport/providers/pipecat)

---

daily

import "github.com/lookatitude/beluga-ai/v2/voice/transport/providers/daily"

Package daily registers the “daily” transport name for Daily.co rooms in the Beluga AI voice pipeline.

Status: not implemented

This provider is a placeholder. It does not connect to Daily: New (and transport.New("daily", ...)) always returns an error wrapping transport.ErrNotImplemented, which in turn wraps errors.ErrUnsupported. Earlier releases returned a transport that silently discarded audio and never produced frames; that fake success was removed.

Use the built-in “websocket” transport (voice/transport) until a real Daily integration is available.

import _ "github.com/lookatitude/beluga-ai/v2/voice/transport/providers/daily"

_, err := transport.New("daily", transport.Config{URL: "..."})
if errors.Is(err, transport.ErrNotImplemented) {
    // fall back to another transport
}

---

livekit

import "github.com/lookatitude/beluga-ai/v2/voice/transport/providers/livekit"

Package livekit registers the “livekit” transport name for LiveKit WebRTC rooms in the Beluga AI voice pipeline.

Status: not implemented

This provider is a placeholder. It does not connect to LiveKit: New (and transport.New("livekit", ...)) always returns an error wrapping transport.ErrNotImplemented, which in turn wraps errors.ErrUnsupported. Earlier releases returned a transport that silently discarded audio and never produced frames; that fake success was removed.

Use the built-in “websocket” transport (voice/transport) until a real LiveKit integration is available.

import _ "github.com/lookatitude/beluga-ai/v2/voice/transport/providers/livekit"

_, err := transport.New("livekit", transport.Config{URL: "..."})
if errors.Is(err, transport.ErrNotImplemented) {
    // fall back to another transport
}

---

pipecat

import "github.com/lookatitude/beluga-ai/v2/voice/transport/providers/pipecat"

Package pipecat registers the “pipecat” transport name for Pipecat servers in the Beluga AI voice pipeline.

Status: not implemented

This provider is a placeholder. It does not connect to Pipecat: New (and transport.New("pipecat", ...)) always returns an error wrapping transport.ErrNotImplemented, which in turn wraps errors.ErrUnsupported. Earlier releases returned a transport that silently discarded audio and never produced frames; that fake success was removed.

Use the built-in “websocket” transport (voice/transport) until a real Pipecat integration is available.

import _ "github.com/lookatitude/beluga-ai/v2/voice/transport/providers/pipecat"

_, err := transport.New("pipecat", transport.Config{URL: "..."})
if errors.Is(err, transport.ErrNotImplemented) {
    // fall back to another transport
}