Operation Modes
VEXYL AI Voice Gateway supports three distinct operation modes, each optimized for different use cases.
Comparison
| Mode | Description | Latency | Best For |
|---|---|---|---|
| Standard | STT → LLM → TTS pipeline | 2-5s | Complex workflows, Custom LLMs |
| Gateway | Direct to OpenAI/ElevenLabs | 0.5-1.5s | Speed, Natural conversation |
| WebSocket | Browser-based widget | 2-5s | Web Apps, Kiosks |
Standard Mode
The default mode. Processes audio through a modular pipeline where you can mix and match any provider.
GATEWAY_MODE=false
LLM_PROVIDER=n8n
STT_PROVIDER=sarvam
TTS_PROVIDER=deepgramGateway Mode
Bypasses the internal pipeline to connect directly to ultra-low latency providers like OpenAI Realtime API.
GATEWAY_MODE=true
LLM_WEBSOCKET_URL=wss://api.openai.com/v1/realtime...Switching Modes
Simply update your environment variables and restart the container. No code changes required.