v2.0 Release Available

Enterprise Voice AI
You Own & Control

Transform customer communications with AI that runs on your infrastructure. Sub-200ms latency, 17+ providers, and complete data privacy.

Source Available
Self-Hosted
Enterprise Ready
bash - 80x24
~ docker run -d vexyl/vexyl-voice-gateway
Unable to find image 'vexyl/vexyl-voice-gateway:latest' locally
latest: Pulling from vexyl/vexyl-voice-gateway
Digest: sha256:8b0a...4f2e
Status: Downloaded newer image for vexyl/vexyl-voice-gateway:latest
7f3b2a1c9d8e
~ curl localhost:8081/health
{ "status": "ok", "version": "2.0.0" }
~ _
Status
Online

Powering connections with world-class AI

AWS Google Cloud OpenAI Deepgram ElevenLabs Groq

Engineered for Performance

Don't compromise. Get the flexibility of self-hosting with the reliability of enterprise software.

17 AI Providers

Mix and match 7 TTS, 5 STT, and 5 LLM providers. Swap Sarvam for local accents or Groq for speed instantly without code changes.

Sub-200ms Latency

Industry-leading response times. Gateway Mode pipes audio directly to OpenAI Realtime for conversations faster than human thought.

Seamless PBX Integration

Built for telephony. Connects via AudioSocket (TCP) directly to Asterisk dialplans. Supports barge-in, DTMF, and call transfers.

The Voice Stack Architecture

A highly modular pipeline that decouples telephony from intelligence. Swap components without rewriting code.

1

Input Layer

Handles raw PCM streams from Asterisk (TCP) or Web Browsers (WebSocket).

2

Intelligence Layer

Routes audio to STT, manages conversation state with LLM (n8n/GPT-4), and caches TTS.

3

Output Layer

Streams synthesized audio back to the caller with smart buffering for jitter-free playback.


Asterisk / Browser
Gateway Core
STT
LLM
TTS

AI Providers

Redis Cache

Ecosystem Integrations

Ears

Speech Recognition

Sarvam Deepgram Groq OpenAI

Mind

Intelligence & Logic

n8n Flowise GPT-4o Claude

Voice

Synthesis

ElevenLabs Azure Google Aura

Deploy in Minutes

Choose your preferred deployment method.

FASTEST

Quick Install (Script)

Automated script for standalone Binary installation.

bash
curl -fsSL https://vexyl.ai/downloads/install_gateway.sh | sudo bash
RECOMMENDED

Docker

The fastest way to get started. Works on any system with Docker installed.

1. Pull Image
docker pull vexyl/vexyl-voice-gateway:latest
2. Run Container
docker run -d -p 8080:8080 -e SARVAM_API_KEY=key ...

Binary

Perfect for bare-metal servers or air-gapped environments without Docker.

Download Archive (.zip)
1. Download & Extract
wget -O gateway.zip https://vexyl.ai/downloads/vexyl-gateway.zip && unzip gateway.zip
2. Permission & Run
chmod +x vexyl-gateway && ./vexyl-gateway
3. Create Config
echo "SARVAM_API_KEY=..." > .env

Kubernetes

For high-availability enterprise deployments with auto-scaling.

1. Create Secrets
kubectl create secret generic...
2. Apply Manifest
kubectl apply -f vexyl-deployment.yaml

Comprehensive Documentation