Open voice infrastructure for AI agents.
A Python framework that bridges telephony infrastructure (Asterisk, FreeSWITCH, LiveKit) with AI voice agents (STT, LLM, TTS). It lets developers build AI-powered call centers without needing to understand telecom internals.
Telephony in, AI out.
Voxtra sits between the PBX and the model providers. It hands you a session API (say, listen, agent) and orchestrates the STT → LLM → TTS pipeline behind the scenes. Every layer is a registry — swap providers without touching the rest.
Layered by design.
Voxtra is a stack of small Python packages, each owning one concern. Core handles app lifecycle and session routing. Telephony adapts ARI / SIP. Audio runs the AudioSocket TCP transport. Media bridges sessions into the pipeline. AI exposes the STT / LLM / TTS / VAD provider registry. Pipeline orchestrates all of it per call.
Two files, one running call center.
Configure providers in YAML; describe the call flow in Python. voxtra start wires the rest.
from voxtra import VoxtraApp
app = VoxtraApp.from_yaml("voxtra.yaml")
@app.route(extension="1000")
async def support_call(session):
await session.answer()
await session.say("Hello, how can I help you?")
text = await session.listen()
reply = await session.agent.respond(text)
await session.say(reply.text)
app.run()