Nari plugin for LiveKit Agents

Streaming speech-to-text and text-to-speech with the Nari API. Maintained by Nari Labs. Python 3.11+ · Tested with LiveKit Agents 1.8.2.

Installation

Private preview — install from this checkout:

uv sync --extra examples
cp .env.example .env

Set NARI_API_KEY in .env. The example also needs OPENROUTER_API_KEY and OPENROUTER_MODEL.

Usage

Add Nari to an existing agent with your LLM:

from livekit.agents import AgentSession
from livekit.plugins import nari, silero

recognizer = nari.STT()
session = AgentSession(
    stt=recognizer,
    tts=nari.TTS(voice="diana"),
    llm=your_llm,
    vad=silero.VAD.load(min_silence_duration=0.2),
    turn_handling={"turn_detection": "vad"},
)
recognizer.bind(session)  # Required before session.start().

Defaults: qwen3-asr-fast and qwen3-tts-fast. diana is an English voice. Create one recognizer per session; LiveKit handles VAD, turns and interruptions. Custom applications can pass api_key= or export NARI_API_KEY.

Run the example

The voice agent connects Nari STT → OpenRouter → Nari TTS:

uv run --extra examples python examples/agent.py console

For room mode, set the LIVEKIT_* credentials in .env and replace console with dev.

Documentation

• Configuration, streaming, metrics and recovery
• Tests and measured results
• Changelog · MIT license