fastrtc_canary 0.0.1

The realtime communication library for Python - fastrtc with Nvidia's Canary STT

FastRTC Canary

The Real-Time Communication Library for Python with Canary STT

Turn any python function into a real-time audio and video stream over WebRTC or WebSockets.

Installation

Assume you have fastrtc installed - fastrtc

pip install fastrtc_canary

Features

• 🎯 Direct integration with Nvidia's Canary STT model
• 🔄 Seamless compatibility with all FastRTC features
• 🚀 Real-time audio transcription
• 🔌 Drop-in replacement for FastRTC's STT components

Quick Start

from fastrtc_canary import get_stt_model as get_stt_model_canary
from fastrtc import (ReplyOnPause, Stream, get_tts_model, list_stt_models)
from groq import Groq

print(f"Available STT models: {list_stt_models()}")  
# Available STT models: ['moonshine/base', 'moonshine/tiny', 'canary/1b']

client = Groq()
stt_model = get_stt_model_canary(
                        "canary/1b",
                        lang="en"  # Optional, defaults to "en"
                    )
tts_model = get_tts_model()

def echo(audio):
    prompt = stt_model.stt(audio)

    response = (
        client.chat.completions.create(
            model="llama-3.1-8b-instant",
            max_tokens=200,
            messages=[
                    {"role": "system", "content": "You are a helpful assistant."},
                    {"role": "user", "content": prompt}
                ]
        )
        .choices[0]
        .message.content
    )

    for audio_chunk in tts_model.stream_tts_sync(response):
        yield audio_chunk

stream = Stream(ReplyOnPause(echo), modality="audio", mode="send-receive")

stream.ui.launch()

Requirements

• Python >= 3.10
• FastRTC >= 0.0.14
• Groq >= 0.20.0

Documentation

For more detailed information about the base FastRTC package, visit https://fastrtc.org

License

MIT