This project implements a real-time AI voice assistant using LiveKit Agents, integrating STT, LLM, and TTS for natural conversations. It supports multilingual speech, handles interruptions, and logs detailed session metrics for performance analysis.
From the README
🤖 LiveKit AI Voice Assistant – Internship Task This project implements a real-time AI voice assistant using LiveKit Agents, combining STT, LLM, and TTS in a seamless pipeline. The assistant supports multilingual speech, logs performance metrics, handles conversation interruptions, and responds with natural-sounding speech. --- 🚀 Features • 🎙️ Real-time Speech-to-Text (STT) via Deepgram • 🧠 LLM inference via Groq using llama3-8b-8192 • 🗣️ High-quality Text-to-Speech (TTS) using Cartesia • 🌐 Multilingual voice input support • 🔁 Turn detection and robust VAD • 🔇 Noise cancellation with LiveKit BVC • 📊 Metric logging: EOU delay, TTFT, TTFB, Total Latency • 📄 Exports call...
Skills
- Python
