🎙️ AI Voice Agent Engineer – Interview Exercise

Welcome!
This exercise is designed to test your real-time coding ability, systems thinking, and how you collaborate while building an AI voice agent.
We’ll start small and extend together as we go. Don’t worry about perfection—focus on clarity, iteration, and communication.

---

🥅 Goal

Build a local voice agent using LiveKit Agents (Python).
The agent should:

• Listen to microphone input.
• Transcribe speech into text (with VAD/turn detection).
• Decide how to respond (using a realtime LLM).
• Speak back to the user (TTS).
• Support barge-in (stop talking if the user interrupts).
• Keep short-term memory (e.g., user’s name).
• Provide at least one tool (like telling the current time or setting a timer).

---

⚙️ Setup

1. Install dependencies

# Create and activate virtual environment
python -m venv .venv && source .venv/bin/activate

# Install LiveKit agents and plugins
pip install "livekit-agents[openai,silero,turn-detector]~=1.2" \
            "livekit-plugins-noise-cancellation~=0.2" python-dotenv