Module 4 · Project 3: Build Your Own Voice Agent with LiveKit (web + phone, inbound + outbound)

OBJECTIVE: build the same receptionist yourself in Python with LiveKit Agents — talk to it in the browser, then connect it to real phone numbers for inbound and outbound calls — and understand and control every part: STT, LLM, TTS, VAD, turn detection, tools, hand-offs, recording and latency. WHAT WE COVER (in project order): • LiveKit concepts: rooms, participants, tracks, agent workers • Your first voice agent: STT + LLM + TTS + VAD + turn detection • Tools (function calling) and call state • Multiple agents in one call (hand-offs) • Knowledge (RAG) and Indian-language support • Testing in the console and browser playground • Telephony: SIP trunk, inbound calls, outbound calls, transfer to a human • Speech-to-speech (realtime) model option • Recording, transcripts and latency measurement AFTER THIS MODULE YOU CAN: • Build, run and debug a voice agent in code • Connect it to phone numbers for inbound and outbound calls • Measure and reduce latency yourself TOOLS: Python, LiveKit Agents framework, LiveKit Cloud (free tier) or self-hosted LiveKit server, STT/LLM/TTS provider keys, a SIP trunk (Twilio, Telnyx, Plivo or Exotel).