Build and deploy a real-time voice agent | LiveKit + Rime

Описание: Nudge voice agent gives teams a deployable foundation for automating document follow-up while preserving human approval. Built with LiveKit and Rime, it combines natural, streaming speech with explicit workflow checks and interruption recovery. The modular reference implementation includes tests and latency measurements, helping engineering leaders assess behavior and integration effort before connecting real account data and business systems.

Read the full tutorial:
https://www.rime.ai/resources/build-a...

The stack uses LiveKit Agents, Deepgram speech recognition, Gemini, and Rime Coda model with the Wawona voice. The demo uses a fictional account and records a request preview. It does not approve an extension or write to a production backend.

The tutorial covers local setup, credentials, an extension-request tool, streamed speech, measuring time to first audio, preserving pending acknowledgments after interruption, and routing a phone call to a deployed worker. The 138 ms median and 146 ms P95 measurements cover ready-text-to-client audio on a warmed TTS connection; they are not end-to-end phone-call latency.

CHAPTERS
0:00 Live phone demo
1:06 Build Nudge with LiveKit and Rime
1:34 Document follow-up use cases
1:54 Project setup and credentials
2:30 Run locally
2:43 Build the extension-request tool
3:32 Inspect the workflow output
4:08 Follow the request through the stack
4:24 Stream speech with Rime Coda
4:59 Hear Wawona in the phone call
5:28 Measure TTS latency and the full turn
6:08 Handle interruptions and recover state
6:48 Deploy and connect a phone number
7:07 Next steps

RESOURCES
Rime: https://www.rime.ai/
Rime documentation: https://docs.rime.ai/
LiveKit Agents: https://docs.livekit.io/agents/

Presented by Ankit Khare, Lead Developer Evangelist at Rime.
English closed captions are available through the CC button.

MUSIC
Synapse by Shane Ivers
https://www.silvermansound.com/free-m...
Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).
https://creativecommons.org/licenses/...
Excerpted, looped, filtered, faded and level adjusted for this video.

#VoiceAI #Rime #LiveKit