THE AI solution that customers need...

I now use Livekit / +Agents 100% for voice and chat, as it is a closed ecosystem and 100% open source. A RAG DB (in my case, a qdrant) typically serves as the basis for the agents. The data from any ERP is “extracted” via n8n automation and stored in the qdrant in the correct format.

This enables the following scenario (out of the box):

Standardized provision of the complete Livekit environment via Unix server template
Agent templating and provisioning through a central (agent) code platform (local or cloud)
Minimal agent code adaptation for various scenarios such as medical, financial, support agent, etc.
Tool and function calling for transferring collected data to other agents
SIP integration for classic phone calls (incoming and outgoing)

Advantages of Livekit:

100% open source = 100% local installation
Multi-agent capable (e.g., agents per department)
Multi-tenant capable (e.g., multiple customers per instance)
Voice and vision capable (LLM support required)
API interface (connection to ERP, etc.)
Full RAG support (e.g., qdrant)
Full STT/TTS support
Integration of local/cloud LLMs
Standardized code based on Python
Modern front-end solutions for web and mobile

Which companies use Livekit for their “AI”?

OpenAI

HeyGen

Elevenlabs

Tinder

Here a presentation of a medical agent, realized by us with livekit local install (german Language presentation): https://www.youtube.com/watch?v=vH2WIr0cFf8

12 comments