How can we implement multi-agent concepts with GPT-realtime?
I want to create an HR interview bot with multiple agents, each responsible for a different topic. The challenge is managing one communication stream with GPT-realtime instead of each agent having its own stream. Any ideas?
Thanks.
the real problem here is that gpt-realtime doesn’t offer multiple source for each session id
using webRTC you are able to connect twice to the same session (one from the server and one from the client) but is still acts as separate sessions.
multi source session would be usefull to enable other agents to send signals (messages) to the gpt-realtime model with the same user session. (without adding a proxy server)