Hi all,
I’m using GPT-4’s Realtime/Streaming API over WebRTC to enable live, two-way audio conversations in a browser. Everything is working, but I’m finding that the AI sometimes references previous conversation context across new sessions.
I need each new WebRTC connection to start with a fresh context. Does anyone know if there’s a parameter or recommended way to ensure each session does not inherit old conversation memory (e.g., something like session_id, conversation_id, or a specific “reset context” flag)?
Thanks