To interrupt a session with the OpenAI realtime model and update the agent gracefully without encountering errors like "AgentSession is closing, cannot use generate_reply()", follow these steps: 1. Interrupt the Session: Use the

interrupt()

method of

AgentSession

to stop the current speech and update the chat context. This method returns an

asyncio.Future

that completes once the interruption is fully processed. 2. Await the Interruption: Ensure you await the returned

Future

from

interrupt()

to allow the operation to complete before proceeding. 3. Check Session Activity: After the interruption, verify if the session activity is still running. If it is, you can safely call

generate_reply()

to produce a new response. Here’s an example implementation:

async def handle_interruption(session: AgentSession):
    try:
        # Interrupt the current session
        await session.interrupt()
    except RuntimeError as e:
        logger.warning(f"Interrupting the session failed: {e}")
        return

    # Check if the session is still active
    if session._activity is None:
        logger.warning("Session activity is closed, cannot generate reply.")
        return

    # Generate a new reply after interruption
    try:
        handle = session.generate_reply()
        await handle
    except RuntimeError as e:
        logger.warning(f"Generating reply after interruption failed: {e}")

This ensures the session is interrupted properly and the agent can continue processing without errors. Sources: livekit-plugins/livekit-plugins-openai/livekit/plugins/openai/realtime/real...