The Interruptible Version

In order to make this truly interactive, we need to add the ability to interrupt the bot. There's a method in the transport for that. Here's what the end of our app looks like now:

        async def run_conversation():
            messages = [
                {
                    "role": "system",
                    "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be converted to audio. Respond to what the user said in a creative and helpful way.",
                },
            ]

            await transport.run_interruptible_pipeline(
                pipeline,
                post_processor=LLMResponseAggregator(messages),
                pre_processor=UserResponseAggregator(messages),
            )

        await asyncio.gather(transport.run(), run_conversation())

You can see the full version in examples/foundational/07-interruptible.py. Instead of passing the pipeline into the transport, you'll need to run it as a separate coroutine, and use asyncio.gather to wait for both of them.

You'll need to set enable_vad=True when you initialize your transport object. When you do that, the transport will run Voice Activity Detection (VAD) on incoming audio in a separate thread. When it detects that the user has started speaking, it will emit a UserStartedSpeakingFrame. The run_interruptible_pipeline method runs your pipeline in an asyncio.task while watching for those frames. When it receives a UserStartedSpeakingFrame, it cancels the currently running task and empties its receive queue, which basically stops everything. Then, the function starts a new asyncio.task with your pipeline again.

To move beyond this example, you'll want to start implementing your own custom pipeline services. We'll see how to do that in the next section.