Best approach to adding an INLINE audio processing block between a user call and AWS Connect Agent?

A user calls in to an AWS Connect number and a flow is triggered. An agent answers the call. So far this is the normal flow.

Now, we wish to "process" this audio stream that is being sent to/from the user, i.e insert our code INLINE to the audio stream

I have seen flows where a copy of the audio can be split and sent off elsewhere. This is different in that here we wish to process the caller audio INLINE, and then forward it to the rest of the flow (As an example, the use case could be use ASR and take some action on the stream before handing it off to the agent) Same on the way out. After agent has spoken, we want to manipulate the stream before sending it off to the customer)

Is it possible to do within AWS Connect, or will we have to add an external block using Chime or another external programmable voice service in front of AWS Connect which would then forward the call to AWS Connect?

Think of it as a voice bot that forwards/relays to AWS Connect (rather than transfer)