Mostafa,
you're right that the real challenge is to forward RTP stream to ASR engine. TTS is not so complicated because you could synthesize a text, trasforming it in a .wav and play back it. It is not very efficient but, in many circumstances, it could work. The very big problem is related to ASR because, if you save the speech utterance by the customer in a file to process later by ASR, you could get a definitely bad experience.
The only alternative you could have is to develop a MRCP interface that could invoke ASR (and TTS) services through the interface you have (websocket). It's not an easy task but you can do it, even because there should be some free MRCP code available on internet.
But, at the end, the most important thing: you're talking about
IBM ASR/TTS. As far as I know, they should provide a MRCP interface and, maybe, also a gRPC interface (you need to upgrade MCP to v9 to use that). We should start a trial with them very soon but they told me they already integrated GVP through a MRCP interface. I would suggest to double check this with them, because it could be the easier solution.
Hope this helps you a bit.
------------------------------
Giuliano Ferri
Indra Italia spa
------------------------------