These past days I am trying to figure out how to implement the IPC (inter process communication) between Python (where the AI model will be made) and the C++ “megaclient” that acts as 10 clients that connect to the game server. From what I’ve seen one of the easiest ways to implement it is through binary standard streams because I’ll have to move 14KB of data per frame, and because I will unlock the speed of the server physics simulation, it will be like thousands of frames per second, which will add up pretty fast. But the problem isn’t the actual size of data, but the latency, if I would just save it into a txt file my SSD will be hit with thousands of requests every second, so keeping everything in the RAM is why I chose to try and implement the IPC through binary standard streams.
But I didn’t yet get it working and my brain hurts a little cause for the first time since the beginning of this project I have absolutely no idea what to do or how to actually solve this issue and I kinda feel lost and a bit demotivated, that’s why this week I managed a bit over 1hr of coding per day.
But at the same time I’ve been watching more and more videos about how to build AI models and for the first version I’ll build a standard Backprop model as I’ve read in a paper that Predictive Coding approximates Backprop. The problem is this will have to come after the IPC is implemented successfully, which currently it is not.

If anyone sees and reads this I would happily accept any suggestions on how to implement the IPC between Python and C++.