Thank you! That shows the server can start, but the model is ~6gb and it's trying (and failing) to cram it onto 3gb of VRAM, so it's failing to offload the rest to ram to CPU. I'm going to make some updates to hopefully fix you're issue. Apologies and thank you for reporting!
I didn't account for a mixed GPU+CPU fallback but it should be implemented now. I tested virtually on my machine, limiting the VRAM to 3gb and it seemed like it worked so I uploaded the fixes. I also added a "No Model" download, grab that one instead of the zip so you dont have to redownload the 5GB model lol. Once you unzip that one, you'll have to grab the model from the models folder in the first download you did and put it in the models folder for the No Models download. Let me know if you have questions or other issues!
I pushed a quick update for error logging, this won't fix anything, but could you grab the latest version and run it again? Either the normal one or the no model, if you grab the no model make sure you grab the 5gb file out of the other model folder and put it in the new model folder. It should be named "Gemma-4-E4B-Uncensored-HauhauCS-Aggressive-Q5_K_P.gguf" and put it directly in the models folder of the new one. Once you do that, can you launch the game from that new download and wait till it errors again, then close it and send me the log/logs from this folder: %APPDATA%\Monster Girl Chat\logs\
If that doesn't work, the log should be in C drive -> users -> AppData (this is a hidden folder) -> Roaming -> Monster Girl Chat -> logs
The log might be kind of long, you can email it to me if it's too big for here: freerangeslop@gmail.com