Hello! I ran into some issues with koboldcpp - namely, my dual Xeon CPU system does not support AVX2. That crashes the precomplied binaries that are automatically downloaded off of GitHub. I recomplied it from sources and it now runs fine on my system. Unfortunately, even using a wrapper script in place of the original binary, the game fails to load the local GGUF model. Would it be feasable to add an option for the game to just accept a URL to the KoboldCpp API?
I'm not sure what you mean by wrapper script, but you can change the binary the game downloads by replacing koboldUrl in AppData\LocalLow\Three Eyes Software\Silverpine\settings.json. In your case the Cuda11 + AVX1 version should work.
"koboldUrl": "https://github.com/LostRuins/koboldcpp/releases/download/v1.103/koboldcpp-oldpc.exe", "linuxKoboldUrl": "https://github.com/LostRuins/koboldcpp/releases/download/v1.103/koboldcpp-linux-x64-oldpc",
If you're on Windows, the game should also detect any already running Kobold process before the setup dialog and start communicating with it on port 5001. In that case, the game will assume the Kobold process it detected is running the last selected local model, so confirm that modelName in settings.json is correctly set to "Nemo" or "Mistral-Small-3.2".
I'm on Linux. Good to know I can just change the URL in the config! koboldcpp 1.108.1 runs really smoothly.
What I meant by the wrapper script: compiling by hand builds no single binary, just a koboldcpp.py script and library files, so I created a bash wrapper with the expected name to start it up and altered the seal file to match the file size. I'm happy to say i don't need that now! Still, just putting an API URL in the config to use what I already have instead of downloading another miniconda all-in-one package would be more efficient.
