llamacpp-for-kobold, a zero dependency KoboldAI compatible REST API interfacing with llama.cpp via ctypes bindings

[LostRuins]

Now that it's somewhat usable, I'd like to share a thing I made: https://github.com/LostRuins/llamacpp-for-kobold

This fork does 2 things mainly:

• Adds ctypes python bindings allowing llama.cpp to load models and generate text directly from python code, without messing with stdout pipes or command line flags, and wrapping it all in a nice llamacpp.dll
• Emulates a KoboldAI compatible HTTP server, allowing it to be used as a custom API endpoint from within Kobold, which provides an excellent UI for text generation.
• Not use any dependencies like Flask or Pybind11 requiring basically 0 installation.
• You can either run the local kobold client or the KoboldAI Lite web client, and connect directly to this server, allowing you to enjoy local text generation with full editing and file saving features.

There is only one pressing flaw related to prompt processing latency, and that is #229 which hopefully someone can solve.