[FEATURE] KoboldCpp as a local OpenAI API provider in the Docs

[BadlyDrawnBoy]

Describe the feature you'd like I would like to suggest to include a reference to KoboldCpp (https://github.com/LostRuins/koboldcpp) as an API provider for a local LLM in the docs. In my case with dolphin-2.7-mixtral-8x7b.Q5_K_M on my Lenovo Legion 5 pro with a rtx3070 dGPU running Debian 12, it's much more performant and easier than setting up LocalAI, where I spent almost a weekend trying to compile it from source, then trying it with docker and cuda, only to find out that the model gives (mostly) bad results or/and is slow like a c64... (in the way that you can brew and drink a coffee between answers, even though cuda and the gpu are used and you tried to enable and optimize all possible features in the model configuration...)