undreamai/LLMUnity

Integrates a RAG (Retrieval-Augmented Generation) system for semantic search across custom data to enhance character knowledge. Built on llama.cpp through the LlamaLib C++/C# wrapper, it supports CPU and GPU inference (Nvidia, AMD, Metal) across PC, mobile, and VR platforms. Runs entirely locally without internet, making it suitable for offline games while supporting both single-instance and remote server architectures.

1,532 stars. Actively maintained with 16 commits in the last 30 days.