Loading local LLM's - own implementation

[jacobdab]

Hello,

I did own implementation of LoadManager that can load many different LLM's into one GPU, managing VRAM utilizing HuggingFace transfomers, safesensors, bitsandbytes, airllm. In docs I didn't find any information to using own loader instead of Ollama, HuggingFaceEndpoint, LM Studio etc.

How I can use my loader and integrate it into CrewAI?