Improve the inference speed by use of llama.cpp or optimizing the ollama

[avarshvir]

we can improve speed by optimizing the code logic or use of llama.cpp