fix: changing control.py so that it doesnt execute unecessary functions and also changing llm.py to better prompts and better responses

test from page 9 to 13

trying with another torch version

changed the llama2 version to the hf version

token in the tokenizer

added the token to the environ and the llm agent

updating to real llama2

fix: added sqlite in requirements

maybe

first push from pages 9 to 25