improved query from user prompt

remove one suggested question

fixes for better inference, model selection and ingestion

more fixes and suggested questions

working perfectly on docker locally

running HF streamlit with files