randomly select from attraction list for better suggestion to user

handle not to save messages and not using history in LLMs

some improvements in prompt and validations

handle city which is not in exist in db

Use the city which is existing in the user question

UI enhancement

optimize prompt

UI enhancement

use classifier.py before search for attractions to remove irrelevant data in prompt

don't send response to websocket when system ask question from llm

format attractions to be readable for llm and optimize and limit tokens for openai prompt

find nearest attractions to user's destination

fix color of chat message tail

generate string instead of tokens in llm responses

Update README.md