bot.py

from flask import Flask, render_template, request
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores import Chroma
from langchain_openai import ChatOpenAI
from langchain.chains import ConversationalRetrievalChain
from langchain.memory import ConversationBufferMemory
from langchain.prompts import PromptTemplate
from bs4 import BeautifulSoup
import dotenv
import os
import re
import time
import warnings
import sys
maxInt = sys.maxsize

# Suppress UserWarnings
warnings.filterwarnings("ignore", category=UserWarning)

app = Flask(__name__)

dotenv.load_dotenv()

CHROMA_PATH = 'chroma_data'

# Prompt template for our conversational retrieval chain
custom_prompt_template = """
You are an AI chatbot named UT BOT.

Use the following pieces of information to answer the user's question.
Your primary goal is to provide accurate information about the University of Texas at Tyler provided the context.
Be respectful, and format the answer well.
Add contect info if the response contains a faculty member.
Also add helpful links to the responses when useful.
Do not make up links that don't exist, only use the links you are provided with.
If you don't know the answer, just say that you don't know and give them a helpful
contact information that will help them or a link that will help, don't try to make up an answer.

Context: {context} 
Question: {question}

Only return the helpful answer and nothing else.
Helpful answer: 
"""

## Creating our prompt
prompt = PromptTemplate(template=custom_prompt_template,
                            input_variables=['context', 'question'])

#  Initialize our LLM model
llm = ChatOpenAI(temperature=0, model="gpt-3.5-turbo-0125")
#  Memory for Chat History
memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True, output_key='answer')

# Initialize embeddings
embeddings = OpenAIEmbeddings(model='text-embedding-3-large', show_progress_bar=True)

# Initialize our Chroma vector store
vectorstore =  Chroma(persist_directory=CHROMA_PATH
                ,embedding_function=embeddings)

# Create our conversational chain
qa_chain = ConversationalRetrievalChain.from_llm(
    llm = llm,
    retriever = vectorstore.as_retriever(search_type="similarity",search_kwargs={'k': 2}),
    memory = memory,
    combine_docs_chain_kwargs={'prompt': prompt},
    chain_type="stuff",
    return_source_documents=True,
    )  

## RUN THE APPLICATION
@app.route('/')
def home():
    return render_template('index.html')

@app.route('/ask', methods=['POST'])
def ask():
    user_input = request.form['question']
    result = qa_chain.invoke({"question": user_input, "chat_history": {}})
    bot_response = result['answer']

    # Remove labels like '[Label]' from the answer
    answer = re.sub(r'\[[^\[\]]+\]', '', bot_response)

    # Check if the answer contains iframe HTML
    if 'iframe' in answer:
        return render_template('iframe.html', iframe_html=answer)
    else:
        # Remove unnecessary characters such as parentheses
        cleaned_answer = re.sub(r'[()\[\]]', '', answer)

        # Initialize Beautiful Soup
        soup = BeautifulSoup(cleaned_answer, 'html.parser')

        # Find all URLs and email addresses in the text
        urls = re.findall(r'\bhttps?://\S+\b', str(soup))
        emails = re.findall(r'\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b', str(soup))

        # Replace each URL with an anchor tag
        for url in urls:
            # Create a new anchor tag
            new_tag = soup.new_tag('a', href=url, target='_blank', rel='noopener noreferrer')
            # Append icon to the link
            icon_tag = BeautifulSoup('<i class="fa-solid fa-arrow-up-right-from-square" style="margin-left: 10px;"></i>', 'html.parser')
            new_tag.append('Click here ')
            new_tag.append(icon_tag)
            # Replace the URL with the anchor tag
            soup = BeautifulSoup(str(soup).replace(url, str(new_tag)), 'html.parser')

        # Replace each email address with a mailto link
        for email in emails:
            # Create a new anchor tag
            new_tag = soup.new_tag('a', href='mailto:' + email)
            new_tag.append('Contact ')
            # Append icon to the link
            icon_tag = BeautifulSoup('<i class="fa-solid fa-envelope" style="margin-left: 10px;"></i>', 'html.parser')
            new_tag.append(icon_tag)
            # Replace the email with the anchor tag
            email_tag_str = str(new_tag)
            soup = BeautifulSoup(str(soup).replace(email, email_tag_str), 'html.parser')

        # Convert back to string and remove any loose characters after links
        answer_with_links = str(soup).strip().rstrip('/. ')

        # Add line breaks
        answer_with_line_breaks = answer_with_links.replace('\n', '<br>')

        # Check if bullet points are needed
        if '•' in answer_with_line_breaks:
            # Split the answer into lines and wrap each line with <li> tags
            lines = answer_with_line_breaks.split('\n')
            bulleted_lines = '<ul>' + ''.join([f'<li>{line}</li>' for line in lines]) + '</ul>'
            return bulleted_lines
        else:
            return answer_with_line_breaks

# Run the Flask app
if __name__ == '__main__':
    app.run(debug=False)