a.py

import streamlit as st
import pandas as pd
from langchain_community.llms import LlamaCpp
from langchain_core.callbacks import StreamingStdOutCallbackHandler
from langchain_core.prompts import PromptTemplate

# Load the CSV file for Kendra Locator
df = pd.read_csv('location.csv', encoding='Windows-1252')

# Initialize session state for selected service and chatbot history
if 'selected_service' not in st.session_state:
    st.session_state.selected_service = "Kendra Locator"
if 'user_input' not in st.session_state:
    st.session_state['user_input'] = ''


st.set_page_config(layout="centered", initial_sidebar_state="expanded")

st.sidebar.title("KENDR LOCATOR")
st.sidebar.write("Find One Near You!")

display_option = st.sidebar.selectbox("Select:", ["Address", "Email"])
pin_code_input = st.sidebar.text_input("Enter Pin Code:")

if st.sidebar.button("Locate"):
    if pin_code_input:

        result = df[df['Pin'].astype(str) == pin_code_input]
        if not result.empty:
            st.sidebar.write(f"**Name**: {result['Name'].values[0]}")

            if display_option == "Address":
                st.sidebar.write(f"**Address**: {result['Address'].values[0]}")
            elif display_option == "Email":
                st.sidebar.write(f"**Email**: {result['Email'].values[0]}")
        else:
            st.sidebar.write("No results found.")
    else:
        st.sidebar.write("Please enter a pin code.")


llm = LlamaCpp(
    model_path="model.gguf",
    temperature=0.3,
    max_tokens=512,
    top_p=1,
    callbacks=[StreamingStdOutCallbackHandler()],
    verbose=False,
    stop=["###"]
)

template = """You are a knowledgeable, polite, and contextually aware assistant. Your goal is to provide accurate, relevant, and easy-to-understand answers to users' questions. Always maintain a professional tone and avoid providing offensive or inappropriate responses.

### Question:
{input}

### Response:
{response}"""

prompt = PromptTemplate.from_template(template)

PROFANE_WORDS = [
    "damn", "shit", "fuck", "bitch", "asshole", "dick", "piss", "crap", "cunt",
    "twat", "slut", "whore", "faggot", "nigger", "kike", "chink", "gook", "spic",
    "dyke", "suck", "cock", "pussy", "motherfucker", "bastard", "prick", "wanker",
    "bollocks", "arse", "bloody", "bugger", "tosser", "git", "slag", "pillock",
    "knob", "knobhead", "wazzock", "clit", "scrotum", "fanny", "ass", "freak",
    "bimbo", "dumbass", "jackass", "wimp", "idiot", "moron", "loser", "fool",
    "retard", "cocksucker", "shag", "shagger", "piss off", "go to hell",
    "hell", "dammit", "son of a bitch", "jerk", "puke", "chut", "chutiyah",
    "bhosdike", "bhenchod", "madarchod", "gandu", "gand", "bhancho",
    "saala", "kameena", "bhenji", "bhadwa", "kothi", "aankhmar", "launda",
    "bhikari", "sala", "billi", "bhosdika", "kothi", "sundar", "langda",
    "kaamchor", "gaddha", "bakra", "chudiya", "gando", "bhencod", "lanat",
    "bhoot", "chakkar", "chutak", "haramkhor", "bandar", "banda", "bakwas",
    "nikamma", "pagal", "nalayak", "pagal", "khota", "madharchod"
]


def contains_profanity(text):
    """Check if the text contains any profane words."""
    return any(word in text.lower() for word in PROFANE_WORDS)


def truncate_at_full_stop(text, max_length=512):
    if len(text) <= max_length:
        return text

    truncated = text[:max_length]
    print(f"Truncated text: {truncated}")

    last_period = truncated.rfind('.')
    print(f"Last period index: {last_period}")

    if last_period != -1:
        return truncated[:last_period + 1]

    return truncated


if st.session_state.selected_service == "Kendra Locator":
    st.title("MedAI")

    user_input = st.text_input("Your Queries:", key='temp_user_input')

    if st.button("Ask Away"):
        if user_input:
            if contains_profanity(user_input):
                st.markdown("<span style='color: red;'>Mind The Language Dear!</span>", unsafe_allow_html=True)
            else:
                formatted_prompt = prompt.format(
                    input=user_input,
                    response=""
                )

                response = llm.invoke(formatted_prompt)

                truncated_response = truncate_at_full_stop(response)
                st.markdown(f"**You:** {user_input}", unsafe_allow_html=False)
                st.markdown(f"**MedAI:** {truncated_response}.", unsafe_allow_html=False)


st.warning("Developer's notice : Responses are generated by AI and maybe inaccurate or inappropriate. Any received medical or financial consult is not a substitute for professional advice.")