app.py

import os
import streamlit as st
from typing import List, Dict, Any
from serpapi import GoogleSearch
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

def search_serpapi(query: str, loc: str, api_key: str) -> List[Dict[str, Any]]:
    """
    Search using SerpAPI for the given query and return the results.
    """
    try:
        search = GoogleSearch({
            "q": query,
            "location": loc,
            "api_key": api_key
        })
        results = search.get_dict()
        return results.get("organic_results", [])
    except Exception as e:
        raise Exception(f"An error occurred: {e}")

def convert_to_md_table(data):
    md_table = "| Title | Link |\n| :--- | :--- |\n"
    for item in data:
        title = item['title']
        link = item['link']
        md_table += f"| {title} | [Link]({link}) |\n"
    return md_table

# Load model directly
tokenizer = AutoTokenizer.from_pretrained("jy46604790/Fake-News-Bert-Detect")
model = AutoModelForSequenceClassification.from_pretrained("jy46604790/Fake-News-Bert-Detect")

def call_classifier(text: str):
    inputs = tokenizer(text, return_tensors="pt")
    with torch.no_grad():
        outputs = model(**inputs)
    logits = outputs.logits
    probabilities = torch.nn.functional.softmax(logits, dim=-1)
    label = torch.argmax(probabilities, dim=1).item()
    score = probabilities[0][label].item()
    return {"label": label, "score": score}

# Initialize session state
if 'history' not in st.session_state:
    st.session_state.history = []

if 'user_input' not in st.session_state:
    st.session_state.user_input = ""

if 'score' not in st.session_state:
    st.session_state.score = {'label': 'LABEL_0', 'score': 0.0}

# Streamlit app layout
st.title("Chatbot News Search")

# User input
st.session_state.user_input = st.text_input("What news do you want to search?", st.session_state.user_input)

# Threshold
threshold = 0.7

# Main logic
if st.session_state.user_input:
    st.session_state.history.append(f"User: {st.session_state.user_input}")
    
    if st.session_state.score['score'] > threshold:
        query = st.session_state.user_input
        SERPAPI_API_KEY = os.getenv("SERPAPI_API_KEY")  # Use environment variable for SerpAPI key
        if not SERPAPI_API_KEY:
            st.error("SerpAPI API key not found. Please set the SERPAPI_API_KEY environment variable.")
        else:
            news_results = search_serpapi(query, "New York", SERPAPI_API_KEY)
            formatted_news = convert_to_md_table(news_results)
            st.session_state.history.append(f"Chatbot: Here are the latest news results:\n{formatted_news}")

            user_continue = st.radio("Are you okay with this?", ('Y', 'E'))
            if user_continue == 'E':
                st.session_state.history.append("User exited the conversation.")
            else:
                new_score = call_classifier(query)
                while new_score['score'] < st.session_state.score['score']:
                    st.session_state.history.append("Run the SerpAPI again.")
                    new_score = call_classifier(query)['score']
                    st.session_state.history.append(f"New score: {new_score}")

                    user_continue = st.radio("Are you okay with this?", ('Y', 'E'))
                    if user_continue == 'E':
                        st.session_state.history.append("User exited the conversation.")
                        break

                st.session_state.score = new_score

    else:
        st.session_state.history.append(f"Chatbot: Current score: {st.session_state.score['score']}")
        st.session_state.user_input = st.text_input("Please provide more information to refine the news search:", st.session_state.user_input)
        st.session_state.score = call_classifier(st.session_state.user_input)
        st.session_state.history.append(f"New score: {st.session_state.score['score']}")

# Display chat history
for message in st.session_state.history:
    st.write(message)

st.write("If you want to finish the conversation, please enter 'exit'.")