import streamlit as st
from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification
import torch

# Define the summarization pipeline
summarizer_ntg = pipeline("summarization", model="mrm8488/t5-base-finetuned-summarize-news")

# Load the tokenizer and model for classification
tokenizer_bb = AutoTokenizer.from_pretrained("your-username/your-model-name")
model_bb = AutoModelForSequenceClassification.from_pretrained("your-username/your-model-name")

# Streamlit application title
st.title("News Article Summarizer and Classifier")
st.write("Enter a news article text to get its summary and category.")

# Text input for user to enter the news article text
text = st.text_area("Enter the news article text here:")

# Perform summarization and classification when the user clicks the "Classify" button
if st.button("Classify"):
    # Perform text summarization
    summary = summarizer_ntg(text)[0]['summary_text']
    
    # Tokenize the summarized text
    inputs = tokenizer_bb(summary, return_tensors="pt", truncation=True, padding=True, max_length=512)
    
    # Move inputs and model to the same device (GPU or CPU)
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    inputs = {k: v.to(device) for k, v in inputs.items()}
    model_bb.to(device)
    
    # Perform text classification
    with torch.no_grad():
        outputs = model_bb(**inputs)
    
    # Get the predicted label
    predicted_label_id = torch.argmax(outputs.logits, dim=-1).item()
    label_mapping = model_bb.config.id2label
    predicted_label = label_mapping[predicted_label_id]
    
    # Display the summary and classification result
    st.write("Summary:", summary)
    st.write("Category:", predicted_label)