import streamlit as st
from transformers import T5Tokenizer, AutoModelForSeq2SeqLM

# Load the Hugging Face model with SentencePiece tokenizer
@st.cache_resource
def load_model():
    tokenizer = T5Tokenizer.from_pretrained("Vamsi/T5_Paraphrase_Paws")
    model = AutoModelForSeq2SeqLM.from_pretrained("Vamsi/T5_Paraphrase_Paws")
    return tokenizer, model

# Load the model and tokenizer
tokenizer, model = load_model()

# Streamlit app interface
st.title("Paraphrasing Tool - AI to Human")
st.write("Paste your AI-generated text below, and the tool will humanize it:")

# Input text box
input_text = st.text_area("Enter text here (no word limit):")

if st.button("Paraphrase"):
    if input_text.strip():
        with st.spinner("Paraphrasing... Please wait."):
            try:
                # Prepare input for the model
                inputs = tokenizer.encode("paraphrase: " + input_text, 
                                          return_tensors="pt")
                
                # Generate paraphrased output
                outputs = model.generate(
                    inputs, 
                    num_beams=5, 
                    temperature=0.7, 
                    early_stopping=True
                )
                paraphrased_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
                st.success("Here is the paraphrased text:")
                st.write(paraphrased_text)
            except Exception as e:
                st.error(f"An error occurred: {e}")
    else:
        st.error("Please enter some text to paraphrase.")