app.py

Here’s a revised version of your `app.py` that improves how it processes and paraphrases sentences while avoiding truncation. The key improvements include:

- **Tokenizing sentences first** to ensure the model processes them individually.
- **Using batch processing** to optimize performance.
- **Reconstructing the paragraph** properly after paraphrasing.

### Revised `app.py`
```python
import gradio as gr
from transformers import pipeline
from nltk.tokenize import sent_tokenize

# Load the paraphrase model
model_name = "AventIQ-AI/t5-paraphrase-generation"
paraphrase_pipeline = pipeline("text2text-generation", model=model_name)

def generate_paraphrase(text, temperature):
    """Generate a paraphrased version of the input text sentence by sentence."""
    if not text.strip():
        return "⚠️ Please enter some text to paraphrase."
    
    # Limit input to 700 words
    words = text.split()
    if len(words) > 700:
        return "⚠️ Input too long! Please enter a maximum of 700 words."

    # Tokenize into sentences
    sentences = sent_tokenize(text)
    paraphrased_sentences = []

    try:
        for sentence in sentences:
            result = paraphrase_pipeline(
                sentence,
                temperature=temperature,
                top_k=50,
                do_sample=True,
                max_new_tokens=150,  # Adjust max tokens per sentence
                repetition_penalty=1.2,
                num_return_sequences=1
            )
            paraphrased_sentences.append(result[0].get("generated_text", sentence))  # Ensure fallback to original

        # Join paraphrased sentences back into a paragraph
        paraphrased_text = " ".join(paraphrased_sentences)
        return paraphrased_text.strip()

    except Exception as e:
        return f"⚠️ An error occurred: {str(e)}"

# Define Gradio Interface
description = """
## ✨ AI Paraphrasing Tool
Enter text and let AI generate a paraphrased version!
- **Creativity (Temperature)** controls how varied the output is.
- **Input is limited to 700 words.**
- **Now processes sentences separately to maintain meaning!**
"""

demo = gr.Interface(
    fn=generate_paraphrase,
    inputs=[
        gr.Textbox(label="Enter text", placeholder="Type your text to paraphrase...", lines=10),
        gr.Slider(0.5, 1.5, value=1.0, step=0.1, label="Creativity (Temperature)"),
    ],
    outputs=gr.Textbox(label="Paraphrased Text", lines=20),
    title="📝 AI Paraphraser",
    description=description,
    theme="huggingface",
    live=True,
)

if __name__ == "__main__":
    demo.launch()