app.py

from transformers import pipeline, AutoTokenizer
import gradio as gr

# Load tokenizer for truncation
tokenizer = AutoTokenizer.from_pretrained("sshleifer/distilbart-cnn-6-6")

# Load summarization pipeline
summarizer = pipeline(
    "summarization",
    model="sshleifer/distilbart-cnn-6-6",
)

# Truncation helper 🫶
def safe_truncate(text, max_tokens=1024):
    tokens = tokenizer.encode(text, truncation=True, max_length=max_tokens)
    return tokenizer.decode(tokens, skip_special_tokens=True)

# Summary function with truncation applied
def summarize_article(text):
    short_text = safe_truncate(text)  # prevent model from breaking!
    summary = summarizer(
        short_text,
        max_length=250,
        min_length=100,
        do_sample=False,
    )
    return summary[0]['summary_text']

# Default example
default_article = """New York (CNN)When Liana Barrientos was 23 years old, she got married..."""  # [shortened for demo]
default_summary = summarize_article(default_article)

# Gradio Interface
iface = gr.Interface(
    fn=summarize_article,
    inputs=gr.Textbox(lines=20, label="Article (Read Only)", value=default_article, interactive=False),
    outputs=gr.Textbox(label="Summary (Read Only)", value=default_summary, interactive=False),
    title="⚡ Fast Article Summarizer (CPU Optimized)",
    description="Fast summarization with longer output using CPU only. Inputs and outputs are read-only."
)

iface.launch()