import re
import gradio as gr
from huggingface_hub import InferenceClient

MODEL = "facebook/bart-large-cnn"
client = InferenceClient()


def split_sentences(text):
    return [p.strip() for p in re.split(r"(?<=[.!?])\s+", text.strip()) if p.strip()]


def chunk_text(text, max_words=600):
    chunks, cur, n = [], [], 0
    for s in split_sentences(text):
        w = len(s.split())
        if cur and n + w > max_words:
            chunks.append(" ".join(cur)); cur, n = [], 0
        cur.append(s); n += w
    if cur:
        chunks.append(" ".join(cur))
    return chunks


def summarize_one(t):
    out = client.summarization(t, model=MODEL)
    return (getattr(out, "summary_text", None) or out["summary_text"]).strip()


def run(text):
    text = (text or "").strip()
    if len(text.split()) < 30:
        return "Please paste a longer text (at least ~30 words)."
    try:
        parts = [summarize_one(c) for c in chunk_text(text)]
        combined = " ".join(parts)
        if len(parts) > 1 and len(combined.split()) > 130:
            combined = summarize_one(combined)
        return combined
    except Exception as e:
        return f"The summarisation service is busy or unavailable right now. Please try again in a moment.\n\n({e})"


demo = gr.Interface(
    fn=run,
    inputs=gr.Textbox(lines=12, label="Paste a long text (article, report, notes)"),
    outputs=gr.Textbox(lines=6, label="Summary"),
    title="Text Summarizer",
    description="Abstractive summarisation (BART) via the Hugging Face Inference API. Long inputs are chunked, so it handles full articles.",
    article="Code: https://github.com/delcenjo/text-summarizer",
)

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