app/analyzers/summary.py

"""
Extractive summarization via TextRank (sumy).
Ringan, tanpa model berat.
"""
from typing import List, Dict
import re


def summarize(text: str, num_sentences: int = 3) -> Dict:
    text = (text or "").strip()
    if not text:
        return {"summary": "", "sentences": []}

    try:
        from sumy.parsers.plaintext import PlaintextParser
        from sumy.nlp.tokenizers import Tokenizer
        from sumy.summarizers.text_rank import TextRankSummarizer

        parser = PlaintextParser.from_string(text, Tokenizer("english"))
        summarizer = TextRankSummarizer()
        sentences = [str(s) for s in summarizer(parser.document, num_sentences)]
        if sentences:
            return {"summary": " ".join(sentences), "sentences": sentences}
    except Exception:
        pass

    # Fallback: ambil N kalimat terpanjang (proxy informativeness)
    raw = re.split(r"(?<=[.!?])\s+", text)
    raw = [s.strip() for s in raw if len(s.strip()) > 40]
    raw.sort(key=len, reverse=True)
    picked = raw[:num_sentences]
    return {"summary": " ".join(picked), "sentences": picked}