import argparse, json, re, sys, os from pathlib import Path from collections import Counter from tqdm import tqdm
# app.py from flask import Flask, request, jsonify from safe_summary import process_pdf # <-- the file above (rename to safe_summary.py) Malayalam Kambi Kadakal Amma.pdfl
It is geared toward PDFs that may contain adult‑oriented Malayalam text (e.g. Malayalam Kambi Kadakal Amma.pdf ), but it – it only returns a short, neutral summary, language detection, and a “content‑warning” flag. import argparse, json, re, sys, os from pathlib
| Step | Code snippet | Explanation | |------|--------------|-------------| | | extract_text_from_pdf() | Uses pdfplumber for text‑based PDFs; falls back to pytesseract when the page looks scanned. | | Detect language | detect_language() | langdetect + a quick Malayalam‑character ratio check (ensures we don’t mis‑classify English‑heavy PDFs). | | Adult‑flag | is_adult_content() | Normalises every token, counts hits against the curated adult‑keyword set. | | Summarise | summarise() | Embeds each sentence with a multilingual MiniLM model, selects the most “central” sentences – these tend to be plot‑related, not explicit. | | Translate (optional) | translate() | Leverages Google‑Translate API (free, no key required). Swap in any LLM‑based translation if you prefer. | | Output | JSON | Easy to pipe into a front‑end, store in a DB, or feed to another micro‑service. | | | Detect language | detect_language() | langdetect
# ------------------------------------------------------------ # 8️⃣ Main orchestration # ------------------------------------------------------------ def process_pdf(pdf_path: Path, translate_to: str = None) -> dict: raw_text = extract_text_from_pdf(pdf_path)