Summarises a PDF in the browser with a local extractive algorithm — no AI service is called and nothing is uploaded. Sentences are scored by TF-IDF over the document’s own vocabulary, position and overlap with the title and headings, then picked with maximal marginal relevance so the summary does not repeat itself. You also get key terms, optional per-section summaries and the dates, amounts, percentages, emails and links the document mentions. English and Indonesian stop words are recognised; scanned PDFs need OCR first.