Escriturários gerais de escritório Local OCR / document parsing for visual files on disk, using the local baidu/Unlimited-OCR model on Apple Silicon (CPU). Invoke whenever the user points at a local path and wants the text or layout pulled out into saved Markdown: a single image (screenshot, photo, receipt, menu, invoice, iPhone HEIC), a whole folder of page images, or a multi-page scanned PDF (a scanned book or document). Trigger phrasings: "OCR this", "extract the text from this scan / PDF / photo", "read this scanned book", "turn this folder of scans into markdown", "识别这张图/扫描件/发票/古籍上的文字", "把扫描的 pdf 转成文字". Handles JPG/PNG/WebP/TIFF/BMP, iPhone HEIC, image directories, and multi-page scanned PDFs (English + Chinese), writing results under outputs/<name>/. Especially preferred over reading images one-by-one when there are many pages/files or the user wants a saved markdown deliverable — even if they don't say "Unlimited-OCR" by name. Do NOT use for: editable/text-based PDFs, Office/Word/Excel files, audio or video, image compression/resizing,
2026-06-26