# Docling CLI Reference Complete CLI options for docling. ## Input/Output Formats ### `--from` Input format. Options: `docx`, `pptx`, `html`, `image`, `pdf`, `asciidoc`, `md`, `csv`, `xlsx`, `xml_uspto`, `xml_jats`, `mets_gbs`, `json_docling`, `audio`, `vtt`, `latex` Default: all formats (auto-detect) ### `--to` Output format. Options: `md`, `json`, `yaml`, `html`, `html_split_page`, `text`, `doctags` Default: `md` (Markdown) ### `--output` Output directory path. Default: current directory (`.`) ## Processing Options ### `--device` Accelerator device. Options: `auto`, `cpu`, `cuda`, `mps`, `xpu` Default: `auto` ### `--ocr` / `--no-ocr` Enable OCR for bitmap content. Default: `True` ### `--force-ocr` / `--no-force-ocr` Replace existing text with OCR. Default: `False` ### `--tables` / `--no-tables` Extract table structure. Default: `True` ### `--table-mode` Table extraction mode: `fast` or `accurate` Default: `accurate` ### `--ocr-engine` OCR engine: `auto`, `easyocr`, `ocrmac`, `rapidocr`, `tesserocr`, `tesseract` Default: `auto` ### `--ocr-lang` Comma-separated language codes for OCR. Example: `--ocr-lang en,pt` ## PDF Options ### `--pdf-backend` PDF backend: `pypdfium2`, `dlparse_v1`, `dlparse_v2`, `dlparse_v4` Default: `dlparse_v4` ### `--pdf-password` Password for protected PDFs. ### `--page-batch-size` Pages processed per batch. Default: `4` ## Pipeline Options ### `--pipeline` Processing pipeline: `legacy`, `standard`, `vlm`, `asr` Default: `standard` ### `--vlm-model` Vision-Language Model preset for PDFs/images. Options: `smoldocling`, `granite_docling`, `deepseek_ocr`, `granite_vision`, `pixtral`, `got_ocr`, `phi4`, `qwen`, `gemma_12b`, `gemma_27b`, `dolphin` Default: `granite_docling` ### `--asr-model` ASR model for audio/video. Options: `whisper_tiny`, `whisper_small`, `whisper_medium`, `whisper_base`, `whisper_large`, `whisper_turbo`, and `_mlx`/`_native` variants. Default: `whisper_tiny` ## Enrichment Options ### `--enrich-code` Enable code enrichment model. Default: `False` ### `--enrich-formula` Enable formula enrichment model. Default: `False` ### `--enrich-picture-classes` Enable picture classification. Default: `False` ### `--enrich-picture-description` Enable picture description model. Default: `False` ### `--enrich-chart-extraction` Convert bar/pie/line charts to tables. Default: `False` ## HTTP Options ### `--headers` JSON string of HTTP headers for URL fetching. Example: `--headers '{"User-Agent": "Mozilla/5.0"}'` ## Output Options ### `--image-export-mode` Image handling: `placeholder`, `embedded`, `referenced` Default: `embedded` ### `--show-layout` Show bounding boxes on page images. Default: `False` ## Debug Options ### `--verbose` / `-v` Verbosity level. `-v` for info, `-vv` for debug. ### `--debug-visualize-cells` Visualize PDF cells. ### `--debug-visualize-ocr` Visualize OCR cells. ### `--debug-visualize-layout` Visualize layout clusters. ### `--debug-visualize-tables` Visualize table cells. ### `--profiling` Summarize profiling details. ### `--save-profiling` Save profiling to JSON. ## Other Options ### `--document-timeout` Timeout per document (seconds). ### `--num-threads` Number of threads. Default: `4` ### `--abort-on-error` Abort on first error. Default: `False` ### `--artifacts-path` Location of model artifacts. ### `--enable-remote-services` Enable remote model services. Default: `False` ### `--allow-external-plugins` Load third-party plugins. Default: `False` ### `--version` Show version info. ### `--help` Show help message.