LiteParse

Fast, local PDF & document parsing with OCR, screenshots, and markdown output — powered by LiteParse

Actions2

Parse Document

AI-generated

Summary

Parse each input item's document from the configured binary field by invoking the LiteParse CLI. Return extracted content as Markdown, plain text, or a reduced JSON page summary, with optional OCR and page, image, and link settings.

Inputs

  • Input Binary Field (required) — Binary property containing the document on each input item; defaults to data.
  • Output Format — Choose Markdown (default), Text, or JSON. Markdown can include image placeholders and rendered links; JSON returns extracted text plus page count and per-page number, dimensions, and text.
  • Enable OCR — Run OCR for scanned documents and images; enabled by default. When enabled, configure OCR Language and choose built-in Tesseract or an external OCR server.
  • OCR Language — Tesseract language code, default eng; supports combined codes such as ara+eng. Used when OCR is enabled.
  • Use External OCR Server — Use an HTTP OCR server instead of built-in Tesseract; disabled by default. When enabled, provide its URL.
  • OCR Server URL — URL of the external server implementing the LiteParse OCR API; used when external OCR is enabled.
  • Tessdata Path — Directory containing Tesseract .traineddata files; used with built-in OCR when provided.
  • Target Pages — Optional page selection such as 1-5,10; empty means parse all pages.
  • Max Pages — Maximum pages to parse per document; defaults to 1000.
  • DPI — Rendering resolution for OCR; defaults to 150 DPI.
  • Image Mode — For Markdown output, either emit image references in reading order (default) or strip images.
  • Extract Links — For Markdown output, render hyperlink annotations as [text](url); enabled by default.
  • Password — Password for an encrypted or protected document; empty by default.

Output shape

One output item per successfully parsed input item, containing extracted content in json and the original input binary data.

Markdown and Text return {text, outputFormat}. JSON returns {text, pages, outputFormat, pageData}, where pageData contains each page's number, width, height, and text; other JSON fields from the CLI are not included. With Continue On Fail enabled, a failed item instead returns {error} and is paired with its input item.

Links

Discussion