LiteParse — локальный парсинг документов без облака

★ 12.2K

LiteParse is an open-source document parsing library that extracts text with spatial bounding boxes and runs entirely on your own machine — no cloud calls, no proprietary LLM dependencies. Reach for it when you need to parse large batches of PDFs for RAG pipelines or search indexes without sending sensitive files outside your infrastructure — contracts, financial reports, internal documents stay local — or when you want to cheaply check a document's complexity before committing to a heavier pipeline. The core is written in Rust and uses PDFium for text extraction; supported input formats include PDF, DOCX, XLSX, PPTX, and images. Built-in OCR via Tesseract requires zero setup; external OCR servers (EasyOCR, PaddleOCR, custom HTTP) can be plugged in via a standard API. Worker Pool mode enables true parallelism with per-parse timeouts in Python and Node.js. Output formats are Markdown (with headings, tables, and links), structured JSON with bounding boxes, and plain text. Bindings are available for Python (PyPI), Node.js/TypeScript (npm), browser (WASM), and CLI. Runs on Linux, macOS (Intel/ARM), and Windows. The tool extracts and structures text from documents — it does not generate