run-llama/liteparse

Features

• Document Parsing and Extraction - Core document parser that extracts text and bounding boxes from PDFs and other formats into structured output.
• Cost-Optimized Parsers - Routes each page to the cheapest suitable parsing tier automatically to balance accuracy and expense without manual configuration.
• Open-Source Document Parsers - An open-source document parser that extracts text, tables, and layout from PDFs and office files into Markdown or JSON.
• Content Parsing Prompts - Provides prompt-based parsing customization that steers extraction results using natural-language instructions or structured schemas.
• Output Schema Instructions - Ships prompt-driven output shaping that accepts natural-language instructions or structured schemas to steer extraction results.
• Document Page Routing Optimizers - Provides automatic per-page routing to the cheapest suitable parsing tier for cost-efficient document extraction.
• Structured Document Extraction - Converts PDFs and office documents into structured Markdown or JSON with spatial layout for direct use by language models.
• Document Text Extractors - Parses documents to retrieve text alongside precise positional coordinates for each extracted element.
• Spatial Text Extractors - Parses documents to retrieve text alongside precise positional coordinates for each extracted element.
• OCR Document Parsers - Applies optical character recognition using a bundled engine or external HTTP server.
• PDF Text Extractors - Parses PDF files and extracts text with spatial bounding boxes, returning structured Markdown, JSON, or plain text.
• Text Extraction and OCR - Applies OCR to scanned or image-based PDFs to extract text with optional language selection.
• Office and Documents - Extracts text from DOCX, XLSX, PPTX, PNG, JPG, and other file formats via automatic conversion.
• Document Generation from Markdown - Reconstructs headings, tables, lists, images, and links from a PDF's spatial layout into structured Markdown.
• Document Format Converters - Automatically converts over 130 file types including office documents and images into PDF before extracting text and layout.
• Document to Markdown Converters - Reconstructs headings, tables, lists, images, and links from spatial layout for LLM and RAG pipelines.
• LLM-Ready Markdown Converters - Converts PDFs and office documents into structured Markdown optimized for language model and RAG pipeline consumption.
• PDF to Markdown Conversion - Converts PDF documents into structured Markdown preserving headings, tables, lists, images, and links.
• Document Table Extractors - Recovers table data from PDFs, scans, and images with cell structure intact for downstream use.
• Browser-Based Parsers - Runs the entire parsing engine and OCR inside a web browser using WebAssembly for offline or serverless document extraction.
• PDF Spatial Layout Parsers - Extracts text from PDFs while preserving exact position on each page including bounding boxes for every line.
• Multi-Format Document Ingestion - Handles PDF, DOCX, PPTX, XLSX, HTML, JPEG, PNG, XML, EPUB, and many other formats for flexible document ingestion.
• Layout-Aware Extraction - Combines spatial layout analysis with OCR to extract text, tables, and charts preserving document structure.
• Layout Preservation - Extracts text, tables, and images from PDFs and office documents while preserving spatial layout and structure.
• Multi-Format Document Parsing - Converts over 130 file types including office documents and images into PDF before extracting text and layout.
• Document Page Cost Optimizers - Automatically routes each page to the cheapest suitable parsing tier, reserving premium accuracy for complex layouts.
• Document Layout Bounding Box Extractors - Returns precise coordinates for every text line and table cell, preserving document layout for downstream analysis.
• OCR Document Conversion - Extracts text, tables, and charts from PDFs while preserving spatial layout and structure.
• Document Page Cost Optimizers - Automatically routes each page to the cheapest suitable parsing tier, reserving premium accuracy for complex layouts.
• Document Spatial Coordinate Outputs - Extracts text items from a PDF and returns them with spatial coordinates for precise layout analysis.
• Document Bounding Box Extractors - Returns spatial coordinates for every line of text extracted from documents for visualization or processing.
• Document Bounding Box Extractors - Returns spatial bounding boxes for each text line, enabling visualization or further geometric processing.
• Document JSON Bounding Box Outputs - Extracts text with bounding boxes from a PDF and outputs the result as structured JSON.
• Document Output Shapers - Accepts natural-language instructions or structured schemas to steer document extraction results toward desired formats.
• Custom OCR Backend Registrations - Accepts a user-defined OCR engine with a recognize method for custom text extraction.
• Markdown RAG Pipeline Outputs - Reconstructs headings, tables, lists, images, and links from spatial layout for direct use in LLMs and RAG pipelines.
• Multi-Runtime Libraries - Provides library APIs and CLI for Rust, Node.js/TypeScript, Python, and browser WASM environments.
• REST and SDK Parsing Interfaces - Provides REST, Python, and TypeScript interfaces to upload documents and retrieve parsed results programmatically.
• Command-Line Document Processors - Processes files from the command line with options for format, page range, and remote URLs.
• Document Parsing Services - Integrates document parsing into applications through a library API accepting file paths or raw byte buffers.
• WASM-Based PDF Parsers - Parses PDF documents entirely in the browser using WebAssembly, requiring no server or cloud calls.
• Diagram Structure Parsing - Converts visual data from charts, plots, and diagrams into structured formats for numerical reasoning by LLMs.
• Document Chart Parsers - Extracts charts, plots, and diagrams from documents into structured data for numerical reasoning by LLMs.
• Browser-Based OCR Engines - Provides a JavaScript-side OCR engine with a recognize method for text extraction in WASM environments.
• OCR REST API Servers - Sends OCR requests to remote HTTP services for higher accuracy or performance.
• Document Page Rendering - Converts document pages into raster images for LLM agents to extract visual information.
• PDF Page Image Generators - Renders PDF pages as raster images for use in LLM agents or visual workflows.
• Browser Screenshot Capture - Generates page images as PNG byte buffers for use with LLMs or disk storage.
• Browser-Based Runtimes - Runs the entire parsing engine and OCR inside a web browser using WebAssembly for offline document extraction.
• Document Directory Parsers - Parses all documents in a given input folder and writes results to a specified output directory.
• Batch Document Processing - Processes entire directories of documents efficiently with a single command, reusing the parsing engine.
• Document Page Screenshot Capturers - Renders pages as high-quality PNG images to capture visual information that text alone cannot convey.

Star history