最后更新：9/23/2026最后核验：2026-09-23
<p>LiteParse 面向需要快速、可在本机运行的文档处理流程。它不仅能提取 PDF 文本，还能保留文字在页面上的位置、生成带标题、表格、列表、图片和链接的结构化 Markdown，并获取嵌入图片、矢量图、注释和表单字段等 PDF 数据。对于扫描件，可使用内置 Tesseract 或外部 OCR 服务；在真正解析前，<code>is-complex</code> 还能识别扫描页、乱码文本、多栏、表格和密集图形，帮助应用把困难文档路由到更合适的处理链路。LiteParse 以 Apache-2.0 开源，可通过 CLI 或 Node.js、Python、Rust、WASM 库接入，本身不依赖云端 API 或 LLM。</p><ul><li>将 PDF、Office 文档和图片解析为带布局信息的文本与结构化 Markdown</li><li>提取文字边界框、图片、矢量图、注释、表单字段等文档数据</li><li>支持内置或外接 OCR，并用复杂度检测识别扫描件、多栏和表格页面</li><li>提供 CLI、Node.js、Python、Rust 与浏览器 WASM 接口，可本地运行</li></ul>
详情
平台
网页版macOSWindowsLinux
功能特性
- Parses PDFs, Office documents, and images into spatially grounded text and structured Markdown
- Extracts bounding boxes, embedded images, vector graphics, annotations, form fields, and tagged PDF structures
- Runs built-in or external OCR and detects scanned, multi-column, table-heavy, and otherwise complex pages before routing
- Provides CLI and library APIs for Node.js/TypeScript, Python, Rust, and browser-based WASM workflows
支持语言
en
适合人群
Developers building local RAG, coding-agent, or real-time document-processing workflowsTeams needing structured Markdown, layout coordinates, OCR, and PDF artifacts without cloud API dependenciesApplications that route scanned or layout-heavy documents based on a fast complexity check
已知限制
- LiteParse runs locally and does not add cloud models or LLM reasoning, so complex documents can require routing to a more advanced parsing pipeline
- OCR quality and support depend on the configured built-in Tesseract or external OCR server
- Document-complexity scoring identifies difficult layouts but does not itself resolve extraction errors
- The project documentation and primary interfaces are currently English-first
替代工具
LlamaParse — Hosted LlamaIndex parsing platform for more advanced cloud document-processing workflows
MarkItDown — Open-source document-to-Markdown converter with a different extraction and layout focus
Unstructured — Document-ingestion platform with hosted and self-managed processing options
