Firecrawl released anydoc, an open source Rust library that converts 14 document formats across eight file types into GitHub-flavored Markdown, according to the project's GitHub repository. Supported types include Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV and PDF.
The library detects file format by reading content markers rather than trusting file extensions, and routes every format through what the project calls a unified document model so output stays consistent regardless of source format, the repository says. Median conversion time is under 5 milliseconds per document, according to the project's published benchmarks.
Anydoc ships with bindings for Node.js, Python, WebAssembly and native Rust, plus a command-line tool and a browser-based demo that converts files locally without uploading them, the repository says. It's released under the MIT license and had passed 16,000 GitHub stars as of Thursday.
For anyone feeding documents into an LLM pipeline, a sub-5-millisecond converter that treats fourteen formats as one output shape removes a chunk of the format-detection and cleanup code most retrieval pipelines end up writing themselves.