Skip to content

Changelog

0.1.0 (2026-11-14)

First public release.

  • readanydoc.parse and readanydoc.extract as a Python library; extract takes a pydantic class or JSON Schema and returns typed values with citations.
  • Two modes: engine (default) and model-only, both on Claude Haiku 5.5 at high effort; handwritten pages go to Claude Sonnet 5.5.
  • Every model the pipeline uses can be chosen, with its effort and any provider request options. Anthropic, OpenAI and OpenAI-compatible endpoints.
  • Reducto-compatible HTTP server: parse, extract, split, classify, edit, async jobs, webhooks, citations.
  • Benchmarks against Reducto r-1 on olmOCR-bench, OmniDocBench, RD-TableBench, FUNSD and long-document extraction (BENCHMARKS.md).