tesseract-rs

A pure-Rust transcode of Tesseract OCR — not a binding to the C++ library.

The original antimatter15 FFI wrapper (tesseract-sys / tesseract-plumbing) was removed on 2026-06-18 per the operator directive: transcode Tesseract into Rust, do not wrap libtesseract. The OCR is rebuilt leaf-by-leaf in pure Rust, each leaf byte-parity-proven against the C++ original (source: AdaWorldAPI/Tesseract) before it lands. Transcoded primitives ride the OGAR Core (lance-graph-contract) per the Core-First transcode doctrine.

Layout

• crates/tesseract-core — the pure-Rust OCR core. It consumes the UNICHARSET / UNICHAR primitives from the OGAR Core (lance-graph-contract), where each is byte-parity-proven against a libtesseract oracle. Landed so far: the id↔unichar bijection (112/112) and the unichar UTF-8 codec (268/268), plus the per-id properties, script, and other_case accessors (112/112 each), and ids_to_text.

Start with CLAUDE.md (the session contract: Core-First doctrine, the proven self-validating-oracle method, iron rules, and the next leaf). See .claude/plans/ for the transcode plan (tesseract-rs-ast-dll-codegen-v1, tesseract-rs-receive-contract-v1).