Crate gaze_document 

Document ingestion + safe-bundle generation for the Gaze runtime.

gaze-document turns a single image (PNG / JPG) or single-page PDF into a SafeBundle: tokenized Markdown, a restorable gaze::Manifest, and a structured OCR + PII BundleReport. PII detection flows through the standard gaze::Pipeline so the manifest stays canonical and reversible (Axis 2 reversibility).

Quickstart

use std::path::Path;

let bundle = gaze_document::clean(
    Path::new("invoice.pdf"),
    Path::new("./safe-out"),
)?;
assert!(!bundle.clean_markdown.is_empty());

Runtime requirements

• tesseract binary on PATH (Tesseract 4.x or 5.x).
• For PDF input: a pdfium dynamic library available to the process. See the crate README for per-OS install instructions.

Feature flags

Flag	Default	What it enables
ocr-tesseract	yes	Tesseract subprocess OCR backend.
pdf-input	yes	Single-page PDF rasterization via pdfium-render.
serde	yes	Serialize / Deserialize for BundleReport.
extract-docling	no	Reserved — future Docling layout adapter (no impl yet).
render-image	no	Reserved — future redacted-preview renderer (no impl yet).

Re-exports

pub use bundle::clean;ocr-tesseract

pub use bundle::BundleReport;

pub use bundle::ClassCount;

pub use bundle::LayoutSummary;

pub use bundle::SafeBundle;

pub use bundle::BUNDLE_VERSION;

pub use layout::ReadingOrder;

pub use ocr::OcrAdapter;

pub use render::Renderer;

Modules

bundle

SafeBundle generation: OCR + Gaze redact → on-disk artifacts.

extract

Input extraction backends.

layout

Layout / reading-order contract surface.

mcpmcp

MCP tool adapters for gaze-document.

ocr

OCR adapter contract surface and concrete backends.

render

Renderer contract surface.

Enums

DocumentError

Crate-level error type for gaze-document.