Crate harumi 

harumi

Pure-Rust PDF library — CJK font embedding (Chinese/Japanese/Korean), OCR text overlay, text extraction, HTML→PDF, page merge/split. Zero C/C++ dependencies. WASM-compatible.

Use cases

Scenario	Key API
OCR invisible text layer	add_invisible_text · add_invisible_text_runs
AI / RAG text extraction	extract_text_runs · extract_text_chunks · extract_as_markdown
PDF watermark / stamp	add_text · add_text_with_rotation
Scanned PDF → searchable	add_invisible_text + hOCR helpers (ocr feature)
HTML → PDF	render_html_to_pdf (html feature)
PDF text replacement	replace_text · replace_text_resubset
Page merge / split	merge_from · extract_pages
WASM / Edge / Lambda	All APIs — zero C/C++ dependencies

Motivation

Rust lacks a high-level, zero-C-dependency library for injecting text into existing PDFs. Low-level crates like lopdf expose the raw PDF object graph and require manual CID font assembly. harumi wraps that complexity behind a simple, ergonomic API.

Quick start

use harumi::{Document, TextRun};

let mut doc = Document::from_file("scanned.pdf")?;
let font = doc.embed_font(include_bytes!("../tests/fixtures/NotoSansJP-Regular.ttf"))?;

// Invisible OCR text layer
doc.page(1)?.add_invisible_text("日本語テキスト", font, [72.0, 700.0], 12.0)?;

// Visible red label
doc.page(1)?.add_text("CONFIDENTIAL", font, [72.0, 750.0], 18.0, [0.8, 0.0, 0.0])?;

doc.save("output.pdf")?;

Coordinate system

All coordinates are in PDF points (1 pt = 1/72 inch). The origin is at the bottom-left of the page. Use page.size() to query the page dimensions and position text relative to them.

Font subsetting

embed_font stores the raw TTF bytes without processing. At save time, harumi collects every character used across all pages, runs a single subset per font, and embeds the result. This means subsetting overhead is paid once regardless of how many pages or text runs reference the same font.

Feature flags

Flag	Enables	Extra deps
ocr	hOCR pixel→PDF coordinate helpers	none
draw	Shapes: rect, line, ellipse, polygon, path	none
image	JPEG/PNG embed + extraction; enables draw	png crate
flow	FlowDocument auto-pagination builder + headers/footers	none
html	HTML→PDF renderer; enables flow	none (internal tokenizer)

Re-exports

pub use signature::SignatureInfo;

Modules

signature

Digital signature verification for PDFs.

Structs

Document

An existing PDF document that can be annotated with text overlays.

FontHandle

Opaque handle to a font registered with crate::Document::embed_font.

FormField

A PDF form field returned by Document::form_fields.

PageHandle

A handle to a specific page for queuing text overlays.

PdfMetadata

PDF /Info dictionary fields.

TextChunk

A semantic text chunk extracted from a page.

TextFieldOptions

Options for creating a text field via Document::add_text_field.

TextFragment

A text fragment extracted from a page content stream.

TextRun

A single text placement descriptor for use with PageHandle::add_invisible_text_runs.

Enums

ChunkType

The semantic type of a text chunk.

Color

A color value that can be either RGB or CMYK.

Error

Errors returned by harumi operations.

FieldType

The type of a PDF form field.

VerticalAlign

Vertical alignment for PageHandle::add_text_box_aligned.

Functions

sort_by_reading_order

Sort text fragments by reading order: top-to-bottom, then left-to-right.

Type Aliases

Result

Alias for std::result::Result<T, harumi::Error>.