pretext-pdf 0.8.0 → 0.8.2

package/CHANGELOG.md

@@ -1,462 +1,489 @@
-# Changelog
-
-<!-- markdownlint-disable MD024 -->
-
-All notable changes to pretext-pdf are documented here.
-Format: [Keep a Changelog 1.1.0](https://keepachangelog.com/en/1.1.0/)
-
----
-
-## [0.8.0] — 2026-04-19
-
-### Added
-
-- **`qr-code` element** — generate QR codes as inline PDF content using the `qrcode` optional peer dependency. Supports `data`, `size`, `errorCorrectionLevel` (L/M/Q/H), `foreground`/`background` hex colours, `margin`, `align`, `spaceBefore`/`spaceAfter`. Fully serverless — pure JS, no canvas required.
-- **`barcode` element** — generate 100+ barcode symbologies (EAN-13, Code128, PDF417, QR, DataMatrix, etc.) via the `bwip-js` optional peer dependency. Supports `symbology`, `data`, `width`, `height`, `includeText`, `align`, `spaceBefore`/`spaceAfter`. Pure JS, Lambda/Edge safe.
-- **`chart` element** — embed Vega-Lite charts as vector SVG using `vega` + `vega-lite` optional peer deps. Accepts any Vega-Lite `spec`, `width`, `height`, `caption`, `align`. Rendered with `renderer: 'none'` — zero canvas/puppeteer dependency.
-- **`pretext-pdf/markdown` entry point** — `markdownToContent(md, options?)` converts a Markdown string to `ContentElement[]`. Requires optional `marked` peer dep. Supports headings, bold/italic/links (→ rich-paragraph), lists (2 levels), blockquotes, code blocks, and HR.
-- **`pretext-pdf/templates` entry point** — three typed template functions with zero extra dependencies: `createInvoice(data)` (generic invoice with currency, tax, discount, QR payment), `createGstInvoice(data)` (GST-compliant Indian tax invoice with IGST/CGST+SGST, UPI QR, bank details, amount in words), `createReport(data)` (structured business report with optional TOC).
-- **New error codes** — `QR_DEP_MISSING`, `QR_GENERATE_FAILED`, `BARCODE_DEP_MISSING`, `BARCODE_GENERATE_FAILED`, `BARCODE_SYMBOLOGY_INVALID`, `CHART_DEP_MISSING`, `CHART_SPEC_INVALID`, `CHART_RENDER_FAILED`, `MARKDOWN_DEP_MISSING`.
-
----
-
-## [0.7.1] — 2026-04-19
-
-### Changed
-
-- **Upstream pretext pinned to `f2014338487a`** — picks up unreleased CJK opening-bracket annotation fix, Hangul jamo line-walker alignment fix, and two internal line-object churn reductions. No public API changes.
-
-### Fixed
-
-- **List nesting depth enforced at validation** — `ListItem.items` (2nd-level items) now correctly rejects any further `.items` property, matching the documented 2-level maximum. Previously the validation silently passed 3-level data which could cause undefined render behaviour.
-- **3 phase-11 list tests corrected** — test data incorrectly contained 3-level nesting while named "2-level"; data trimmed to match documented contract.
-
----
-
-## [Unreleased]
-
-### Added (Phase 11 — Cross-cutting Enhancements)
-
-- **`floatSpans` on image elements** — rich-text alternative to plain `floatText`. Accepts `InlineSpan[]` for mixed bold/italic/color/link captions beside float images. Mutually exclusive with `floatText` (validated).
-- **2-level list nesting** — `ListItem.items` now supports one further level of nesting (depth 0 → 1 → 2). Unordered marker: `▪`. Ordered: inherits parent counter or restarts via `nestedNumberingStyle: 'restart'`.
-- **Table `rowspan`** — `TableCell.rowspan` spans a cell across multiple rows. Works alongside `colspan`. Origin cell draws background over full span height; continuation rows automatically receive placeholder cells.
-- **`onFormFieldError` callback** — `doc.onFormFieldError: (name, err) => 'skip' | 'throw'` mirrors `onImageLoadError`. Controls render behaviour when a form field fails.
-- **`createFootnoteSet(defs)`** — helper exported from `pretext-pdf` that generates footnote definition/reference pairs with globally unique IDs. Returns `Array<{ id, def }>`.
-- **`renderDate` field** — `doc.renderDate: Date | string` overrides the PDF creation date. Useful for reproducible builds and testing.
-- **`{{date}}` and `{{author}}` tokens** in header/footer text — join existing `{{pageNumber}}` / `{{totalPages}}`. `{{date}}` resolves from `renderDate`; `{{author}}` resolves from `doc.metadata.author`.
-- **`tabularNumbers`** on `rich-paragraph` — digits rendered at uniform slot width (widest digit in font), so columns of numbers align without OpenType TNUM feature.
-- **`smallCaps` + `letterSpacing` per span** — `InlineSpan.smallCaps` and `InlineSpan.letterSpacing` now respected in `rich-paragraph` rendering.
-- **Per-span `fontSize`** — `InlineSpan.fontSize` overrides the element-level font size for that span. Enables mixed-size text in a single paragraph.
-
-### Fixed (Phase 11)
-
-- `resolveTokens()` used `.replace()` (replaces first occurrence only) — changed to `.replaceAll()` for all four tokens.
-- Table span grid: continuation-row cursor was advancing by 1 instead of `colspan` when skipping a spanned column — now advances by full span width.
-- Font family names now validated for safe characters (`/^[a-zA-Z0-9 _-]+$/`) in `requireFamily()` — rejects null bytes and control characters.
-- Annotation `color` and `author` fields now validated in `validateElement()` for both `paragraph` and `heading` annotations.
-- `buildOutlineTree` memoizes `parentIdxOf()` into a pre-computed array — eliminates O(n²) scan for documents with large heading counts.
-- Table grid-line renderer pre-computes active boundary set — eliminates O(rows × cols) inner loop for large tables.
-- `addLinkAnnotation()` re-validates URL scheme at render time (defense-in-depth; `validate.ts` is the primary gate).
-
-### Planned (Phase 9+)
-
-- Phase 9A: Digital signatures (cryptographic PKCS#7 via `@signpdf/signpdf`)
-- Phase 9B: Image floats (text flowing alongside images — requires paginator rewrite)
-- Phase 9C: Font subsetting pre-computation (explicit glyph hints for icon fonts)
-
----
-
-## [0.7.0] — 2026-04-17
-
-### Added
-
-- **6 production templates** (`templates/`) — GST invoice, international invoice, resume, multi-section report, NDA, and meeting minutes. Each is a self-contained `.ts` file outputting a valid PDF. Smoke-tested in Phase 2F Block D.
-- **StackBlitz live demo** (`demo/stackblitz/`) — 4-tab UI (Invoice, Report, Resume, Custom) backed by a Node.js render server. Edit JSON and generate PDFs instantly, no install required. Accessible at the StackBlitz link in the README.
-- **`## Performance` section in README** — measured render times and PDF sizes for 1-page, 10-page, and mixed-element documents. Font subsetting behaviour documented.
-- **Stress tests and benchmarks** (`test/phase-2f-stress.test.ts`) — 32 tests across 4 blocks: large document stress (400-element, 200-row table), edge case stress (CJK, RTL, empty arrays, extreme sizes), timing benchmarks (1-page < 500 ms, 10-page < 5,000 ms), and template smoke tests.
-- **Error code coverage** — new tests for `COLUMN_WIDTH_TOO_NARROW`, `IMAGE_LOAD_FAILED`, `SVG_LOAD_FAILED`, and `ASSEMBLY_FAILED`. 16 of 19 error codes now have direct test coverage.
-
-### Changed
-
-- **`as any` audit** — eliminated 10 casts in `validate.ts` by introducing a typed `FormFieldElement` local binding. The remaining 8 instances (pdf-lib interop, dynamic import, internal back-references) are now documented with one-line comments explaining the constraint.
-- **Comparison article** (`docs/articles/pretext-pdf-vs-pdfmake-2026.md`) — 2,200-word draft covering feature matrix, typography quality, API design, performance, and migration quick-start. Marked `published: false` pending live demo.
-- **Migration guide** (`docs/MIGRATION_FROM_PDFMAKE.md`) — 30+ pdfmake → pretext-pdf mappings, complete before/after invoice example, and a quick-start checklist. Linked from README.
-
-### Fixed
-
-- **Phase 2F test types** — `fontWeight: 700 as 700` cast in pre-constructed rows array; removed non-existent `creationDate` from `DocumentMetadata`; replaced `allowCopying: false` with correct `encryption: { permissions: { copying: false } }`.
-- **StackBlitz integration** — added `.stackblitzrc` so WebContainer auto-runs `npm start` and opens the browser preview on port 3000.
-
----
-
-## [0.5.3] — 2026-04-16
-
-### Changed
-
-- **Upgraded `@chenglou/pretext` from 0.0.3 to 0.0.5** — picks up improved text analysis accuracy (~35% larger analysis module), better measurement precision, extracted bidi-data module for cleaner tree-shaking, and new `rich-inline` export (not yet used by pretext-pdf). No breaking changes — `prepareWithSegments()` and `layoutWithLines()` APIs are unchanged. All 223 tests pass, 3 example PDFs visually verified (RTL, TOC, hyperlinks).
-
----
-
-## [0.5.2] — 2026-04-13
-
-### Added
-
-- **`onImageLoadError` callback on `PdfDocument`** — gives callers control over image load failures. Return `'skip'` to silently omit the image (preserves existing default behavior). Return `'throw'` to abort rendering with the original error. Previously, all image failures were silently downgraded to `console.warn` with no way to detect them programmatically.
-
-  ```typescript
-  await render({
-    content: [...],
-    onImageLoadError: (src, error) => {
-      myLogger.warn('Image skipped', { src, error })
-      return 'skip'   // or 'throw' to abort
-    }
-  })
-  ```
-
----
-
-## [0.4.0] — 2026-04-08
-
-### Breaking Changes
-
-- **Migrated from `pdf-lib` to `@cantoo/pdf-lib`** — `@cantoo/pdf-lib` is now a direct `dependency` (always installed). Previously it was an optional peer dependency required only for encryption. This removes the `ENCRYPTION_NOT_AVAILABLE` error code and the separate `npm install @cantoo/pdf-lib` installation step. Encryption now works out of the box.
-- **`ENCRYPTION_NOT_AVAILABLE` error code removed** — encryption is now always available. Update any `switch` statements that handled this code.
-
-### Why this change
-
-`pdf-lib` (the original) has not received a meaningful commit since November 2021. `@cantoo/pdf-lib` is the actively maintained fork (v2.6.5, 107+ releases, MIT license). pretext-pdf was already using `@cantoo/pdf-lib` for encryption — this commit makes it the single source of truth for all PDF operations.
-
-### Added
-
-- `test/pretext-api-contract.test.ts` — canary test that asserts `@chenglou/pretext` exports the exact functions pretext-pdf depends on. Breaks loudly if pretext changes its API.
-- `docs/ROADMAP.md` — public multi-phase development plan
-
-### Changed
-
-- `@chenglou/pretext` version pinned to exact `0.0.3` (no caret) — prevents surprise breaking changes from upstream auto-updates
-- `test:contract` script added — runs the pretext API contract test before the full test suite
-- All internal comments updated from `pdf-lib` to `@cantoo/pdf-lib`
-
----
-
-## [0.3.1] — 2026-04-08
-
-### Fixed
-
-- **Critical: Font resolution when installed as npm package** — `@fontsource/inter` is now resolved via `createRequire(import.meta.url)` instead of a hardcoded relative path. Previously, `path.join(__dirname, '..', 'node_modules', '@fontsource', 'inter', ...)` failed when npm hoisted the dependency to the consumer's top-level `node_modules`, causing `FONT_LOAD_FAILED` on every install. Now resolves correctly regardless of npm hoisting behavior.
-
----
-
-## [0.3.0] — 2026-04-08
-
-### Added (Phase 8B — Interactive Forms)
-
-- New `form-field` element type — creates interactive AcroForm fields in PDFs
-- `fieldType: 'text' | 'checkbox' | 'radio' | 'dropdown' | 'button'`
-- `label` renders above the field as static text
-- Text fields: `defaultValue`, `multiline`, `placeholder`, `maxLength`
-- Checkboxes: `checked` initial state
-- Radio groups and dropdowns: `options` array, `defaultSelected`
-- `doc.flattenForms: true` — bakes all fields into static content
-- Custom `borderColor`, `backgroundColor`, `width`, `height`, `fontSize` per field
-- New error codes: `FORM_FIELD_NAME_DUPLICATE` (duplicate `name` across fields), `FORM_FLATTEN_FAILED`
-- Post-render `form.updateFieldAppearances()` ensures proper display in all PDF readers
-- 10 comprehensive tests covering all form field types
-
-### Added (Phase 8E — Signature Placeholder)
-
-- `doc.signature` — visual signature box drawn on a specified page
-- Fields: `signerName`, `reason`, `location`, `x`, `y`, `width`, `height`, `page`, `borderColor`, `fontSize`
-- Draws signature line, date line, and optional text inside a bordered rectangle
-- `page` is 0-indexed, defaults to last page, clamps gracefully if out of range
-- 6 comprehensive tests
-
-### Added (Phase 8D — Callout Boxes)
-
-- New `callout` element type — styled highlight box with optional title
-- Preset styles: `style: 'info'` (blue), `'warning'` (amber), `'tip'` (green), `'note'` (gray)
-- Optional `title` rendered bold above content with left border accent
-- Fully customizable: `backgroundColor`, `borderColor`, `color`, `titleColor`, `padding`
-- Paginates correctly across pages (reuses blockquote pagination logic)
-- 8 comprehensive tests
-
-### Added (Phase 8F — Document Metadata Extensions)
-
-- `doc.metadata.language` — sets PDF `/Lang` catalog entry (BCP47 tag e.g. `'en-US'`, `'hi'`)
-- `doc.metadata.producer` — sets PDF producer field (e.g. `'MyApp v2.1'`)
-- Both fields validate as non-empty strings
-- 5 comprehensive tests
-
----
-
-## [0.2.0] — 2026-04-08
-
-### Added (Phase 8H — Inline Formatting)
-
-- `verticalAlign: 'superscript' | 'subscript'` on `InlineSpan` in rich-paragraphs
-- Superscript renders at 65% font size, baseline shifted up by 40% of font size
-- Subscript renders at 65% font size, baseline shifted down by 20% of font size
-- `letterSpacing?: number` on `ParagraphElement`, `HeadingElement`, `RichParagraphElement` — extra pt between characters
-- `smallCaps?: boolean` on those same three element types — simulated via uppercase + 80% fontSize
-- Character-by-character rendering for letterSpacing (pdf-lib has no native spacing param)
-- 8 comprehensive tests covering all inline formatting functionality
-
-### Added (Phase 8A — Annotations/Comments)
-
-- New `comment` element type — sticky note annotation at position in document
-- `annotation?: AnnotationSpec` on `ParagraphElement` and `HeadingElement` — attach note to element
-- Supports: `contents`, `author`, `color` (hex), `open` (popup default state)
-- Uses PDF `Subtype: 'Text'` annotation (sticky note icon in PDF viewers)
-- 8 comprehensive tests covering all annotation functionality
-
-### Added (Phase 8C — Document Assembly)
-
-- New `merge(pdfs: Uint8Array[])` exported function — combine pre-rendered PDFs
-- New `assemble(parts: AssemblyPart[])` exported function — mix rendered docs + existing PDFs
-- `AssemblyPart` interface: `{ doc?: PdfDocument, pdf?: Uint8Array }`
-- New error codes: `ASSEMBLY_EMPTY`, `ASSEMBLY_FAILED`
-- 8 comprehensive tests covering all assembly functionality
-
-### Fixed
-
-- **CI case-sensitivity bug**: `test/phase-7-integration.test.ts` used `'en-US'` (uppercase) for hyphenation language. On Linux CI (case-sensitive filesystem) this failed with `UNSUPPORTED_LANGUAGE`. Changed to `'en-us'` to match package name `hyphenation.en-us`.
-
----
-
-## [0.1.1] — 2026-04-08
-
-### Added
-
-- **Phase 8G: Hyperlinks** — Complete link annotation support:
-  - `paragraph.url` for external URI links on paragraphs
-  - `heading.url` for external URI links on headings
-  - `heading.anchor` for named PDF destinations (internal cross-references)
-  - `InlineSpan.href` for external and internal `#anchorId` links in rich-paragraphs
-  - `mailto:` scheme support for email links
-  - GoTo annotations for internal anchor references
-  - 9 comprehensive tests covering all hyperlink functionality
-
-### Fixed
-
-- **Memory leak in test suite**: Removed module-level `_hypherCache` in `src/measure.ts` that accumulated ~188KB per language across 255+ test runs. Changed from cached Hypher instances to fresh instances per call (negligible performance impact, massive memory savings).
-- **Node.js version compatibility**: Replaced `--experimental-strip-types` with `tsx` runner to support Node.js 18.x, 20.x, and 22.x in CI
-- **Broken CI examples**: Removed references to non-existent Phase 8 example scripts from GitHub Actions workflow
-- **README examples mismatch**: Updated Examples section to only list 5 existing Phase 7 examples (watermark, bookmarks, toc, rtl, encryption)
-- **Test suite OOM issues**: Split large test files (paginate.test.ts) into paginate-basic.test.ts to work around Node.js `--experimental-strip-types` heap exhaustion bug on files >17KB
-
-### Changed
-
-- `test:unit` now runs only `test/paginate-basic.test.ts` (fast, no canvas overhead)
-- Reorganized test scripts: `test:unit`, `test:validate`, `test:e2e`, `test:phases` for better memory management
-- Moved internal planning documentation to archive (preserved, not published)
-- `devDependencies`: Added `@napi-rs/canvas` explicitly (was missing, causing CI failures)
-
-### Added
-
-- `CONTRIBUTING.md`: Development setup, TDD workflow, PR process guide
-- `CHENG_LOU_EMAIL_DRAFT.md`: Template for requesting endorsement from pretext creator
-- `examples/comparison-pdfmake.ts`: pdfmake version of invoice for typography comparison
-
----
-
-## [0.1.0] — 2026-04-07
-
-### Added (Phase 7G — Encryption)
-
-- `doc.encryption` configuration for password-protecting PDFs
-- User password and owner password support
-- Granular permission restrictions: printing, copying, modifying, annotating
-- Lazy-loads `@cantoo/pdf-lib` (optional peer dependency) — zero cost when not used
-- Error code: `ENCRYPTION_NOT_AVAILABLE` when encryption is requested but dependency not installed
-
-### Added (Phase 7F — RTL Text Support)
-
-- Right-to-left text support for Arabic, Hebrew, and other RTL languages
-- Unicode bidirectional text algorithm via `bidi-js`
-- `dir` attribute on text elements: `'ltr'` | `'rtl'` | `'auto'` for per-element control
-- RTL text works with headings, paragraphs, lists, tables, and all text elements
-- Automatic detection of mixed LTR/RTL content
-
-### Added (Phase 7E — SVG Support)
-
-- `{ type: 'svg', svg: '<...' }` element for embedding SVG graphics
-- SVG rasterization via `@napi-rs/canvas`
-- ViewBox auto-sizing: automatic height calculation from viewBox aspect ratio
-- Explicit sizing: `width` and `height` parameters for precise control
-- Alignment options: `align: 'left' | 'center' | 'right'`
-- Multi-page support: SVGs paginate correctly across page breaks
-- Error code: `SVG_RENDER_FAILED` for SVG rasterization errors
-
-### Added (Phase 7D — Table of Contents)
-
-- `{ type: 'toc' }` element for automatic TOC generation
-- Two-pass rendering pipeline ensures accurate page numbers
-- Configurable: `title`, `showTitle`, `minLevel`/`maxLevel`, dot leaders, level indentation
-- Auto-indexed from heading structure (H1, H2, H3, etc.)
-- Supports custom formatting via `fontSize`, `color`, `spaceAfter` parameters
-
-### Added (Phase 7C — Hyphenation)
-
-- Automatic word hyphenation for better justified text layout
-- `doc.hyphenation: { language: 'en-US' }` for document-level config
-- Liang's algorithm via `hypher` package for accurate break points
-- Configurable: `minWordLength`, `leftMin`, `rightMin`, per-element `hyphenate: false` opt-out
-- Language support: includes `hyphenation.en-us` (additional languages via npm packages)
-- Error code: `UNSUPPORTED_LANGUAGE` when language not available
-
-### Added (Phase 7B — Watermarks)
-
-- `doc.watermark` for text or image watermarks on every page
-- Text watermarks: `text`, `fontSize`, `fontWeight`, `color`, `opacity`, `rotation`
-- Image watermarks: `image` (Uint8Array), `opacity`, `rotation`, `color` (tint)
-- Watermarks render behind content (lower z-index)
-- Rotation bounds: -360 ≤ rotation ≤ 360 degrees
-- Validation: must provide either text or image, never both required
-
-### Added (Phase 7A — Bookmarks / PDF Outline)
-
-- PDF sidebar bookmarks auto-generated from heading structure
-- Enabled by default: `bookmarks: true` or `bookmarks: { minLevel: 1, maxLevel: 3 }`
-- Level filtering: include/exclude heading levels from outline
-- Per-heading opt-out: `bookmark: false` on heading elements
-- Keyboard navigation: Cmd/Ctrl+Opt/Alt+O in PDF readers to toggle bookmark sidebar
-
-### Added (Phase 6 — Advanced Features)
-
-- Header and footer support with {{pageNumber}} and {{totalPages}} tokens
-- Text decoration: strikethrough, underline
-- Text alignment: left, center, right, justify
-- Line height control: custom line-height multipliers
-- Column layout with multi-column content flow
-- Tables with colspan/rowspan support
-
-### Added (Phase 5 — Rich Text / Builder API)
-
-- Fluent builder API for programmatic document construction
-- Rich text element with nested formatting (bold, italic, links)
-- Inline code and code blocks with syntax highlighting
-- Block quotes with custom styling
-- Horizontal rules (hr element)
-- Numbered and bulleted lists with nesting
-
-### Added (Phases 1–4 — Core Engine)
-
-- Core PDF generation via `pdf-lib`
-- Element types: paragraph, heading, table, image, list, code, blockquote
-- Font support: Inter bundled, custom TTF embedding
-- Document metadata: title, author, subject, keywords, created date
-- Page sizing: A4, A3, A5, Letter, Legal, or custom dimensions
-- Margins: top, bottom, left, right per page
-- Multi-page pagination with orphan/widow control
-- Image formats: PNG, JPG, WebP
-- Table features: custom column widths, cell padding, borders, header styling
-- Colors: hex color codes throughout (text, backgrounds, borders)
-
----
-
-## Features by Phase
-
-| Phase | Feature | Status | Tests | Version |
-|-------|---------|--------|-------|---------|
-| 7A | Bookmarks / PDF Outline | ✅ Complete | 8 | 0.1.0 |
-| 7B | Watermarks | ✅ Complete | 8 | 0.1.0 |
-| 7C | Hyphenation | ✅ Complete | 10 | 0.1.0 |
-| 7D | Table of Contents | ✅ Complete | 10 | 0.1.0 |
-| 7E | SVG Support | ✅ Complete | 8 | 0.1.0 |
-| 7F | RTL Text Support | ✅ Complete | 12 | 0.1.0 |
-| 7G | Encryption | ✅ Complete | 7 | 0.1.0 |
-| Integration | Feature Combinations | ✅ Complete | 6 | 0.1.0 |
-| 6 | Advanced Features | ✅ Complete | — | 0.1.0 |
-| 5 | Rich Text / Builder API | ✅ Complete | — | 0.1.0 |
-| 1–4 | Core Engine | ✅ Complete | — | 0.1.0 |
-
----
-
-## How to Use Examples
-
-All Phase 7 features have working examples in the `examples/` directory. Run them with:
-
-```bash
-npm run example:watermark    # Phase 7B — Watermarks
-npm run example:bookmarks    # Phase 7A — Bookmarks
-npm run example:toc          # Phase 7D — Table of Contents
-npm run example:rtl          # Phase 7F — RTL Text Support
-npm run example:encryption   # Phase 7G — Encryption
-```
-
-PDF output is written to `output/phase7-*.pdf`.
-
----
-
-## Test Coverage
-
-All phases include comprehensive test coverage:
-
-```bash
-npm test                  # Run all 75+ tests
-npm run test:unit        # Unit tests only
-npm run test:e2e         # End-to-end tests
-npm run test:visual      # Visual regression tests
-```
-
----
-
-## Dependencies
-
-### Required
-
-- `@chenglou/pretext` — Pretext text layout engine
-- `pdf-lib` — PDF document manipulation
-- `@fontsource/inter` — Font: Inter (bundled)
-- `bidi-js` — Bidirectional text algorithm (Phase 7F)
-- `hypher` — Hyphenation algorithm (Phase 7C)
-- `hyphenation.en-us` — English hyphenation patterns (Phase 7C)
-
-### Optional (Peer)
-
-- `@napi-rs/canvas` — SVG rasterization (Phase 7E)
-- `@cantoo/pdf-lib` — PDF encryption (Phase 7G)
-
----
-
-## Architecture
-
-pretext-pdf uses a modular, layered architecture:
-
-```
-render(doc) → validate → layout → paginate → render-pages → PDF bytes
-```
-
-Each phase adds features orthogonally:
-- Phase 1–4: Core engine, pagination, typography
-- Phase 5: Rich text and builder patterns
-- Phase 6: Advanced layout and formatting
-- Phase 7: Security, internationalization, accessibility
-
----
-
-## Future Roadmap
-
-Potential Phase 8+ features (not yet implemented):
-- Hyperlinks and anchors
-- Justified text alignment
-- Enhanced text decorations
-- Font subsetting for file size reduction
-- Browser compatibility improvements
-- Performance optimizations
-
----
-
-## Contributing
-
-pretext-pdf is actively maintained. To report issues, request features, or contribute:
-
-1. Check existing issues on the project repo
-2. Write failing tests first (TDD)
-3. Submit pull requests with test coverage ≥80%
-
----
-
-## License
-
-Check LICENSE file in repository root.
+# Changelog
+
+<!-- markdownlint-disable MD024 -->
+
+All notable changes to pretext-pdf are documented here.
+Format: [Keep a Changelog 1.1.0](https://keepachangelog.com/en/1.1.0/)
+
+---
+
+## [0.8.2] — 2026-04-20
+
+### Fixed
+
+- **Rich-paragraph whitespace collapse** — multi-span `rich-paragraph` content rendered with adjacent words overlapping (e.g. `"Founder & CEO" + "  —  Antigravity Systems"` displayed as `"Founder& CEO—AntigravitySystems"`). Root cause: pretext's `layoutWithLines` follows CSS-like behavior and excludes trailing whitespace from line widths, so tokens like `"Hello "` or `"  "` measured to width 0 and downstream fragments overlapped the previous one. `measureTokenWidth` in [src/rich-text.ts](src/rich-text.ts) now uses a sentinel-character technique (append non-whitespace `\u2588`, measure combined string, subtract sentinel width) to recover the true rendered width whenever a token has trailing whitespace. Sentinel width is cached per font config.
+- The fast path (no trailing whitespace) is unchanged — single pretext call. Slow path adds two pretext calls per affected token, with one cached.
+
+### Tests
+
+- Added 3 regression tests in `test/rich-text.test.ts` under `whitespace preservation (v0.8.2 fix)` covering trailing whitespace inside spans, whitespace-only separator spans, and the exact `"Founder & CEO" → "Antigravity Systems"` resume-preset scenario.
+
+---
+
+## [0.8.1] — 2026-04-20
+
+### Fixed
+
+- **Browser support** — `pretext-pdf` now imports cleanly in browsers. Module-init in `src/fonts.ts` previously called `fileURLToPath(import.meta.url)` and `createRequire(import.meta.url)` eagerly, which threw `"The URL must be of scheme file"` whenever the module was loaded from a non-`file://` URL (esm.sh, jsdelivr, Vite dev server). Both calls are now gated on a runtime `IS_NODE` check, and the bundled-Inter `BUNDLED_INTER_PATHS` arrays are constructed only in Node.
+- **Browser font-loading errors** — `loadFontBytes` now throws clear `FONT_LOAD_FAILED` messages when bundled Inter or string font paths are requested in a browser, pointing the consumer at the correct workaround (supply `Uint8Array` bytes via `doc.fonts`).
+
+### Notes for browser users
+
+- Always supply Inter (or your default font) explicitly via `doc.fonts: [{ family: 'Inter', weight: 400, src: <Uint8Array> }, { family: 'Inter', weight: 700, src: <Uint8Array> }]`. The library cannot read local font files in the browser.
+- SVG / chart / qr-code / barcode elements still depend on `@napi-rs/canvas` at runtime; in the browser, the native `OffscreenCanvas` is used instead and the polyfill is skipped automatically.
+
+---
+
+## [0.8.0] — 2026-04-19
+
+### Added
+
+- **`qr-code` element** — generate QR codes as inline PDF content using the `qrcode` optional peer dependency. Supports `data`, `size`, `errorCorrectionLevel` (L/M/Q/H), `foreground`/`background` hex colours, `margin`, `align`, `spaceBefore`/`spaceAfter`. Fully serverless — pure JS, no canvas required.
+- **`barcode` element** — generate 100+ barcode symbologies (EAN-13, Code128, PDF417, QR, DataMatrix, etc.) via the `bwip-js` optional peer dependency. Supports `symbology`, `data`, `width`, `height`, `includeText`, `align`, `spaceBefore`/`spaceAfter`. Pure JS, Lambda/Edge safe.
+- **`chart` element** — embed Vega-Lite charts as vector SVG using `vega` + `vega-lite` optional peer deps. Accepts any Vega-Lite `spec`, `width`, `height`, `caption`, `align`. Rendered with `renderer: 'none'` — zero canvas/puppeteer dependency.
+- **`pretext-pdf/markdown` entry point** — `markdownToContent(md, options?)` converts a Markdown string to `ContentElement[]`. Requires optional `marked` peer dep. Supports headings, bold/italic/links (→ rich-paragraph), lists (2 levels), blockquotes, code blocks, and HR.
+- **`pretext-pdf/templates` entry point** — three typed template functions with zero extra dependencies: `createInvoice(data)` (generic invoice with currency, tax, discount, QR payment), `createGstInvoice(data)` (GST-compliant Indian tax invoice with IGST/CGST+SGST, UPI QR, bank details, amount in words), `createReport(data)` (structured business report with optional TOC).
+- **New error codes** — `QR_DEP_MISSING`, `QR_GENERATE_FAILED`, `BARCODE_DEP_MISSING`, `BARCODE_GENERATE_FAILED`, `BARCODE_SYMBOLOGY_INVALID`, `CHART_DEP_MISSING`, `CHART_SPEC_INVALID`, `CHART_RENDER_FAILED`, `MARKDOWN_DEP_MISSING`.
+
+---
+
+## [0.7.1] — 2026-04-19
+
+### Changed
+
+- **Upstream pretext pinned to `f2014338487a`** — picks up unreleased CJK opening-bracket annotation fix, Hangul jamo line-walker alignment fix, and two internal line-object churn reductions. No public API changes.
+
+### Fixed
+
+- **List nesting depth enforced at validation** — `ListItem.items` (2nd-level items) now correctly rejects any further `.items` property, matching the documented 2-level maximum. Previously the validation silently passed 3-level data which could cause undefined render behaviour.
+- **3 phase-11 list tests corrected** — test data incorrectly contained 3-level nesting while named "2-level"; data trimmed to match documented contract.
+
+---
+
+## [Unreleased]
+
+### Added (Phase 11 — Cross-cutting Enhancements)
+
+- **`floatSpans` on image elements** — rich-text alternative to plain `floatText`. Accepts `InlineSpan[]` for mixed bold/italic/color/link captions beside float images. Mutually exclusive with `floatText` (validated).
+- **2-level list nesting** — `ListItem.items` now supports one further level of nesting (depth 0 → 1 → 2). Unordered marker: `▪`. Ordered: inherits parent counter or restarts via `nestedNumberingStyle: 'restart'`.
+- **Table `rowspan`** — `TableCell.rowspan` spans a cell across multiple rows. Works alongside `colspan`. Origin cell draws background over full span height; continuation rows automatically receive placeholder cells.
+- **`onFormFieldError` callback** — `doc.onFormFieldError: (name, err) => 'skip' | 'throw'` mirrors `onImageLoadError`. Controls render behaviour when a form field fails.
+- **`createFootnoteSet(defs)`** — helper exported from `pretext-pdf` that generates footnote definition/reference pairs with globally unique IDs. Returns `Array<{ id, def }>`.
+- **`renderDate` field** — `doc.renderDate: Date | string` overrides the PDF creation date. Useful for reproducible builds and testing.
+- **`{{date}}` and `{{author}}` tokens** in header/footer text — join existing `{{pageNumber}}` / `{{totalPages}}`. `{{date}}` resolves from `renderDate`; `{{author}}` resolves from `doc.metadata.author`.
+- **`tabularNumbers`** on `rich-paragraph` — digits rendered at uniform slot width (widest digit in font), so columns of numbers align without OpenType TNUM feature.
+- **`smallCaps` + `letterSpacing` per span** — `InlineSpan.smallCaps` and `InlineSpan.letterSpacing` now respected in `rich-paragraph` rendering.
+- **Per-span `fontSize`** — `InlineSpan.fontSize` overrides the element-level font size for that span. Enables mixed-size text in a single paragraph.
+
+### Fixed (Phase 11)
+
+- `resolveTokens()` used `.replace()` (replaces first occurrence only) — changed to `.replaceAll()` for all four tokens.
+- Table span grid: continuation-row cursor was advancing by 1 instead of `colspan` when skipping a spanned column — now advances by full span width.
+- Font family names now validated for safe characters (`/^[a-zA-Z0-9 _-]+$/`) in `requireFamily()` — rejects null bytes and control characters.
+- Annotation `color` and `author` fields now validated in `validateElement()` for both `paragraph` and `heading` annotations.
+- `buildOutlineTree` memoizes `parentIdxOf()` into a pre-computed array — eliminates O(n²) scan for documents with large heading counts.
+- Table grid-line renderer pre-computes active boundary set — eliminates O(rows × cols) inner loop for large tables.
+- `addLinkAnnotation()` re-validates URL scheme at render time (defense-in-depth; `validate.ts` is the primary gate).
+
+### Planned (Phase 9+)
+
+- Phase 9A: Digital signatures (cryptographic PKCS#7 via `@signpdf/signpdf`)
+- Phase 9B: Image floats (text flowing alongside images — requires paginator rewrite)
+- Phase 9C: Font subsetting pre-computation (explicit glyph hints for icon fonts)
+
+---
+
+## [0.7.0] — 2026-04-17
+
+### Added
+
+- **6 production templates** (`templates/`) — GST invoice, international invoice, resume, multi-section report, NDA, and meeting minutes. Each is a self-contained `.ts` file outputting a valid PDF. Smoke-tested in Phase 2F Block D.
+- **StackBlitz live demo** (`demo/stackblitz/`) — 4-tab UI (Invoice, Report, Resume, Custom) backed by a Node.js render server. Edit JSON and generate PDFs instantly, no install required. Accessible at the StackBlitz link in the README.
+- **`## Performance` section in README** — measured render times and PDF sizes for 1-page, 10-page, and mixed-element documents. Font subsetting behaviour documented.
+- **Stress tests and benchmarks** (`test/phase-2f-stress.test.ts`) — 32 tests across 4 blocks: large document stress (400-element, 200-row table), edge case stress (CJK, RTL, empty arrays, extreme sizes), timing benchmarks (1-page < 500 ms, 10-page < 5,000 ms), and template smoke tests.
+- **Error code coverage** — new tests for `COLUMN_WIDTH_TOO_NARROW`, `IMAGE_LOAD_FAILED`, `SVG_LOAD_FAILED`, and `ASSEMBLY_FAILED`. 16 of 19 error codes now have direct test coverage.
+
+### Changed
+
+- **`as any` audit** — eliminated 10 casts in `validate.ts` by introducing a typed `FormFieldElement` local binding. The remaining 8 instances (pdf-lib interop, dynamic import, internal back-references) are now documented with one-line comments explaining the constraint.
+- **Comparison article** (`docs/articles/pretext-pdf-vs-pdfmake-2026.md`) — 2,200-word draft covering feature matrix, typography quality, API design, performance, and migration quick-start. Marked `published: false` pending live demo.
+- **Migration guide** (`docs/MIGRATION_FROM_PDFMAKE.md`) — 30+ pdfmake → pretext-pdf mappings, complete before/after invoice example, and a quick-start checklist. Linked from README.
+
+### Fixed
+
+- **Phase 2F test types** — `fontWeight: 700 as 700` cast in pre-constructed rows array; removed non-existent `creationDate` from `DocumentMetadata`; replaced `allowCopying: false` with correct `encryption: { permissions: { copying: false } }`.
+- **StackBlitz integration** — added `.stackblitzrc` so WebContainer auto-runs `npm start` and opens the browser preview on port 3000.
+
+---
+
+## [0.5.3] — 2026-04-16
+
+### Changed
+
+- **Upgraded `@chenglou/pretext` from 0.0.3 to 0.0.5** — picks up improved text analysis accuracy (~35% larger analysis module), better measurement precision, extracted bidi-data module for cleaner tree-shaking, and new `rich-inline` export (not yet used by pretext-pdf). No breaking changes — `prepareWithSegments()` and `layoutWithLines()` APIs are unchanged. All 223 tests pass, 3 example PDFs visually verified (RTL, TOC, hyperlinks).
+
+---
+
+## [0.5.2] — 2026-04-13
+
+### Added
+
+- **`onImageLoadError` callback on `PdfDocument`** — gives callers control over image load failures. Return `'skip'` to silently omit the image (preserves existing default behavior). Return `'throw'` to abort rendering with the original error. Previously, all image failures were silently downgraded to `console.warn` with no way to detect them programmatically.
+
+  ```typescript
+  await render({
+    content: [...],
+    onImageLoadError: (src, error) => {
+      myLogger.warn('Image skipped', { src, error })
+      return 'skip'   // or 'throw' to abort
+    }
+  })
+  ```
+
+---
+
+## [0.4.0] — 2026-04-08
+
+### Breaking Changes
+
+- **Migrated from `pdf-lib` to `@cantoo/pdf-lib`** — `@cantoo/pdf-lib` is now a direct `dependency` (always installed). Previously it was an optional peer dependency required only for encryption. This removes the `ENCRYPTION_NOT_AVAILABLE` error code and the separate `npm install @cantoo/pdf-lib` installation step. Encryption now works out of the box.
+- **`ENCRYPTION_NOT_AVAILABLE` error code removed** — encryption is now always available. Update any `switch` statements that handled this code.
+
+### Why this change
+
+`pdf-lib` (the original) has not received a meaningful commit since November 2021. `@cantoo/pdf-lib` is the actively maintained fork (v2.6.5, 107+ releases, MIT license). pretext-pdf was already using `@cantoo/pdf-lib` for encryption — this commit makes it the single source of truth for all PDF operations.
+
+### Added
+
+- `test/pretext-api-contract.test.ts` — canary test that asserts `@chenglou/pretext` exports the exact functions pretext-pdf depends on. Breaks loudly if pretext changes its API.
+- `docs/ROADMAP.md` — public multi-phase development plan
+
+### Changed
+
+- `@chenglou/pretext` version pinned to exact `0.0.3` (no caret) — prevents surprise breaking changes from upstream auto-updates
+- `test:contract` script added — runs the pretext API contract test before the full test suite
+- All internal comments updated from `pdf-lib` to `@cantoo/pdf-lib`
+
+---
+
+## [0.3.1] — 2026-04-08
+
+### Fixed
+
+- **Critical: Font resolution when installed as npm package** — `@fontsource/inter` is now resolved via `createRequire(import.meta.url)` instead of a hardcoded relative path. Previously, `path.join(__dirname, '..', 'node_modules', '@fontsource', 'inter', ...)` failed when npm hoisted the dependency to the consumer's top-level `node_modules`, causing `FONT_LOAD_FAILED` on every install. Now resolves correctly regardless of npm hoisting behavior.
+
+---
+
+## [0.3.0] — 2026-04-08
+
+### Added (Phase 8B — Interactive Forms)
+
+- New `form-field` element type — creates interactive AcroForm fields in PDFs
+- `fieldType: 'text' | 'checkbox' | 'radio' | 'dropdown' | 'button'`
+- `label` renders above the field as static text
+- Text fields: `defaultValue`, `multiline`, `placeholder`, `maxLength`
+- Checkboxes: `checked` initial state
+- Radio groups and dropdowns: `options` array, `defaultSelected`
+- `doc.flattenForms: true` — bakes all fields into static content
+- Custom `borderColor`, `backgroundColor`, `width`, `height`, `fontSize` per field
+- New error codes: `FORM_FIELD_NAME_DUPLICATE` (duplicate `name` across fields), `FORM_FLATTEN_FAILED`
+- Post-render `form.updateFieldAppearances()` ensures proper display in all PDF readers
+- 10 comprehensive tests covering all form field types
+
+### Added (Phase 8E — Signature Placeholder)
+
+- `doc.signature` — visual signature box drawn on a specified page
+- Fields: `signerName`, `reason`, `location`, `x`, `y`, `width`, `height`, `page`, `borderColor`, `fontSize`
+- Draws signature line, date line, and optional text inside a bordered rectangle
+- `page` is 0-indexed, defaults to last page, clamps gracefully if out of range
+- 6 comprehensive tests
+
+### Added (Phase 8D — Callout Boxes)
+
+- New `callout` element type — styled highlight box with optional title
+- Preset styles: `style: 'info'` (blue), `'warning'` (amber), `'tip'` (green), `'note'` (gray)
+- Optional `title` rendered bold above content with left border accent
+- Fully customizable: `backgroundColor`, `borderColor`, `color`, `titleColor`, `padding`
+- Paginates correctly across pages (reuses blockquote pagination logic)
+- 8 comprehensive tests
+
+### Added (Phase 8F — Document Metadata Extensions)
+
+- `doc.metadata.language` — sets PDF `/Lang` catalog entry (BCP47 tag e.g. `'en-US'`, `'hi'`)
+- `doc.metadata.producer` — sets PDF producer field (e.g. `'MyApp v2.1'`)
+- Both fields validate as non-empty strings
+- 5 comprehensive tests
+
+---
+
+## [0.2.0] — 2026-04-08
+
+### Added (Phase 8H — Inline Formatting)
+
+- `verticalAlign: 'superscript' | 'subscript'` on `InlineSpan` in rich-paragraphs
+- Superscript renders at 65% font size, baseline shifted up by 40% of font size
+- Subscript renders at 65% font size, baseline shifted down by 20% of font size
+- `letterSpacing?: number` on `ParagraphElement`, `HeadingElement`, `RichParagraphElement` — extra pt between characters
+- `smallCaps?: boolean` on those same three element types — simulated via uppercase + 80% fontSize
+- Character-by-character rendering for letterSpacing (pdf-lib has no native spacing param)
+- 8 comprehensive tests covering all inline formatting functionality
+
+### Added (Phase 8A — Annotations/Comments)
+
+- New `comment` element type — sticky note annotation at position in document
+- `annotation?: AnnotationSpec` on `ParagraphElement` and `HeadingElement` — attach note to element
+- Supports: `contents`, `author`, `color` (hex), `open` (popup default state)
+- Uses PDF `Subtype: 'Text'` annotation (sticky note icon in PDF viewers)
+- 8 comprehensive tests covering all annotation functionality
+
+### Added (Phase 8C — Document Assembly)
+
+- New `merge(pdfs: Uint8Array[])` exported function — combine pre-rendered PDFs
+- New `assemble(parts: AssemblyPart[])` exported function — mix rendered docs + existing PDFs
+- `AssemblyPart` interface: `{ doc?: PdfDocument, pdf?: Uint8Array }`
+- New error codes: `ASSEMBLY_EMPTY`, `ASSEMBLY_FAILED`
+- 8 comprehensive tests covering all assembly functionality
+
+### Fixed
+
+- **CI case-sensitivity bug**: `test/phase-7-integration.test.ts` used `'en-US'` (uppercase) for hyphenation language. On Linux CI (case-sensitive filesystem) this failed with `UNSUPPORTED_LANGUAGE`. Changed to `'en-us'` to match package name `hyphenation.en-us`.
+
+---
+
+## [0.1.1] — 2026-04-08
+
+### Added
+
+- **Phase 8G: Hyperlinks** — Complete link annotation support:
+  - `paragraph.url` for external URI links on paragraphs
+  - `heading.url` for external URI links on headings
+  - `heading.anchor` for named PDF destinations (internal cross-references)
+  - `InlineSpan.href` for external and internal `#anchorId` links in rich-paragraphs
+  - `mailto:` scheme support for email links
+  - GoTo annotations for internal anchor references
+  - 9 comprehensive tests covering all hyperlink functionality
+
+### Fixed
+
+- **Memory leak in test suite**: Removed module-level `_hypherCache` in `src/measure.ts` that accumulated ~188KB per language across 255+ test runs. Changed from cached Hypher instances to fresh instances per call (negligible performance impact, massive memory savings).
+- **Node.js version compatibility**: Replaced `--experimental-strip-types` with `tsx` runner to support Node.js 18.x, 20.x, and 22.x in CI
+- **Broken CI examples**: Removed references to non-existent Phase 8 example scripts from GitHub Actions workflow
+- **README examples mismatch**: Updated Examples section to only list 5 existing Phase 7 examples (watermark, bookmarks, toc, rtl, encryption)
+- **Test suite OOM issues**: Split large test files (paginate.test.ts) into paginate-basic.test.ts to work around Node.js `--experimental-strip-types` heap exhaustion bug on files >17KB
+
+### Changed
+
+- `test:unit` now runs only `test/paginate-basic.test.ts` (fast, no canvas overhead)
+- Reorganized test scripts: `test:unit`, `test:validate`, `test:e2e`, `test:phases` for better memory management
+- Moved internal planning documentation to archive (preserved, not published)
+- `devDependencies`: Added `@napi-rs/canvas` explicitly (was missing, causing CI failures)
+
+### Added
+
+- `CONTRIBUTING.md`: Development setup, TDD workflow, PR process guide
+- `CHENG_LOU_EMAIL_DRAFT.md`: Template for requesting endorsement from pretext creator
+- `examples/comparison-pdfmake.ts`: pdfmake version of invoice for typography comparison
+
+---
+
+## [0.1.0] — 2026-04-07
+
+### Added (Phase 7G — Encryption)
+
+- `doc.encryption` configuration for password-protecting PDFs
+- User password and owner password support
+- Granular permission restrictions: printing, copying, modifying, annotating
+- Lazy-loads `@cantoo/pdf-lib` (optional peer dependency) — zero cost when not used
+- Error code: `ENCRYPTION_NOT_AVAILABLE` when encryption is requested but dependency not installed
+
+### Added (Phase 7F — RTL Text Support)
+
+- Right-to-left text support for Arabic, Hebrew, and other RTL languages
+- Unicode bidirectional text algorithm via `bidi-js`
+- `dir` attribute on text elements: `'ltr'` | `'rtl'` | `'auto'` for per-element control
+- RTL text works with headings, paragraphs, lists, tables, and all text elements
+- Automatic detection of mixed LTR/RTL content
+
+### Added (Phase 7E — SVG Support)
+
+- `{ type: 'svg', svg: '<...' }` element for embedding SVG graphics
+- SVG rasterization via `@napi-rs/canvas`
+- ViewBox auto-sizing: automatic height calculation from viewBox aspect ratio
+- Explicit sizing: `width` and `height` parameters for precise control
+- Alignment options: `align: 'left' | 'center' | 'right'`
+- Multi-page support: SVGs paginate correctly across page breaks
+- Error code: `SVG_RENDER_FAILED` for SVG rasterization errors
+
+### Added (Phase 7D — Table of Contents)
+
+- `{ type: 'toc' }` element for automatic TOC generation
+- Two-pass rendering pipeline ensures accurate page numbers
+- Configurable: `title`, `showTitle`, `minLevel`/`maxLevel`, dot leaders, level indentation
+- Auto-indexed from heading structure (H1, H2, H3, etc.)
+- Supports custom formatting via `fontSize`, `color`, `spaceAfter` parameters
+
+### Added (Phase 7C — Hyphenation)
+
+- Automatic word hyphenation for better justified text layout
+- `doc.hyphenation: { language: 'en-US' }` for document-level config
+- Liang's algorithm via `hypher` package for accurate break points
+- Configurable: `minWordLength`, `leftMin`, `rightMin`, per-element `hyphenate: false` opt-out
+- Language support: includes `hyphenation.en-us` (additional languages via npm packages)
+- Error code: `UNSUPPORTED_LANGUAGE` when language not available
+
+### Added (Phase 7B — Watermarks)
+
+- `doc.watermark` for text or image watermarks on every page
+- Text watermarks: `text`, `fontSize`, `fontWeight`, `color`, `opacity`, `rotation`
+- Image watermarks: `image` (Uint8Array), `opacity`, `rotation`, `color` (tint)
+- Watermarks render behind content (lower z-index)
+- Rotation bounds: -360 ≤ rotation ≤ 360 degrees
+- Validation: must provide either text or image, never both required
+
+### Added (Phase 7A — Bookmarks / PDF Outline)
+
+- PDF sidebar bookmarks auto-generated from heading structure
+- Enabled by default: `bookmarks: true` or `bookmarks: { minLevel: 1, maxLevel: 3 }`
+- Level filtering: include/exclude heading levels from outline
+- Per-heading opt-out: `bookmark: false` on heading elements
+- Keyboard navigation: Cmd/Ctrl+Opt/Alt+O in PDF readers to toggle bookmark sidebar
+
+### Added (Phase 6 — Advanced Features)
+
+- Header and footer support with {{pageNumber}} and {{totalPages}} tokens
+- Text decoration: strikethrough, underline
+- Text alignment: left, center, right, justify
+- Line height control: custom line-height multipliers
+- Column layout with multi-column content flow
+- Tables with colspan/rowspan support
+
+### Added (Phase 5 — Rich Text / Builder API)
+
+- Fluent builder API for programmatic document construction
+- Rich text element with nested formatting (bold, italic, links)
+- Inline code and code blocks with syntax highlighting
+- Block quotes with custom styling
+- Horizontal rules (hr element)
+- Numbered and bulleted lists with nesting
+
+### Added (Phases 1–4 — Core Engine)
+
+- Core PDF generation via `pdf-lib`
+- Element types: paragraph, heading, table, image, list, code, blockquote
+- Font support: Inter bundled, custom TTF embedding
+- Document metadata: title, author, subject, keywords, created date
+- Page sizing: A4, A3, A5, Letter, Legal, or custom dimensions
+- Margins: top, bottom, left, right per page
+- Multi-page pagination with orphan/widow control
+- Image formats: PNG, JPG, WebP
+- Table features: custom column widths, cell padding, borders, header styling
+- Colors: hex color codes throughout (text, backgrounds, borders)
+
+---
+
+## Features by Phase
+
+| Phase | Feature | Status | Tests | Version |
+|-------|---------|--------|-------|---------|
+| 7A | Bookmarks / PDF Outline | ✅ Complete | 8 | 0.1.0 |
+| 7B | Watermarks | ✅ Complete | 8 | 0.1.0 |
+| 7C | Hyphenation | ✅ Complete | 10 | 0.1.0 |
+| 7D | Table of Contents | ✅ Complete | 10 | 0.1.0 |
+| 7E | SVG Support | ✅ Complete | 8 | 0.1.0 |
+| 7F | RTL Text Support | ✅ Complete | 12 | 0.1.0 |
+| 7G | Encryption | ✅ Complete | 7 | 0.1.0 |
+| Integration | Feature Combinations | ✅ Complete | 6 | 0.1.0 |
+| 6 | Advanced Features | ✅ Complete | — | 0.1.0 |
+| 5 | Rich Text / Builder API | ✅ Complete | — | 0.1.0 |
+| 1–4 | Core Engine | ✅ Complete | — | 0.1.0 |
+
+---
+
+## How to Use Examples
+
+All Phase 7 features have working examples in the `examples/` directory. Run them with:
+
+```bash
+npm run example:watermark    # Phase 7B — Watermarks
+npm run example:bookmarks    # Phase 7A — Bookmarks
+npm run example:toc          # Phase 7D — Table of Contents
+npm run example:rtl          # Phase 7F — RTL Text Support
+npm run example:encryption   # Phase 7G — Encryption
+```
+
+PDF output is written to `output/phase7-*.pdf`.
+
+---
+
+## Test Coverage
+
+All phases include comprehensive test coverage:
+
+```bash
+npm test                  # Run all 75+ tests
+npm run test:unit        # Unit tests only
+npm run test:e2e         # End-to-end tests
+npm run test:visual      # Visual regression tests
+```
+
+---
+
+## Dependencies
+
+### Required
+
+- `@chenglou/pretext` — Pretext text layout engine
+- `pdf-lib` — PDF document manipulation
+- `@fontsource/inter` — Font: Inter (bundled)
+- `bidi-js` — Bidirectional text algorithm (Phase 7F)
+- `hypher` — Hyphenation algorithm (Phase 7C)
+- `hyphenation.en-us` — English hyphenation patterns (Phase 7C)
+
+### Optional (Peer)
+
+- `@napi-rs/canvas` — SVG rasterization (Phase 7E)
+- `@cantoo/pdf-lib` — PDF encryption (Phase 7G)
+
+---
+
+## Architecture
+
+pretext-pdf uses a modular, layered architecture:
+
+```
+render(doc) → validate → layout → paginate → render-pages → PDF bytes
+```
+
+Each phase adds features orthogonally:
+- Phase 1–4: Core engine, pagination, typography
+- Phase 5: Rich text and builder patterns
+- Phase 6: Advanced layout and formatting
+- Phase 7: Security, internationalization, accessibility
+
+---
+
+## Future Roadmap
+
+Potential Phase 8+ features (not yet implemented):
+- Hyperlinks and anchors
+- Justified text alignment
+- Enhanced text decorations
+- Font subsetting for file size reduction
+- Browser compatibility improvements
+- Performance optimizations
+
+---
+
+## Contributing
+
+pretext-pdf is actively maintained. To report issues, request features, or contribute:
+
+1. Check existing issues on the project repo
+2. Write failing tests first (TDD)
+3. Submit pull requests with test coverage ≥80%
+
+---
+
+## License
+
+Check LICENSE file in repository root.