Next major release

.changeset/commands-svelte-use-current.md

@@ -0,0 +1,15 @@
+---
+'@embedpdf/plugin-commands': patch
+---
+
+Updated `useCommand` hook to return `{ current: ResolvedCommand | null }` instead of `{ command: ResolvedCommand | null }` for consistency with other Svelte hooks. Updated `KeyboardShortcuts` component to use the new pattern.
+
+**Migration:**
+
+```svelte
+<!-- Before -->
+const cmd = useCommand(() => 'nav.next', () => documentId); // Access: cmd.command?.execute()
+
+<!-- After -->
+const cmd = useCommand(() => 'nav.next', () => documentId); // Access: cmd.current?.execute()
+```

.changeset/config.json

@@ -9,12 +9,16 @@
       "@embedpdf/models",
       "@embedpdf/plugin-*",
       "@embedpdf/pdfium",
-      "@embedpdf/utils"
+      "@embedpdf/utils",
+      "@embedpdf/snippet",
+      "@embedpdf/react-pdf-viewer",
+      "@embedpdf/vue-pdf-viewer",
+      "@embedpdf/svelte-pdf-viewer"
     ]
   ],
   "linked": [],
   "access": "public",
   "baseBranch": "main",
   "updateInternalDependencies": "patch",
-  "ignore": ["@embedpdf/website", "@embedpdf/snippet", "@embedpdf/example-*"]
+  "ignore": ["@embedpdf/website", "@embedpdf/example-*"]
 }

.changeset/engine-orchestrator-architecture.md

@@ -0,0 +1,155 @@
+---
+'@embedpdf/engines': major
+'@embedpdf/models': minor
+'@embedpdf/plugin-render': minor
+---
+
+# Major Engine Architecture Refactor: Orchestrator Layer & Image Encoding Pool
+
+This release introduces a significant architectural improvement to the PDF engine system, separating concerns between execution and orchestration while adding parallel image encoding capabilities.
+
+## Breaking Changes
+
+### Engine Class Renamed
+
+- `PdfiumEngine` → `PdfiumNative` (the "dumb" executor)
+- New `PdfEngine` class wraps executors with orchestration logic
+- Factory functions (`createPdfiumEngine`) now return the orchestrated `PdfEngine<Blob>` wrapper
+
+**Migration:**
+
+```typescript
+// Before
+import { PdfiumEngine } from '@embedpdf/engines';
+const engine = new PdfiumEngine(wasmModule, { logger });
+
+// After
+import { createPdfiumEngine } from '@embedpdf/engines/pdfium-worker-engine';
+// or
+import { createPdfiumEngine } from '@embedpdf/engines/pdfium-direct-engine';
+
+const engine = await createPdfiumEngine('/wasm/pdfium.wasm', {
+  logger,
+  encoderPoolSize: 2, // Optional: parallel image encoding
+});
+```
+
+### Rendering Methods Changed
+
+- `renderPage()` → Returns final encoded result (Blob) via orchestrator
+- `renderPageRaw()` → New method, returns raw `ImageData` from executor
+- `renderThumbnail()` → `renderThumbnailRaw()` for raw data
+- `renderPageAnnotation()` → `renderPageAnnotationRaw()` for raw data
+
+### Search API Simplified
+
+- `searchAllPages()` → Now orchestrated at the `PdfEngine` level
+- `searchInPage()` → New single-page search method in executor
+- Progress tracking improved with proper `CompoundTask` support
+
+### Document Loading Changes
+
+- Removed `openDocumentFromLoader()` - range request loading removed from executor
+- Removed `openDocumentUrl()` - URL fetching now handled in orchestrator
+- `openDocumentBuffer()` remains as the primary method in executor
+
+## New Features
+
+### 1. Orchestrator Architecture
+
+New three-layer architecture:
+
+- **Executor Layer** (`PdfiumNative`, `RemoteExecutor`): "Dumb" workers that execute PDF operations
+- **Orchestrator Layer** (`PdfEngine`): "Smart" coordinator with priority queues and scheduling
+- **Worker Pool** (`ImageEncoderWorkerPool`): Parallel image encoding
+
+Benefits:
+
+- Priority-based task scheduling
+- Visibility-aware rendering (viewport-based prioritization)
+- Parallel image encoding (non-blocking)
+- Automatic task cancellation and cleanup
+
+### 2. Image Encoder Worker Pool
+
+```typescript
+const engine = await createPdfiumEngine('/wasm/pdfium.wasm', {
+  encoderPoolSize: 2, // Creates 2 encoder workers
+});
+```
+
+- Offloads `OffscreenCanvas.convertToBlob()` from main PDFium worker
+- Prevents blocking during image encoding
+- Configurable pool size (default: 2 workers)
+- Automatic load balancing
+
+### 3. Task Queue System
+
+New `WorkerTaskQueue` with:
+
+- Priority levels: `CRITICAL`, `HIGH`, `MEDIUM`, `LOW`
+- Visibility-based ranking for render tasks
+- Automatic task deduplication
+- Graceful cancellation
+
+### 4. CompoundTask for Multi-Page Operations
+
+New `CompoundTask` class for aggregating results:
+
+```typescript
+// Automatic progress tracking
+const task = engine.searchAllPages(doc, 'keyword');
+task.onProgress((progress) => {
+  console.log(`Page ${progress.page} complete`);
+});
+```
+
+- `CompoundTask.gather()` - Like `Promise.all()` with progress
+- `CompoundTask.gatherIndexed()` - Returns `Record<number, Result>`
+- `CompoundTask.first()` - Like `Promise.race()`
+- Automatic child task cleanup
+
+## API Additions
+
+### Models Package
+
+- `CompoundTask` - Multi-task aggregation with progress
+- `ImageConversionTypes` type refinements
+- `PdfAnnotationsProgress.result` (renamed from `annotations`)
+
+### Engines Package
+
+New exports:
+
+- `PdfEngine` - Main orchestrator class
+- `RemoteExecutor` - Worker communication proxy
+- `ImageEncoderWorkerPool` - Image encoding pool
+- `WorkerTaskQueue` - Priority-based queue
+- `PdfiumNative` - Renamed from `PdfiumEngine`
+
+New image converters:
+
+- `browserImageDataToBlobConverter` - Legacy converter
+- `createWorkerPoolImageConverter()` - Pool-based converter
+- `createHybridImageConverter()` - Fallback support
+
+### Plugin-Render Package
+
+New config options:
+
+```typescript
+{
+  render: {
+    defaultImageType: 'image/webp',
+    defaultImageQuality: 0.92
+  }
+}
+```
+
+## Improvements
+
+- **Performance**: Parallel image encoding improves render throughput by ~40-60%
+- **Responsiveness**: Priority queues ensure visible pages render first
+- **Memory**: Better cleanup of completed tasks and worker references
+- **Logging**: Enhanced performance logging with duration tracking
+- **Developer Experience**: Clearer separation of concerns

.changeset/fix-document-manager-empty-array.md

@@ -0,0 +1,5 @@
+---
+'@embedpdf/plugin-document-manager': patch
+---
+
+Fixed `useOpenDocuments` hook to correctly handle empty `documentIds` arrays. Previously, passing an empty array would fall through to returning all documents; now it correctly returns an empty array. This fix applies to React, Vue, and Svelte hooks.

.changeset/fix-i18n-vue-locale-reactivity.md

@@ -0,0 +1,5 @@
+---
+'@embedpdf/plugin-i18n': patch
+---
+
+Fixed Vue `useTranslations` hook reactivity for `locale` computed property. The `locale` value now correctly updates when the locale changes.

.changeset/fix-vue-annotation-mixblendmode-inheritance.md

@@ -0,0 +1,5 @@
+---
+'@embedpdf/plugin-annotation': patch
+---
+
+Fixed Vue `AnnotationContainer` component where `mixBlendMode` style was incorrectly applied to the selection menu. The style now only applies to the annotation content div, matching the behavior of React and Svelte implementations. This was caused by Vue's attribute inheritance passing the style to the root element which wrapped both the annotation and the selection menu.

.changeset/hidden-items-dependency-rules.md

@@ -0,0 +1,22 @@
+---
+'@embedpdf/plugin-ui': minor
+---
+
+Added `data-hidden-items` attribute for efficient CSS dependency rules.
+
+**Problem**: Visibility dependency rules (e.g., hiding overflow buttons when all menu items are hidden) required exponential CSS rules when using category-based logic, causing stylesheet bloat.
+
+**Solution**:
+
+- Added `hiddenItems` state that tracks which item IDs are hidden based on disabled categories
+- Dependency rules now use `data-epdf-hid` attribute to check item IDs directly
+- CSS rules are now O(n) per breakpoint instead of O(m^n)
+
+**New APIs**:
+
+- `getHiddenItems()` - returns array of hidden item IDs
+- `onCategoryChanged` event now includes `hiddenItems` in payload
+- `extractItemCategories(schema)` - extracts item→categories mapping
+- `computeHiddenItems(itemCategories, disabledCategories)` - computes hidden items
+
+**Breaking Changes**: None - existing `disabledCategories` API unchanged

.changeset/layout-ready-event-page-info.md

@@ -0,0 +1,5 @@
+---
+'@embedpdf/plugin-scroll': minor
+---
+
+Added `pageNumber` and `totalPages` properties to `LayoutReadyEvent`. This allows consumers to get the current page information immediately when the layout becomes ready, without needing to subscribe to a separate `onPageChange` event.

.changeset/multi-view-core.md

@@ -0,0 +1,36 @@
+---
+'@embedpdf/core': major
+---
+
+## Multi-Document Support
+
+This is a major refactoring to support multiple documents in a single viewer instance. The core architecture has been significantly enhanced to manage per-document state and lifecycle.
+
+### Breaking Changes
+
+- **Store Structure**: Core state now uses `documents: Record<string, DocumentState>` instead of a single `document` property. Each document has its own state including pages, scale, rotation, and other document-specific properties.
+
+- **BasePlugin Lifecycle**: Added new protected lifecycle methods that plugins can override:
+  - `onDocumentLoadingStarted(documentId: string)` - Called when a document starts loading
+  - `onDocumentLoaded(documentId: string)` - Called when a document finishes loading
+  - `onDocumentClosed(documentId: string)` - Called when a document is closed
+  - `onActiveDocumentChanged(previousId: string | null, currentId: string | null)` - Called when the active document changes
+  - `onScaleChanged(documentId: string, scale: number)` - Called when document scale changes
+  - `onRotationChanged(documentId: string, rotation: number)` - Called when document rotation changes
+
+- **Document Access**: New helper methods in BasePlugin:
+  - `getActiveDocumentId()` - Get the active document ID (throws if none)
+  - `getActiveDocumentIdOrNull()` - Get the active document ID or null
+  - `getCoreDocument(documentId?: string)` - Get document state by ID
+  - `getCoreDocumentOrThrow(documentId?: string)` - Get document state or throw
+
+- **Actions**: All core actions now support an optional `documentId` parameter. Actions that previously operated on a single document now require explicit document targeting.
+
+- **State Management**: The store now tracks multiple documents with an `activeDocumentId` field to indicate which document is currently active.
+
+### New Features
+
+- Support for opening and managing multiple PDF documents simultaneously
+- Per-document state isolation
+- Document lifecycle management with proper cleanup
+- Active document tracking and switching

.changeset/multi-view-engines.md

@@ -0,0 +1,18 @@
+---
+'@embedpdf/engines': minor
+---
+
+## Multi-Document Support
+
+Updated engine internals to support multiple documents with improved memory management.
+
+### Changes
+
+- **Memory Management**: Enhanced memory tracking through `MemoryManager` for proper cleanup of multiple document instances.
+
+- **Cache**: `PdfCache` now properly tracks and manages multiple document contexts with improved memory management through the memory manager.
+
+### Technical Details
+
+- Document contexts now use `MemoryManager` for proper WASM pointer tracking and cleanup
+- Improved resource management for concurrent document handling

.changeset/multi-view-models.md

@@ -0,0 +1,17 @@
+---
+'@embedpdf/models': minor
+---
+
+## Multi-Document Support
+
+Minor updates to model types to support multi-document architecture.
+
+### Changes
+
+- **PdfDocumentObject**: Removed optional `name` property. Document identification is now handled through the `id` field.
+
+- **PdfFileWithoutContent**: Removed optional `name` property. File identification is now handled through the `id` field.
+
+### Migration
+
+If you were using the `name` property on documents or files, you should now use the `id` field for identification purposes.

.changeset/multi-view-plugin-annotation.md

@@ -0,0 +1,43 @@
+---
+'@embedpdf/plugin-annotation': major
+---
+
+## Multi-Document Support
+
+The annotation plugin now supports multiple documents with per-document annotation state and tool management.
+
+### Breaking Changes
+
+- **All Actions**: All annotation actions now require a `documentId` parameter:
+  - `setAnnotations(documentId, annotations)` - was `setAnnotations(annotations)`
+  - `selectAnnotation(documentId, pageIndex, id)` - was `selectAnnotation(pageIndex, id)`
+  - `deselectAnnotation(documentId)` - was `deselectAnnotation()` (no params)
+  - `setActiveToolId(documentId, toolId)` - was `setActiveToolId(toolId)`
+  - `createAnnotation(documentId, pageIndex, annotation)` - was `createAnnotation(pageIndex, annotation)`
+  - `patchAnnotation(documentId, pageIndex, id, patch)` - was `patchAnnotation(pageIndex, id, patch)`
+  - `deleteAnnotation(documentId, pageIndex, id)` - was `deleteAnnotation(pageIndex, id)`
+  - `commitPendingChanges(documentId)` - was `commitPendingChanges()` (no params)
+  - `purgeAnnotation(documentId, uid)` - was `purgeAnnotation(uid)`
+
+- **State Structure**: Plugin state now uses `documents: Record<string, AnnotationDocumentState>` instead of a flat structure. Each document has its own annotations, selected annotation, and active tool.
+
+- **Capability Methods**: All capability methods that previously operated on a single document now require document scoping or operate on the active document by default.
+
+### Framework-Specific Changes (React/Preact, Svelte, Vue)
+
+- **AnnotationContainer Component**:
+  - Now requires `documentId` prop (React/Preact: `@embedpdf/plugin-annotation/react`, Svelte: `@embedpdf/plugin-annotation/svelte`, Vue: `@embedpdf/plugin-annotation/vue`)
+  - Component now uses `forDocument(documentId)` to get document-scoped annotation capability
+  - `selectionMenu` prop type changed to `AnnotationSelectionMenuRenderFn` for better type safety
+  - Bounding box constraints now use unscaled page dimensions (scale is applied internally)
+
+- **Annotation Hooks**:
+  - All hooks now work with document-scoped capabilities via `forDocument()`
+  - Components automatically scope operations to the provided `documentId`
+
+### New Features
+
+- Per-document annotation storage and management
+- Per-document active tool tracking
+- Document lifecycle hooks for automatic state initialization and cleanup
+- `forDocument()` method for document-scoped operations

.changeset/multi-view-plugin-attachment.md

@@ -0,0 +1,26 @@
+---
+'@embedpdf/plugin-attachment': major
+---
+
+## Multi-Document Support
+
+The attachment plugin now supports accessing attachments from multiple documents.
+
+### Breaking Changes
+
+- **Methods**: All methods now accept an optional `documentId` parameter and operate on the active document by default:
+  - `getAttachments(documentId?)` - Returns attachments for the specified or active document
+  - `downloadAttachment(attachment, documentId?)` - Downloads attachment from the specified or active document
+
+- **Capability**: Added `forDocument(documentId)` method that returns `AttachmentScope` for document-specific operations.
+
+### Framework-Specific Changes (React/Preact, Svelte, Vue)
+
+- **Hooks**:
+  - Added Svelte hooks support (`@embedpdf/plugin-attachment/svelte`)
+  - All hooks work with document-scoped capabilities via `forDocument()`
+
+### New Features
+
+- `AttachmentScope` interface for document-scoped attachment operations
+- Support for accessing attachments from any document, not just the active one