@laminapdf/wasm 0.2.0-alpha.3 → 0.2.0-alpha.4

package/dist/lamina_pdf_base.d.ts

@@ -32,7 +32,7 @@ export interface ProgressJs {
      */
     total?: number;
     /**
-     * The phase (`parse`, `layout`, `extract`, `render`, `write`, `sign`).
+     * The phase (`parse`, `layout`, `extract`, `render`, `write`, `sign`, `model`).
      */
     phase: string;
 }
@@ -891,7 +891,8 @@ export class WasmPageBuilder {
     [Symbol.dispose](): void;
     /**
      * Opens an artifact until the matching `endTag`: `kind` is `layout`, `page`, `pagination`
-     * or `background`.
+     * or `background`, or a pagination artifact that says what it is: `header` (attached to the
+     * top edge), `footer` (the bottom) or `watermark`.
      *
      * # Errors
      * `invalid-argument` for another kind.

package/dist/lamina_pdf_base.js

@@ -674,7 +674,8 @@ export class WasmPageBuilder {
     }
     /**
      * Opens an artifact until the matching `endTag`: `kind` is `layout`, `page`, `pagination`
-     * or `background`.
+     * or `background`, or a pagination artifact that says what it is: `header` (attached to the
+     * top edge), `footer` (the bottom) or `watermark`.
      *
      * # Errors
      * `invalid-argument` for another kind.

package/dist/lamina_pdf_extract.d.ts

@@ -50,11 +50,35 @@ export interface ProgressJs {
      */
     total?: number;
     /**
-     * The phase (`parse`, `layout`, `extract`, `render`, `write`, `sign`).
+     * The phase (`parse`, `layout`, `extract`, `render`, `write`, `sign`, `model`).
      */
     phase: string;
 }
 
+/**
+ * A kid of a structure element: an element, or marked content on a page.
+ */
+export type StructureKidJs = ({ type: "node" } & StructureNodeJs) | { type: "content"; page: number; mcid: number };
+
+/**
+ * A layout region, its box `[x, y, width, height]` in top-left points (ADR 0027), as the
+ * output's boxes are.
+ */
+export interface RegionInit {
+    /**
+     * What it is.
+     */
+    label: RegionLabelName;
+    /**
+     * `[x, y, width, height]`.
+     */
+    bbox: [number, number, number, number];
+    /**
+     * The model's confidence, 0 to 1.
+     */
+    score: number;
+}
+
 /**
  * A list of ops with its schema version ([ADR 0006] §6): `{ "v": 1, "ops": [ … ] }`, used
  * for saved op logs, batch processing and sync messages. Unknown fields are rejected, and so
@@ -76,11 +100,114 @@ export interface OpLog {
     ops: Op[];
 }
 
+/**
+ * A model result the job waits for.
+ */
+export interface ModelRequestJs {
+    /**
+     * `layout` (regions) or `ocr` (words).
+     */
+    kind: string;
+    /**
+     * The 0-based page to render.
+     */
+    page: number;
+    /**
+     * The resolution to render it at.
+     */
+    dpi: number;
+    /**
+     * OCR's languages (BCP 47); empty for the engine's default.
+     */
+    langs: string[];
+}
+
+/**
+ * A model the host runs, as the output's `generator.models` names it.
+ */
+export interface ModelInfoInit {
+    /**
+     * Its name.
+     */
+    name: string;
+    /**
+     * Its version or revision.
+     */
+    version: string;
+    /**
+     * Its weights' SHA-256, lowercase hex.
+     */
+    sha256: string;
+    /**
+     * Its weights' SPDX licence.
+     */
+    licence: string;
+}
+
+/**
+ * A page and how to show it.
+ */
+export interface TargetJs {
+    /**
+     * The page's 0-based index.
+     */
+    page: number;
+    /**
+     * How to show it, in page space.
+     */
+    view: TargetViewJs;
+}
+
 /**
  * A page-label numbering style.
  */
 export type LabelStyleInit = "decimal" | "upperRoman" | "lowerRoman" | "upperLetters" | "lowerLetters";
 
+/**
+ * A region's kind.
+ */
+export type RegionLabelName = "text" | "title" | "section-header" | "list-item" | "table" | "picture" | "caption" | "footnote" | "formula" | "page-header" | "page-footer" | "other";
+
+/**
+ * A step's progress, and the model result the job waits for, if any.
+ */
+export interface ExtractProgressJs {
+    /**
+     * Pages done.
+     */
+    done: number;
+    /**
+     * Pages asked for.
+     */
+    total?: number;
+    /**
+     * `extract`, or `model` while a request is open.
+     */
+    phase: string;
+    /**
+     * The result the job waits for: hand it to `provide`.
+     */
+    request?: ModelRequestJs;
+}
+
+/**
+ * An OCR word, its box as a region's.
+ */
+export interface WordInit {
+    /**
+     * Its text.
+     */
+    text: string;
+    /**
+     * `[x, y, width, height]`.
+     */
+    bbox: [number, number, number, number];
+    /**
+     * The engine's confidence, 0 to 1.
+     */
+    confidence: number;
+}
+
 /**
  * An attached file's options.
  */
@@ -181,6 +308,11 @@ export interface SetInfo {
  */
 export type FitInit = { type: "page" } | { type: "width"; top?: number } | { type: "height"; left?: number } | { type: "xyz"; left?: number; top?: number; zoom?: number } | { type: "rect"; x: number; y: number; width: number; height: number } | { type: "bounds" };
 
+/**
+ * How a target shows its page, in page space; an unset coordinate is left out.
+ */
+export type TargetViewJs = { kind: "at"; x?: number; y?: number; zoom?: number } | { kind: "fit" } | { kind: "fit-width"; y?: number } | { kind: "fit-height"; x?: number } | { kind: "fit-rect"; x0: number; y0: number; x1: number; y1: number };
+
 /**
  * How a viewer lays pages out.
  */
@@ -237,6 +369,32 @@ export interface ViewerPreferencesInit {
  */
 export type RenumberJs = "keep" | "compact";
 
+/**
+ * One element of the structure tree (`laminapdf::StructureNode`; its strings are document text).
+ */
+export interface StructureNodeJs {
+    /**
+     * Its standard structure type (`H1`, `P`, `LI`, `Figure` …), through the `/RoleMap`.
+     */
+    kind: string;
+    /**
+     * `/Alt`: alternate text.
+     */
+    alt?: string;
+    /**
+     * `/Lang`: the language of its content.
+     */
+    lang?: string;
+    /**
+     * `/ActualText`: the text its content stands for.
+     */
+    actualText?: string;
+    /**
+     * Its kids, in reading order.
+     */
+    kids: StructureKidJs[];
+}
+
 /**
  * One outline item and its children.
  */
@@ -271,6 +429,28 @@ export interface OutlineEntryInit {
     italic?: boolean;
 }
 
+/**
+ * One outline item.
+ */
+export interface OutlineNodeJs {
+    /**
+     * Its title (document text).
+     */
+    title: string;
+    /**
+     * Whether it opens with its children shown.
+     */
+    open: boolean;
+    /**
+     * Where it goes, when its destination names a page of this document.
+     */
+    target?: TargetJs;
+    /**
+     * Its children, in order.
+     */
+    children: OutlineNodeJs[];
+}
+
 /**
  * One page's facts (`laminapdf::PageInfo`).
  */
@@ -402,6 +582,10 @@ export interface ExtractInit {
      * How chunks are cut (`chunks`, and `json` to include them).
      */
     chunk?: ChunkInit;
+    /**
+     * Models the host runs (`extractJob` only): the job asks for their results.
+     */
+    models?: ModelsInit;
 }
 
 /**
@@ -609,6 +793,28 @@ export interface MetadataInit {
  */
 export type ModificationClass = "metadata" | "formFill" | "annotation" | "signature" | "pageStructure" | "content" | "document";
 
+/**
+ * The models the host runs for an extraction job.
+ */
+export interface ModelsInit {
+    /**
+     * A layout model: every page's regions are asked for.
+     */
+    layout?: ModelInfoInit;
+    /**
+     * An OCR engine: pages with no visible text are asked for.
+     */
+    ocr?: ModelInfoInit;
+    /**
+     * The resolution pages are asked for at; 150 by default.
+     */
+    dpi?: number;
+    /**
+     * OCR's languages (BCP 47).
+     */
+    langs?: string[];
+}
+
 /**
  * The pages an edit touched.
  */
@@ -634,6 +840,11 @@ export type SaveModeJs = "auto" | "incremental" | "fullRewrite";
  */
 export type ExtractFormat = "markdown" | "json" | "text" | "chunks";
 
+/**
+ * What `provide` takes.
+ */
+export type ModelResultInit = { kind: "layout"; regions: RegionInit[] } | { kind: "ocr"; words: WordInit[] };
+
 /**
  * What a save wrote (`laminapdf::SaveReport`).
  */
@@ -873,6 +1084,16 @@ export class WasmDocument {
      * error (nothing is applied).
      */
     apply(ops: Op[]): EditedJs;
+    /**
+     * An extraction stepped a page at a time, whose models the host runs: each step returns
+     * its progress, with the result it waits for under `request` (render that page at `dpi`,
+     * run the model, hand the result to `provide`), and once every page is done, the output in
+     * `options.format` (the extract and text modules).
+     *
+     * # Errors
+     * As `extract`, for the options and the pages asked for.
+     */
+    extractJob(options: ExtractInit): WasmExtractJob;
     /**
      * The document's text for retrieval and LLMs, as one string: Markdown, plain text, the
      * `laminapdf.extract/1` JSON, or the chunks as a JSON array (the extract and text modules).
@@ -893,6 +1114,21 @@ export class WasmDocument {
      * Reading errors.
      */
     info(): DocumentInfoJs;
+    /**
+     * Whether the document says it is tagged (`/MarkInfo /Marked true`; the extract and text modules).
+     *
+     * # Errors
+     * Reading errors.
+     */
+    isTagged(): boolean;
+    /**
+     * The named destination `name`, or `undefined` when the document has none by that name or
+     * it names no page here (the extract and text modules).
+     *
+     * # Errors
+     * `data-pending` and the reading errors.
+     */
+    namedDestination(name: string): TargetJs | undefined;
     /**
      * Opens a document from its bytes.
      *
@@ -910,6 +1146,14 @@ export class WasmDocument {
      * As `openBytes`, and `data-pending`.
      */
     static openSource(source: WasmSparseSource, options: OpenOptionsInit | undefined): WasmDocument;
+    /**
+     * The outline's top-level items, each with its children; empty without an outline (the
+     * extract and text modules).
+     *
+     * # Errors
+     * `data-pending` and the reading errors.
+     */
+    outline(): OutlineNodeJs[];
     /**
      * A handle on the page with `id` (`p12.0`).
      *
@@ -973,6 +1217,14 @@ export class WasmDocument {
      * The save's errors (`would-invalidate-signatures`, `requires-rewrite`, …).
      */
     save(options: SaveOptionsInit | undefined): { bytes: Uint8Array<ArrayBuffer>; report: SaveReportJs };
+    /**
+     * The structure tree's top-level elements in reading order, or `undefined` without one (the
+     * extract and text modules, beside `WasmPage.text`, whose runs it links by MCID).
+     *
+     * # Errors
+     * Reading errors.
+     */
+    structureTree(): StructureNodeJs[] | undefined;
     /**
      * Undoes the newest transaction; `undefined` when there is nothing to undo.
      *
@@ -986,6 +1238,38 @@ export class WasmDocument {
     warnings(): WasmWarningCursor;
 }
 
+/**
+ * An extraction in progress, from `WasmDocument.extractJob`; it holds its document weakly.
+ */
+export class WasmExtractJob {
+    private constructor();
+    free(): void;
+    [Symbol.dispose](): void;
+    /**
+     * Stops the job: its next step fails with `cancelled`.
+     */
+    cancel(): void;
+    /**
+     * Hands the job the result of its open request: `{ kind: "layout", regions }` or
+     * `{ kind: "ocr", words }`, boxes as `[x, y, width, height]` in top-left points.
+     *
+     * # Errors
+     * `model-result-unexpected` when nothing was asked for or `result` is the other kind;
+     * `invalid-argument` / `unknown-option` for a malformed result.
+     */
+    provide(result: ModelResultInit): void;
+    /**
+     * Extracts up to `budget` pages: `{ state: "progress", done, total, phase, request? }` or
+     * `{ state: "done", output }`, the output in the job's format. While `request` is set, a
+     * step does nothing until `provide` answers it.
+     *
+     * # Errors
+     * `cancelled`, `document-changed`, `job-finished`, `disposed` when the document is freed,
+     * and the reading errors.
+     */
+    step(budget: number): ({ state: "progress" } & ExtractProgressJs) | { state: "done"; output: string };
+}
+
 /**
  * An image to draw (decoded or passed through once, shared by every use).
  */
@@ -1071,7 +1355,8 @@ export class WasmPageBuilder {
     [Symbol.dispose](): void;
     /**
      * Opens an artifact until the matching `endTag`: `kind` is `layout`, `page`, `pagination`
-     * or `background`.
+     * or `background`, or a pagination artifact that says what it is: `header` (attached to the
+     * top edge), `footer` (the bottom) or `watermark`.
      *
      * # Errors
      * `invalid-argument` for another kind.
@@ -1340,6 +1625,7 @@ export interface InitOutput {
     readonly memory: WebAssembly.Memory;
     readonly __wbg_wasmbuilder_free: (a: number, b: number) => void;
     readonly __wbg_wasmdocument_free: (a: number, b: number) => void;
+    readonly __wbg_wasmextractjob_free: (a: number, b: number) => void;
     readonly __wbg_wasmimage_free: (a: number, b: number) => void;
     readonly __wbg_wasmpage_free: (a: number, b: number) => void;
     readonly __wbg_wasmpagebuilder_free: (a: number, b: number) => void;
@@ -1362,10 +1648,14 @@ export interface InitOutput {
     readonly wasmdocument_apply: (a: number, b: any) => [number, number, number];
     readonly wasmdocument_applyLog: (a: number, b: number, c: number) => [number, number, number];
     readonly wasmdocument_extract: (a: number, b: any) => [number, number, number, number];
+    readonly wasmdocument_extractJob: (a: number, b: any) => [number, number, number];
     readonly wasmdocument_hasUnsavedChanges: (a: number) => number;
     readonly wasmdocument_info: (a: number) => [number, number, number];
+    readonly wasmdocument_isTagged: (a: number) => [number, number, number];
+    readonly wasmdocument_namedDestination: (a: number, b: number, c: number) => [number, number, number];
     readonly wasmdocument_openBytes: (a: number, b: number, c: any) => [number, number, number];
     readonly wasmdocument_openSource: (a: number, b: any) => [number, number, number];
+    readonly wasmdocument_outline: (a: number) => [number, number, number];
     readonly wasmdocument_page: (a: number, b: number) => [number, number, number];
     readonly wasmdocument_pageById: (a: number, b: number, c: number) => [number, number, number];
     readonly wasmdocument_pageCount: (a: number) => [number, number, number];
@@ -1375,8 +1665,12 @@ export interface InitOutput {
     readonly wasmdocument_redo: (a: number) => [number, number, number];
     readonly wasmdocument_save: (a: number, b: any) => [number, number, number];
     readonly wasmdocument_saveJob: (a: number, b: any) => [number, number, number];
+    readonly wasmdocument_structureTree: (a: number) => [number, number, number];
     readonly wasmdocument_undo: (a: number) => [number, number, number];
     readonly wasmdocument_warnings: (a: number) => number;
+    readonly wasmextractjob_cancel: (a: number) => void;
+    readonly wasmextractjob_provide: (a: number, b: any) => [number, number];
+    readonly wasmextractjob_step: (a: number, b: number) => [number, number, number];
     readonly wasmimage_height: (a: number) => number;
     readonly wasmimage_jpeg: (a: number, b: number) => [number, number, number];
     readonly wasmimage_png: (a: number, b: number) => [number, number, number];

package/dist/lamina_pdf_extract.js

@@ -240,6 +240,25 @@ export class WasmDocument {
         }
         return takeFromExternrefTable0(ret[0]);
     }
+    /**
+     * An extraction stepped a page at a time, whose models the host runs: each step returns
+     * its progress, with the result it waits for under `request` (render that page at `dpi`,
+     * run the model, hand the result to `provide`), and once every page is done, the output in
+     * `options.format` (the extract and text modules).
+     *
+     * # Errors
+     * As `extract`, for the options and the pages asked for.
+     * @param {ExtractInit} options
+     * @returns {WasmExtractJob}
+     */
+    extractJob(options) {
+        let ret;
+        ret = wasm.wasmdocument_extractJob(this.__wbg_ptr, options);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return WasmExtractJob.__wrap(ret[0]);
+    }
     /**
      * The document's text for retrieval and LLMs, as one string: Markdown, plain text, the
      * `laminapdf.extract/1` JSON, or the chunks as a JSON array (the extract and text modules).
@@ -293,6 +312,40 @@ export class WasmDocument {
         }
         return takeFromExternrefTable0(ret[0]);
     }
+    /**
+     * Whether the document says it is tagged (`/MarkInfo /Marked true`; the extract and text modules).
+     *
+     * # Errors
+     * Reading errors.
+     * @returns {boolean}
+     */
+    isTagged() {
+        let ret;
+        ret = wasm.wasmdocument_isTagged(this.__wbg_ptr);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return ret[0] !== 0;
+    }
+    /**
+     * The named destination `name`, or `undefined` when the document has none by that name or
+     * it names no page here (the extract and text modules).
+     *
+     * # Errors
+     * `data-pending` and the reading errors.
+     * @param {string} name
+     * @returns {TargetJs | undefined}
+     */
+    namedDestination(name) {
+        const ptr0 = passStringToWasm0(name, wasm.__wbindgen_malloc, wasm.__wbindgen_realloc);
+        const len0 = WASM_VECTOR_LEN;
+        let ret;
+        ret = wasm.wasmdocument_namedDestination(this.__wbg_ptr, ptr0, len0);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return takeFromExternrefTable0(ret[0]);
+    }
     /**
      * Opens a document from its bytes.
      *
@@ -333,6 +386,22 @@ export class WasmDocument {
         }
         return WasmDocument.__wrap(ret[0]);
     }
+    /**
+     * The outline's top-level items, each with its children; empty without an outline (the
+     * extract and text modules).
+     *
+     * # Errors
+     * `data-pending` and the reading errors.
+     * @returns {OutlineNodeJs[]}
+     */
+    outline() {
+        let ret;
+        ret = wasm.wasmdocument_outline(this.__wbg_ptr);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return takeFromExternrefTable0(ret[0]);
+    }
     /**
      * A handle on the page with `id` (`p12.0`).
      *
@@ -476,6 +545,22 @@ export class WasmDocument {
         }
         return takeFromExternrefTable0(ret[0]);
     }
+    /**
+     * The structure tree's top-level elements in reading order, or `undefined` without one (the
+     * extract and text modules, beside `WasmPage.text`, whose runs it links by MCID).
+     *
+     * # Errors
+     * Reading errors.
+     * @returns {StructureNodeJs[] | undefined}
+     */
+    structureTree() {
+        let ret;
+        ret = wasm.wasmdocument_structureTree(this.__wbg_ptr);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return takeFromExternrefTable0(ret[0]);
+    }
     /**
      * Undoes the newest transaction; `undefined` when there is nothing to undo.
      *
@@ -503,6 +588,73 @@ export class WasmDocument {
 }
 if (Symbol.dispose) WasmDocument.prototype[Symbol.dispose] = WasmDocument.prototype.free;
 
+/**
+ * An extraction in progress, from `WasmDocument.extractJob`; it holds its document weakly.
+ */
+export class WasmExtractJob {
+    static __wrap(ptr) {
+        const obj = Object.create(WasmExtractJob.prototype);
+        obj.__wbg_ptr = ptr;
+        Object.defineProperty(obj, '__wbg_inst', { value: __wbg_instance_id, writable: true });
+
+        WasmExtractJobFinalization.register(obj, { ptr, instance: __wbg_instance_id }, obj);
+        return obj;
+    }
+    __destroy_into_raw() {
+        const ptr = this.__wbg_ptr;
+        this.__wbg_ptr = 0;
+        WasmExtractJobFinalization.unregister(this);
+        return ptr;
+    }
+    free() {
+        const ptr = this.__destroy_into_raw();
+        wasm.__wbg_wasmextractjob_free(ptr, 0);
+
+    }
+    /**
+     * Stops the job: its next step fails with `cancelled`.
+     */
+    cancel() {
+        wasm.wasmextractjob_cancel(this.__wbg_ptr);
+    }
+    /**
+     * Hands the job the result of its open request: `{ kind: "layout", regions }` or
+     * `{ kind: "ocr", words }`, boxes as `[x, y, width, height]` in top-left points.
+     *
+     * # Errors
+     * `model-result-unexpected` when nothing was asked for or `result` is the other kind;
+     * `invalid-argument` / `unknown-option` for a malformed result.
+     * @param {ModelResultInit} result
+     */
+    provide(result) {
+        let ret;
+        ret = wasm.wasmextractjob_provide(this.__wbg_ptr, result);
+        if (ret[1]) {
+            throw takeFromExternrefTable0(ret[0]);
+        }
+    }
+    /**
+     * Extracts up to `budget` pages: `{ state: "progress", done, total, phase, request? }` or
+     * `{ state: "done", output }`, the output in the job's format. While `request` is set, a
+     * step does nothing until `provide` answers it.
+     *
+     * # Errors
+     * `cancelled`, `document-changed`, `job-finished`, `disposed` when the document is freed,
+     * and the reading errors.
+     * @param {number} budget
+     * @returns {({ state: "progress" } & ExtractProgressJs) | { state: "done"; output: string }}
+     */
+    step(budget) {
+        let ret;
+        ret = wasm.wasmextractjob_step(this.__wbg_ptr, budget);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return takeFromExternrefTable0(ret[0]);
+    }
+}
+if (Symbol.dispose) WasmExtractJob.prototype[Symbol.dispose] = WasmExtractJob.prototype.free;
+
 /**
  * An image to draw (decoded or passed through once, shared by every use).
  */
@@ -721,7 +873,8 @@ export class WasmPageBuilder {
     }
     /**
      * Opens an artifact until the matching `endTag`: `kind` is `layout`, `page`, `pagination`
-     * or `background`.
+     * or `background`, or a pagination artifact that says what it is: `header` (attached to the
+     * top edge), `footer` (the bottom) or `watermark`.
      *
      * # Errors
      * `invalid-argument` for another kind.
@@ -1488,6 +1641,11 @@ const WasmDocumentFinalization = (typeof FinalizationRegistry === 'undefined')
     : new FinalizationRegistry(({ ptr, instance }) => {
     if (instance === __wbg_instance_id) wasm.__wbg_wasmdocument_free(ptr, 1);
 });
+const WasmExtractJobFinalization = (typeof FinalizationRegistry === 'undefined')
+    ? { register: () => {}, unregister: () => {} }
+    : new FinalizationRegistry(({ ptr, instance }) => {
+    if (instance === __wbg_instance_id) wasm.__wbg_wasmextractjob_free(ptr, 1);
+});
 const WasmImageFinalization = (typeof FinalizationRegistry === 'undefined')
     ? { register: () => {}, unregister: () => {} }
     : new FinalizationRegistry(({ ptr, instance }) => {

package/dist/lamina_pdf_text.d.ts

@@ -55,11 +55,35 @@ export interface ProgressJs {
      */
     total?: number;
     /**
-     * The phase (`parse`, `layout`, `extract`, `render`, `write`, `sign`).
+     * The phase (`parse`, `layout`, `extract`, `render`, `write`, `sign`, `model`).
      */
     phase: string;
 }
 
+/**
+ * A kid of a structure element: an element, or marked content on a page.
+ */
+export type StructureKidJs = ({ type: "node" } & StructureNodeJs) | { type: "content"; page: number; mcid: number };
+
+/**
+ * A layout region, its box `[x, y, width, height]` in top-left points (ADR 0027), as the
+ * output's boxes are.
+ */
+export interface RegionInit {
+    /**
+     * What it is.
+     */
+    label: RegionLabelName;
+    /**
+     * `[x, y, width, height]`.
+     */
+    bbox: [number, number, number, number];
+    /**
+     * The model's confidence, 0 to 1.
+     */
+    score: number;
+}
+
 /**
  * A link over drawn text.
  */
@@ -95,6 +119,64 @@ export interface OpLog {
     ops: Op[];
 }
 
+/**
+ * A model result the job waits for.
+ */
+export interface ModelRequestJs {
+    /**
+     * `layout` (regions) or `ocr` (words).
+     */
+    kind: string;
+    /**
+     * The 0-based page to render.
+     */
+    page: number;
+    /**
+     * The resolution to render it at.
+     */
+    dpi: number;
+    /**
+     * OCR's languages (BCP 47); empty for the engine's default.
+     */
+    langs: string[];
+}
+
+/**
+ * A model the host runs, as the output's `generator.models` names it.
+ */
+export interface ModelInfoInit {
+    /**
+     * Its name.
+     */
+    name: string;
+    /**
+     * Its version or revision.
+     */
+    version: string;
+    /**
+     * Its weights' SHA-256, lowercase hex.
+     */
+    sha256: string;
+    /**
+     * Its weights' SPDX licence.
+     */
+    licence: string;
+}
+
+/**
+ * A page and how to show it.
+ */
+export interface TargetJs {
+    /**
+     * The page's 0-based index.
+     */
+    page: number;
+    /**
+     * How to show it, in page space.
+     */
+    view: TargetViewJs;
+}
+
 /**
  * A page-label numbering style.
  */
@@ -122,11 +204,56 @@ export interface RectJs {
     height: number;
 }
 
+/**
+ * A region's kind.
+ */
+export type RegionLabelName = "text" | "title" | "section-header" | "list-item" | "table" | "picture" | "caption" | "footnote" | "formula" | "page-header" | "page-footer" | "other";
+
+/**
+ * A step's progress, and the model result the job waits for, if any.
+ */
+export interface ExtractProgressJs {
+    /**
+     * Pages done.
+     */
+    done: number;
+    /**
+     * Pages asked for.
+     */
+    total?: number;
+    /**
+     * `extract`, or `model` while a request is open.
+     */
+    phase: string;
+    /**
+     * The result the job waits for: hand it to `provide`.
+     */
+    request?: ModelRequestJs;
+}
+
 /**
  * Alignment of a text box's lines.
  */
 export type AlignInit = "start" | "end" | "left" | "right" | "center";
 
+/**
+ * An OCR word, its box as a region's.
+ */
+export interface WordInit {
+    /**
+     * Its text.
+     */
+    text: string;
+    /**
+     * `[x, y, width, height]`.
+     */
+    bbox: [number, number, number, number];
+    /**
+     * The engine's confidence, 0 to 1.
+     */
+    confidence: number;
+}
+
 /**
  * An attached file's options.
  */
@@ -257,6 +384,11 @@ export interface TextOptionsInit {
     direction?: DirectionInit;
 }
 
+/**
+ * How a target shows its page, in page space; an unset coordinate is left out.
+ */
+export type TargetViewJs = { kind: "at"; x?: number; y?: number; zoom?: number } | { kind: "fit" } | { kind: "fit-width"; y?: number } | { kind: "fit-height"; x?: number } | { kind: "fit-rect"; x0: number; y0: number; x1: number; y1: number };
+
 /**
  * How a text box laid its text out.
  */
@@ -266,9 +398,15 @@ export interface TextLayoutJs {
      */
     lines: LineBoxJs[];
     /**
-     * Where the text that didn't fit starts (UTF-16 offset), if any.
+     * Where the text that didn't fit starts (UTF-16 offset), if any; `0` when no line fitted.
      */
     overflow?: number;
+    /**
+     * Where the next element can start: the bottom of the last drawn line that draws
+     * something, before the surface's transform. Left out when no line does (nothing, only
+     * white space and control characters, or a box too short).
+     */
+    bottom?: number;
 }
 
 /**
@@ -354,6 +492,32 @@ export interface LineBoxJs {
     baseline: number;
 }
 
+/**
+ * One element of the structure tree (`laminapdf::StructureNode`; its strings are document text).
+ */
+export interface StructureNodeJs {
+    /**
+     * Its standard structure type (`H1`, `P`, `LI`, `Figure` …), through the `/RoleMap`.
+     */
+    kind: string;
+    /**
+     * `/Alt`: alternate text.
+     */
+    alt?: string;
+    /**
+     * `/Lang`: the language of its content.
+     */
+    lang?: string;
+    /**
+     * `/ActualText`: the text its content stands for.
+     */
+    actualText?: string;
+    /**
+     * Its kids, in reading order.
+     */
+    kids: StructureKidJs[];
+}
+
 /**
  * One outline item and its children.
  */
@@ -388,6 +552,28 @@ export interface OutlineEntryInit {
     italic?: boolean;
 }
 
+/**
+ * One outline item.
+ */
+export interface OutlineNodeJs {
+    /**
+     * Its title (document text).
+     */
+    title: string;
+    /**
+     * Whether it opens with its children shown.
+     */
+    open: boolean;
+    /**
+     * Where it goes, when its destination names a page of this document.
+     */
+    target?: TargetJs;
+    /**
+     * Its children, in order.
+     */
+    children: OutlineNodeJs[];
+}
+
 /**
  * One page's facts (`laminapdf::PageInfo`).
  */
@@ -519,6 +705,10 @@ export interface ExtractInit {
      * How chunks are cut (`chunks`, and `json` to include them).
      */
     chunk?: ChunkInit;
+    /**
+     * Models the host runs (`extractJob` only): the job asks for their results.
+     */
+    models?: ModelsInit;
 }
 
 /**
@@ -766,6 +956,28 @@ export interface MetadataInit {
  */
 export type ModificationClass = "metadata" | "formFill" | "annotation" | "signature" | "pageStructure" | "content" | "document";
 
+/**
+ * The models the host runs for an extraction job.
+ */
+export interface ModelsInit {
+    /**
+     * A layout model: every page's regions are asked for.
+     */
+    layout?: ModelInfoInit;
+    /**
+     * An OCR engine: pages with no visible text are asked for.
+     */
+    ocr?: ModelInfoInit;
+    /**
+     * The resolution pages are asked for at; 150 by default.
+     */
+    dpi?: number;
+    /**
+     * OCR's languages (BCP 47).
+     */
+    langs?: string[];
+}
+
 /**
  * The pages an edit touched.
  */
@@ -791,6 +1003,11 @@ export type SaveModeJs = "auto" | "incremental" | "fullRewrite";
  */
 export type ExtractFormat = "markdown" | "json" | "text" | "chunks";
 
+/**
+ * What `provide` takes.
+ */
+export type ModelResultInit = { kind: "layout"; regions: RegionInit[] } | { kind: "ocr"; words: WordInit[] };
+
 /**
  * What a run measures.
  */
@@ -807,6 +1024,12 @@ export interface TextRunJs {
      * Its lines (a run breaks only at line breaks in the text).
      */
     lines: number;
+    /**
+     * Where the next element can start: the bottom of the descent below its last line that
+     * draws something, before the surface's transform. Left out when nothing is drawn (empty
+     * text, or only white space and control characters); invisible text counts.
+     */
+    bottom?: number;
 }
 
 /**
@@ -1069,6 +1292,16 @@ export class WasmDocument {
      * error (nothing is applied).
      */
     apply(ops: Op[]): EditedJs;
+    /**
+     * An extraction stepped a page at a time, whose models the host runs: each step returns
+     * its progress, with the result it waits for under `request` (render that page at `dpi`,
+     * run the model, hand the result to `provide`), and once every page is done, the output in
+     * `options.format` (the extract and text modules).
+     *
+     * # Errors
+     * As `extract`, for the options and the pages asked for.
+     */
+    extractJob(options: ExtractInit): WasmExtractJob;
     /**
      * The document's text for retrieval and LLMs, as one string: Markdown, plain text, the
      * `laminapdf.extract/1` JSON, or the chunks as a JSON array (the extract and text modules).
@@ -1089,6 +1322,21 @@ export class WasmDocument {
      * Reading errors.
      */
     info(): DocumentInfoJs;
+    /**
+     * Whether the document says it is tagged (`/MarkInfo /Marked true`; the extract and text modules).
+     *
+     * # Errors
+     * Reading errors.
+     */
+    isTagged(): boolean;
+    /**
+     * The named destination `name`, or `undefined` when the document has none by that name or
+     * it names no page here (the extract and text modules).
+     *
+     * # Errors
+     * `data-pending` and the reading errors.
+     */
+    namedDestination(name: string): TargetJs | undefined;
     /**
      * Opens a document from its bytes.
      *
@@ -1106,6 +1354,14 @@ export class WasmDocument {
      * As `openBytes`, and `data-pending`.
      */
     static openSource(source: WasmSparseSource, options: OpenOptionsInit | undefined): WasmDocument;
+    /**
+     * The outline's top-level items, each with its children; empty without an outline (the
+     * extract and text modules).
+     *
+     * # Errors
+     * `data-pending` and the reading errors.
+     */
+    outline(): OutlineNodeJs[];
     /**
      * A handle on the page with `id` (`p12.0`).
      *
@@ -1169,6 +1425,14 @@ export class WasmDocument {
      * The save's errors (`would-invalidate-signatures`, `requires-rewrite`, …).
      */
     save(options: SaveOptionsInit | undefined): { bytes: Uint8Array<ArrayBuffer>; report: SaveReportJs };
+    /**
+     * The structure tree's top-level elements in reading order, or `undefined` without one (the
+     * extract and text modules, beside `WasmPage.text`, whose runs it links by MCID).
+     *
+     * # Errors
+     * Reading errors.
+     */
+    structureTree(): StructureNodeJs[] | undefined;
     /**
      * Undoes the newest transaction; `undefined` when there is nothing to undo.
      *
@@ -1182,6 +1446,38 @@ export class WasmDocument {
     warnings(): WasmWarningCursor;
 }
 
+/**
+ * An extraction in progress, from `WasmDocument.extractJob`; it holds its document weakly.
+ */
+export class WasmExtractJob {
+    private constructor();
+    free(): void;
+    [Symbol.dispose](): void;
+    /**
+     * Stops the job: its next step fails with `cancelled`.
+     */
+    cancel(): void;
+    /**
+     * Hands the job the result of its open request: `{ kind: "layout", regions }` or
+     * `{ kind: "ocr", words }`, boxes as `[x, y, width, height]` in top-left points.
+     *
+     * # Errors
+     * `model-result-unexpected` when nothing was asked for or `result` is the other kind;
+     * `invalid-argument` / `unknown-option` for a malformed result.
+     */
+    provide(result: ModelResultInit): void;
+    /**
+     * Extracts up to `budget` pages: `{ state: "progress", done, total, phase, request? }` or
+     * `{ state: "done", output }`, the output in the job's format. While `request` is set, a
+     * step does nothing until `provide` answers it.
+     *
+     * # Errors
+     * `cancelled`, `document-changed`, `job-finished`, `disposed` when the document is freed,
+     * and the reading errors.
+     */
+    step(budget: number): ({ state: "progress" } & ExtractProgressJs) | { state: "done"; output: string };
+}
+
 /**
  * A font registered with a [`WasmBuilder`].
  */
@@ -1288,7 +1584,8 @@ export class WasmPageBuilder {
     [Symbol.dispose](): void;
     /**
      * Opens an artifact until the matching `endTag`: `kind` is `layout`, `page`, `pagination`
-     * or `background`.
+     * or `background`, or a pagination artifact that says what it is: `header` (attached to the
+     * top edge), `footer` (the bottom) or `watermark`.
      *
      * # Errors
      * `invalid-argument` for another kind.
@@ -1574,6 +1871,7 @@ export interface InitOutput {
     readonly memory: WebAssembly.Memory;
     readonly __wbg_wasmbuilder_free: (a: number, b: number) => void;
     readonly __wbg_wasmdocument_free: (a: number, b: number) => void;
+    readonly __wbg_wasmextractjob_free: (a: number, b: number) => void;
     readonly __wbg_wasmfont_free: (a: number, b: number) => void;
     readonly __wbg_wasmimage_free: (a: number, b: number) => void;
     readonly __wbg_wasmpage_free: (a: number, b: number) => void;
@@ -1599,10 +1897,14 @@ export interface InitOutput {
     readonly wasmdocument_apply: (a: number, b: any) => [number, number, number];
     readonly wasmdocument_applyLog: (a: number, b: number, c: number) => [number, number, number];
     readonly wasmdocument_extract: (a: number, b: any) => [number, number, number, number];
+    readonly wasmdocument_extractJob: (a: number, b: any) => [number, number, number];
     readonly wasmdocument_hasUnsavedChanges: (a: number) => number;
     readonly wasmdocument_info: (a: number) => [number, number, number];
+    readonly wasmdocument_isTagged: (a: number) => [number, number, number];
+    readonly wasmdocument_namedDestination: (a: number, b: number, c: number) => [number, number, number];
     readonly wasmdocument_openBytes: (a: number, b: number, c: any) => [number, number, number];
     readonly wasmdocument_openSource: (a: number, b: any) => [number, number, number];
+    readonly wasmdocument_outline: (a: number) => [number, number, number];
     readonly wasmdocument_page: (a: number, b: number) => [number, number, number];
     readonly wasmdocument_pageById: (a: number, b: number, c: number) => [number, number, number];
     readonly wasmdocument_pageCount: (a: number) => [number, number, number];
@@ -1612,8 +1914,12 @@ export interface InitOutput {
     readonly wasmdocument_redo: (a: number) => [number, number, number];
     readonly wasmdocument_save: (a: number, b: any) => [number, number, number];
     readonly wasmdocument_saveJob: (a: number, b: any) => [number, number, number];
+    readonly wasmdocument_structureTree: (a: number) => [number, number, number];
     readonly wasmdocument_undo: (a: number) => [number, number, number];
     readonly wasmdocument_warnings: (a: number) => number;
+    readonly wasmextractjob_cancel: (a: number) => void;
+    readonly wasmextractjob_provide: (a: number, b: any) => [number, number];
+    readonly wasmextractjob_step: (a: number, b: number) => [number, number, number];
     readonly wasmfont_family: (a: number) => [number, number];
     readonly wasmfont_isStandard: (a: number) => number;
     readonly wasmfont_postscriptName: (a: number) => [number, number];

package/dist/lamina_pdf_text.js

@@ -278,6 +278,25 @@ export class WasmDocument {
         }
         return takeFromExternrefTable0(ret[0]);
     }
+    /**
+     * An extraction stepped a page at a time, whose models the host runs: each step returns
+     * its progress, with the result it waits for under `request` (render that page at `dpi`,
+     * run the model, hand the result to `provide`), and once every page is done, the output in
+     * `options.format` (the extract and text modules).
+     *
+     * # Errors
+     * As `extract`, for the options and the pages asked for.
+     * @param {ExtractInit} options
+     * @returns {WasmExtractJob}
+     */
+    extractJob(options) {
+        let ret;
+        ret = wasm.wasmdocument_extractJob(this.__wbg_ptr, options);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return WasmExtractJob.__wrap(ret[0]);
+    }
     /**
      * The document's text for retrieval and LLMs, as one string: Markdown, plain text, the
      * `laminapdf.extract/1` JSON, or the chunks as a JSON array (the extract and text modules).
@@ -331,6 +350,40 @@ export class WasmDocument {
         }
         return takeFromExternrefTable0(ret[0]);
     }
+    /**
+     * Whether the document says it is tagged (`/MarkInfo /Marked true`; the extract and text modules).
+     *
+     * # Errors
+     * Reading errors.
+     * @returns {boolean}
+     */
+    isTagged() {
+        let ret;
+        ret = wasm.wasmdocument_isTagged(this.__wbg_ptr);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return ret[0] !== 0;
+    }
+    /**
+     * The named destination `name`, or `undefined` when the document has none by that name or
+     * it names no page here (the extract and text modules).
+     *
+     * # Errors
+     * `data-pending` and the reading errors.
+     * @param {string} name
+     * @returns {TargetJs | undefined}
+     */
+    namedDestination(name) {
+        const ptr0 = passStringToWasm0(name, wasm.__wbindgen_malloc, wasm.__wbindgen_realloc);
+        const len0 = WASM_VECTOR_LEN;
+        let ret;
+        ret = wasm.wasmdocument_namedDestination(this.__wbg_ptr, ptr0, len0);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return takeFromExternrefTable0(ret[0]);
+    }
     /**
      * Opens a document from its bytes.
      *
@@ -371,6 +424,22 @@ export class WasmDocument {
         }
         return WasmDocument.__wrap(ret[0]);
     }
+    /**
+     * The outline's top-level items, each with its children; empty without an outline (the
+     * extract and text modules).
+     *
+     * # Errors
+     * `data-pending` and the reading errors.
+     * @returns {OutlineNodeJs[]}
+     */
+    outline() {
+        let ret;
+        ret = wasm.wasmdocument_outline(this.__wbg_ptr);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return takeFromExternrefTable0(ret[0]);
+    }
     /**
      * A handle on the page with `id` (`p12.0`).
      *
@@ -514,6 +583,22 @@ export class WasmDocument {
         }
         return takeFromExternrefTable0(ret[0]);
     }
+    /**
+     * The structure tree's top-level elements in reading order, or `undefined` without one (the
+     * extract and text modules, beside `WasmPage.text`, whose runs it links by MCID).
+     *
+     * # Errors
+     * Reading errors.
+     * @returns {StructureNodeJs[] | undefined}
+     */
+    structureTree() {
+        let ret;
+        ret = wasm.wasmdocument_structureTree(this.__wbg_ptr);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return takeFromExternrefTable0(ret[0]);
+    }
     /**
      * Undoes the newest transaction; `undefined` when there is nothing to undo.
      *
@@ -541,6 +626,73 @@ export class WasmDocument {
 }
 if (Symbol.dispose) WasmDocument.prototype[Symbol.dispose] = WasmDocument.prototype.free;
 
+/**
+ * An extraction in progress, from `WasmDocument.extractJob`; it holds its document weakly.
+ */
+export class WasmExtractJob {
+    static __wrap(ptr) {
+        const obj = Object.create(WasmExtractJob.prototype);
+        obj.__wbg_ptr = ptr;
+        Object.defineProperty(obj, '__wbg_inst', { value: __wbg_instance_id, writable: true });
+
+        WasmExtractJobFinalization.register(obj, { ptr, instance: __wbg_instance_id }, obj);
+        return obj;
+    }
+    __destroy_into_raw() {
+        const ptr = this.__wbg_ptr;
+        this.__wbg_ptr = 0;
+        WasmExtractJobFinalization.unregister(this);
+        return ptr;
+    }
+    free() {
+        const ptr = this.__destroy_into_raw();
+        wasm.__wbg_wasmextractjob_free(ptr, 0);
+
+    }
+    /**
+     * Stops the job: its next step fails with `cancelled`.
+     */
+    cancel() {
+        wasm.wasmextractjob_cancel(this.__wbg_ptr);
+    }
+    /**
+     * Hands the job the result of its open request: `{ kind: "layout", regions }` or
+     * `{ kind: "ocr", words }`, boxes as `[x, y, width, height]` in top-left points.
+     *
+     * # Errors
+     * `model-result-unexpected` when nothing was asked for or `result` is the other kind;
+     * `invalid-argument` / `unknown-option` for a malformed result.
+     * @param {ModelResultInit} result
+     */
+    provide(result) {
+        let ret;
+        ret = wasm.wasmextractjob_provide(this.__wbg_ptr, result);
+        if (ret[1]) {
+            throw takeFromExternrefTable0(ret[0]);
+        }
+    }
+    /**
+     * Extracts up to `budget` pages: `{ state: "progress", done, total, phase, request? }` or
+     * `{ state: "done", output }`, the output in the job's format. While `request` is set, a
+     * step does nothing until `provide` answers it.
+     *
+     * # Errors
+     * `cancelled`, `document-changed`, `job-finished`, `disposed` when the document is freed,
+     * and the reading errors.
+     * @param {number} budget
+     * @returns {({ state: "progress" } & ExtractProgressJs) | { state: "done"; output: string }}
+     */
+    step(budget) {
+        let ret;
+        ret = wasm.wasmextractjob_step(this.__wbg_ptr, budget);
+        if (ret[2]) {
+            throw takeFromExternrefTable0(ret[1]);
+        }
+        return takeFromExternrefTable0(ret[0]);
+    }
+}
+if (Symbol.dispose) WasmExtractJob.prototype[Symbol.dispose] = WasmExtractJob.prototype.free;
+
 /**
  * A font registered with a [`WasmBuilder`].
  */
@@ -828,7 +980,8 @@ export class WasmPageBuilder {
     }
     /**
      * Opens an artifact until the matching `endTag`: `kind` is `layout`, `page`, `pagination`
-     * or `background`.
+     * or `background`, or a pagination artifact that says what it is: `header` (attached to the
+     * top edge), `footer` (the bottom) or `watermark`.
      *
      * # Errors
      * `invalid-argument` for another kind.
@@ -1648,6 +1801,11 @@ const WasmDocumentFinalization = (typeof FinalizationRegistry === 'undefined')
     : new FinalizationRegistry(({ ptr, instance }) => {
     if (instance === __wbg_instance_id) wasm.__wbg_wasmdocument_free(ptr, 1);
 });
+const WasmExtractJobFinalization = (typeof FinalizationRegistry === 'undefined')
+    ? { register: () => {}, unregister: () => {} }
+    : new FinalizationRegistry(({ ptr, instance }) => {
+    if (instance === __wbg_instance_id) wasm.__wbg_wasmextractjob_free(ptr, 1);
+});
 const WasmFontFinalization = (typeof FinalizationRegistry === 'undefined')
     ? { register: () => {}, unregister: () => {} }
     : new FinalizationRegistry(({ ptr, instance }) => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@laminapdf/wasm",
-  "version": "0.2.0-alpha.3",
+  "version": "0.2.0-alpha.4",
   "description": "Lamina PDF's WebAssembly modules and their wasm-bindgen glue: lamina_pdf_base (open, read, edit, save, vector and image pages), lamina_pdf_text (all of base plus fonts and text) and lamina_pdf_extract (all of base plus text extraction); and lamina_pdf_render_hayro, laminapdf-render's hayro adapter, which @laminapdf/render-hayro wraps. Generated into dist/ by `cargo xtask wasm`; @laminapdf/sdk is the API to use.",
   "license": "MIT OR Apache-2.0",
   "homepage": "https://laminapdf.com",