extract-pdf 0.1.114 → 0.1.115

package/README.md

@@ -54,10 +54,12 @@ const { html } = await convertPDFToHTML(buffer, { addPageNumbers: true });
 
 ### Options
 
-| Option           | Default | Description                                                                                |
-| ---------------- | ------- | ------------------------------------------------------------------------------------------ |
-| `addPageNumbers` | `false` | Inserts `[n]` markers at each page boundary                                                |
-| `addCitation`    | `true`  | Reads PDF metadata and first-page heading to populate `title`/`author` in the return value |
+| Option            | Default               | Description                                                                                |
+| ----------------- | ---------------------- | ------------------------------------------------------------------------------------------ |
+| `addPageNumbers`  | `false`                | Inserts `[n]` markers at each page boundary                                                |
+| `addCitation`     | `true`                 | Reads PDF metadata and first-page heading to populate `title`/`author` in the return value |
+| `method`          | `"ts-block-algorithm"` | Parsing engine — `"ts-block-algorithm"` or `"liteparse"` (see below)                       |
+| `liteParseOptions`| `{}`                   | Passed through to LiteParse's constructor when `method: "liteparse"`                       |
 
 ### Return value
 
@@ -65,6 +67,63 @@ const { html } = await convertPDFToHTML(buffer, { addPageNumbers: true });
 { html: string, title?: string, author?: string, format: "pdf" }
 ```
 
+## Parse methods
+
+`convertPDFToHTML` supports two interchangeable parsing engines via `options.method`:
+
+| Method                              | Engine                                                              | Environments                       | OCR |
+| ------------------------------------ | -------------------------------------------------------------------- | ----------------------------------- | --- |
+| `"ts-block-algorithm"` (default)     | The pure-TS pipeline documented below (this package)                  | Node.js, Cloudflare Workers, browser | No  |
+| `"liteparse"`                        | [LiteParse](https://github.com/run-llama/liteparse) (native, `@llamaindex/liteparse`) | Node.js only                        | Optional |
+
+```ts
+import { convertPDFToHTML } from "extract-pdf";
+
+const { html } = await convertPDFToHTML(buffer, { method: "liteparse" });
+```
+
+LiteParse ships a native (napi) addon, so `method: "liteparse"` only runs in
+Node.js — it is not bundled into browser or Cloudflare Workers builds. Install
+it explicitly (`bun add @llamaindex/liteparse`) since it's an optional
+dependency; if it isn't installed, `convertPDFToHTML` returns `{ error }`
+instead of throwing.
+
+By default the LiteParse path runs with OCR disabled (`ocrEnabled: false`) —
+matching this package's "instant, no backend" philosophy. Use
+`detectPdfNeedsOcr` (below) to decide when a document is worth re-parsing with
+`liteParseOptions: { ocrEnabled: true }`.
+
+## Detecting whether a PDF needs OCR
+
+Before committing to a full (and potentially slow) OCR parse, `detectPdfNeedsOcr`
+runs a cheap, text-layer-only pass and reports whether each page needs OCR or
+other heavy parsing — useful for routing documents to different pipelines
+(fast path vs. OCR vs. screenshots vs. a heavier parser like LlamaParse or
+Docling):
+
+```ts
+import { detectPdfNeedsOcr, convertPDFToHTML } from "extract-pdf";
+
+const assessment = await detectPdfNeedsOcr(buffer);
+// { needsOcr: boolean, pages: PageComplexityStats[], reasons: string[] }
+
+if (!assessment.needsOcr) {
+  const { html } = await convertPDFToHTML(buffer, { method: "liteparse" });
+} else {
+  console.log("Needs OCR:", assessment.reasons); // e.g. ["scanned", "sparse-text"]
+  // Route to an OCR-enabled pipeline, e.g.:
+  const { html } = await convertPDFToHTML(buffer, {
+    method: "liteparse",
+    liteParseOptions: { ocrEnabled: true },
+  });
+}
+```
+
+`reasons` collects every distinct signal found across pages: `"scanned"`,
+`"no-text"`, `"sparse-text"`, `"embedded-images"`, `"garbled"`, or
+`"vector-text"`. Like `method: "liteparse"`, this is Node.js only and requires
+`@llamaindex/liteparse`.
+
 ## Pipeline
 
 The conversion runs a sequential chain of transformations on a `ParseResult` (pages → items):

package/dist/detect-needs-ocr.d.ts

@@ -0,0 +1,23 @@
+import { LiteParseConfig, PageComplexityStats } from '@llamaindex/liteparse';
+export interface DetectPdfNeedsOcrOptions {
+    liteParseOptions?: Partial<LiteParseConfig>;
+}
+export interface PdfOcrAssessment {
+    /** True when any page needs OCR or other advanced parsing. */
+    needsOcr: boolean;
+    /** Per-page complexity signals, one entry per page. */
+    pages: PageComplexityStats[];
+    /**
+     * Every distinct reason found across all pages, e.g. `"scanned"`,
+     * `"no-text"`, `"sparse-text"`, `"embedded-images"`, `"garbled"`,
+     * `"vector-text"`. Empty when `needsOcr` is false.
+     */
+    reasons: string[];
+}
+/**
+ * Determines whether a PDF (URL or ArrayBuffer) needs OCR to extract usable
+ * text, without running a full parse.
+ * @param pdfURLOrBuffer - URL to a PDF file or buffer from fs.readFile
+ * @category Extract
+ */
+export declare function detectPdfNeedsOcr(pdfURLOrBuffer: any, options?: DetectPdfNeedsOcrOptions): Promise<PdfOcrAssessment>;

package/dist/liteparse-to-html.d.ts

@@ -0,0 +1,31 @@
+import { LiteParseConfig } from '@llamaindex/liteparse';
+export interface LiteParseHTMLOptions {
+    addPageNumbers?: boolean;
+    addCitation?: boolean;
+    liteParseOptions?: Partial<LiteParseConfig>;
+}
+/**
+ * Converts a PDF (URL or ArrayBuffer) into HTML using LiteParse's spatial text
+ * extraction, mirroring the return shape of `convertPDFToHTML`.
+ * @param pdfURLOrBuffer - URL to a PDF file or buffer from fs.readFile
+ * @param options.addPageNumbers default=false - Adds `[n]` markers at each page boundary
+ * @param options.addCitation default=true - Populates `title`/`author` from PDF metadata
+ * @param options.liteParseOptions - Passed through to the `LiteParse` constructor;
+ *   defaults to `{ ocrEnabled: false, ocrFailureFatal: false }` (use detectPdfNeedsOcr
+ *   to decide when a document is worth re-parsing with `ocrEnabled: true`)
+ * @returns `{ html, title, author, format: "pdf" }`, or `{ error }` on failure
+ * @category Extract
+ */
+export declare function convertPDFToHTMLWithLiteParse(pdfURLOrBuffer: any, options?: LiteParseHTMLOptions): Promise<{
+    error: any;
+    author?: undefined;
+    title?: undefined;
+    html?: undefined;
+    format?: undefined;
+} | {
+    author: string;
+    title: string;
+    html: any;
+    format: string;
+    error?: undefined;
+}>;

package/dist/pdf-to-html.cjs.js

@@ -1 +1 @@
-Object.defineProperty(exports,Symbol.toStringTag,{value:"Module"});function t(t){for(;32===t.charCodeAt(0);)t=t.substring(1,t.length);return t}function e(t){for(;32===t.charCodeAt(t.length-1);)t=t.substring(0,t.length-1);return t}function s(t){return t>=48&&t<=57}function n(t){for(var e=0;e<t.length;e++)if(!s(t.charCodeAt(e)))return!1;return!0}function o(t,e){const s=e.charCodeAt(0);for(var n=0;n<t.length;n++)if(t.charCodeAt(n)!==s)return!1;return!0}function r(t){return function(t){const e=[];for(var s=0;s<t.length;s++)e.push(t.charCodeAt(s));return e}(t=t.toUpperCase()).filter(t=>32!==t&&9!==t&&46!==t)}function a(t){if(t.length>1)return!1;const e=t.charAt(0);return"-"===e||"•"===e||"–"===e}function i(t){return/^[\s]*[-•–][\s].*$/g.test(t)}function h(t,e){const s=new Set(t.toUpperCase().split(" ")),n=new Set(e.toUpperCase().split(" "));return new Set([...s].filter(t=>n.has(t))).size/Math.max(s.size,n.size)}var l=(t,e,s)=>Math.floor(t/e*s),c=(s,o,r)=>{for(const{str:a}of s){const s=e(t(a));n(s)&&(o[r]||(o[r]=[]),o[r].push(Number(s)))}return o},m=(t,e,s)=>{const n=l(1,6,s.length),o=l(5,6,s.length),r=c(s.slice(0,n),t,e);return c(s.slice(o),r,e)},u=t=>{let e=0;const s=Object.keys(t);if(0!==s.length&&1!==s.length)for(let n=0;n<s.length-1;n++){const o=t[s[n]],r=t[s[n+1]],a=e;for(let t=0;t<o.length&&e<2;t++)for(let a=0;a<r.length&&e<2;a++){const i=Number(s[n+1])-Number(s[n]);(o[t]+1===r[a]||i>1&&o[t]+i===r[a])&&e++}let i=n>0?Object.entries(t)[n-1]:Object.entries(t)[n];if(a===e)e=0,i=Object.entries(t)[n];else if(e>=2)return{pageIndex:Number(i[0]),pageNum:i[1][0]}}},f=(t,e)=>{const s={items:[...t.items]},n=l(1,6,s.items.length),o=l(5,6,s.items.length);return s.items=s.items.filter((t,r)=>{const a=r>0&&r<n,i=r>o&&r<s.items.length;return a||i?Number(t.str)!==Number(e):t}),s},p=class t{type;annotation;parsedElements;constructor(e){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");this.type=e.type??null,this.annotation=e.annotation??null,this.parsedElements=e.parsedElements??null}},d=class extends p{x;y;width;height;text;font;lineFormat;unopenedFormat;unclosedFormat;constructor(t){super(t),this.x=t.x,this.y=t.y,this.width=t.width,this.height=t.height,this.text=t.text,this.font=t.font,this.lineFormat=t.lineFormat??null,this.unopenedFormat=t.unopenedFormat??null,this.unclosedFormat=t.unclosedFormat??null}},g=class{index;items;constructor(t){this.index=t.index,this.items=t.items||[]}},y=class t{name;itemType;constructor(e,s){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");if(this.transform===t.prototype.transform)throw new TypeError("Please implement abstract method 'transform()'.");this.name=e,this.itemType=s}transform(t){throw new TypeError("Do not call abstract method foo from child.")}completeTransform(t){return t.messages=[],t}},w=class{category;color;constructor(t){this.category=t.category,this.color=t.color}},x=new w({category:"Added",color:"green"}),E=new w({category:"Removed",color:"red"}),T=new w({category:"Unchanged",color:"brown"}),v=new w({category:"Detected",color:"green"}),b=new w({category:"Modified",color:"green"}),L=class t extends y{constructor(e){if(super(e,d.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==E),t.items.forEach(t=>t.annotation=null)}),t}},I=class{pages;globals;messages;constructor(t){this.pages=t.pages||[],this.globals=t.globals||{},this.messages=t.messages||[]}},O=class{string;type;format;constructor(t){this.string=t.string,this.type=t.type??null,this.format=t.format??null}},S=class extends p{x;y;width;height;words;constructor(t){super(t),this.x=t.x??0,this.y=t.y??0,this.width=t.width??0,this.height=t.height??0,this.words=t.words||[],t.text&&!t.words&&(this.words=t.text.split(" ").filter(t=>t.trim().length>0).map(t=>new O({string:t})))}text(){return this.wordStrings().join(" ")}wordStrings(){return this.words.map(t=>t.string)}},k=class t{results;stash;constructor(){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");this.results=[],this.stash=[]}consumeAll(t){t.forEach(t=>this.consume(t))}consume(t){this.shouldStash(t)?(this.matchesStash(t)||this.flushStash(),this.pushOnStash(t)):(this.stash.length>0&&this.flushStash(),this.results.push(t))}pushOnStash(t){this.onPushOnStash(t),this.stash.push(t)}complete(){return this.stash.length>0&&this.flushStash(),this.results}matchesStash(t){if(0===this.stash.length)return!0;const e=this.stash[this.stash.length-1];return this.doMatchesStash(e,t)}flushStash(){this.stash.length>0&&(this.doFlushStash(this.stash,this.results),this.stash=[])}onPushOnStash(t){}shouldStash(t){throw new TypeError(" Do not call abstract method foo from child."+t)}doMatchesStash(t,e){throw new TypeError(" Do not call abstract method foo from child."+t+e)}doFlushStash(t,e){throw new TypeError(" Do not call abstract method foo from child."+t+e)}},C=class{footnoteLinks;footnotes;containLinks;formattedWords;constructor(t){this.footnoteLinks=t.footnoteLinks||[],this.footnotes=t.footnotes||[],this.containLinks=t.containLinks??!1,this.formattedWords=t.formattedWords??0}add(t){this.footnoteLinks=this.footnoteLinks.concat(t.footnoteLinks),this.footnotes=this.footnotes.concat(t.footnotes),this.containLinks=this.containLinks||t.containLinks,this.formattedWords+=t.formattedWords}};function F(t){t.sort((t,e)=>t.x-e.x)}var N={BOLD:{name:"BOLD",startSymbol:"<strong>",endSymbol:"</strong>"},OBLIQUE:{name:"OBLIQUE",startSymbol:"<em>",endSymbol:"</em>"},BOLD_OBLIQUE:{name:"BOLD_OBLIQUE",startSymbol:"<strong><em>",endSymbol:"</em></strong>"}},D={LINK:{name:"LINK",toText:t=>`<a href="${t}">${t}</a>`},FOOTNOTE_LINK:{name:"FOOTNOTE_LINK",attachWithoutWhitespace:!0,plainTextFormat:!0,toText:t=>`<sup><a href="#${t}">${t}</a></sup>`},FOOTNOTE:{name:"FOOTNOTE",toText:t=>`<p id="${t}">^${t}</p>`}},H=class{fontToFormats;constructor(t){this.fontToFormats=t}compact(t){F(t);const e=new M(this.fontToFormats);e.consumeAll(t.map(t=>new d({...t})));const s=e.complete();var n=0,o=0;return t.forEach(t=>{n=Math.max(n,t.height),o+=t.width}),new S({x:t[0].x,y:t[0].y,height:n,width:o,words:s,parsedElements:new C({footnoteLinks:e.footnoteLinks,footnotes:e.footnotes,containLinks:e.containLinks,formattedWords:e.formattedWords})})}},M=class extends k{fontToFormats;footnoteLinks;footnotes;formattedWords;containLinks;stashedNumber;firstY;currentItem;constructor(t){super(),this.fontToFormats=t,this.footnoteLinks=[],this.footnotes=[],this.formattedWords=0,this.containLinks=!1,this.stashedNumber=!1,this.currentItem=null}shouldStash(t){return this.firstY||(this.firstY=t.y),this.currentItem=t,!0}onPushOnStash(t){this.stashedNumber=n(t.text.trim())}doMatchesStash(t,e){if(this.fontToFormats.get(t.font)!==this.fontToFormats.get(e.font))return!1;const s=n(e.text.trim());return this.stashedNumber===s}doFlushStash(t,e){if(this.stashedNumber){const s=t.map(t=>t.text).join("").trim();t[0].y>this.firstY?(e.push(new O({string:`${s}`,type:D.FOOTNOTE_LINK})),this.footnoteLinks.push(parseInt(s))):this.currentItem&&this.currentItem.y<t[0].y?(e.push(new O({string:`${s}`,type:D.FOOTNOTE})),this.footnotes.push(s)):this.copyStashItemsAsText(t,e)}else this.copyStashItemsAsText(t,e)}copyStashItemsAsText(t,e){const s=this.fontToFormats.get(t[0].font);e.push(...this.itemsToWords(t,s))}itemsToWords(t,e){const s=(n=t,o="",r=null,n.forEach(t=>{var e=t.text;o.endsWith(" ")||e.startsWith(" ")||(r?t.x-r.x-r.width>5&&(o+=" "):a(t.text)&&(e+=" ")),o+=e,r=t}),o).split(" ");var n,o,r;const i=e?N[e]:null;return s.filter(t=>t.trim().length>0).map(t=>{var e=null;return t.startsWith("http:")?(this.containLinks=!0,e=D.LINK):t.startsWith("www.")&&(this.containLinks=!0,t=`http://${t}`,e=D.LINK),i&&this.formattedWords++,new O({string:t,type:e,format:i})})}};var B=class extends L{fontMap;constructor(t){super("Calculate Global Stats"),this.fontMap=t}transform(t){const e={},s={};var n,o=0;t.pages.forEach(t=>{t.items.forEach(t=>{t.height&&(e[t.height]=e[t.height]?e[t.height]+1:1,s[t.font]=s[t.font]?s[t.font]+1:1,t.height>o&&(o=t.height,n=t.font))})});const r=parseInt(U(e)),a=U(s),i={};t.pages.forEach(t=>{var e;t.items.forEach(t=>{if(t.height===r&&t.text.trim().length>0){if(e&&t.y!==e.y){const s=e.y-t.y;s>0&&(i[s]=i[s]?i[s]+1:1)}e=t}else e=null})});const h=parseInt(U(i)),l=[],c=new Map;this.fontMap?.forEach(function(t,e){l.push(e+" = "+t.name);const s=t.name.toLowerCase();var o=null;e===a?o=null:s.includes("bold")&&(s.includes("oblique")||s.includes("italic"))?o=N.BOLD_OBLIQUE:s.includes("bold")?o=N.BOLD:s.includes("oblique")||s.includes("italic")?o=N.OBLIQUE:s===n&&(o=N.BOLD),o&&c.set(e,o.name)}),l.sort();const m=t.pages.map(t=>({...t,items:t.items.map(t=>({...t}))}));return new I({...t,pages:m,globals:{mostUsedHeight:r,mostUsedFont:a,mostUsedDistance:h,maxHeight:o,maxHeightFont:n,fontToFormats:c},messages:["Items per height: "+JSON.stringify(e),"Items per font: "+JSON.stringify(s),"Items per distance: "+JSON.stringify(i),"Fonts:"+JSON.stringify(l)]})}};function U(t){var e=0,s="";return Object.keys(t).map(n=>{(!s||t[n]>e)&&(e=t[n],s=n)}),s}var A=class t extends y{constructor(e){if(super(e,S.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==E),t.items.forEach(t=>t.annotation=null)}),t}},W=class{mostUsedDistance;constructor(t){this.mostUsedDistance=t.mostUsedDistance||12}group(t){const e=[];var s=[];return t.forEach(t=>{s.length>0&&Math.abs(s[0].y-t.y)>=this.mostUsedDistance/2&&(e.push(s),s=[]),s.push(t)}),e.push(s),e.forEach(t=>{F(t)}),e}},P={H1:{name:"H1",headline:!0,headlineLevel:1,toText:t=>"<h1>"+$(t.items,!0)+"</h1>"},H2:{name:"H2",headline:!0,headlineLevel:2,toText:t=>"<h2>"+$(t.items,!0)+"</h2>"},H3:{name:"H3",headline:!0,headlineLevel:3,toText:t=>"<h3>"+$(t.items,!0)+"</h3>"},H4:{name:"H4",headline:!0,headlineLevel:4,toText:t=>"<h4>"+$(t.items,!0)+"</h4>"},H5:{name:"H5",headline:!0,headlineLevel:5,toText:t=>"<h5>"+$(t.items,!0)+"</h5>"},H6:{name:"H6",headline:!0,headlineLevel:6,toText:t=>"<h6>"+$(t.items,!0)+"</h6>"},TOC:{name:"TOC",mergeToBlock:!0,toText:t=>$(t.items,!0)},FOOTNOTES:{name:"FOOTNOTES",mergeToBlock:!0,mergeFollowingNonTypedItems:!0,toText:t=>"<p>"+$(t.items,!1)+"</p>"},CODE:{name:"CODE",mergeToBlock:!0,toText:t=>"<code>"+$(t.items,!0)+"</code>"},LIST:{name:"LIST",mergeToBlock:!1,mergeFollowingNonTypedItemsWithSmallDistance:!0,toText:t=>"<ul>\n"+$(t.items,!1)+"</ul>"},PARAGRAPH:{name:"PARAGRAPH",toText:t=>"<p>"+$(t.items,!1)+"</p>"}};function $(t,e){var s,n="";const o=()=>{n+=s.endSymbol,s=null};return t.forEach((r,a)=>{var i;r&&(r.words.forEach((t,r)=>{const a=t.type,i=t.format;var h;!s||i&&i===s||o(),r>0&&(!a||!a.attachWithoutWhitespace)&&(1!==(h=t.string).length||"."!==h[0]&&"!"!==h[0]&&"?"!==h[0])&&(n+=" "),!i||s||e||(n+=(s=i).startSymbol),!a||e&&!a.plainTextFormat?n+=t.string:n+=a.toText(t.string)}),!s||a!==t.length-1&&(0===(i=t[a+1]).words.length?null:i.words[0].format)===s||o(),n+="\n")}),n}P.enumValueOf=function(t){return P[t]},P.isHeadline=function(t){return t&&2===t.name.length&&"H"===t.name[0]},P.blockToText=function(t){return t.type?t.type.toText(t):$(t.items,!1)},P.headlineByLevel=function(t){return 1===t?P.H1:2===t?P.H2:3===t?P.H3:4===t?P.H4:5===t?P.H5:(t>6&&console.warn("Unsupported headline level: "+t+" (supported are 1-6), defaulting to level 6"),P.H6)};var j=class extends A{constructor(){super("Compact To Lines")}transform(t){const{mostUsedDistance:e,fontToFormats:s}=t.globals,n=[],o=[];var r=0,a=0;const i=new W({mostUsedDistance:e}),h=new H(s);return t.pages.forEach(t=>{if(t.items.length>0){const e=[];i.group(t.items).forEach(s=>{const i=h.compact(s);if(s.length>1&&(i.annotation=x,s.forEach(t=>{t.annotation=E,e.push(new S({...t}))})),0===i.words.length&&(i.annotation=E),e.push(i),i.parsedElements.formattedWords&&(a+=i.parsedElements.formattedWords),i.parsedElements.containLinks&&r++,i.parsedElements.footnoteLinks.length>0){const e=i.parsedElements.footnoteLinks.map(e=>({footnoteLink:e,page:t.index+1}));o.push.apply(o,e)}if(i.parsedElements.footnotes.length>0){i.type=P.FOOTNOTES;const e=i.parsedElements.footnotes.map(e=>({footnote:e,page:t.index+1}));n.push.apply(n,e)}}),t.items=e}}),new I({...t,messages:["Detected "+a+" formatted words","Found "+r+" links","Detected "+o.length+" footnotes links","Detected "+n.length+" footnotes"]})}};function q(t){var e=0;if(0===t.trim().length)return e;for(var n=0;n<t.length;n++){const o=t.charCodeAt(n);s(o)||32===o||160===o||(e=(e<<5)-e+o,e|=0)}return e}var Y=class extends A{constructor(){super("Remove Repetitive Elements")}transform(t){const e=[],s={},n={};t.pages.forEach(t=>{const o=t.items.reduce((t,e)=>(e.y<t.minY?(t.minElements=[e],t.minY=e.y):e.y===t.minY&&t.minElements.push(e),e.y>t.maxY?(t.maxElements=[e],t.maxY=e.y):e.y===t.maxY&&t.maxElements.push(e),t),{minY:999,maxY:0,minElements:[],maxElements:[]}),r=q(o.minElements.reduce((t,e)=>t+e.text().toUpperCase(),"")),a=q(o.maxElements.reduce((t,e)=>t+e.text().toUpperCase(),""));e.push({minElements:o.minElements,maxElements:o.maxElements,minLineHash:r,maxLineHash:a}),s[r]=s[r]?s[r]+1:1,n[a]=n[a]?n[a]+1:1});var o=0,r=0;return t.pages.forEach((a,i)=>{s[e[i].minLineHash]>=Math.max(3,2*t.pages.length/3)&&(e[i].minElements.forEach(t=>{t.annotation=E}),r++),n[e[i].maxLineHash]>=Math.max(3,2*t.pages.length/3)&&(e[i].maxElements.forEach(t=>{t.annotation=E}),o++)}),new I({...t,messages:["Removed Header: "+o,"Removed Footers: "+r]})}},R=class extends A{constructor(){super("Vertical to Horizontal Text")}transform(t){var e=0;return t.pages.forEach(t=>{const s=new V;s.consumeAll(t.items),t.items=s.complete(),e+=s.foundVerticals}),new I({...t,messages:["Converted "+e+" verticals"]})}},V=class extends k{foundVerticals;constructor(){super(),this.foundVerticals=0}shouldStash(t){return 1===t.words.length&&1===t.words[0].string.length}doMatchesStash(t,e){return t.y-e.y>5&&t.words[0].type===e.words[0].type}doFlushStash(t,e){if(t.length>5){var s=[],n=999,o=0,r=0,a=0;t.forEach(t=>{t.annotation=E,e.push(t),s.push(t.words[0]),n=Math.min(n,t.x),o=Math.max(o,t.y),r+=t.width,a=Math.max(a,t.height)}),e.push(new S({...t[0],x:n,y:o,width:r,height:a,words:s,annotation:x})),this.foundVerticals++}else e.push(...t)}},K=class{headlineCharCodes;stackedLineItems;stackedChars;constructor(t){this.headlineCharCodes=r(t.headline),this.stackedLineItems=[],this.stackedChars=0}consume(t){const e=r(t.text());if(this.matchAll(e)){if(this.stackedLineItems.push(t),this.stackedChars+=e.length,this.stackedChars===this.headlineCharCodes.length)return this.stackedLineItems}else this.stackedChars>0&&(this.stackedChars=0,this.stackedLineItems=[],this.consume(t));return null}matchAll(t){for(var e=0;e<t.length;e++){const s=this.headlineCharCodes[this.stackedChars+e];if(t[e]!==s)return!1}return!0}},J=class extends A{constructor(){super("Detect TOC")}transform(t){const e=[],r=Math.min(20,t.pages.length),a=new z;var i,l,c=[];t.pages.slice(0,r).forEach(t=>{var r=0;const h=new Set,m=[];var u,f;if(t.items.filter(t=>t.words.includes(t=>o(t.string,"."))).forEach(t=>{var e=t.words.filter(t=>!o(t.string,"."));const a=[];for(;e.length>0&&n(e[e.length-1].string);){const t=e.pop();a.unshift(t.string)}if(0===a.length&&e.length>0){const t=e[e.length-1];for(;s(t.string.charCodeAt(t.string.length-1));)a.unshift(t.string.charAt(t.string.length-1)),t.string=t.string.substring(0,t.string.length-1)}var i=a.length>0;i?(i=!0,u&&(e.push(...u),u=null),m.push(new G({pageNumber:parseInt(a.join("")),lineItem:new S({...t,words:e})})),r++):l?(u&&h.add(f),u=e,f=t):l=t}),100*r/t.items.length>75){e.push(t.index+1),i=t,a.levelPageItems(m),c.push(...m);const s=[];t.items.forEach(t=>{h.has(t)||(t.annotation=E),s.push(t),t===l&&s.push(new S({...t,type:P.H2,annotation:x}))}),t.items=s}else l=null});var m=c.length;const u=[],f=[],p={};if(e.length>0){c.forEach(t=>{i.items.push(new S({words:[new O({string:" ".repeat(3*t.level)+"-"})].concat(t.lineItem.words),type:P.TOC,annotation:x}))});const e=function(t,e){for(var s of e){const e=Q(t,s.lineItem.text());if(e)return e.index-s.pageNumber}return null}(t.pages.filter(t=>t.index>i.index),c);c.forEach(s=>{var n,o=t.pages[s.pageNumber+e];o&&((n=_(o,s.lineItem.text()))||(o=t.pages[s.pageNumber+e+1])&&(n=_(o,s.lineItem.text()))),n?function(t,e,s,n){s.headlineItems.forEach(t=>t.annotation=E);const o=P.headlineByLevel(e.level+2),r=s.headlineItems.reduce((t,e)=>Math.max(t,e.height),0);t.items.splice(s.lineIndex+1,0,new S({...s.headlineItems[0],words:e.lineItem.words,height:r,type:o,annotation:x}));var a=n[o.name];a?(a.min=Math.min(a.min,r),a.max=Math.max(a.max,r)):(a={min:r,max:r},n[o.name]=a)}(o,s,n,p):u.push(s)});var d=i.index+2,g=[];const s=e=>{g.length>0&&(g.forEach(s=>{const n=P.headlineByLevel(s.level+2),o=p[n.name];if(o){const[r,a]=function(t,e,s,n,o){const r=e.lineItem.text().toUpperCase();for(var a=n;a<=o;a++){const e=t[a-1];if(e){const t=e.items.findIndex(t=>!t.type&&!t.annotation&&t.height>=s.min&&t.height<=s.max&&h(r,t.text())>=.5);if(t>-1)return[a-1,t]}}return[-1,-1]}(t.pages,s,o,d,e);if(a>-1){const e=t.pages[r];e.items[a].annotation=E,e.items.splice(a+1,0,new S({...s.lineItem,type:n,annotation:x})),f.push(s)}}}),g=[])};u.length>0&&(c.forEach(t=>{u.includes(t)?g.push(t):(s(t.pageNumber),d=t.pageNumber)}),g.length>0&&s(t.pages.length))}const y=[];return y.push("Detected "+e.length+" table of content pages"),e.length>0&&(y.push("TOC headline heights: "+JSON.stringify(p)),y.push("Found TOC headlines: "+(m-u.length+f.length)+"/"+m)),u.length>0&&(y.push("Found TOC headlines (by size): "+f.map(t=>t.lineItem.text())),y.push("Missing TOC headlines: "+u.filter(t=>!f.includes(t)).map(t=>t.lineItem.text()+"=>"+t.pageNumber))),new I({...t,globals:{...t.globals,tocPages:e,headlineTypeToHeightRange:p},messages:y})}};function Q(t,e){for(var s of t)if(_(s,e))return s;return null}function _(t,e){const s=new K({headline:e});var n=0;for(var o of t.items){const t=s.consume(o);if(t)return{lineIndex:n,headlineItems:t};n++}return null}var z=class{levelByMethod;uniqueFonts;constructor(){this.levelByMethod=null,this.uniqueFonts=[]}levelPageItems(t){if(!this.levelByMethod)if(this.calculateUniqueX(t).length>1)this.levelByMethod=this.levelByXDiff;else{const e=this.calculateUniqueFonts(t);e.length>1?(this.uniqueFonts=e,this.levelByMethod=this.levelByFont):this.levelByMethod=this.levelToZero}this.levelByMethod(t)}levelByXDiff(t){const e=this.calculateUniqueX(t);t.forEach(t=>{t.level=e.indexOf(t.lineItem.x)})}levelByFont(t){t.forEach(t=>{t.level=this.uniqueFonts.indexOf(t.lineItem.font)})}levelToZero(t){t.forEach(t=>{t.level=0})}calculateUniqueX(t){var e=t.reduce(function(t,e){return t.indexOf(e.lineItem.x)<0&&t.push(e.lineItem.x),t},[]);return e.sort((t,e)=>t-e),e}calculateUniqueFonts(t){return t.reduce(function(t,e){return t.indexOf(e.lineItem.font)<0&&t.push(e.lineItem.font),t},[])}},G=class{lineItem;pageNumber;level;constructor(t){this.lineItem=t.lineItem,this.pageNumber=t.pageNumber,this.level=0}},X=class extends A{constructor(){super("Detect List Items")}transform(t){var e=0,s=0;return t.pages.forEach(t=>{const n=[];t.items.forEach(t=>{if(n.push(t),!t.type){var o=t.text();if(a(t.words[0].string))if(e++,"-"===t.words[0].string)t.annotation=v,t.type=P.LIST;else{t.annotation=E;const e=t.words.map(t=>new O({...t}));e[0].string="-",n.push(new S({...t,words:e,annotation:x,type:P.LIST}))}else/^[\s]*[\d]*[.][\s].*$/g.test(o)&&(s++,t.annotation=v,t.type=P.LIST)}}),t.items=n}),new I({...t,messages:["Detected "+e+" plain list items.","Detected "+s+" numbered list items."]})}},Z=class extends A{constructor(){super("Detect Headers")}transform(t){const{tocPages:e,headlineTypeToHeightRange:s,mostUsedHeight:n,mostUsedDistance:o,mostUsedFont:r,maxHeight:a}=t.globals,h=e.length>0;var l=0;const c=function(t,e){const s=new Set;return t.forEach(t=>{t.items.forEach(n=>{n.type||n.height!==e||s.add(t)})}),s}(t.pages,a),m=n+(a-n)/4;if(c.forEach(t=>{t.items.forEach(t=>{const e=t.height;!t.type&&e>m&&(t.type=e===a?P.H1:P.H2,t.annotation=v,l++)})}),h)Object.keys(s).forEach(e=>{var o=s[e];o.max>n&&t.pages.forEach(t=>{t.items.forEach(t=>{t.type||t.height!==o.max||(t.annotation=v,t.type=P.enumValueOf(e),l++)})})});else{const e=[];t.pages.forEach(t=>{t.items.forEach(t=>{!t.type&&t.height>n&&!i(t.text())&&(e.includes(t.height)||e.push(t.height))})}),e.sort((t,e)=>e-t),e.forEach((e,s)=>{if(s+2<=6){const n=P.headlineByLevel(2+s);t.pages.forEach(t=>{t.items.forEach(t=>{t.type||t.height!==e||i(t.text())||(l++,t.annotation=v,t.type=n)})})}})}var u=1;if(t.pages.forEach(t=>{t.items.forEach(t=>{t.type&&t.type.headline&&(u=Math.max(u,t.type.headlineLevel))})}),u<6){const e=P.headlineByLevel(u+1);t.pages.forEach(t=>{var s;t.items.forEach(t=>{!t.type&&t.height===n&&t.font!==r&&(!s||s.y<t.y||s.type&&s.type.headline||s.y-t.y>2*o)&&t.text()===t.text().toUpperCase()&&(l++,t.annotation=v,t.type=e),s=t})})}return new I({...t,messages:["Detected "+l+" headlines."]})}};var tt=class extends p{items;constructor(t){super(t),this.items=[],t.items&&t.items.forEach(t=>this.addItem(t))}addItem(t){if(this.type&&t.type&&this.type!==t.type)throw new Error(`Adding item of type ${t.type.name} to block of type ${this.type.name}`);this.type||(this.type=t.type),t.parsedElements&&(this.parsedElements?this.parsedElements.add(t.parsedElements):this.parsedElements=t.parsedElements);const e=new S({...t});e.type=null,this.items.push(e)}},et=class t extends y{constructor(e){if(super(e,tt.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==E),t.items.forEach(t=>t.annotation=null)}),t}},st=class extends et{constructor(){super("Gather Blocks")}transform(t){const{mostUsedDistance:e}=t.globals;var s=0,n=0;return t.pages.map(t=>{n+=t.items.length;const o=[];var r=new tt({});const a=()=>{r.items.length>1&&(r.annotation=v),o.push(r),r=new tt({}),s++};var i=function(t){var e=999;return t.forEach(t=>{e=Math.min(e,t.x)}),999===e?null:e}(t.items);t.items.forEach(t=>{r.items.length>0&&function(t,e,s,n){if(t.type&&t.type.mergeFollowingNonTypedItems&&!e.type)return!1;const o=t.items[t.items.length-1],r=function(t,e,s,n){const o=t.y-e.y;if(o<0-n/2)return!0;var r=n+1;t.x>s&&e.x>s&&(r=n+n/2);return o>r}(o,e,s,n);if(t.type&&t.type.mergeFollowingNonTypedItemsWithSmallDistance&&!e.type&&!r)return!1;if(e.type!==t.type)return!0;return e.type?!e.type.mergeToBlock:r}(r,t,i,e)&&a(),r.addItem(t)}),r.items.length>0&&a(),t.items=o}),new I({...t,messages:["Gathered "+s+" blocks out of "+n+" line items"]})}};var nt=class extends et{constructor(){super("$1")}transform(t){const e=t.globals.mostUsedHeight??0;var s=0;return t.pages.forEach(t=>{var n=function(t){var e=999;return t.forEach(t=>{t.items.forEach(t=>{e=Math.min(e,t.x)})}),999===e?null:e}(t.items);t.items.forEach(t=>{!t.type&&function(t,e,s){if(0===e.length)return!1;if(1===e.length)return e[0].x>t&&e[0].height<=s+1;for(var n of e)if(n.x===t)return!1;return!0}(n,t.items,e)&&(t.annotation=v,t.type=P.CODE,s++)})}),new I({...t,messages:["Detected "+s+" code/quote items."]})}};var ot=class extends et{constructor(){super("Level Lists")}transform(t){var e=0,s=0;return t.pages.forEach(t=>{t.items.filter(t=>t.type===P.LIST).forEach(t=>{var n,o=0;const r={};var a=!1;t.items.forEach(t=>{n?t.x>n?(o++,r[t.x]=o):t.x<n&&(o=r[t.x]):r[t.x]=0,o>0&&(t.words=[new O({string:" ".repeat(3*o)})].concat(t.words),a=!0),n=t.x}),e++,a?(s++,t.annotation=b):t.annotation=T})}),new I({...t,messages:["Modified "+s+" / "+e+" list blocks."]})}},rt=class extends y{constructor(){super("To Text Blocks","TextBlock")}transform(t){return t.pages.forEach(t=>{const e=[];t.items.forEach(t=>{const s=t.type?t.type.name:"Unknown";e.push({category:s,text:P.blockToText(t)})}),t.items=e}),new I({...t})}},at=class extends y{constructor(){super("To HTML","String")}transform(t){return t.pages.forEach(t=>{var e="";t.items.forEach(t=>{let s;s="TOC"===t.category?t.text:t.text.replace(/(\r\n|\n|\r)/gm,"\n"),"LIST"!==t.category&&(s=s.split("- ").join("")),"CODE"===t.category&&(s=s.split("<code>").join("").split("</code>").join("")),e+=`<p>${s}</p>\n\n`}),t.items=[e]}),new I({...t})}};exports.convertPDFToHTML=async function(t,e={}){var{addPageNumbers:s=!1,addCitation:n=!0}=e,o="string"==typeof t?await async function(t,e={}){const s=e.timeout?1e3*e.timeout:1e4,n=new AbortController,o=setTimeout(()=>n.abort(),s);try{const s=await fetch(t,{signal:n.signal});if(clearTimeout(o),!s.ok)throw new Error(`HTTP ${s.status}`);return"arraybuffer"===e.responseType?await s.arrayBuffer():await s.text()}catch(r){throw clearTimeout(o),r}}(t,{responseType:"arraybuffer",timeout:10}):t;let r;try{let{resolvePDFJS:t}=await import("https://cdn.jsdelivr.net/npm/pdfjs-serverless@1.1.0/+esm");const{getDocument:e}=await t();r=await e({data:new Uint8Array(o),useSystemFonts:!0,verbosity:0}).promise}catch(E){return{error:E.message}}const a=[...Array(r.numPages).keys()].map(t=>new g({index:t}));let i,h={};for(let f=1;f<=r.numPages;f++){const t=await r.getPage(f),e=await t.getTextContent();if(!(Object.keys(h).length<10)){i=u(h);break}h=m(h,t.pageNumber-1,e.items)}let l=i?i.pageNum:0;for(let m=1;m<=r.numPages;m++){const t=await r.getPage(m);await t.getOperatorList();const e=t.getViewport({scale:1});let s=await t.getTextContent();i&&t.pageIndex>=i.pageIndex&&(s=f(s,l),l++);const n=s.items.map(t=>{const s=[1,0,0,1,0,0];for(let r=0;r<6;r++)s[r]+=t.transform[r]*e.transform[r%2?3:0],r%2&&(s[r+1]+=t.transform[r]*e.transform[1]);const n=Math.sqrt(s[2]*s[2]+s[3]*s[3]),o=t.height/n;return new d({x:Math.round(t.transform[4]),y:Math.round(t.transform[5]),width:Math.round(t.width),height:Math.round(o<=1?t.height:o),text:t.str,font:t.fontName})});a[t.pageNumber-1].items=n}var c=new I({pages:a});let p;[new B,new j,new Y,new R,new J,new Z,new X,new st,new nt,new ot,new rt,new at].forEach(t=>{p&&(c=p.completeTransform(c)),c=t.transform(c),p=t});var y=c.pages.reduce((t,e,n)=>t+`<p id="page-${n+1}">${s?` [${n+1}] `:""}${e.items.join('</p><p id="page-'+n+'">')}</p>`,"");if(n){var{Author:w,Title:x}=(await r.getMetadata()).info;x=y.slice(0,400).match(/<h[0-9]>(.*?)<\/h[0-9]>/)?.[1]||x}return{author:w,title:x,html:y,format:"pdf"}};
+Object.defineProperty(exports,Symbol.toStringTag,{value:"Module"});function t(t){for(;32===t.charCodeAt(0);)t=t.substring(1,t.length);return t}function e(t){for(;32===t.charCodeAt(t.length-1);)t=t.substring(0,t.length-1);return t}function s(t){return t>=48&&t<=57}function n(t){for(var e=0;e<t.length;e++)if(!s(t.charCodeAt(e)))return!1;return!0}function o(t,e){const s=e.charCodeAt(0);for(var n=0;n<t.length;n++)if(t.charCodeAt(n)!==s)return!1;return!0}function r(t){return function(t){const e=[];for(var s=0;s<t.length;s++)e.push(t.charCodeAt(s));return e}(t=t.toUpperCase()).filter(t=>32!==t&&9!==t&&46!==t)}function a(t){if(t.length>1)return!1;const e=t.charAt(0);return"-"===e||"•"===e||"–"===e}function i(t){return/^[\s]*[-•–][\s].*$/g.test(t)}function h(t,e){const s=new Set(t.toUpperCase().split(" ")),n=new Set(e.toUpperCase().split(" "));return new Set([...s].filter(t=>n.has(t))).size/Math.max(s.size,n.size)}var l=(t,e,s)=>Math.floor(t/e*s),c=(s,o,r)=>{for(const{str:a}of s){const s=e(t(a));n(s)&&(o[r]||(o[r]=[]),o[r].push(Number(s)))}return o},m=(t,e,s)=>{const n=l(1,6,s.length),o=l(5,6,s.length),r=c(s.slice(0,n),t,e);return c(s.slice(o),r,e)},u=t=>{let e=0;const s=Object.keys(t);if(0!==s.length&&1!==s.length)for(let n=0;n<s.length-1;n++){const o=t[s[n]],r=t[s[n+1]],a=e;for(let t=0;t<o.length&&e<2;t++)for(let a=0;a<r.length&&e<2;a++){const i=Number(s[n+1])-Number(s[n]);(o[t]+1===r[a]||i>1&&o[t]+i===r[a])&&e++}let i=n>0?Object.entries(t)[n-1]:Object.entries(t)[n];if(a===e)e=0,i=Object.entries(t)[n];else if(e>=2)return{pageIndex:Number(i[0]),pageNum:i[1][0]}}},p=(t,e)=>{const s={items:[...t.items]},n=l(1,6,s.items.length),o=l(5,6,s.items.length);return s.items=s.items.filter((t,r)=>{const a=r>0&&r<n,i=r>o&&r<s.items.length;return a||i?Number(t.str)!==Number(e):t}),s},f=class t{type;annotation;parsedElements;constructor(e){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");this.type=e.type??null,this.annotation=e.annotation??null,this.parsedElements=e.parsedElements??null}},d=class extends f{x;y;width;height;text;font;lineFormat;unopenedFormat;unclosedFormat;constructor(t){super(t),this.x=t.x,this.y=t.y,this.width=t.width,this.height=t.height,this.text=t.text,this.font=t.font,this.lineFormat=t.lineFormat??null,this.unopenedFormat=t.unopenedFormat??null,this.unclosedFormat=t.unclosedFormat??null}},g=class{index;items;constructor(t){this.index=t.index,this.items=t.items||[]}};async function y(t,e={}){const s=e.timeout?1e3*e.timeout:1e4,n=new AbortController,o=setTimeout(()=>n.abort(),s);try{const s=await fetch(t,{signal:n.signal});if(clearTimeout(o),!s.ok)throw new Error(`HTTP ${s.status}`);return"arraybuffer"===e.responseType?await s.arrayBuffer():await s.text()}catch(r){throw clearTimeout(o),r}}var w=class t{name;itemType;constructor(e,s){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");if(this.transform===t.prototype.transform)throw new TypeError("Please implement abstract method 'transform()'.");this.name=e,this.itemType=s}transform(t){throw new TypeError("Do not call abstract method foo from child.")}completeTransform(t){return t.messages=[],t}},x=class{category;color;constructor(t){this.category=t.category,this.color=t.color}},E=new x({category:"Added",color:"green"}),T=new x({category:"Removed",color:"red"}),v=new x({category:"Unchanged",color:"brown"}),b=new x({category:"Detected",color:"green"}),L=new x({category:"Modified",color:"green"}),O=class t extends w{constructor(e){if(super(e,d.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==T),t.items.forEach(t=>t.annotation=null)}),t}},I=class{pages;globals;messages;constructor(t){this.pages=t.pages||[],this.globals=t.globals||{},this.messages=t.messages||[]}},S=class{string;type;format;constructor(t){this.string=t.string,this.type=t.type??null,this.format=t.format??null}},k=class extends f{x;y;width;height;words;constructor(t){super(t),this.x=t.x??0,this.y=t.y??0,this.width=t.width??0,this.height=t.height??0,this.words=t.words||[],t.text&&!t.words&&(this.words=t.text.split(" ").filter(t=>t.trim().length>0).map(t=>new S({string:t})))}text(){return this.wordStrings().join(" ")}wordStrings(){return this.words.map(t=>t.string)}},F=class t{results;stash;constructor(){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");this.results=[],this.stash=[]}consumeAll(t){t.forEach(t=>this.consume(t))}consume(t){this.shouldStash(t)?(this.matchesStash(t)||this.flushStash(),this.pushOnStash(t)):(this.stash.length>0&&this.flushStash(),this.results.push(t))}pushOnStash(t){this.onPushOnStash(t),this.stash.push(t)}complete(){return this.stash.length>0&&this.flushStash(),this.results}matchesStash(t){if(0===this.stash.length)return!0;const e=this.stash[this.stash.length-1];return this.doMatchesStash(e,t)}flushStash(){this.stash.length>0&&(this.doFlushStash(this.stash,this.results),this.stash=[])}onPushOnStash(t){}shouldStash(t){throw new TypeError(" Do not call abstract method foo from child."+t)}doMatchesStash(t,e){throw new TypeError(" Do not call abstract method foo from child."+t+e)}doFlushStash(t,e){throw new TypeError(" Do not call abstract method foo from child."+t+e)}},C=class{footnoteLinks;footnotes;containLinks;formattedWords;constructor(t){this.footnoteLinks=t.footnoteLinks||[],this.footnotes=t.footnotes||[],this.containLinks=t.containLinks??!1,this.formattedWords=t.formattedWords??0}add(t){this.footnoteLinks=this.footnoteLinks.concat(t.footnoteLinks),this.footnotes=this.footnotes.concat(t.footnotes),this.containLinks=this.containLinks||t.containLinks,this.formattedWords+=t.formattedWords}};function N(t){t.sort((t,e)=>t.x-e.x)}var M={BOLD:{name:"BOLD",startSymbol:"<strong>",endSymbol:"</strong>"},OBLIQUE:{name:"OBLIQUE",startSymbol:"<em>",endSymbol:"</em>"},BOLD_OBLIQUE:{name:"BOLD_OBLIQUE",startSymbol:"<strong><em>",endSymbol:"</em></strong>"}},D={LINK:{name:"LINK",toText:t=>`<a href="${t}">${t}</a>`},FOOTNOTE_LINK:{name:"FOOTNOTE_LINK",attachWithoutWhitespace:!0,plainTextFormat:!0,toText:t=>`<sup><a href="#${t}">${t}</a></sup>`},FOOTNOTE:{name:"FOOTNOTE",toText:t=>`<p id="${t}">^${t}</p>`}},H=class{fontToFormats;constructor(t){this.fontToFormats=t}compact(t){N(t);const e=new B(this.fontToFormats);e.consumeAll(t.map(t=>new d({...t})));const s=e.complete();var n=0,o=0;return t.forEach(t=>{n=Math.max(n,t.height),o+=t.width}),new k({x:t[0].x,y:t[0].y,height:n,width:o,words:s,parsedElements:new C({footnoteLinks:e.footnoteLinks,footnotes:e.footnotes,containLinks:e.containLinks,formattedWords:e.formattedWords})})}},B=class extends F{fontToFormats;footnoteLinks;footnotes;formattedWords;containLinks;stashedNumber;firstY;currentItem;constructor(t){super(),this.fontToFormats=t,this.footnoteLinks=[],this.footnotes=[],this.formattedWords=0,this.containLinks=!1,this.stashedNumber=!1,this.currentItem=null}shouldStash(t){return this.firstY||(this.firstY=t.y),this.currentItem=t,!0}onPushOnStash(t){this.stashedNumber=n(t.text.trim())}doMatchesStash(t,e){if(this.fontToFormats.get(t.font)!==this.fontToFormats.get(e.font))return!1;const s=n(e.text.trim());return this.stashedNumber===s}doFlushStash(t,e){if(this.stashedNumber){const s=t.map(t=>t.text).join("").trim();t[0].y>this.firstY?(e.push(new S({string:`${s}`,type:D.FOOTNOTE_LINK})),this.footnoteLinks.push(parseInt(s))):this.currentItem&&this.currentItem.y<t[0].y?(e.push(new S({string:`${s}`,type:D.FOOTNOTE})),this.footnotes.push(s)):this.copyStashItemsAsText(t,e)}else this.copyStashItemsAsText(t,e)}copyStashItemsAsText(t,e){const s=this.fontToFormats.get(t[0].font);e.push(...this.itemsToWords(t,s))}itemsToWords(t,e){const s=(n=t,o="",r=null,n.forEach(t=>{var e=t.text;o.endsWith(" ")||e.startsWith(" ")||(r?t.x-r.x-r.width>5&&(o+=" "):a(t.text)&&(e+=" ")),o+=e,r=t}),o).split(" ");var n,o,r;const i=e?M[e]:null;return s.filter(t=>t.trim().length>0).map(t=>{var e=null;return t.startsWith("http:")?(this.containLinks=!0,e=D.LINK):t.startsWith("www.")&&(this.containLinks=!0,t=`http://${t}`,e=D.LINK),i&&this.formattedWords++,new S({string:t,type:e,format:i})})}};var U=class extends O{fontMap;constructor(t){super("Calculate Global Stats"),this.fontMap=t}transform(t){const e={},s={};var n,o=0;t.pages.forEach(t=>{t.items.forEach(t=>{t.height&&(e[t.height]=e[t.height]?e[t.height]+1:1,s[t.font]=s[t.font]?s[t.font]+1:1,t.height>o&&(o=t.height,n=t.font))})});const r=parseInt(A(e)),a=A(s),i={};t.pages.forEach(t=>{var e;t.items.forEach(t=>{if(t.height===r&&t.text.trim().length>0){if(e&&t.y!==e.y){const s=e.y-t.y;s>0&&(i[s]=i[s]?i[s]+1:1)}e=t}else e=null})});const h=parseInt(A(i)),l=[],c=new Map;this.fontMap?.forEach(function(t,e){l.push(e+" = "+t.name);const s=t.name.toLowerCase();var o=null;e===a?o=null:s.includes("bold")&&(s.includes("oblique")||s.includes("italic"))?o=M.BOLD_OBLIQUE:s.includes("bold")?o=M.BOLD:s.includes("oblique")||s.includes("italic")?o=M.OBLIQUE:s===n&&(o=M.BOLD),o&&c.set(e,o.name)}),l.sort();const m=t.pages.map(t=>({...t,items:t.items.map(t=>({...t}))}));return new I({...t,pages:m,globals:{mostUsedHeight:r,mostUsedFont:a,mostUsedDistance:h,maxHeight:o,maxHeightFont:n,fontToFormats:c},messages:["Items per height: "+JSON.stringify(e),"Items per font: "+JSON.stringify(s),"Items per distance: "+JSON.stringify(i),"Fonts:"+JSON.stringify(l)]})}};function A(t){var e=0,s="";return Object.keys(t).map(n=>{(!s||t[n]>e)&&(e=t[n],s=n)}),s}var P=class t extends w{constructor(e){if(super(e,k.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==T),t.items.forEach(t=>t.annotation=null)}),t}},W=class{mostUsedDistance;constructor(t){this.mostUsedDistance=t.mostUsedDistance||12}group(t){const e=[];var s=[];return t.forEach(t=>{s.length>0&&Math.abs(s[0].y-t.y)>=this.mostUsedDistance/2&&(e.push(s),s=[]),s.push(t)}),e.push(s),e.forEach(t=>{N(t)}),e}},$={H1:{name:"H1",headline:!0,headlineLevel:1,toText:t=>"<h1>"+j(t.items,!0)+"</h1>"},H2:{name:"H2",headline:!0,headlineLevel:2,toText:t=>"<h2>"+j(t.items,!0)+"</h2>"},H3:{name:"H3",headline:!0,headlineLevel:3,toText:t=>"<h3>"+j(t.items,!0)+"</h3>"},H4:{name:"H4",headline:!0,headlineLevel:4,toText:t=>"<h4>"+j(t.items,!0)+"</h4>"},H5:{name:"H5",headline:!0,headlineLevel:5,toText:t=>"<h5>"+j(t.items,!0)+"</h5>"},H6:{name:"H6",headline:!0,headlineLevel:6,toText:t=>"<h6>"+j(t.items,!0)+"</h6>"},TOC:{name:"TOC",mergeToBlock:!0,toText:t=>j(t.items,!0)},FOOTNOTES:{name:"FOOTNOTES",mergeToBlock:!0,mergeFollowingNonTypedItems:!0,toText:t=>"<p>"+j(t.items,!1)+"</p>"},CODE:{name:"CODE",mergeToBlock:!0,toText:t=>"<code>"+j(t.items,!0)+"</code>"},LIST:{name:"LIST",mergeToBlock:!1,mergeFollowingNonTypedItemsWithSmallDistance:!0,toText:t=>"<ul>\n"+j(t.items,!1)+"</ul>"},PARAGRAPH:{name:"PARAGRAPH",toText:t=>"<p>"+j(t.items,!1)+"</p>"}};function j(t,e){var s,n="";const o=()=>{n+=s.endSymbol,s=null};return t.forEach((r,a)=>{var i;r&&(r.words.forEach((t,r)=>{const a=t.type,i=t.format;var h;!s||i&&i===s||o(),r>0&&(!a||!a.attachWithoutWhitespace)&&(1!==(h=t.string).length||"."!==h[0]&&"!"!==h[0]&&"?"!==h[0])&&(n+=" "),!i||s||e||(n+=(s=i).startSymbol),!a||e&&!a.plainTextFormat?n+=t.string:n+=a.toText(t.string)}),!s||a!==t.length-1&&(0===(i=t[a+1]).words.length?null:i.words[0].format)===s||o(),n+="\n")}),n}$.enumValueOf=function(t){return $[t]},$.isHeadline=function(t){return t&&2===t.name.length&&"H"===t.name[0]},$.blockToText=function(t){return t.type?t.type.toText(t):j(t.items,!1)},$.headlineByLevel=function(t){return 1===t?$.H1:2===t?$.H2:3===t?$.H3:4===t?$.H4:5===t?$.H5:(t>6&&console.warn("Unsupported headline level: "+t+" (supported are 1-6), defaulting to level 6"),$.H6)};var q=class extends P{constructor(){super("Compact To Lines")}transform(t){const{mostUsedDistance:e,fontToFormats:s}=t.globals,n=[],o=[];var r=0,a=0;const i=new W({mostUsedDistance:e}),h=new H(s);return t.pages.forEach(t=>{if(t.items.length>0){const e=[];i.group(t.items).forEach(s=>{const i=h.compact(s);if(s.length>1&&(i.annotation=E,s.forEach(t=>{t.annotation=T,e.push(new k({...t}))})),0===i.words.length&&(i.annotation=T),e.push(i),i.parsedElements.formattedWords&&(a+=i.parsedElements.formattedWords),i.parsedElements.containLinks&&r++,i.parsedElements.footnoteLinks.length>0){const e=i.parsedElements.footnoteLinks.map(e=>({footnoteLink:e,page:t.index+1}));o.push.apply(o,e)}if(i.parsedElements.footnotes.length>0){i.type=$.FOOTNOTES;const e=i.parsedElements.footnotes.map(e=>({footnote:e,page:t.index+1}));n.push.apply(n,e)}}),t.items=e}}),new I({...t,messages:["Detected "+a+" formatted words","Found "+r+" links","Detected "+o.length+" footnotes links","Detected "+n.length+" footnotes"]})}};function Y(t){var e=0;if(0===t.trim().length)return e;for(var n=0;n<t.length;n++){const o=t.charCodeAt(n);s(o)||32===o||160===o||(e=(e<<5)-e+o,e|=0)}return e}var R=class extends P{constructor(){super("Remove Repetitive Elements")}transform(t){const e=[],s={},n={};t.pages.forEach(t=>{const o=t.items.reduce((t,e)=>(e.y<t.minY?(t.minElements=[e],t.minY=e.y):e.y===t.minY&&t.minElements.push(e),e.y>t.maxY?(t.maxElements=[e],t.maxY=e.y):e.y===t.maxY&&t.maxElements.push(e),t),{minY:999,maxY:0,minElements:[],maxElements:[]}),r=Y(o.minElements.reduce((t,e)=>t+e.text().toUpperCase(),"")),a=Y(o.maxElements.reduce((t,e)=>t+e.text().toUpperCase(),""));e.push({minElements:o.minElements,maxElements:o.maxElements,minLineHash:r,maxLineHash:a}),s[r]=s[r]?s[r]+1:1,n[a]=n[a]?n[a]+1:1});var o=0,r=0;return t.pages.forEach((a,i)=>{s[e[i].minLineHash]>=Math.max(3,2*t.pages.length/3)&&(e[i].minElements.forEach(t=>{t.annotation=T}),r++),n[e[i].maxLineHash]>=Math.max(3,2*t.pages.length/3)&&(e[i].maxElements.forEach(t=>{t.annotation=T}),o++)}),new I({...t,messages:["Removed Header: "+o,"Removed Footers: "+r]})}},V=class extends P{constructor(){super("Vertical to Horizontal Text")}transform(t){var e=0;return t.pages.forEach(t=>{const s=new K;s.consumeAll(t.items),t.items=s.complete(),e+=s.foundVerticals}),new I({...t,messages:["Converted "+e+" verticals"]})}},K=class extends F{foundVerticals;constructor(){super(),this.foundVerticals=0}shouldStash(t){return 1===t.words.length&&1===t.words[0].string.length}doMatchesStash(t,e){return t.y-e.y>5&&t.words[0].type===e.words[0].type}doFlushStash(t,e){if(t.length>5){var s=[],n=999,o=0,r=0,a=0;t.forEach(t=>{t.annotation=T,e.push(t),s.push(t.words[0]),n=Math.min(n,t.x),o=Math.max(o,t.y),r+=t.width,a=Math.max(a,t.height)}),e.push(new k({...t[0],x:n,y:o,width:r,height:a,words:s,annotation:E})),this.foundVerticals++}else e.push(...t)}},J=class{headlineCharCodes;stackedLineItems;stackedChars;constructor(t){this.headlineCharCodes=r(t.headline),this.stackedLineItems=[],this.stackedChars=0}consume(t){const e=r(t.text());if(this.matchAll(e)){if(this.stackedLineItems.push(t),this.stackedChars+=e.length,this.stackedChars===this.headlineCharCodes.length)return this.stackedLineItems}else this.stackedChars>0&&(this.stackedChars=0,this.stackedLineItems=[],this.consume(t));return null}matchAll(t){for(var e=0;e<t.length;e++){const s=this.headlineCharCodes[this.stackedChars+e];if(t[e]!==s)return!1}return!0}},Q=class extends P{constructor(){super("Detect TOC")}transform(t){const e=[],r=Math.min(20,t.pages.length),a=new G;var i,l,c=[];t.pages.slice(0,r).forEach(t=>{var r=0;const h=new Set,m=[];var u,p;if(t.items.filter(t=>t.words.includes(t=>o(t.string,"."))).forEach(t=>{var e=t.words.filter(t=>!o(t.string,"."));const a=[];for(;e.length>0&&n(e[e.length-1].string);){const t=e.pop();a.unshift(t.string)}if(0===a.length&&e.length>0){const t=e[e.length-1];for(;s(t.string.charCodeAt(t.string.length-1));)a.unshift(t.string.charAt(t.string.length-1)),t.string=t.string.substring(0,t.string.length-1)}var i=a.length>0;i?(i=!0,u&&(e.push(...u),u=null),m.push(new X({pageNumber:parseInt(a.join("")),lineItem:new k({...t,words:e})})),r++):l?(u&&h.add(p),u=e,p=t):l=t}),100*r/t.items.length>75){e.push(t.index+1),i=t,a.levelPageItems(m),c.push(...m);const s=[];t.items.forEach(t=>{h.has(t)||(t.annotation=T),s.push(t),t===l&&s.push(new k({...t,type:$.H2,annotation:E}))}),t.items=s}else l=null});var m=c.length;const u=[],p=[],f={};if(e.length>0){c.forEach(t=>{i.items.push(new k({words:[new S({string:" ".repeat(3*t.level)+"-"})].concat(t.lineItem.words),type:$.TOC,annotation:E}))});const e=function(t,e){for(var s of e){const e=_(t,s.lineItem.text());if(e)return e.index-s.pageNumber}return null}(t.pages.filter(t=>t.index>i.index),c);c.forEach(s=>{var n,o=t.pages[s.pageNumber+e];o&&((n=z(o,s.lineItem.text()))||(o=t.pages[s.pageNumber+e+1])&&(n=z(o,s.lineItem.text()))),n?function(t,e,s,n){s.headlineItems.forEach(t=>t.annotation=T);const o=$.headlineByLevel(e.level+2),r=s.headlineItems.reduce((t,e)=>Math.max(t,e.height),0);t.items.splice(s.lineIndex+1,0,new k({...s.headlineItems[0],words:e.lineItem.words,height:r,type:o,annotation:E}));var a=n[o.name];a?(a.min=Math.min(a.min,r),a.max=Math.max(a.max,r)):(a={min:r,max:r},n[o.name]=a)}(o,s,n,f):u.push(s)});var d=i.index+2,g=[];const s=e=>{g.length>0&&(g.forEach(s=>{const n=$.headlineByLevel(s.level+2),o=f[n.name];if(o){const[r,a]=function(t,e,s,n,o){const r=e.lineItem.text().toUpperCase();for(var a=n;a<=o;a++){const e=t[a-1];if(e){const t=e.items.findIndex(t=>!t.type&&!t.annotation&&t.height>=s.min&&t.height<=s.max&&h(r,t.text())>=.5);if(t>-1)return[a-1,t]}}return[-1,-1]}(t.pages,s,o,d,e);if(a>-1){const e=t.pages[r];e.items[a].annotation=T,e.items.splice(a+1,0,new k({...s.lineItem,type:n,annotation:E})),p.push(s)}}}),g=[])};u.length>0&&(c.forEach(t=>{u.includes(t)?g.push(t):(s(t.pageNumber),d=t.pageNumber)}),g.length>0&&s(t.pages.length))}const y=[];return y.push("Detected "+e.length+" table of content pages"),e.length>0&&(y.push("TOC headline heights: "+JSON.stringify(f)),y.push("Found TOC headlines: "+(m-u.length+p.length)+"/"+m)),u.length>0&&(y.push("Found TOC headlines (by size): "+p.map(t=>t.lineItem.text())),y.push("Missing TOC headlines: "+u.filter(t=>!p.includes(t)).map(t=>t.lineItem.text()+"=>"+t.pageNumber))),new I({...t,globals:{...t.globals,tocPages:e,headlineTypeToHeightRange:f},messages:y})}};function _(t,e){for(var s of t)if(z(s,e))return s;return null}function z(t,e){const s=new J({headline:e});var n=0;for(var o of t.items){const t=s.consume(o);if(t)return{lineIndex:n,headlineItems:t};n++}return null}var G=class{levelByMethod;uniqueFonts;constructor(){this.levelByMethod=null,this.uniqueFonts=[]}levelPageItems(t){if(!this.levelByMethod)if(this.calculateUniqueX(t).length>1)this.levelByMethod=this.levelByXDiff;else{const e=this.calculateUniqueFonts(t);e.length>1?(this.uniqueFonts=e,this.levelByMethod=this.levelByFont):this.levelByMethod=this.levelToZero}this.levelByMethod(t)}levelByXDiff(t){const e=this.calculateUniqueX(t);t.forEach(t=>{t.level=e.indexOf(t.lineItem.x)})}levelByFont(t){t.forEach(t=>{t.level=this.uniqueFonts.indexOf(t.lineItem.font)})}levelToZero(t){t.forEach(t=>{t.level=0})}calculateUniqueX(t){var e=t.reduce(function(t,e){return t.indexOf(e.lineItem.x)<0&&t.push(e.lineItem.x),t},[]);return e.sort((t,e)=>t-e),e}calculateUniqueFonts(t){return t.reduce(function(t,e){return t.indexOf(e.lineItem.font)<0&&t.push(e.lineItem.font),t},[])}},X=class{lineItem;pageNumber;level;constructor(t){this.lineItem=t.lineItem,this.pageNumber=t.pageNumber,this.level=0}},Z=class extends P{constructor(){super("Detect List Items")}transform(t){var e=0,s=0;return t.pages.forEach(t=>{const n=[];t.items.forEach(t=>{if(n.push(t),!t.type){var o=t.text();if(a(t.words[0].string))if(e++,"-"===t.words[0].string)t.annotation=b,t.type=$.LIST;else{t.annotation=T;const e=t.words.map(t=>new S({...t}));e[0].string="-",n.push(new k({...t,words:e,annotation:E,type:$.LIST}))}else/^[\s]*[\d]*[.][\s].*$/g.test(o)&&(s++,t.annotation=b,t.type=$.LIST)}}),t.items=n}),new I({...t,messages:["Detected "+e+" plain list items.","Detected "+s+" numbered list items."]})}},tt=class extends P{constructor(){super("Detect Headers")}transform(t){const{tocPages:e,headlineTypeToHeightRange:s,mostUsedHeight:n,mostUsedDistance:o,mostUsedFont:r,maxHeight:a}=t.globals,h=e.length>0;var l=0;const c=function(t,e){const s=new Set;return t.forEach(t=>{t.items.forEach(n=>{n.type||n.height!==e||s.add(t)})}),s}(t.pages,a),m=n+(a-n)/4;if(c.forEach(t=>{t.items.forEach(t=>{const e=t.height;!t.type&&e>m&&(t.type=e===a?$.H1:$.H2,t.annotation=b,l++)})}),h)Object.keys(s).forEach(e=>{var o=s[e];o.max>n&&t.pages.forEach(t=>{t.items.forEach(t=>{t.type||t.height!==o.max||(t.annotation=b,t.type=$.enumValueOf(e),l++)})})});else{const e=[];t.pages.forEach(t=>{t.items.forEach(t=>{!t.type&&t.height>n&&!i(t.text())&&(e.includes(t.height)||e.push(t.height))})}),e.sort((t,e)=>e-t),e.forEach((e,s)=>{if(s+2<=6){const n=$.headlineByLevel(2+s);t.pages.forEach(t=>{t.items.forEach(t=>{t.type||t.height!==e||i(t.text())||(l++,t.annotation=b,t.type=n)})})}})}var u=1;if(t.pages.forEach(t=>{t.items.forEach(t=>{t.type&&t.type.headline&&(u=Math.max(u,t.type.headlineLevel))})}),u<6){const e=$.headlineByLevel(u+1);t.pages.forEach(t=>{var s;t.items.forEach(t=>{!t.type&&t.height===n&&t.font!==r&&(!s||s.y<t.y||s.type&&s.type.headline||s.y-t.y>2*o)&&t.text()===t.text().toUpperCase()&&(l++,t.annotation=b,t.type=e),s=t})})}return new I({...t,messages:["Detected "+l+" headlines."]})}};var et=class extends f{items;constructor(t){super(t),this.items=[],t.items&&t.items.forEach(t=>this.addItem(t))}addItem(t){if(this.type&&t.type&&this.type!==t.type)throw new Error(`Adding item of type ${t.type.name} to block of type ${this.type.name}`);this.type||(this.type=t.type),t.parsedElements&&(this.parsedElements?this.parsedElements.add(t.parsedElements):this.parsedElements=t.parsedElements);const e=new k({...t});e.type=null,this.items.push(e)}},st=class t extends w{constructor(e){if(super(e,et.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==T),t.items.forEach(t=>t.annotation=null)}),t}},nt=class extends st{constructor(){super("Gather Blocks")}transform(t){const{mostUsedDistance:e}=t.globals;var s=0,n=0;return t.pages.map(t=>{n+=t.items.length;const o=[];var r=new et({});const a=()=>{r.items.length>1&&(r.annotation=b),o.push(r),r=new et({}),s++};var i=function(t){var e=999;return t.forEach(t=>{e=Math.min(e,t.x)}),999===e?null:e}(t.items);t.items.forEach(t=>{r.items.length>0&&function(t,e,s,n){if(t.type&&t.type.mergeFollowingNonTypedItems&&!e.type)return!1;const o=t.items[t.items.length-1],r=function(t,e,s,n){const o=t.y-e.y;if(o<0-n/2)return!0;var r=n+1;t.x>s&&e.x>s&&(r=n+n/2);return o>r}(o,e,s,n);if(t.type&&t.type.mergeFollowingNonTypedItemsWithSmallDistance&&!e.type&&!r)return!1;if(e.type!==t.type)return!0;return e.type?!e.type.mergeToBlock:r}(r,t,i,e)&&a(),r.addItem(t)}),r.items.length>0&&a(),t.items=o}),new I({...t,messages:["Gathered "+s+" blocks out of "+n+" line items"]})}};var ot=class extends st{constructor(){super("$1")}transform(t){const e=t.globals.mostUsedHeight??0;var s=0;return t.pages.forEach(t=>{var n=function(t){var e=999;return t.forEach(t=>{t.items.forEach(t=>{e=Math.min(e,t.x)})}),999===e?null:e}(t.items);t.items.forEach(t=>{!t.type&&function(t,e,s){if(0===e.length)return!1;if(1===e.length)return e[0].x>t&&e[0].height<=s+1;for(var n of e)if(n.x===t)return!1;return!0}(n,t.items,e)&&(t.annotation=b,t.type=$.CODE,s++)})}),new I({...t,messages:["Detected "+s+" code/quote items."]})}};var rt=class extends st{constructor(){super("Level Lists")}transform(t){var e=0,s=0;return t.pages.forEach(t=>{t.items.filter(t=>t.type===$.LIST).forEach(t=>{var n,o=0;const r={};var a=!1;t.items.forEach(t=>{n?t.x>n?(o++,r[t.x]=o):t.x<n&&(o=r[t.x]):r[t.x]=0,o>0&&(t.words=[new S({string:" ".repeat(3*o)})].concat(t.words),a=!0),n=t.x}),e++,a?(s++,t.annotation=L):t.annotation=v})}),new I({...t,messages:["Modified "+s+" / "+e+" list blocks."]})}},at=class extends w{constructor(){super("To Text Blocks","TextBlock")}transform(t){return t.pages.forEach(t=>{const e=[];t.items.forEach(t=>{const s=t.type?t.type.name:"Unknown";e.push({category:s,text:$.blockToText(t)})}),t.items=e}),new I({...t})}},it=class extends w{constructor(){super("To HTML","String")}transform(t){return t.pages.forEach(t=>{var e="";t.items.forEach(t=>{let s;s="TOC"===t.category?t.text:t.text.replace(/(\r\n|\n|\r)/gm,"\n"),"LIST"!==t.category&&(s=s.split("- ").join("")),"CODE"===t.category&&(s=s.split("<code>").join("").split("</code>").join("")),e+=`<p>${s}</p>\n\n`}),t.items=[e]}),new I({...t})}};async function ht(t,e={}){const{addPageNumbers:s=!1,addCitation:n=!0,liteParseOptions:o={}}=e,r="string"==typeof t?await y(t,{responseType:"arraybuffer",timeout:10}):t;let a;try{({LiteParse:a}=await import("@llamaindex/liteparse"))}catch(u){return{error:`method: 'liteparse' requires the optional @llamaindex/liteparse dependency (Node.js only): ${u.message}`}}const i=new a({quiet:!0,ocrEnabled:!1,ocrFailureFatal:!1,...o});let h;try{h=await i.parse(new Uint8Array(r))}catch(u){return{error:u.message}}const l=h.pages.reduce((t,e)=>{const n=e.text.split(/\n{2,}/).filter(t=>t.trim().length>0).map(t=>{return(e=t,e.replace(/&/g,"&amp;").replace(/</g,"&lt;").replace(/>/g,"&gt;")).replace(/\n/g,"<br/>");var e}).join("</p><p>");return t+`<p id="page-${e.pageNum}">${s?` [${e.pageNum}] `:""}${n}</p>`},"");let c,m;return n&&(m=h.creator||h.producer,c=h.pages[0]?.text.split("\n").find(t=>t.trim().length>0)),{author:m,title:c,html:l,format:"pdf"}}exports.convertPDFToHTML=async function(t,e={}){if("liteparse"===e.method){const{method:s,...n}=e;return ht(t,n)}var{addPageNumbers:s=!1,addCitation:n=!0}=e,o="string"==typeof t?await y(t,{responseType:"arraybuffer",timeout:10}):t;let r;try{let{resolvePDFJS:t}=await import("https://cdn.jsdelivr.net/npm/pdfjs-serverless@1.1.0/+esm");const{getDocument:e}=await t();r=await e({data:new Uint8Array(o),useSystemFonts:!0,verbosity:0}).promise}catch(T){return{error:T.message}}const a=[...Array(r.numPages).keys()].map(t=>new g({index:t}));let i,h={};for(let p=1;p<=r.numPages;p++){const t=await r.getPage(p),e=await t.getTextContent();if(!(Object.keys(h).length<10)){i=u(h);break}h=m(h,t.pageNumber-1,e.items)}let l=i?i.pageNum:0;for(let m=1;m<=r.numPages;m++){const t=await r.getPage(m);await t.getOperatorList();const e=t.getViewport({scale:1});let s=await t.getTextContent();i&&t.pageIndex>=i.pageIndex&&(s=p(s,l),l++);const n=s.items.map(t=>{const s=[1,0,0,1,0,0];for(let r=0;r<6;r++)s[r]+=t.transform[r]*e.transform[r%2?3:0],r%2&&(s[r+1]+=t.transform[r]*e.transform[1]);const n=Math.sqrt(s[2]*s[2]+s[3]*s[3]),o=t.height/n;return new d({x:Math.round(t.transform[4]),y:Math.round(t.transform[5]),width:Math.round(t.width),height:Math.round(o<=1?t.height:o),text:t.str,font:t.fontName})});a[t.pageNumber-1].items=n}var c=new I({pages:a});let f;[new U,new q,new R,new V,new Q,new tt,new Z,new nt,new ot,new rt,new at,new it].forEach(t=>{f&&(c=f.completeTransform(c)),c=t.transform(c),f=t});var w=c.pages.reduce((t,e,n)=>t+`<p id="page-${n+1}">${s?` [${n+1}] `:""}${e.items.join('</p><p id="page-'+n+'">')}</p>`,"");if(n){var{Author:x,Title:E}=(await r.getMetadata()).info;E=w.slice(0,400).match(/<h[0-9]>(.*?)<\/h[0-9]>/)?.[1]||E}return{author:x,title:E,html:w,format:"pdf"}},exports.convertPDFToHTMLWithLiteParse=ht,exports.detectPdfNeedsOcr=async function(t,e={}){const s="string"==typeof t?await y(t,{responseType:"arraybuffer",timeout:10}):t,{LiteParse:n}=await import("@llamaindex/liteparse"),o=await new n({quiet:!0,...e.liteParseOptions}).isComplex(new Uint8Array(s));return{needsOcr:o.some(t=>t.needsOcr),pages:o,reasons:[...new Set(o.flatMap(t=>t.reasons))]}};

package/dist/pdf-to-html.d.ts

@@ -1,3 +1,14 @@
+import { convertPDFToHTMLWithLiteParse, LiteParseHTMLOptions } from './liteparse-to-html';
+/**
+ * Which parsing engine {@link convertPDFToHTML} runs.
+ * - `"ts-block-algorithm"` (default) — the pure-TS pipeline in this file: groups
+ *   pdfjs text spans into lines/blocks and renders headings/lists/code blocks.
+ *   Works in Node.js, Cloudflare Workers, and browsers.
+ * - `"liteparse"` — delegates to [LiteParse](https://github.com/run-llama/liteparse),
+ *   a native/OCR-capable parser. Node.js only (ships a napi addon), and requires
+ *   the optional `@llamaindex/liteparse` dependency to be installed.
+ */
+export type ParseMethod = "ts-block-algorithm" | "liteparse";
 /**
  * Extracts formatted text from PDF with parsing of linebreaks ,
  * page headers, footnotes, and section headings. Supports fonts, links, bold,
@@ -12,6 +23,8 @@
  * @param {Object} [options]
  * @param {boolean} options.addPageNumbers default=false - Adds  #  to end of each page
  * @param {boolean} options.removePageHeaders default=true - Removes repeated headers found on each page
+ * @param {ParseMethod} options.method default="ts-block-algorithm" - Parsing engine to use;
+ *   `"liteparse"` delegates to LiteParse (Node.js only, see {@link ParseMethod})
  * @returns {string|Object} HTML formatted text
  * @category Extract
  * @author [vtempest (2025)](https://github.com/vtempest),
@@ -21,16 +34,26 @@
 export declare function convertPDFToHTML(pdfURLOrBuffer: any, options?: {
     addPageNumbers?: boolean;
     addCitation?: boolean;
-}): Promise<{
+    method?: ParseMethod;
+} & Pick<LiteParseHTMLOptions, "liteParseOptions">): Promise<{
     error: any;
     author?: undefined;
     title?: undefined;
     html?: undefined;
     format?: undefined;
+} | {
+    author: string;
+    title: string;
+    html: any;
+    format: string;
+    error?: undefined;
 } | {
     author: any;
     title: any;
     html: string;
     format: string;
-    error?: undefined;
 }>;
+export { convertPDFToHTMLWithLiteParse };
+export type { LiteParseHTMLOptions } from './liteparse-to-html';
+export { detectPdfNeedsOcr } from './detect-needs-ocr';
+export type { DetectPdfNeedsOcrOptions, PdfOcrAssessment, } from './detect-needs-ocr';

package/dist/pdf-to-html.es.js

@@ -1 +1 @@
-function t(t){for(;32===t.charCodeAt(0);)t=t.substring(1,t.length);return t}function e(t){for(;32===t.charCodeAt(t.length-1);)t=t.substring(0,t.length-1);return t}function s(t){return t>=48&&t<=57}function n(t){for(var e=0;e<t.length;e++)if(!s(t.charCodeAt(e)))return!1;return!0}function o(t,e){const s=e.charCodeAt(0);for(var n=0;n<t.length;n++)if(t.charCodeAt(n)!==s)return!1;return!0}function r(t){return function(t){const e=[];for(var s=0;s<t.length;s++)e.push(t.charCodeAt(s));return e}(t=t.toUpperCase()).filter(t=>32!==t&&9!==t&&46!==t)}function a(t){if(t.length>1)return!1;const e=t.charAt(0);return"-"===e||"•"===e||"–"===e}function i(t){return/^[\s]*[-•–][\s].*$/g.test(t)}function h(t,e){const s=new Set(t.toUpperCase().split(" ")),n=new Set(e.toUpperCase().split(" "));return new Set([...s].filter(t=>n.has(t))).size/Math.max(s.size,n.size)}var l=(t,e,s)=>Math.floor(t/e*s),c=(s,o,r)=>{for(const{str:a}of s){const s=e(t(a));n(s)&&(o[r]||(o[r]=[]),o[r].push(Number(s)))}return o},m=(t,e,s)=>{const n=l(1,6,s.length),o=l(5,6,s.length),r=c(s.slice(0,n),t,e);return c(s.slice(o),r,e)},u=t=>{let e=0;const s=Object.keys(t);if(0!==s.length&&1!==s.length)for(let n=0;n<s.length-1;n++){const o=t[s[n]],r=t[s[n+1]],a=e;for(let t=0;t<o.length&&e<2;t++)for(let a=0;a<r.length&&e<2;a++){const i=Number(s[n+1])-Number(s[n]);(o[t]+1===r[a]||i>1&&o[t]+i===r[a])&&e++}let i=n>0?Object.entries(t)[n-1]:Object.entries(t)[n];if(a===e)e=0,i=Object.entries(t)[n];else if(e>=2)return{pageIndex:Number(i[0]),pageNum:i[1][0]}}},f=(t,e)=>{const s={items:[...t.items]},n=l(1,6,s.items.length),o=l(5,6,s.items.length);return s.items=s.items.filter((t,r)=>{const a=r>0&&r<n,i=r>o&&r<s.items.length;return a||i?Number(t.str)!==Number(e):t}),s},p=class t{type;annotation;parsedElements;constructor(e){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");this.type=e.type??null,this.annotation=e.annotation??null,this.parsedElements=e.parsedElements??null}},d=class extends p{x;y;width;height;text;font;lineFormat;unopenedFormat;unclosedFormat;constructor(t){super(t),this.x=t.x,this.y=t.y,this.width=t.width,this.height=t.height,this.text=t.text,this.font=t.font,this.lineFormat=t.lineFormat??null,this.unopenedFormat=t.unopenedFormat??null,this.unclosedFormat=t.unclosedFormat??null}},g=class{index;items;constructor(t){this.index=t.index,this.items=t.items||[]}},y=class t{name;itemType;constructor(e,s){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");if(this.transform===t.prototype.transform)throw new TypeError("Please implement abstract method 'transform()'.");this.name=e,this.itemType=s}transform(t){throw new TypeError("Do not call abstract method foo from child.")}completeTransform(t){return t.messages=[],t}},w=class{category;color;constructor(t){this.category=t.category,this.color=t.color}},x=new w({category:"Added",color:"green"}),E=new w({category:"Removed",color:"red"}),T=new w({category:"Unchanged",color:"brown"}),v=new w({category:"Detected",color:"green"}),b=new w({category:"Modified",color:"green"}),L=class t extends y{constructor(e){if(super(e,d.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==E),t.items.forEach(t=>t.annotation=null)}),t}},I=class{pages;globals;messages;constructor(t){this.pages=t.pages||[],this.globals=t.globals||{},this.messages=t.messages||[]}},O=class{string;type;format;constructor(t){this.string=t.string,this.type=t.type??null,this.format=t.format??null}},k=class extends p{x;y;width;height;words;constructor(t){super(t),this.x=t.x??0,this.y=t.y??0,this.width=t.width??0,this.height=t.height??0,this.words=t.words||[],t.text&&!t.words&&(this.words=t.text.split(" ").filter(t=>t.trim().length>0).map(t=>new O({string:t})))}text(){return this.wordStrings().join(" ")}wordStrings(){return this.words.map(t=>t.string)}},S=class t{results;stash;constructor(){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");this.results=[],this.stash=[]}consumeAll(t){t.forEach(t=>this.consume(t))}consume(t){this.shouldStash(t)?(this.matchesStash(t)||this.flushStash(),this.pushOnStash(t)):(this.stash.length>0&&this.flushStash(),this.results.push(t))}pushOnStash(t){this.onPushOnStash(t),this.stash.push(t)}complete(){return this.stash.length>0&&this.flushStash(),this.results}matchesStash(t){if(0===this.stash.length)return!0;const e=this.stash[this.stash.length-1];return this.doMatchesStash(e,t)}flushStash(){this.stash.length>0&&(this.doFlushStash(this.stash,this.results),this.stash=[])}onPushOnStash(t){}shouldStash(t){throw new TypeError(" Do not call abstract method foo from child."+t)}doMatchesStash(t,e){throw new TypeError(" Do not call abstract method foo from child."+t+e)}doFlushStash(t,e){throw new TypeError(" Do not call abstract method foo from child."+t+e)}},C=class{footnoteLinks;footnotes;containLinks;formattedWords;constructor(t){this.footnoteLinks=t.footnoteLinks||[],this.footnotes=t.footnotes||[],this.containLinks=t.containLinks??!1,this.formattedWords=t.formattedWords??0}add(t){this.footnoteLinks=this.footnoteLinks.concat(t.footnoteLinks),this.footnotes=this.footnotes.concat(t.footnotes),this.containLinks=this.containLinks||t.containLinks,this.formattedWords+=t.formattedWords}};function F(t){t.sort((t,e)=>t.x-e.x)}var N={BOLD:{name:"BOLD",startSymbol:"<strong>",endSymbol:"</strong>"},OBLIQUE:{name:"OBLIQUE",startSymbol:"<em>",endSymbol:"</em>"},BOLD_OBLIQUE:{name:"BOLD_OBLIQUE",startSymbol:"<strong><em>",endSymbol:"</em></strong>"}},D={LINK:{name:"LINK",toText:t=>`<a href="${t}">${t}</a>`},FOOTNOTE_LINK:{name:"FOOTNOTE_LINK",attachWithoutWhitespace:!0,plainTextFormat:!0,toText:t=>`<sup><a href="#${t}">${t}</a></sup>`},FOOTNOTE:{name:"FOOTNOTE",toText:t=>`<p id="${t}">^${t}</p>`}},H=class{fontToFormats;constructor(t){this.fontToFormats=t}compact(t){F(t);const e=new M(this.fontToFormats);e.consumeAll(t.map(t=>new d({...t})));const s=e.complete();var n=0,o=0;return t.forEach(t=>{n=Math.max(n,t.height),o+=t.width}),new k({x:t[0].x,y:t[0].y,height:n,width:o,words:s,parsedElements:new C({footnoteLinks:e.footnoteLinks,footnotes:e.footnotes,containLinks:e.containLinks,formattedWords:e.formattedWords})})}},M=class extends S{fontToFormats;footnoteLinks;footnotes;formattedWords;containLinks;stashedNumber;firstY;currentItem;constructor(t){super(),this.fontToFormats=t,this.footnoteLinks=[],this.footnotes=[],this.formattedWords=0,this.containLinks=!1,this.stashedNumber=!1,this.currentItem=null}shouldStash(t){return this.firstY||(this.firstY=t.y),this.currentItem=t,!0}onPushOnStash(t){this.stashedNumber=n(t.text.trim())}doMatchesStash(t,e){if(this.fontToFormats.get(t.font)!==this.fontToFormats.get(e.font))return!1;const s=n(e.text.trim());return this.stashedNumber===s}doFlushStash(t,e){if(this.stashedNumber){const s=t.map(t=>t.text).join("").trim();t[0].y>this.firstY?(e.push(new O({string:`${s}`,type:D.FOOTNOTE_LINK})),this.footnoteLinks.push(parseInt(s))):this.currentItem&&this.currentItem.y<t[0].y?(e.push(new O({string:`${s}`,type:D.FOOTNOTE})),this.footnotes.push(s)):this.copyStashItemsAsText(t,e)}else this.copyStashItemsAsText(t,e)}copyStashItemsAsText(t,e){const s=this.fontToFormats.get(t[0].font);e.push(...this.itemsToWords(t,s))}itemsToWords(t,e){const s=(n=t,o="",r=null,n.forEach(t=>{var e=t.text;o.endsWith(" ")||e.startsWith(" ")||(r?t.x-r.x-r.width>5&&(o+=" "):a(t.text)&&(e+=" ")),o+=e,r=t}),o).split(" ");var n,o,r;const i=e?N[e]:null;return s.filter(t=>t.trim().length>0).map(t=>{var e=null;return t.startsWith("http:")?(this.containLinks=!0,e=D.LINK):t.startsWith("www.")&&(this.containLinks=!0,t=`http://${t}`,e=D.LINK),i&&this.formattedWords++,new O({string:t,type:e,format:i})})}};var B=class extends L{fontMap;constructor(t){super("Calculate Global Stats"),this.fontMap=t}transform(t){const e={},s={};var n,o=0;t.pages.forEach(t=>{t.items.forEach(t=>{t.height&&(e[t.height]=e[t.height]?e[t.height]+1:1,s[t.font]=s[t.font]?s[t.font]+1:1,t.height>o&&(o=t.height,n=t.font))})});const r=parseInt(U(e)),a=U(s),i={};t.pages.forEach(t=>{var e;t.items.forEach(t=>{if(t.height===r&&t.text.trim().length>0){if(e&&t.y!==e.y){const s=e.y-t.y;s>0&&(i[s]=i[s]?i[s]+1:1)}e=t}else e=null})});const h=parseInt(U(i)),l=[],c=/* @__PURE__ */new Map;this.fontMap?.forEach(function(t,e){l.push(e+" = "+t.name);const s=t.name.toLowerCase();var o=null;e===a?o=null:s.includes("bold")&&(s.includes("oblique")||s.includes("italic"))?o=N.BOLD_OBLIQUE:s.includes("bold")?o=N.BOLD:s.includes("oblique")||s.includes("italic")?o=N.OBLIQUE:s===n&&(o=N.BOLD),o&&c.set(e,o.name)}),l.sort();const m=t.pages.map(t=>({...t,items:t.items.map(t=>({...t}))}));return new I({...t,pages:m,globals:{mostUsedHeight:r,mostUsedFont:a,mostUsedDistance:h,maxHeight:o,maxHeightFont:n,fontToFormats:c},messages:["Items per height: "+JSON.stringify(e),"Items per font: "+JSON.stringify(s),"Items per distance: "+JSON.stringify(i),"Fonts:"+JSON.stringify(l)]})}};function U(t){var e=0,s="";return Object.keys(t).map(n=>{(!s||t[n]>e)&&(e=t[n],s=n)}),s}var A=class t extends y{constructor(e){if(super(e,k.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==E),t.items.forEach(t=>t.annotation=null)}),t}},W=class{mostUsedDistance;constructor(t){this.mostUsedDistance=t.mostUsedDistance||12}group(t){const e=[];var s=[];return t.forEach(t=>{s.length>0&&Math.abs(s[0].y-t.y)>=this.mostUsedDistance/2&&(e.push(s),s=[]),s.push(t)}),e.push(s),e.forEach(t=>{F(t)}),e}},P={H1:{name:"H1",headline:!0,headlineLevel:1,toText:t=>"<h1>"+$(t.items,!0)+"</h1>"},H2:{name:"H2",headline:!0,headlineLevel:2,toText:t=>"<h2>"+$(t.items,!0)+"</h2>"},H3:{name:"H3",headline:!0,headlineLevel:3,toText:t=>"<h3>"+$(t.items,!0)+"</h3>"},H4:{name:"H4",headline:!0,headlineLevel:4,toText:t=>"<h4>"+$(t.items,!0)+"</h4>"},H5:{name:"H5",headline:!0,headlineLevel:5,toText:t=>"<h5>"+$(t.items,!0)+"</h5>"},H6:{name:"H6",headline:!0,headlineLevel:6,toText:t=>"<h6>"+$(t.items,!0)+"</h6>"},TOC:{name:"TOC",mergeToBlock:!0,toText:t=>$(t.items,!0)},FOOTNOTES:{name:"FOOTNOTES",mergeToBlock:!0,mergeFollowingNonTypedItems:!0,toText:t=>"<p>"+$(t.items,!1)+"</p>"},CODE:{name:"CODE",mergeToBlock:!0,toText:t=>"<code>"+$(t.items,!0)+"</code>"},LIST:{name:"LIST",mergeToBlock:!1,mergeFollowingNonTypedItemsWithSmallDistance:!0,toText:t=>"<ul>\n"+$(t.items,!1)+"</ul>"},PARAGRAPH:{name:"PARAGRAPH",toText:t=>"<p>"+$(t.items,!1)+"</p>"}};function $(t,e){var s,n="";const o=()=>{n+=s.endSymbol,s=null};return t.forEach((r,a)=>{var i;r&&(r.words.forEach((t,r)=>{const a=t.type,i=t.format;var h;!s||i&&i===s||o(),r>0&&(!a||!a.attachWithoutWhitespace)&&(1!==(h=t.string).length||"."!==h[0]&&"!"!==h[0]&&"?"!==h[0])&&(n+=" "),!i||s||e||(n+=(s=i).startSymbol),!a||e&&!a.plainTextFormat?n+=t.string:n+=a.toText(t.string)}),!s||a!==t.length-1&&(0===(i=t[a+1]).words.length?null:i.words[0].format)===s||o(),n+="\n")}),n}P.enumValueOf=function(t){return P[t]},P.isHeadline=function(t){return t&&2===t.name.length&&"H"===t.name[0]},P.blockToText=function(t){return t.type?t.type.toText(t):$(t.items,!1)},P.headlineByLevel=function(t){return 1===t?P.H1:2===t?P.H2:3===t?P.H3:4===t?P.H4:5===t?P.H5:(t>6&&console.warn("Unsupported headline level: "+t+" (supported are 1-6), defaulting to level 6"),P.H6)};var j=class extends A{constructor(){super("Compact To Lines")}transform(t){const{mostUsedDistance:e,fontToFormats:s}=t.globals,n=[],o=[];var r=0,a=0;const i=new W({mostUsedDistance:e}),h=new H(s);return t.pages.forEach(t=>{if(t.items.length>0){const e=[];i.group(t.items).forEach(s=>{const i=h.compact(s);if(s.length>1&&(i.annotation=x,s.forEach(t=>{t.annotation=E,e.push(new k({...t}))})),0===i.words.length&&(i.annotation=E),e.push(i),i.parsedElements.formattedWords&&(a+=i.parsedElements.formattedWords),i.parsedElements.containLinks&&r++,i.parsedElements.footnoteLinks.length>0){const e=i.parsedElements.footnoteLinks.map(e=>({footnoteLink:e,page:t.index+1}));o.push.apply(o,e)}if(i.parsedElements.footnotes.length>0){i.type=P.FOOTNOTES;const e=i.parsedElements.footnotes.map(e=>({footnote:e,page:t.index+1}));n.push.apply(n,e)}}),t.items=e}}),new I({...t,messages:["Detected "+a+" formatted words","Found "+r+" links","Detected "+o.length+" footnotes links","Detected "+n.length+" footnotes"]})}};function q(t){var e=0;if(0===t.trim().length)return e;for(var n=0;n<t.length;n++){const o=t.charCodeAt(n);s(o)||32===o||160===o||(e=(e<<5)-e+o,e|=0)}return e}var Y=class extends A{constructor(){super("Remove Repetitive Elements")}transform(t){const e=[],s={},n={};t.pages.forEach(t=>{const o=t.items.reduce((t,e)=>(e.y<t.minY?(t.minElements=[e],t.minY=e.y):e.y===t.minY&&t.minElements.push(e),e.y>t.maxY?(t.maxElements=[e],t.maxY=e.y):e.y===t.maxY&&t.maxElements.push(e),t),{minY:999,maxY:0,minElements:[],maxElements:[]}),r=q(o.minElements.reduce((t,e)=>t+e.text().toUpperCase(),"")),a=q(o.maxElements.reduce((t,e)=>t+e.text().toUpperCase(),""));e.push({minElements:o.minElements,maxElements:o.maxElements,minLineHash:r,maxLineHash:a}),s[r]=s[r]?s[r]+1:1,n[a]=n[a]?n[a]+1:1});var o=0,r=0;return t.pages.forEach((a,i)=>{s[e[i].minLineHash]>=Math.max(3,2*t.pages.length/3)&&(e[i].minElements.forEach(t=>{t.annotation=E}),r++),n[e[i].maxLineHash]>=Math.max(3,2*t.pages.length/3)&&(e[i].maxElements.forEach(t=>{t.annotation=E}),o++)}),new I({...t,messages:["Removed Header: "+o,"Removed Footers: "+r]})}},R=class extends A{constructor(){super("Vertical to Horizontal Text")}transform(t){var e=0;return t.pages.forEach(t=>{const s=new V;s.consumeAll(t.items),t.items=s.complete(),e+=s.foundVerticals}),new I({...t,messages:["Converted "+e+" verticals"]})}},V=class extends S{foundVerticals;constructor(){super(),this.foundVerticals=0}shouldStash(t){return 1===t.words.length&&1===t.words[0].string.length}doMatchesStash(t,e){return t.y-e.y>5&&t.words[0].type===e.words[0].type}doFlushStash(t,e){if(t.length>5){var s=[],n=999,o=0,r=0,a=0;t.forEach(t=>{t.annotation=E,e.push(t),s.push(t.words[0]),n=Math.min(n,t.x),o=Math.max(o,t.y),r+=t.width,a=Math.max(a,t.height)}),e.push(new k({...t[0],x:n,y:o,width:r,height:a,words:s,annotation:x})),this.foundVerticals++}else e.push(...t)}},K=class{headlineCharCodes;stackedLineItems;stackedChars;constructor(t){this.headlineCharCodes=r(t.headline),this.stackedLineItems=[],this.stackedChars=0}consume(t){const e=r(t.text());if(this.matchAll(e)){if(this.stackedLineItems.push(t),this.stackedChars+=e.length,this.stackedChars===this.headlineCharCodes.length)return this.stackedLineItems}else this.stackedChars>0&&(this.stackedChars=0,this.stackedLineItems=[],this.consume(t));return null}matchAll(t){for(var e=0;e<t.length;e++){const s=this.headlineCharCodes[this.stackedChars+e];if(t[e]!==s)return!1}return!0}},J=class extends A{constructor(){super("Detect TOC")}transform(t){const e=[],r=Math.min(20,t.pages.length),a=new z;var i,l,c=[];t.pages.slice(0,r).forEach(t=>{var r=0;const h=/* @__PURE__ */new Set,m=[];var u,f;if(t.items.filter(t=>t.words.includes(t=>o(t.string,"."))).forEach(t=>{var e=t.words.filter(t=>!o(t.string,"."));const a=[];for(;e.length>0&&n(e[e.length-1].string);){const t=e.pop();a.unshift(t.string)}if(0===a.length&&e.length>0){const t=e[e.length-1];for(;s(t.string.charCodeAt(t.string.length-1));)a.unshift(t.string.charAt(t.string.length-1)),t.string=t.string.substring(0,t.string.length-1)}var i=a.length>0;i?(i=!0,u&&(e.push(...u),u=null),m.push(new G({pageNumber:parseInt(a.join("")),lineItem:new k({...t,words:e})})),r++):l?(u&&h.add(f),u=e,f=t):l=t}),100*r/t.items.length>75){e.push(t.index+1),i=t,a.levelPageItems(m),c.push(...m);const s=[];t.items.forEach(t=>{h.has(t)||(t.annotation=E),s.push(t),t===l&&s.push(new k({...t,type:P.H2,annotation:x}))}),t.items=s}else l=null});var m=c.length;const u=[],f=[],p={};if(e.length>0){c.forEach(t=>{i.items.push(new k({words:[new O({string:" ".repeat(3*t.level)+"-"})].concat(t.lineItem.words),type:P.TOC,annotation:x}))});const e=function(t,e){for(var s of e){const e=Q(t,s.lineItem.text());if(e)return e.index-s.pageNumber}return null}(t.pages.filter(t=>t.index>i.index),c);c.forEach(s=>{var n,o=t.pages[s.pageNumber+e];o&&((n=_(o,s.lineItem.text()))||(o=t.pages[s.pageNumber+e+1])&&(n=_(o,s.lineItem.text()))),n?function(t,e,s,n){s.headlineItems.forEach(t=>t.annotation=E);const o=P.headlineByLevel(e.level+2),r=s.headlineItems.reduce((t,e)=>Math.max(t,e.height),0);t.items.splice(s.lineIndex+1,0,new k({...s.headlineItems[0],words:e.lineItem.words,height:r,type:o,annotation:x}));var a=n[o.name];a?(a.min=Math.min(a.min,r),a.max=Math.max(a.max,r)):(a={min:r,max:r},n[o.name]=a)}(o,s,n,p):u.push(s)});var d=i.index+2,g=[];const s=e=>{g.length>0&&(g.forEach(s=>{const n=P.headlineByLevel(s.level+2),o=p[n.name];if(o){const[r,a]=function(t,e,s,n,o){const r=e.lineItem.text().toUpperCase();for(var a=n;a<=o;a++){const e=t[a-1];if(e){const t=e.items.findIndex(t=>!t.type&&!t.annotation&&t.height>=s.min&&t.height<=s.max&&h(r,t.text())>=.5);if(t>-1)return[a-1,t]}}return[-1,-1]}(t.pages,s,o,d,e);if(a>-1){const e=t.pages[r];e.items[a].annotation=E,e.items.splice(a+1,0,new k({...s.lineItem,type:n,annotation:x})),f.push(s)}}}),g=[])};u.length>0&&(c.forEach(t=>{u.includes(t)?g.push(t):(s(t.pageNumber),d=t.pageNumber)}),g.length>0&&s(t.pages.length))}const y=[];return y.push("Detected "+e.length+" table of content pages"),e.length>0&&(y.push("TOC headline heights: "+JSON.stringify(p)),y.push("Found TOC headlines: "+(m-u.length+f.length)+"/"+m)),u.length>0&&(y.push("Found TOC headlines (by size): "+f.map(t=>t.lineItem.text())),y.push("Missing TOC headlines: "+u.filter(t=>!f.includes(t)).map(t=>t.lineItem.text()+"=>"+t.pageNumber))),new I({...t,globals:{...t.globals,tocPages:e,headlineTypeToHeightRange:p},messages:y})}};function Q(t,e){for(var s of t)if(_(s,e))return s;return null}function _(t,e){const s=new K({headline:e});var n=0;for(var o of t.items){const t=s.consume(o);if(t)return{lineIndex:n,headlineItems:t};n++}return null}var z=class{levelByMethod;uniqueFonts;constructor(){this.levelByMethod=null,this.uniqueFonts=[]}levelPageItems(t){if(!this.levelByMethod)if(this.calculateUniqueX(t).length>1)this.levelByMethod=this.levelByXDiff;else{const e=this.calculateUniqueFonts(t);e.length>1?(this.uniqueFonts=e,this.levelByMethod=this.levelByFont):this.levelByMethod=this.levelToZero}this.levelByMethod(t)}levelByXDiff(t){const e=this.calculateUniqueX(t);t.forEach(t=>{t.level=e.indexOf(t.lineItem.x)})}levelByFont(t){t.forEach(t=>{t.level=this.uniqueFonts.indexOf(t.lineItem.font)})}levelToZero(t){t.forEach(t=>{t.level=0})}calculateUniqueX(t){var e=t.reduce(function(t,e){return t.indexOf(e.lineItem.x)<0&&t.push(e.lineItem.x),t},[]);return e.sort((t,e)=>t-e),e}calculateUniqueFonts(t){return t.reduce(function(t,e){return t.indexOf(e.lineItem.font)<0&&t.push(e.lineItem.font),t},[])}},G=class{lineItem;pageNumber;level;constructor(t){this.lineItem=t.lineItem,this.pageNumber=t.pageNumber,this.level=0}},X=class extends A{constructor(){super("Detect List Items")}transform(t){var e=0,s=0;return t.pages.forEach(t=>{const n=[];t.items.forEach(t=>{if(n.push(t),!t.type){var o=t.text();if(a(t.words[0].string))if(e++,"-"===t.words[0].string)t.annotation=v,t.type=P.LIST;else{t.annotation=E;const e=t.words.map(t=>new O({...t}));e[0].string="-",n.push(new k({...t,words:e,annotation:x,type:P.LIST}))}else/^[\s]*[\d]*[.][\s].*$/g.test(o)&&(s++,t.annotation=v,t.type=P.LIST)}}),t.items=n}),new I({...t,messages:["Detected "+e+" plain list items.","Detected "+s+" numbered list items."]})}},Z=class extends A{constructor(){super("Detect Headers")}transform(t){const{tocPages:e,headlineTypeToHeightRange:s,mostUsedHeight:n,mostUsedDistance:o,mostUsedFont:r,maxHeight:a}=t.globals,h=e.length>0;var l=0;const c=function(t,e){const s=/* @__PURE__ */new Set;return t.forEach(t=>{t.items.forEach(n=>{n.type||n.height!==e||s.add(t)})}),s}(t.pages,a),m=n+(a-n)/4;if(c.forEach(t=>{t.items.forEach(t=>{const e=t.height;!t.type&&e>m&&(t.type=e===a?P.H1:P.H2,t.annotation=v,l++)})}),h)Object.keys(s).forEach(e=>{var o=s[e];o.max>n&&t.pages.forEach(t=>{t.items.forEach(t=>{t.type||t.height!==o.max||(t.annotation=v,t.type=P.enumValueOf(e),l++)})})});else{const e=[];t.pages.forEach(t=>{t.items.forEach(t=>{!t.type&&t.height>n&&!i(t.text())&&(e.includes(t.height)||e.push(t.height))})}),e.sort((t,e)=>e-t),e.forEach((e,s)=>{if(s+2<=6){const n=P.headlineByLevel(2+s);t.pages.forEach(t=>{t.items.forEach(t=>{t.type||t.height!==e||i(t.text())||(l++,t.annotation=v,t.type=n)})})}})}var u=1;if(t.pages.forEach(t=>{t.items.forEach(t=>{t.type&&t.type.headline&&(u=Math.max(u,t.type.headlineLevel))})}),u<6){const e=P.headlineByLevel(u+1);t.pages.forEach(t=>{var s;t.items.forEach(t=>{!t.type&&t.height===n&&t.font!==r&&(!s||s.y<t.y||s.type&&s.type.headline||s.y-t.y>2*o)&&t.text()===t.text().toUpperCase()&&(l++,t.annotation=v,t.type=e),s=t})})}return new I({...t,messages:["Detected "+l+" headlines."]})}};var tt=class extends p{items;constructor(t){super(t),this.items=[],t.items&&t.items.forEach(t=>this.addItem(t))}addItem(t){if(this.type&&t.type&&this.type!==t.type)throw new Error(`Adding item of type ${t.type.name} to block of type ${this.type.name}`);this.type||(this.type=t.type),t.parsedElements&&(this.parsedElements?this.parsedElements.add(t.parsedElements):this.parsedElements=t.parsedElements);const e=new k({...t});e.type=null,this.items.push(e)}},et=class t extends y{constructor(e){if(super(e,tt.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==E),t.items.forEach(t=>t.annotation=null)}),t}},st=class extends et{constructor(){super("Gather Blocks")}transform(t){const{mostUsedDistance:e}=t.globals;var s=0,n=0;return t.pages.map(t=>{n+=t.items.length;const o=[];var r=new tt({});const a=()=>{r.items.length>1&&(r.annotation=v),o.push(r),r=new tt({}),s++};var i=function(t){var e=999;return t.forEach(t=>{e=Math.min(e,t.x)}),999===e?null:e}(t.items);t.items.forEach(t=>{r.items.length>0&&function(t,e,s,n){if(t.type&&t.type.mergeFollowingNonTypedItems&&!e.type)return!1;const o=t.items[t.items.length-1],r=function(t,e,s,n){const o=t.y-e.y;if(o<0-n/2)return!0;var r=n+1;t.x>s&&e.x>s&&(r=n+n/2);return o>r}(o,e,s,n);if(t.type&&t.type.mergeFollowingNonTypedItemsWithSmallDistance&&!e.type&&!r)return!1;if(e.type!==t.type)return!0;return e.type?!e.type.mergeToBlock:r}(r,t,i,e)&&a(),r.addItem(t)}),r.items.length>0&&a(),t.items=o}),new I({...t,messages:["Gathered "+s+" blocks out of "+n+" line items"]})}};var nt=class extends et{constructor(){super("$1")}transform(t){const e=t.globals.mostUsedHeight??0;var s=0;return t.pages.forEach(t=>{var n=function(t){var e=999;return t.forEach(t=>{t.items.forEach(t=>{e=Math.min(e,t.x)})}),999===e?null:e}(t.items);t.items.forEach(t=>{!t.type&&function(t,e,s){if(0===e.length)return!1;if(1===e.length)return e[0].x>t&&e[0].height<=s+1;for(var n of e)if(n.x===t)return!1;return!0}(n,t.items,e)&&(t.annotation=v,t.type=P.CODE,s++)})}),new I({...t,messages:["Detected "+s+" code/quote items."]})}};var ot=class extends et{constructor(){super("Level Lists")}transform(t){var e=0,s=0;return t.pages.forEach(t=>{t.items.filter(t=>t.type===P.LIST).forEach(t=>{var n,o=0;const r={};var a=!1;t.items.forEach(t=>{n?t.x>n?(o++,r[t.x]=o):t.x<n&&(o=r[t.x]):r[t.x]=0,o>0&&(t.words=[new O({string:" ".repeat(3*o)})].concat(t.words),a=!0),n=t.x}),e++,a?(s++,t.annotation=b):t.annotation=T})}),new I({...t,messages:["Modified "+s+" / "+e+" list blocks."]})}},rt=class extends y{constructor(){super("To Text Blocks","TextBlock")}transform(t){return t.pages.forEach(t=>{const e=[];t.items.forEach(t=>{const s=t.type?t.type.name:"Unknown";e.push({category:s,text:P.blockToText(t)})}),t.items=e}),new I({...t})}},at=class extends y{constructor(){super("To HTML","String")}transform(t){return t.pages.forEach(t=>{var e="";t.items.forEach(t=>{let s;s="TOC"===t.category?t.text:t.text.replace(/(\r\n|\n|\r)/gm,"\n"),"LIST"!==t.category&&(s=s.split("- ").join("")),"CODE"===t.category&&(s=s.split("<code>").join("").split("</code>").join("")),e+=`<p>${s}</p>\n\n`}),t.items=[e]}),new I({...t})}};async function it(t,e={}){var{addPageNumbers:s=!1,addCitation:n=!0}=e,o="string"==typeof t?await async function(t,e={}){const s=e.timeout?1e3*e.timeout:1e4,n=new AbortController,o=setTimeout(()=>n.abort(),s);try{const s=await fetch(t,{signal:n.signal});if(clearTimeout(o),!s.ok)throw new Error(`HTTP ${s.status}`);return"arraybuffer"===e.responseType?await s.arrayBuffer():await s.text()}catch(r){throw clearTimeout(o),r}}(t,{responseType:"arraybuffer",timeout:10}):t;let r;try{let{resolvePDFJS:t}=await import("https://cdn.jsdelivr.net/npm/pdfjs-serverless@1.1.0/+esm");const{getDocument:e}=await t();r=await e({data:new Uint8Array(o),useSystemFonts:!0,verbosity:0}).promise}catch(E){return{error:E.message}}const a=[...Array(r.numPages).keys()].map(t=>new g({index:t}));let i,h={};for(let f=1;f<=r.numPages;f++){const t=await r.getPage(f),e=await t.getTextContent();if(!(Object.keys(h).length<10)){i=u(h);break}h=m(h,t.pageNumber-1,e.items)}let l=i?i.pageNum:0;for(let m=1;m<=r.numPages;m++){const t=await r.getPage(m);await t.getOperatorList();const e=t.getViewport({scale:1});let s=await t.getTextContent();i&&t.pageIndex>=i.pageIndex&&(s=f(s,l),l++);const n=s.items.map(t=>{const s=[1,0,0,1,0,0];for(let r=0;r<6;r++)s[r]+=t.transform[r]*e.transform[r%2?3:0],r%2&&(s[r+1]+=t.transform[r]*e.transform[1]);const n=Math.sqrt(s[2]*s[2]+s[3]*s[3]),o=t.height/n;return new d({x:Math.round(t.transform[4]),y:Math.round(t.transform[5]),width:Math.round(t.width),height:Math.round(o<=1?t.height:o),text:t.str,font:t.fontName})});a[t.pageNumber-1].items=n}var c=new I({pages:a});let p;[new B,new j,new Y,new R,new J,new Z,new X,new st,new nt,new ot,new rt,new at].forEach(t=>{p&&(c=p.completeTransform(c)),c=t.transform(c),p=t});var y=c.pages.reduce((t,e,n)=>t+`<p id="page-${n+1}">${s?` [${n+1}] `:""}${e.items.join('</p><p id="page-'+n+'">')}</p>`,"");if(n){var{Author:w,Title:x}=(await r.getMetadata()).info;x=y.slice(0,400).match(/<h[0-9]>(.*?)<\/h[0-9]>/)?.[1]||x}return{author:w,title:x,html:y,format:"pdf"}}export{it as convertPDFToHTML};
+function t(t){for(;32===t.charCodeAt(0);)t=t.substring(1,t.length);return t}function e(t){for(;32===t.charCodeAt(t.length-1);)t=t.substring(0,t.length-1);return t}function s(t){return t>=48&&t<=57}function n(t){for(var e=0;e<t.length;e++)if(!s(t.charCodeAt(e)))return!1;return!0}function o(t,e){const s=e.charCodeAt(0);for(var n=0;n<t.length;n++)if(t.charCodeAt(n)!==s)return!1;return!0}function r(t){return function(t){const e=[];for(var s=0;s<t.length;s++)e.push(t.charCodeAt(s));return e}(t=t.toUpperCase()).filter(t=>32!==t&&9!==t&&46!==t)}function a(t){if(t.length>1)return!1;const e=t.charAt(0);return"-"===e||"•"===e||"–"===e}function i(t){return/^[\s]*[-•–][\s].*$/g.test(t)}function h(t,e){const s=new Set(t.toUpperCase().split(" ")),n=new Set(e.toUpperCase().split(" "));return new Set([...s].filter(t=>n.has(t))).size/Math.max(s.size,n.size)}var l=(t,e,s)=>Math.floor(t/e*s),c=(s,o,r)=>{for(const{str:a}of s){const s=e(t(a));n(s)&&(o[r]||(o[r]=[]),o[r].push(Number(s)))}return o},m=(t,e,s)=>{const n=l(1,6,s.length),o=l(5,6,s.length),r=c(s.slice(0,n),t,e);return c(s.slice(o),r,e)},u=t=>{let e=0;const s=Object.keys(t);if(0!==s.length&&1!==s.length)for(let n=0;n<s.length-1;n++){const o=t[s[n]],r=t[s[n+1]],a=e;for(let t=0;t<o.length&&e<2;t++)for(let a=0;a<r.length&&e<2;a++){const i=Number(s[n+1])-Number(s[n]);(o[t]+1===r[a]||i>1&&o[t]+i===r[a])&&e++}let i=n>0?Object.entries(t)[n-1]:Object.entries(t)[n];if(a===e)e=0,i=Object.entries(t)[n];else if(e>=2)return{pageIndex:Number(i[0]),pageNum:i[1][0]}}},p=(t,e)=>{const s={items:[...t.items]},n=l(1,6,s.items.length),o=l(5,6,s.items.length);return s.items=s.items.filter((t,r)=>{const a=r>0&&r<n,i=r>o&&r<s.items.length;return a||i?Number(t.str)!==Number(e):t}),s},f=class t{type;annotation;parsedElements;constructor(e){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");this.type=e.type??null,this.annotation=e.annotation??null,this.parsedElements=e.parsedElements??null}},d=class extends f{x;y;width;height;text;font;lineFormat;unopenedFormat;unclosedFormat;constructor(t){super(t),this.x=t.x,this.y=t.y,this.width=t.width,this.height=t.height,this.text=t.text,this.font=t.font,this.lineFormat=t.lineFormat??null,this.unopenedFormat=t.unopenedFormat??null,this.unclosedFormat=t.unclosedFormat??null}},g=class{index;items;constructor(t){this.index=t.index,this.items=t.items||[]}};async function y(t,e={}){const s=e.timeout?1e3*e.timeout:1e4,n=new AbortController,o=setTimeout(()=>n.abort(),s);try{const s=await fetch(t,{signal:n.signal});if(clearTimeout(o),!s.ok)throw new Error(`HTTP ${s.status}`);return"arraybuffer"===e.responseType?await s.arrayBuffer():await s.text()}catch(r){throw clearTimeout(o),r}}var w=class t{name;itemType;constructor(e,s){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");if(this.transform===t.prototype.transform)throw new TypeError("Please implement abstract method 'transform()'.");this.name=e,this.itemType=s}transform(t){throw new TypeError("Do not call abstract method foo from child.")}completeTransform(t){return t.messages=[],t}},x=class{category;color;constructor(t){this.category=t.category,this.color=t.color}},E=new x({category:"Added",color:"green"}),T=new x({category:"Removed",color:"red"}),v=new x({category:"Unchanged",color:"brown"}),b=new x({category:"Detected",color:"green"}),L=new x({category:"Modified",color:"green"}),O=class t extends w{constructor(e){if(super(e,d.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==T),t.items.forEach(t=>t.annotation=null)}),t}},I=class{pages;globals;messages;constructor(t){this.pages=t.pages||[],this.globals=t.globals||{},this.messages=t.messages||[]}},S=class{string;type;format;constructor(t){this.string=t.string,this.type=t.type??null,this.format=t.format??null}},k=class extends f{x;y;width;height;words;constructor(t){super(t),this.x=t.x??0,this.y=t.y??0,this.width=t.width??0,this.height=t.height??0,this.words=t.words||[],t.text&&!t.words&&(this.words=t.text.split(" ").filter(t=>t.trim().length>0).map(t=>new S({string:t})))}text(){return this.wordStrings().join(" ")}wordStrings(){return this.words.map(t=>t.string)}},C=class t{results;stash;constructor(){if(this.constructor===t)throw new TypeError("Can not construct abstract class.");this.results=[],this.stash=[]}consumeAll(t){t.forEach(t=>this.consume(t))}consume(t){this.shouldStash(t)?(this.matchesStash(t)||this.flushStash(),this.pushOnStash(t)):(this.stash.length>0&&this.flushStash(),this.results.push(t))}pushOnStash(t){this.onPushOnStash(t),this.stash.push(t)}complete(){return this.stash.length>0&&this.flushStash(),this.results}matchesStash(t){if(0===this.stash.length)return!0;const e=this.stash[this.stash.length-1];return this.doMatchesStash(e,t)}flushStash(){this.stash.length>0&&(this.doFlushStash(this.stash,this.results),this.stash=[])}onPushOnStash(t){}shouldStash(t){throw new TypeError(" Do not call abstract method foo from child."+t)}doMatchesStash(t,e){throw new TypeError(" Do not call abstract method foo from child."+t+e)}doFlushStash(t,e){throw new TypeError(" Do not call abstract method foo from child."+t+e)}},F=class{footnoteLinks;footnotes;containLinks;formattedWords;constructor(t){this.footnoteLinks=t.footnoteLinks||[],this.footnotes=t.footnotes||[],this.containLinks=t.containLinks??!1,this.formattedWords=t.formattedWords??0}add(t){this.footnoteLinks=this.footnoteLinks.concat(t.footnoteLinks),this.footnotes=this.footnotes.concat(t.footnotes),this.containLinks=this.containLinks||t.containLinks,this.formattedWords+=t.formattedWords}};function N(t){t.sort((t,e)=>t.x-e.x)}var D={BOLD:{name:"BOLD",startSymbol:"<strong>",endSymbol:"</strong>"},OBLIQUE:{name:"OBLIQUE",startSymbol:"<em>",endSymbol:"</em>"},BOLD_OBLIQUE:{name:"BOLD_OBLIQUE",startSymbol:"<strong><em>",endSymbol:"</em></strong>"}},H={LINK:{name:"LINK",toText:t=>`<a href="${t}">${t}</a>`},FOOTNOTE_LINK:{name:"FOOTNOTE_LINK",attachWithoutWhitespace:!0,plainTextFormat:!0,toText:t=>`<sup><a href="#${t}">${t}</a></sup>`},FOOTNOTE:{name:"FOOTNOTE",toText:t=>`<p id="${t}">^${t}</p>`}},M=class{fontToFormats;constructor(t){this.fontToFormats=t}compact(t){N(t);const e=new B(this.fontToFormats);e.consumeAll(t.map(t=>new d({...t})));const s=e.complete();var n=0,o=0;return t.forEach(t=>{n=Math.max(n,t.height),o+=t.width}),new k({x:t[0].x,y:t[0].y,height:n,width:o,words:s,parsedElements:new F({footnoteLinks:e.footnoteLinks,footnotes:e.footnotes,containLinks:e.containLinks,formattedWords:e.formattedWords})})}},B=class extends C{fontToFormats;footnoteLinks;footnotes;formattedWords;containLinks;stashedNumber;firstY;currentItem;constructor(t){super(),this.fontToFormats=t,this.footnoteLinks=[],this.footnotes=[],this.formattedWords=0,this.containLinks=!1,this.stashedNumber=!1,this.currentItem=null}shouldStash(t){return this.firstY||(this.firstY=t.y),this.currentItem=t,!0}onPushOnStash(t){this.stashedNumber=n(t.text.trim())}doMatchesStash(t,e){if(this.fontToFormats.get(t.font)!==this.fontToFormats.get(e.font))return!1;const s=n(e.text.trim());return this.stashedNumber===s}doFlushStash(t,e){if(this.stashedNumber){const s=t.map(t=>t.text).join("").trim();t[0].y>this.firstY?(e.push(new S({string:`${s}`,type:H.FOOTNOTE_LINK})),this.footnoteLinks.push(parseInt(s))):this.currentItem&&this.currentItem.y<t[0].y?(e.push(new S({string:`${s}`,type:H.FOOTNOTE})),this.footnotes.push(s)):this.copyStashItemsAsText(t,e)}else this.copyStashItemsAsText(t,e)}copyStashItemsAsText(t,e){const s=this.fontToFormats.get(t[0].font);e.push(...this.itemsToWords(t,s))}itemsToWords(t,e){const s=(n=t,o="",r=null,n.forEach(t=>{var e=t.text;o.endsWith(" ")||e.startsWith(" ")||(r?t.x-r.x-r.width>5&&(o+=" "):a(t.text)&&(e+=" ")),o+=e,r=t}),o).split(" ");var n,o,r;const i=e?D[e]:null;return s.filter(t=>t.trim().length>0).map(t=>{var e=null;return t.startsWith("http:")?(this.containLinks=!0,e=H.LINK):t.startsWith("www.")&&(this.containLinks=!0,t=`http://${t}`,e=H.LINK),i&&this.formattedWords++,new S({string:t,type:e,format:i})})}};var U=class extends O{fontMap;constructor(t){super("Calculate Global Stats"),this.fontMap=t}transform(t){const e={},s={};var n,o=0;t.pages.forEach(t=>{t.items.forEach(t=>{t.height&&(e[t.height]=e[t.height]?e[t.height]+1:1,s[t.font]=s[t.font]?s[t.font]+1:1,t.height>o&&(o=t.height,n=t.font))})});const r=parseInt(A(e)),a=A(s),i={};t.pages.forEach(t=>{var e;t.items.forEach(t=>{if(t.height===r&&t.text.trim().length>0){if(e&&t.y!==e.y){const s=e.y-t.y;s>0&&(i[s]=i[s]?i[s]+1:1)}e=t}else e=null})});const h=parseInt(A(i)),l=[],c=/* @__PURE__ */new Map;this.fontMap?.forEach(function(t,e){l.push(e+" = "+t.name);const s=t.name.toLowerCase();var o=null;e===a?o=null:s.includes("bold")&&(s.includes("oblique")||s.includes("italic"))?o=D.BOLD_OBLIQUE:s.includes("bold")?o=D.BOLD:s.includes("oblique")||s.includes("italic")?o=D.OBLIQUE:s===n&&(o=D.BOLD),o&&c.set(e,o.name)}),l.sort();const m=t.pages.map(t=>({...t,items:t.items.map(t=>({...t}))}));return new I({...t,pages:m,globals:{mostUsedHeight:r,mostUsedFont:a,mostUsedDistance:h,maxHeight:o,maxHeightFont:n,fontToFormats:c},messages:["Items per height: "+JSON.stringify(e),"Items per font: "+JSON.stringify(s),"Items per distance: "+JSON.stringify(i),"Fonts:"+JSON.stringify(l)]})}};function A(t){var e=0,s="";return Object.keys(t).map(n=>{(!s||t[n]>e)&&(e=t[n],s=n)}),s}var P=class t extends w{constructor(e){if(super(e,k.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==T),t.items.forEach(t=>t.annotation=null)}),t}},$=class{mostUsedDistance;constructor(t){this.mostUsedDistance=t.mostUsedDistance||12}group(t){const e=[];var s=[];return t.forEach(t=>{s.length>0&&Math.abs(s[0].y-t.y)>=this.mostUsedDistance/2&&(e.push(s),s=[]),s.push(t)}),e.push(s),e.forEach(t=>{N(t)}),e}},W={H1:{name:"H1",headline:!0,headlineLevel:1,toText:t=>"<h1>"+j(t.items,!0)+"</h1>"},H2:{name:"H2",headline:!0,headlineLevel:2,toText:t=>"<h2>"+j(t.items,!0)+"</h2>"},H3:{name:"H3",headline:!0,headlineLevel:3,toText:t=>"<h3>"+j(t.items,!0)+"</h3>"},H4:{name:"H4",headline:!0,headlineLevel:4,toText:t=>"<h4>"+j(t.items,!0)+"</h4>"},H5:{name:"H5",headline:!0,headlineLevel:5,toText:t=>"<h5>"+j(t.items,!0)+"</h5>"},H6:{name:"H6",headline:!0,headlineLevel:6,toText:t=>"<h6>"+j(t.items,!0)+"</h6>"},TOC:{name:"TOC",mergeToBlock:!0,toText:t=>j(t.items,!0)},FOOTNOTES:{name:"FOOTNOTES",mergeToBlock:!0,mergeFollowingNonTypedItems:!0,toText:t=>"<p>"+j(t.items,!1)+"</p>"},CODE:{name:"CODE",mergeToBlock:!0,toText:t=>"<code>"+j(t.items,!0)+"</code>"},LIST:{name:"LIST",mergeToBlock:!1,mergeFollowingNonTypedItemsWithSmallDistance:!0,toText:t=>"<ul>\n"+j(t.items,!1)+"</ul>"},PARAGRAPH:{name:"PARAGRAPH",toText:t=>"<p>"+j(t.items,!1)+"</p>"}};function j(t,e){var s,n="";const o=()=>{n+=s.endSymbol,s=null};return t.forEach((r,a)=>{var i;r&&(r.words.forEach((t,r)=>{const a=t.type,i=t.format;var h;!s||i&&i===s||o(),r>0&&(!a||!a.attachWithoutWhitespace)&&(1!==(h=t.string).length||"."!==h[0]&&"!"!==h[0]&&"?"!==h[0])&&(n+=" "),!i||s||e||(n+=(s=i).startSymbol),!a||e&&!a.plainTextFormat?n+=t.string:n+=a.toText(t.string)}),!s||a!==t.length-1&&(0===(i=t[a+1]).words.length?null:i.words[0].format)===s||o(),n+="\n")}),n}W.enumValueOf=function(t){return W[t]},W.isHeadline=function(t){return t&&2===t.name.length&&"H"===t.name[0]},W.blockToText=function(t){return t.type?t.type.toText(t):j(t.items,!1)},W.headlineByLevel=function(t){return 1===t?W.H1:2===t?W.H2:3===t?W.H3:4===t?W.H4:5===t?W.H5:(t>6&&console.warn("Unsupported headline level: "+t+" (supported are 1-6), defaulting to level 6"),W.H6)};var q=class extends P{constructor(){super("Compact To Lines")}transform(t){const{mostUsedDistance:e,fontToFormats:s}=t.globals,n=[],o=[];var r=0,a=0;const i=new $({mostUsedDistance:e}),h=new M(s);return t.pages.forEach(t=>{if(t.items.length>0){const e=[];i.group(t.items).forEach(s=>{const i=h.compact(s);if(s.length>1&&(i.annotation=E,s.forEach(t=>{t.annotation=T,e.push(new k({...t}))})),0===i.words.length&&(i.annotation=T),e.push(i),i.parsedElements.formattedWords&&(a+=i.parsedElements.formattedWords),i.parsedElements.containLinks&&r++,i.parsedElements.footnoteLinks.length>0){const e=i.parsedElements.footnoteLinks.map(e=>({footnoteLink:e,page:t.index+1}));o.push.apply(o,e)}if(i.parsedElements.footnotes.length>0){i.type=W.FOOTNOTES;const e=i.parsedElements.footnotes.map(e=>({footnote:e,page:t.index+1}));n.push.apply(n,e)}}),t.items=e}}),new I({...t,messages:["Detected "+a+" formatted words","Found "+r+" links","Detected "+o.length+" footnotes links","Detected "+n.length+" footnotes"]})}};function Y(t){var e=0;if(0===t.trim().length)return e;for(var n=0;n<t.length;n++){const o=t.charCodeAt(n);s(o)||32===o||160===o||(e=(e<<5)-e+o,e|=0)}return e}var R=class extends P{constructor(){super("Remove Repetitive Elements")}transform(t){const e=[],s={},n={};t.pages.forEach(t=>{const o=t.items.reduce((t,e)=>(e.y<t.minY?(t.minElements=[e],t.minY=e.y):e.y===t.minY&&t.minElements.push(e),e.y>t.maxY?(t.maxElements=[e],t.maxY=e.y):e.y===t.maxY&&t.maxElements.push(e),t),{minY:999,maxY:0,minElements:[],maxElements:[]}),r=Y(o.minElements.reduce((t,e)=>t+e.text().toUpperCase(),"")),a=Y(o.maxElements.reduce((t,e)=>t+e.text().toUpperCase(),""));e.push({minElements:o.minElements,maxElements:o.maxElements,minLineHash:r,maxLineHash:a}),s[r]=s[r]?s[r]+1:1,n[a]=n[a]?n[a]+1:1});var o=0,r=0;return t.pages.forEach((a,i)=>{s[e[i].minLineHash]>=Math.max(3,2*t.pages.length/3)&&(e[i].minElements.forEach(t=>{t.annotation=T}),r++),n[e[i].maxLineHash]>=Math.max(3,2*t.pages.length/3)&&(e[i].maxElements.forEach(t=>{t.annotation=T}),o++)}),new I({...t,messages:["Removed Header: "+o,"Removed Footers: "+r]})}},V=class extends P{constructor(){super("Vertical to Horizontal Text")}transform(t){var e=0;return t.pages.forEach(t=>{const s=new K;s.consumeAll(t.items),t.items=s.complete(),e+=s.foundVerticals}),new I({...t,messages:["Converted "+e+" verticals"]})}},K=class extends C{foundVerticals;constructor(){super(),this.foundVerticals=0}shouldStash(t){return 1===t.words.length&&1===t.words[0].string.length}doMatchesStash(t,e){return t.y-e.y>5&&t.words[0].type===e.words[0].type}doFlushStash(t,e){if(t.length>5){var s=[],n=999,o=0,r=0,a=0;t.forEach(t=>{t.annotation=T,e.push(t),s.push(t.words[0]),n=Math.min(n,t.x),o=Math.max(o,t.y),r+=t.width,a=Math.max(a,t.height)}),e.push(new k({...t[0],x:n,y:o,width:r,height:a,words:s,annotation:E})),this.foundVerticals++}else e.push(...t)}},J=class{headlineCharCodes;stackedLineItems;stackedChars;constructor(t){this.headlineCharCodes=r(t.headline),this.stackedLineItems=[],this.stackedChars=0}consume(t){const e=r(t.text());if(this.matchAll(e)){if(this.stackedLineItems.push(t),this.stackedChars+=e.length,this.stackedChars===this.headlineCharCodes.length)return this.stackedLineItems}else this.stackedChars>0&&(this.stackedChars=0,this.stackedLineItems=[],this.consume(t));return null}matchAll(t){for(var e=0;e<t.length;e++){const s=this.headlineCharCodes[this.stackedChars+e];if(t[e]!==s)return!1}return!0}},Q=class extends P{constructor(){super("Detect TOC")}transform(t){const e=[],r=Math.min(20,t.pages.length),a=new G;var i,l,c=[];t.pages.slice(0,r).forEach(t=>{var r=0;const h=/* @__PURE__ */new Set,m=[];var u,p;if(t.items.filter(t=>t.words.includes(t=>o(t.string,"."))).forEach(t=>{var e=t.words.filter(t=>!o(t.string,"."));const a=[];for(;e.length>0&&n(e[e.length-1].string);){const t=e.pop();a.unshift(t.string)}if(0===a.length&&e.length>0){const t=e[e.length-1];for(;s(t.string.charCodeAt(t.string.length-1));)a.unshift(t.string.charAt(t.string.length-1)),t.string=t.string.substring(0,t.string.length-1)}var i=a.length>0;i?(i=!0,u&&(e.push(...u),u=null),m.push(new X({pageNumber:parseInt(a.join("")),lineItem:new k({...t,words:e})})),r++):l?(u&&h.add(p),u=e,p=t):l=t}),100*r/t.items.length>75){e.push(t.index+1),i=t,a.levelPageItems(m),c.push(...m);const s=[];t.items.forEach(t=>{h.has(t)||(t.annotation=T),s.push(t),t===l&&s.push(new k({...t,type:W.H2,annotation:E}))}),t.items=s}else l=null});var m=c.length;const u=[],p=[],f={};if(e.length>0){c.forEach(t=>{i.items.push(new k({words:[new S({string:" ".repeat(3*t.level)+"-"})].concat(t.lineItem.words),type:W.TOC,annotation:E}))});const e=function(t,e){for(var s of e){const e=_(t,s.lineItem.text());if(e)return e.index-s.pageNumber}return null}(t.pages.filter(t=>t.index>i.index),c);c.forEach(s=>{var n,o=t.pages[s.pageNumber+e];o&&((n=z(o,s.lineItem.text()))||(o=t.pages[s.pageNumber+e+1])&&(n=z(o,s.lineItem.text()))),n?function(t,e,s,n){s.headlineItems.forEach(t=>t.annotation=T);const o=W.headlineByLevel(e.level+2),r=s.headlineItems.reduce((t,e)=>Math.max(t,e.height),0);t.items.splice(s.lineIndex+1,0,new k({...s.headlineItems[0],words:e.lineItem.words,height:r,type:o,annotation:E}));var a=n[o.name];a?(a.min=Math.min(a.min,r),a.max=Math.max(a.max,r)):(a={min:r,max:r},n[o.name]=a)}(o,s,n,f):u.push(s)});var d=i.index+2,g=[];const s=e=>{g.length>0&&(g.forEach(s=>{const n=W.headlineByLevel(s.level+2),o=f[n.name];if(o){const[r,a]=function(t,e,s,n,o){const r=e.lineItem.text().toUpperCase();for(var a=n;a<=o;a++){const e=t[a-1];if(e){const t=e.items.findIndex(t=>!t.type&&!t.annotation&&t.height>=s.min&&t.height<=s.max&&h(r,t.text())>=.5);if(t>-1)return[a-1,t]}}return[-1,-1]}(t.pages,s,o,d,e);if(a>-1){const e=t.pages[r];e.items[a].annotation=T,e.items.splice(a+1,0,new k({...s.lineItem,type:n,annotation:E})),p.push(s)}}}),g=[])};u.length>0&&(c.forEach(t=>{u.includes(t)?g.push(t):(s(t.pageNumber),d=t.pageNumber)}),g.length>0&&s(t.pages.length))}const y=[];return y.push("Detected "+e.length+" table of content pages"),e.length>0&&(y.push("TOC headline heights: "+JSON.stringify(f)),y.push("Found TOC headlines: "+(m-u.length+p.length)+"/"+m)),u.length>0&&(y.push("Found TOC headlines (by size): "+p.map(t=>t.lineItem.text())),y.push("Missing TOC headlines: "+u.filter(t=>!p.includes(t)).map(t=>t.lineItem.text()+"=>"+t.pageNumber))),new I({...t,globals:{...t.globals,tocPages:e,headlineTypeToHeightRange:f},messages:y})}};function _(t,e){for(var s of t)if(z(s,e))return s;return null}function z(t,e){const s=new J({headline:e});var n=0;for(var o of t.items){const t=s.consume(o);if(t)return{lineIndex:n,headlineItems:t};n++}return null}var G=class{levelByMethod;uniqueFonts;constructor(){this.levelByMethod=null,this.uniqueFonts=[]}levelPageItems(t){if(!this.levelByMethod)if(this.calculateUniqueX(t).length>1)this.levelByMethod=this.levelByXDiff;else{const e=this.calculateUniqueFonts(t);e.length>1?(this.uniqueFonts=e,this.levelByMethod=this.levelByFont):this.levelByMethod=this.levelToZero}this.levelByMethod(t)}levelByXDiff(t){const e=this.calculateUniqueX(t);t.forEach(t=>{t.level=e.indexOf(t.lineItem.x)})}levelByFont(t){t.forEach(t=>{t.level=this.uniqueFonts.indexOf(t.lineItem.font)})}levelToZero(t){t.forEach(t=>{t.level=0})}calculateUniqueX(t){var e=t.reduce(function(t,e){return t.indexOf(e.lineItem.x)<0&&t.push(e.lineItem.x),t},[]);return e.sort((t,e)=>t-e),e}calculateUniqueFonts(t){return t.reduce(function(t,e){return t.indexOf(e.lineItem.font)<0&&t.push(e.lineItem.font),t},[])}},X=class{lineItem;pageNumber;level;constructor(t){this.lineItem=t.lineItem,this.pageNumber=t.pageNumber,this.level=0}},Z=class extends P{constructor(){super("Detect List Items")}transform(t){var e=0,s=0;return t.pages.forEach(t=>{const n=[];t.items.forEach(t=>{if(n.push(t),!t.type){var o=t.text();if(a(t.words[0].string))if(e++,"-"===t.words[0].string)t.annotation=b,t.type=W.LIST;else{t.annotation=T;const e=t.words.map(t=>new S({...t}));e[0].string="-",n.push(new k({...t,words:e,annotation:E,type:W.LIST}))}else/^[\s]*[\d]*[.][\s].*$/g.test(o)&&(s++,t.annotation=b,t.type=W.LIST)}}),t.items=n}),new I({...t,messages:["Detected "+e+" plain list items.","Detected "+s+" numbered list items."]})}},tt=class extends P{constructor(){super("Detect Headers")}transform(t){const{tocPages:e,headlineTypeToHeightRange:s,mostUsedHeight:n,mostUsedDistance:o,mostUsedFont:r,maxHeight:a}=t.globals,h=e.length>0;var l=0;const c=function(t,e){const s=/* @__PURE__ */new Set;return t.forEach(t=>{t.items.forEach(n=>{n.type||n.height!==e||s.add(t)})}),s}(t.pages,a),m=n+(a-n)/4;if(c.forEach(t=>{t.items.forEach(t=>{const e=t.height;!t.type&&e>m&&(t.type=e===a?W.H1:W.H2,t.annotation=b,l++)})}),h)Object.keys(s).forEach(e=>{var o=s[e];o.max>n&&t.pages.forEach(t=>{t.items.forEach(t=>{t.type||t.height!==o.max||(t.annotation=b,t.type=W.enumValueOf(e),l++)})})});else{const e=[];t.pages.forEach(t=>{t.items.forEach(t=>{!t.type&&t.height>n&&!i(t.text())&&(e.includes(t.height)||e.push(t.height))})}),e.sort((t,e)=>e-t),e.forEach((e,s)=>{if(s+2<=6){const n=W.headlineByLevel(2+s);t.pages.forEach(t=>{t.items.forEach(t=>{t.type||t.height!==e||i(t.text())||(l++,t.annotation=b,t.type=n)})})}})}var u=1;if(t.pages.forEach(t=>{t.items.forEach(t=>{t.type&&t.type.headline&&(u=Math.max(u,t.type.headlineLevel))})}),u<6){const e=W.headlineByLevel(u+1);t.pages.forEach(t=>{var s;t.items.forEach(t=>{!t.type&&t.height===n&&t.font!==r&&(!s||s.y<t.y||s.type&&s.type.headline||s.y-t.y>2*o)&&t.text()===t.text().toUpperCase()&&(l++,t.annotation=b,t.type=e),s=t})})}return new I({...t,messages:["Detected "+l+" headlines."]})}};var et=class extends f{items;constructor(t){super(t),this.items=[],t.items&&t.items.forEach(t=>this.addItem(t))}addItem(t){if(this.type&&t.type&&this.type!==t.type)throw new Error(`Adding item of type ${t.type.name} to block of type ${this.type.name}`);this.type||(this.type=t.type),t.parsedElements&&(this.parsedElements?this.parsedElements.add(t.parsedElements):this.parsedElements=t.parsedElements);const e=new k({...t});e.type=null,this.items.push(e)}},st=class t extends w{constructor(e){if(super(e,et.name),this.constructor===t)throw new TypeError("Can not construct abstract class.")}completeTransform(t){return t.messages=[],t.pages.forEach(t=>{t.items=t.items.filter(t=>!t.annotation||t.annotation!==T),t.items.forEach(t=>t.annotation=null)}),t}},nt=class extends st{constructor(){super("Gather Blocks")}transform(t){const{mostUsedDistance:e}=t.globals;var s=0,n=0;return t.pages.map(t=>{n+=t.items.length;const o=[];var r=new et({});const a=()=>{r.items.length>1&&(r.annotation=b),o.push(r),r=new et({}),s++};var i=function(t){var e=999;return t.forEach(t=>{e=Math.min(e,t.x)}),999===e?null:e}(t.items);t.items.forEach(t=>{r.items.length>0&&function(t,e,s,n){if(t.type&&t.type.mergeFollowingNonTypedItems&&!e.type)return!1;const o=t.items[t.items.length-1],r=function(t,e,s,n){const o=t.y-e.y;if(o<0-n/2)return!0;var r=n+1;t.x>s&&e.x>s&&(r=n+n/2);return o>r}(o,e,s,n);if(t.type&&t.type.mergeFollowingNonTypedItemsWithSmallDistance&&!e.type&&!r)return!1;if(e.type!==t.type)return!0;return e.type?!e.type.mergeToBlock:r}(r,t,i,e)&&a(),r.addItem(t)}),r.items.length>0&&a(),t.items=o}),new I({...t,messages:["Gathered "+s+" blocks out of "+n+" line items"]})}};var ot=class extends st{constructor(){super("$1")}transform(t){const e=t.globals.mostUsedHeight??0;var s=0;return t.pages.forEach(t=>{var n=function(t){var e=999;return t.forEach(t=>{t.items.forEach(t=>{e=Math.min(e,t.x)})}),999===e?null:e}(t.items);t.items.forEach(t=>{!t.type&&function(t,e,s){if(0===e.length)return!1;if(1===e.length)return e[0].x>t&&e[0].height<=s+1;for(var n of e)if(n.x===t)return!1;return!0}(n,t.items,e)&&(t.annotation=b,t.type=W.CODE,s++)})}),new I({...t,messages:["Detected "+s+" code/quote items."]})}};var rt=class extends st{constructor(){super("Level Lists")}transform(t){var e=0,s=0;return t.pages.forEach(t=>{t.items.filter(t=>t.type===W.LIST).forEach(t=>{var n,o=0;const r={};var a=!1;t.items.forEach(t=>{n?t.x>n?(o++,r[t.x]=o):t.x<n&&(o=r[t.x]):r[t.x]=0,o>0&&(t.words=[new S({string:" ".repeat(3*o)})].concat(t.words),a=!0),n=t.x}),e++,a?(s++,t.annotation=L):t.annotation=v})}),new I({...t,messages:["Modified "+s+" / "+e+" list blocks."]})}},at=class extends w{constructor(){super("To Text Blocks","TextBlock")}transform(t){return t.pages.forEach(t=>{const e=[];t.items.forEach(t=>{const s=t.type?t.type.name:"Unknown";e.push({category:s,text:W.blockToText(t)})}),t.items=e}),new I({...t})}},it=class extends w{constructor(){super("To HTML","String")}transform(t){return t.pages.forEach(t=>{var e="";t.items.forEach(t=>{let s;s="TOC"===t.category?t.text:t.text.replace(/(\r\n|\n|\r)/gm,"\n"),"LIST"!==t.category&&(s=s.split("- ").join("")),"CODE"===t.category&&(s=s.split("<code>").join("").split("</code>").join("")),e+=`<p>${s}</p>\n\n`}),t.items=[e]}),new I({...t})}};async function ht(t,e={}){const{addPageNumbers:s=!1,addCitation:n=!0,liteParseOptions:o={}}=e,r="string"==typeof t?await y(t,{responseType:"arraybuffer",timeout:10}):t;let a;try{({LiteParse:a}=await import("@llamaindex/liteparse"))}catch(u){return{error:`method: 'liteparse' requires the optional @llamaindex/liteparse dependency (Node.js only): ${u.message}`}}const i=new a({quiet:!0,ocrEnabled:!1,ocrFailureFatal:!1,...o});let h;try{h=await i.parse(new Uint8Array(r))}catch(u){return{error:u.message}}const l=h.pages.reduce((t,e)=>{const n=e.text.split(/\n{2,}/).filter(t=>t.trim().length>0).map(t=>{return(e=t,e.replace(/&/g,"&amp;").replace(/</g,"&lt;").replace(/>/g,"&gt;")).replace(/\n/g,"<br/>");var e}).join("</p><p>");return t+`<p id="page-${e.pageNum}">${s?` [${e.pageNum}] `:""}${n}</p>`},"");let c,m;return n&&(m=h.creator||h.producer,c=h.pages[0]?.text.split("\n").find(t=>t.trim().length>0)),{author:m,title:c,html:l,format:"pdf"}}async function lt(t,e={}){const s="string"==typeof t?await y(t,{responseType:"arraybuffer",timeout:10}):t,{LiteParse:n}=await import("@llamaindex/liteparse"),o=await new n({quiet:!0,...e.liteParseOptions}).isComplex(new Uint8Array(s));return{needsOcr:o.some(t=>t.needsOcr),pages:o,reasons:[...new Set(o.flatMap(t=>t.reasons))]}}async function ct(t,e={}){if("liteparse"===e.method){const{method:s,...n}=e;return ht(t,n)}var{addPageNumbers:s=!1,addCitation:n=!0}=e,o="string"==typeof t?await y(t,{responseType:"arraybuffer",timeout:10}):t;let r;try{let{resolvePDFJS:t}=await import("https://cdn.jsdelivr.net/npm/pdfjs-serverless@1.1.0/+esm");const{getDocument:e}=await t();r=await e({data:new Uint8Array(o),useSystemFonts:!0,verbosity:0}).promise}catch(T){return{error:T.message}}const a=[...Array(r.numPages).keys()].map(t=>new g({index:t}));let i,h={};for(let p=1;p<=r.numPages;p++){const t=await r.getPage(p),e=await t.getTextContent();if(!(Object.keys(h).length<10)){i=u(h);break}h=m(h,t.pageNumber-1,e.items)}let l=i?i.pageNum:0;for(let m=1;m<=r.numPages;m++){const t=await r.getPage(m);await t.getOperatorList();const e=t.getViewport({scale:1});let s=await t.getTextContent();i&&t.pageIndex>=i.pageIndex&&(s=p(s,l),l++);const n=s.items.map(t=>{const s=[1,0,0,1,0,0];for(let r=0;r<6;r++)s[r]+=t.transform[r]*e.transform[r%2?3:0],r%2&&(s[r+1]+=t.transform[r]*e.transform[1]);const n=Math.sqrt(s[2]*s[2]+s[3]*s[3]),o=t.height/n;return new d({x:Math.round(t.transform[4]),y:Math.round(t.transform[5]),width:Math.round(t.width),height:Math.round(o<=1?t.height:o),text:t.str,font:t.fontName})});a[t.pageNumber-1].items=n}var c=new I({pages:a});let f;[new U,new q,new R,new V,new Q,new tt,new Z,new nt,new ot,new rt,new at,new it].forEach(t=>{f&&(c=f.completeTransform(c)),c=t.transform(c),f=t});var w=c.pages.reduce((t,e,n)=>t+`<p id="page-${n+1}">${s?` [${n+1}] `:""}${e.items.join('</p><p id="page-'+n+'">')}</p>`,"");if(n){var{Author:x,Title:E}=(await r.getMetadata()).info;E=w.slice(0,400).match(/<h[0-9]>(.*?)<\/h[0-9]>/)?.[1]||E}return{author:x,title:E,html:w,format:"pdf"}}export{ct as convertPDFToHTML,ht as convertPDFToHTMLWithLiteParse,lt as detectPdfNeedsOcr};

package/dist/utils/grab.d.ts

@@ -0,0 +1,8 @@
+/**
+ * Minimal fetch wrapper for grabbing binary/text content, shared by every
+ * parse method (ts-block-algorithm, liteparse) and the OCR-need detector.
+ */
+export declare function grab(url: string, options?: {
+    responseType?: string;
+    timeout?: number;
+}): Promise<string | ArrayBuffer>;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "extract-pdf",
-  "version": "0.1.114",
+  "version": "0.1.115",
   "description": "Convert a PDF (URL or ArrayBuffer) into clean HTML with structural tagging — headings, lists, footnotes, code blocks, bold/italic. Works in Node.js, Cloudflare Workers, and browser environments.",
   "author": "vtempest <grokthiscontact@gmail.com>",
   "license": "rights.institute/PROSPER",
@@ -26,8 +26,9 @@
     "test:coverage": "bun test --coverage --coverage-reporter=lcov --coverage-reporter=text",
     "ship": "npm run build && npx standard-version --release-as patch; rm -f CHANGELOG.md; npm publish"
   },
-  "dependencies": {
-    "grab-url": "^1.6.22"
+  "dependencies": {},
+  "optionalDependencies": {
+    "@llamaindex/liteparse": "^2.10.0"
   },
   "devDependencies": {
     "terser": "^5.46.0",

package/src/detect-needs-ocr.ts

@@ -0,0 +1,52 @@
+/**
+ * Cheap, text-layer-only pass that determines whether a PDF needs OCR (or
+ * other heavy parsing) before committing to a full parse — useful for routing
+ * documents to different pipelines. Backed by LiteParse's `isComplex`, so this
+ * is Node.js only (native napi addon) and requires the optional
+ * `@llamaindex/liteparse` dependency.
+ */
+import { grab } from "./utils/grab";
+import type { LiteParseConfig, PageComplexityStats } from "@llamaindex/liteparse";
+
+export interface DetectPdfNeedsOcrOptions {
+  liteParseOptions?: Partial<LiteParseConfig>;
+}
+
+export interface PdfOcrAssessment {
+  /** True when any page needs OCR or other advanced parsing. */
+  needsOcr: boolean;
+  /** Per-page complexity signals, one entry per page. */
+  pages: PageComplexityStats[];
+  /**
+   * Every distinct reason found across all pages, e.g. `"scanned"`,
+   * `"no-text"`, `"sparse-text"`, `"embedded-images"`, `"garbled"`,
+   * `"vector-text"`. Empty when `needsOcr` is false.
+   */
+  reasons: string[];
+}
+
+/**
+ * Determines whether a PDF (URL or ArrayBuffer) needs OCR to extract usable
+ * text, without running a full parse.
+ * @param pdfURLOrBuffer - URL to a PDF file or buffer from fs.readFile
+ * @category Extract
+ */
+export async function detectPdfNeedsOcr(
+  pdfURLOrBuffer: any,
+  options: DetectPdfNeedsOcrOptions = {},
+): Promise<PdfOcrAssessment> {
+  const buffer =
+    typeof pdfURLOrBuffer === "string"
+      ? await grab(pdfURLOrBuffer, { responseType: "arraybuffer", timeout: 10 })
+      : pdfURLOrBuffer;
+
+  const { LiteParse } = await import("@llamaindex/liteparse");
+  const parser = new LiteParse({ quiet: true, ...options.liteParseOptions });
+  const pages = await parser.isComplex(new Uint8Array(buffer));
+
+  return {
+    needsOcr: pages.some((page) => page.needsOcr),
+    pages,
+    reasons: [...new Set(pages.flatMap((page) => page.reasons))],
+  };
+}

package/src/liteparse-to-html.ts

@@ -0,0 +1,95 @@
+/**
+ * Alternate `ParseMethod` for convertPDFToHTML that delegates to
+ * [LiteParse](https://github.com/run-llama/liteparse) instead of the built-in
+ * ts-block-algorithm pipeline. LiteParse ships a native (napi) addon, so this
+ * path is Node.js only — it will not run in Cloudflare Workers or browsers —
+ * and requires the optional `@llamaindex/liteparse` dependency to be installed.
+ */
+import { grab } from "./utils/grab";
+import type { LiteParseConfig } from "@llamaindex/liteparse";
+
+export interface LiteParseHTMLOptions {
+  addPageNumbers?: boolean;
+  addCitation?: boolean;
+  liteParseOptions?: Partial<LiteParseConfig>;
+}
+
+function escapeHtml(text: string): string {
+  return text.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;");
+}
+
+/**
+ * Converts a PDF (URL or ArrayBuffer) into HTML using LiteParse's spatial text
+ * extraction, mirroring the return shape of `convertPDFToHTML`.
+ * @param pdfURLOrBuffer - URL to a PDF file or buffer from fs.readFile
+ * @param options.addPageNumbers default=false - Adds `[n]` markers at each page boundary
+ * @param options.addCitation default=true - Populates `title`/`author` from PDF metadata
+ * @param options.liteParseOptions - Passed through to the `LiteParse` constructor;
+ *   defaults to `{ ocrEnabled: false, ocrFailureFatal: false }` (use detectPdfNeedsOcr
+ *   to decide when a document is worth re-parsing with `ocrEnabled: true`)
+ * @returns `{ html, title, author, format: "pdf" }`, or `{ error }` on failure
+ * @category Extract
+ */
+export async function convertPDFToHTMLWithLiteParse(
+  pdfURLOrBuffer: any,
+  options: LiteParseHTMLOptions = {},
+) {
+  const { addPageNumbers = false, addCitation = true, liteParseOptions = {} } = options;
+
+  const buffer =
+    typeof pdfURLOrBuffer === "string"
+      ? await grab(pdfURLOrBuffer, { responseType: "arraybuffer", timeout: 10 })
+      : pdfURLOrBuffer;
+
+  let LiteParse: typeof import("@llamaindex/liteparse").LiteParse;
+  try {
+    ({ LiteParse } = await import("@llamaindex/liteparse"));
+  } catch (e: any) {
+    return {
+      error:
+        "method: 'liteparse' requires the optional @llamaindex/liteparse dependency " +
+        `(Node.js only): ${e.message}`,
+    };
+  }
+
+  // Default to no OCR: this method exists for the fast/local path (see
+  // detectPdfNeedsOcr for routing scanned/complex pages elsewhere), and OCR
+  // requires downloading tessdata on first use, which is slow and can fail
+  // offline. `ocrFailureFatal: false` keeps already-recovered native text
+  // instead of discarding the whole parse if a caller opts OCR back in and it
+  // fails on some pages.
+  const parser = new LiteParse({
+    quiet: true,
+    ocrEnabled: false,
+    ocrFailureFatal: false,
+    ...liteParseOptions,
+  });
+
+  let result;
+  try {
+    result = await parser.parse(new Uint8Array(buffer));
+  } catch (e: any) {
+    return { error: e.message };
+  }
+
+  const html = result.pages.reduce((acc, page) => {
+    const body = page.text
+      .split(/\n{2,}/)
+      .filter((paragraph) => paragraph.trim().length > 0)
+      .map((paragraph) => escapeHtml(paragraph).replace(/\n/g, "<br/>"))
+      .join("</p><p>");
+    return (
+      acc +
+      `<p id="page-${page.pageNum}">${addPageNumbers ? ` [${page.pageNum}] ` : ""}${body}</p>`
+    );
+  }, "");
+
+  let title: string | undefined;
+  let author: string | undefined;
+  if (addCitation) {
+    author = result.creator || result.producer;
+    title = result.pages[0]?.text.split("\n").find((line) => line.trim().length > 0);
+  }
+
+  return { author, title, html, format: "pdf" };
+}

package/src/pdf-to-html.ts

@@ -9,34 +9,7 @@ import {
 } from "./utils/page-number-functions";
 import TextItem from "./models/text-item";
 import Page from "./models/page";
-
-/**
- * Fetch wrapper for grabbing binary content
- */
-async function grab(url: string, options: { responseType?: string; timeout?: number } = {}) {
-  const timeout = options.timeout ? options.timeout * 1000 : 10000;
-  const controller = new AbortController();
-  const timeoutId = setTimeout(() => controller.abort(), timeout);
-
-  try {
-    const response = await fetch(url, {
-      signal: controller.signal,
-    });
-    clearTimeout(timeoutId);
-
-    if (!response.ok) {
-      throw new Error(`HTTP ${response.status}`);
-    }
-
-    if (options.responseType === "arraybuffer") {
-      return await response.arrayBuffer();
-    }
-    return await response.text();
-  } catch (error) {
-    clearTimeout(timeoutId);
-    throw error;
-  }
-}
+import { grab } from "./utils/grab";
 
 import CalculateGlobalStats from "./transforms/calculate-global-stats";
 import CompactLines from "./transforms/line-item/compact-lines";
@@ -52,6 +25,21 @@ import DetectListLevels from "./transforms/block/detect-list-levels";
 import ToTextBlocks from "./transforms/to-text-blocks";
 import ToHTML from "./transforms/to-html";
 import ParseResult from "./models/parse-result";
+import {
+  convertPDFToHTMLWithLiteParse,
+  type LiteParseHTMLOptions,
+} from "./liteparse-to-html";
+
+/**
+ * Which parsing engine {@link convertPDFToHTML} runs.
+ * - `"ts-block-algorithm"` (default) — the pure-TS pipeline in this file: groups
+ *   pdfjs text spans into lines/blocks and renders headings/lists/code blocks.
+ *   Works in Node.js, Cloudflare Workers, and browsers.
+ * - `"liteparse"` — delegates to [LiteParse](https://github.com/run-llama/liteparse),
+ *   a native/OCR-capable parser. Node.js only (ships a napi addon), and requires
+ *   the optional `@llamaindex/liteparse` dependency to be installed.
+ */
+export type ParseMethod = "ts-block-algorithm" | "liteparse";
 
 /**
  * Extracts formatted text from PDF with parsing of linebreaks ,
@@ -67,6 +55,8 @@ import ParseResult from "./models/parse-result";
  * @param {Object} [options]
  * @param {boolean} options.addPageNumbers default=false - Adds  #  to end of each page
  * @param {boolean} options.removePageHeaders default=true - Removes repeated headers found on each page
+ * @param {ParseMethod} options.method default="ts-block-algorithm" - Parsing engine to use;
+ *   `"liteparse"` delegates to LiteParse (Node.js only, see {@link ParseMethod})
  * @returns {string|Object} HTML formatted text
  * @category Extract
  * @author [vtempest (2025)](https://github.com/vtempest),
@@ -75,8 +65,17 @@ import ParseResult from "./models/parse-result";
  */
 export async function convertPDFToHTML(
   pdfURLOrBuffer: any,
-  options: { addPageNumbers?: boolean; addCitation?: boolean } = {},
+  options: {
+    addPageNumbers?: boolean;
+    addCitation?: boolean;
+    method?: ParseMethod;
+  } & Pick<LiteParseHTMLOptions, "liteParseOptions"> = {},
 ) {
+  if (options.method === "liteparse") {
+    const { method: _method, ...liteParseOptions } = options;
+    return convertPDFToHTMLWithLiteParse(pdfURLOrBuffer, liteParseOptions);
+  }
+
   // try {
   var { addPageNumbers = false, addCitation = true } = options;
 
@@ -222,4 +221,11 @@ export async function convertPDFToHTML(
   return { author, title, html, format: "pdf" };
 }
 
+export { convertPDFToHTMLWithLiteParse };
+export type { LiteParseHTMLOptions } from "./liteparse-to-html";
+export { detectPdfNeedsOcr } from "./detect-needs-ocr";
+export type {
+  DetectPdfNeedsOcrOptions,
+  PdfOcrAssessment,
+} from "./detect-needs-ocr";
 

package/src/utils/grab.ts

@@ -0,0 +1,31 @@
+/**
+ * Minimal fetch wrapper for grabbing binary/text content, shared by every
+ * parse method (ts-block-algorithm, liteparse) and the OCR-need detector.
+ */
+export async function grab(
+  url: string,
+  options: { responseType?: string; timeout?: number } = {},
+) {
+  const timeout = options.timeout ? options.timeout * 1000 : 10000;
+  const controller = new AbortController();
+  const timeoutId = setTimeout(() => controller.abort(), timeout);
+
+  try {
+    const response = await fetch(url, {
+      signal: controller.signal,
+    });
+    clearTimeout(timeoutId);
+
+    if (!response.ok) {
+      throw new Error(`HTTP ${response.status}`);
+    }
+
+    if (options.responseType === "arraybuffer") {
+      return await response.arrayBuffer();
+    }
+    return await response.text();
+  } catch (error) {
+    clearTimeout(timeoutId);
+    throw error;
+  }
+}