meticulous-ui 3.12.0 → 3.12.1

package/index.d.ts

@@ -1488,28 +1488,32 @@ export declare function getQrAsJsonContent(source: QrSource): Promise<QrJsonResu
 
 export interface RecognizeImageResult {
   success: boolean;
-  /** Whether the TextDetector API is available in this browser. */
-  supported?: boolean;
   /** BCP-47 code of the primary language used for recognition. */
   detectedLanguage?: string;
-  /** Full extracted text, newline-separated by detected region. */
+  /** Full extracted text from OCR, newline-separated by detected region. */
   text?: string;
-  /** OCR confidence score 0–100, or null when not available. */
+  /** OCR confidence score 0–100. */
   confidence?: number | null;
-  /** Raw per-region data including bounding boxes and corner points. */
-  rawJson?: Array<{
-    rawValue: string;
-    boundingBox: { x: number; y: number; width: number; height: number };
-    cornerPoints: Array<{ x: number; y: number }>;
-  }>;
+  /**
+   * All readable lines from the OCR output as a key-value object.
+   * Lines matching `ALL-CAPS KEY : value` use the label as the key.
+   * Lines without a recognisable label are assigned generic keys (`param_1`, `param_2`, …).
+   * Continuation lines (e.g. a wrapped address) are appended to the previous key.
+   */
+  fields?: Record<string, string>;
+  /** Raw Tesseract output including per-word bounding boxes and metadata. */
+  rawJson?: object;
   /** Human-readable error when success is false. */
   error?: string;
 }
 
 /**
- * Detect text in a base64 image using the browser-native TextDetector API.
- * No external dependencies or network requests.
- * Supported in Chrome 70+ and Edge 79+; returns supported:false on other browsers.
+ * Runs OCR on a base64 image in all browsers with no npm dependency.
+ * Powered by Tesseract.js loaded on demand from a CDN (fetched once, then
+ * browser-cached). Preprocesses the image (grayscale + 2× upscale + contrast
+ * boost) before recognition for higher accuracy. Returns structured `fields`
+ * with extracted key-value pairs alongside the raw `text`, `confidence` score,
+ * and `rawJson` with full word-level bounding-box metadata.
  */
 export declare function recognizeImageContent(
   base64Image: string,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "meticulous-ui",
-  "version": "3.12.0",
+  "version": "3.12.1",
   "license": "ISC",
   "description": "A comprehensive React UI component library with a wide range of customizable components, icons, colors, and utilities for building modern web applications.",
   "types": "./index.d.ts",

package/utils/getImageContentAsJson.js

@@ -1,4 +1,62 @@
-const l = {
+const b = (l) => new Promise((n, s) => {
+  const o = new Image();
+  o.onload = () => {
+    const r = document.createElement("canvas");
+    r.width = o.width * 2, r.height = o.height * 2;
+    const e = r.getContext("2d");
+    e.filter = "grayscale(1) contrast(1.8)", e.drawImage(o, 0, 0, r.width, r.height), n(r.toDataURL("image/png"));
+  }, o.onerror = s, o.src = l;
+}), u = (l) => {
+  var r;
+  const n = ((r = l.words) != null ? r : []).map((e) => {
+    var t;
+    return { text: (t = e.text) == null ? void 0 : t.trim(), bbox: e.bbox };
+  }).filter((e) => e.text && e.bbox);
+  if (!n.length) return [];
+  const s = [];
+  for (const e of n) {
+    const t = (e.bbox.y0 + e.bbox.y1) / 2, c = (e.bbox.y1 - e.bbox.y0) / 2;
+    let a = s.find((m) => Math.abs(m.cy - t) < c);
+    a || (a = { cy: t, words: [] }, s.push(a)), a.words.push(e);
+  }
+  const o = [];
+  for (const e of s) {
+    const t = e.words.slice().sort((i, x) => i.bbox.x0 - x.bbox.x0), c = t.slice(1).map((i, x) => i.bbox.x0 - t[x].bbox.x1), a = c.length ? c.reduce((i, x) => i + x, 0) / c.length : 0, m = Math.max(a * 3, (t[0].bbox.y1 - t[0].bbox.y0) * 1.5);
+    let d = [t[0]];
+    for (let i = 1; i < t.length; i++)
+      t[i].bbox.x0 - t[i - 1].bbox.x1 > m && (o.push({
+        text: d.map((x) => x.text).join(" "),
+        y: e.cy,
+        x: d[0].bbox.x0
+      }), d = []), d.push(t[i]);
+    o.push({
+      text: d.map((i) => i.text).join(" "),
+      y: e.cy,
+      x: d[0].bbox.x0
+    });
+  }
+  return o.sort((e, t) => e.y - t.y || e.x - t.x), o.map((e) => e.text);
+}, h = (l) => {
+  const n = {};
+  let s = null, o = 0;
+  const r = (e) => {
+    let t = e.replace(/^[^A-Za-z0-9"'(]+/, "").trim();
+    return t = t.replace(/[^A-Za-z0-9"')]+$/, "").trim(), t = t.replace(/^[A-Za-z]{1,3}\s+(?=\d)/, ""), t;
+  };
+  for (const e of l) {
+    const t = r(e.trim());
+    if (!t) {
+      s = null;
+      continue;
+    }
+    const c = t.match(/\b([A-Z][A-Z.]{0,}(?:\s+[A-Z.]+){0,2})\s*[:©]\s*(.+)/);
+    if (c) {
+      const a = c[1].trim();
+      n[a] = c[2].trim(), s = a;
+    } else s && /^[A-Z0-9][A-Z0-9\s.,'\-]+$/.test(t) ? (n[s] += " " + t, s = null) : (/\w{3,}/.test(t) && (n[`param_${++o}`] = t), s = null);
+  }
+  return n;
+}, g = {
   en: "eng",
   hi: "hin",
   fr: "fra",
@@ -17,37 +75,40 @@ const l = {
   vi: "vie",
   th: "tha",
   id: "ind"
-}, d = (t) => t.map((r) => {
-  var e;
-  return (e = l[r]) != null ? e : r;
+}, f = (l) => l.map((n) => {
+  var s;
+  return (s = g[n]) != null ? s : n;
 }).join("+");
-let s = null;
-const u = () => (s || (s = import(
+let p = null;
+const y = () => (p || (p = import(
   /* webpackIgnore: true */
   /* @vite-ignore */
   "https://esm.sh/tesseract.js@5"
-)), s), p = async (t, r = "en") => {
-  var e;
+)), p), w = async (l, n = "en") => {
+  var s;
   try {
-    const a = t.startsWith("data:") ? t : `data:image/png;base64,${t}`, o = r.split("+").filter(Boolean), { createWorker: i } = await u(), c = await i(d(o));
+    const o = l.startsWith("data:") ? l : `data:image/png;base64,${l}`, r = n.split("+").filter(Boolean), e = await b(o), { createWorker: t } = await y(), c = await t(f(r));
     try {
-      const { data: n } = await c.recognize(a);
+      await c.setParameters({ tessedit_pageseg_mode: "3" });
+      const { data: a } = await c.recognize(e), m = a.text.trim();
       return {
         success: !0,
-        detectedLanguage: (e = o[0]) != null ? e : "en",
-        text: n.text.trim(),
-        confidence: n.confidence,
+        detectedLanguage: (s = r[0]) != null ? s : "en",
+        text: m,
+        confidence: a.confidence,
         // 0–100
-        rawJson: n
+        fields: h(u(a)),
+        // structured key-value extraction, split on bbox gaps
+        rawJson: a
         // full metadata: words, blocks, bounding boxes
       };
     } finally {
       await c.terminate();
     }
-  } catch (a) {
-    return { success: !1, error: a.message };
+  } catch (o) {
+    return { success: !1, error: o.message };
   }
 };
 export {
-  p as default
+  w as default
 };