@arkyc/ocr 1.0.0

package/dist/parsers/mrz.d.mts.map

@@ -0,0 +1 @@
+{"version":3,"file":"mrz.d.mts","names":[],"sources":["../../src/parsers/mrz.ts"],"mappings":";;;;iBAkKgB,SAAA,IAAa,cAAc"}

package/dist/parsers/mrz.mjs

@@ -0,0 +1,149 @@
+//#region src/parsers/mrz.ts
+/**
+* Parser for the ICAO 9303 Machine-Readable Zone (MRZ) found on passports (TD3),
+* ID/residence cards (TD1) and some visas (TD2). The MRZ is standardized across
+* countries, which makes this a strong default parser. Check digits are verified
+* to derive a confidence score.
+*/
+const FILLER = "<";
+/** Numeric value of an MRZ character for check-digit computation. */
+function charValue(ch) {
+	if (ch === FILLER) return 0;
+	if (ch >= "0" && ch <= "9") return ch.charCodeAt(0) - 48;
+	if (ch >= "A" && ch <= "Z") return ch.charCodeAt(0) - 55;
+	return 0;
+}
+/** ICAO 9303 check digit over a field (weights 7,3,1 repeating, mod 10). */
+function checkDigit(field) {
+	const weights = [
+		7,
+		3,
+		1
+	];
+	let sum = 0;
+	for (let i = 0; i < field.length; i += 1) sum += charValue(field[i]) * weights[i % 3];
+	return sum % 10;
+}
+/** Whether `field`'s trailing check digit matches (`expected` is the digit char). */
+function checkOk(field, expected) {
+	if (!/^[0-9]$/.test(expected)) return false;
+	return checkDigit(field) === Number(expected);
+}
+/** Convert an MRZ `YYMMDD` to ISO `YYYY-MM-DD`, or `undefined` if implausible. */
+function mrzDate(yymmdd, kind) {
+	if (!/^[0-9]{6}$/.test(yymmdd)) return void 0;
+	const yy = Number(yymmdd.slice(0, 2));
+	const mm = yymmdd.slice(2, 4);
+	const dd = yymmdd.slice(4, 6);
+	if (Number(mm) < 1 || Number(mm) > 12 || Number(dd) < 1 || Number(dd) > 31) return void 0;
+	const pivot = 70;
+	return `${(kind === "dob" ? yy > pivot ? 1900 : 2e3 : yy < pivot ? 2e3 : 1900) + yy}-${mm}-${dd}`;
+}
+/** Split an MRZ name field (`SURNAME<<GIVEN<NAMES`) into structured names. */
+function parseNameField(field) {
+	const [surnameRaw = "", givenRaw = ""] = field.split("<<");
+	const surname = surnameRaw.replace(/</g, " ").trim();
+	const given = givenRaw.replace(/</g, " ").trim();
+	const firstName = given.split(/\s+/)[0] || void 0;
+	const fullName = [given, surname].filter(Boolean).join(" ") || void 0;
+	return {
+		firstName,
+		lastName: surname || void 0,
+		fullName
+	};
+}
+/** Normalize raw OCR text into candidate MRZ lines (uppercase, MRZ charset). */
+function mrzLines(input) {
+	return (input.lines ?? input.text.split(/\r?\n/)).map((l) => l.toUpperCase().replace(/\s+/g, "").replace(/[^A-Z0-9<]/g, "")).filter((l) => l.length >= 28 && /^[A-Z0-9<]+$/.test(l) && l.includes(FILLER));
+}
+function field(raw, expected) {
+	return {
+		value: raw.replace(/</g, ""),
+		ok: checkOk(raw, expected)
+	};
+}
+function build(name, documentNumber, nationality, dob, expiry) {
+	const fields = {
+		...name,
+		documentNumber: documentNumber.value || void 0,
+		nationality: nationality.replace(/</g, "") || void 0,
+		dateOfBirth: mrzDate(dob.value, "dob"),
+		expiryDate: mrzDate(expiry.value, "expiry")
+	};
+	const checks = [
+		documentNumber.ok,
+		dob.ok,
+		expiry.ok
+	];
+	return {
+		fields,
+		confidence: .5 + .5 * (checks.filter(Boolean).length / checks.length),
+		raw: {
+			format: "mrz",
+			checks: {
+				documentNumber: documentNumber.ok,
+				dob: dob.ok,
+				expiry: expiry.ok
+			}
+		}
+	};
+}
+/** TD3 (passport): two 44-char lines. */
+function parseTd3(l1, l2) {
+	return build(parseNameField(l1.slice(5)), field(l2.slice(0, 9), l2[9]), l2.slice(10, 13), {
+		value: l2.slice(13, 19),
+		ok: checkOk(l2.slice(13, 19), l2[19])
+	}, {
+		value: l2.slice(21, 27),
+		ok: checkOk(l2.slice(21, 27), l2[27])
+	});
+}
+/** TD2: two 36-char lines. */
+function parseTd2(l1, l2) {
+	return build(parseNameField(l1.slice(5)), field(l2.slice(0, 9), l2[9]), l2.slice(10, 13), {
+		value: l2.slice(13, 19),
+		ok: checkOk(l2.slice(13, 19), l2[19])
+	}, {
+		value: l2.slice(21, 27),
+		ok: checkOk(l2.slice(21, 27), l2[27])
+	});
+}
+/** TD1 (ID card): three 30-char lines. */
+function parseTd1(l1, l2, l3) {
+	const documentNumber = field(l1.slice(5, 14), l1[14]);
+	const dob = {
+		value: l2.slice(0, 6),
+		ok: checkOk(l2.slice(0, 6), l2[6])
+	};
+	const expiry = {
+		value: l2.slice(8, 14),
+		ok: checkOk(l2.slice(8, 14), l2[14])
+	};
+	const nationality = l2.slice(15, 18);
+	return build(parseNameField(l3), documentNumber, nationality, dob, expiry);
+}
+var MrzParser = class {
+	name = "mrz";
+	parse(input) {
+		const lines = mrzLines(input);
+		if (lines.length < 2) return null;
+		const td1 = lines.filter((l) => l.length >= 28 && l.length <= 32);
+		if (td1.length >= 3) return parseTd1(pad(td1[0], 30), pad(td1[1], 30), td1[2]);
+		const td3 = lines.filter((l) => l.length >= 42 && l.length <= 46);
+		if (td3.length >= 2) return parseTd3(pad(td3[0], 44), pad(td3[1], 44));
+		const td2 = lines.filter((l) => l.length >= 34 && l.length <= 38);
+		if (td2.length >= 2) return parseTd2(pad(td2[0], 36), pad(td2[1], 36));
+		return null;
+	}
+};
+function pad(line, len) {
+	return line.length >= len ? line : line + FILLER.repeat(len - line.length);
+}
+/** The MRZ parser — a country-agnostic default for machine-readable documents. */
+function mrzParser() {
+	return new MrzParser();
+}
+//#endregion
+export { mrzParser };
+
+//# sourceMappingURL=mrz.mjs.map

package/dist/parsers/mrz.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"mrz.mjs","names":[],"sources":["../../src/parsers/mrz.ts"],"sourcesContent":["import type { IsoDate, OcrFields } from '@arkyc/types'\nimport type { DocumentParser, ParseInput, ParseOutput } from './types'\n\n/**\n * Parser for the ICAO 9303 Machine-Readable Zone (MRZ) found on passports (TD3),\n * ID/residence cards (TD1) and some visas (TD2). The MRZ is standardized across\n * countries, which makes this a strong default parser. Check digits are verified\n * to derive a confidence score.\n */\n\nconst FILLER = '<'\n\n/** Numeric value of an MRZ character for check-digit computation. */\nfunction charValue(ch: string): number {\n  if (ch === FILLER) return 0\n  if (ch >= '0' && ch <= '9') return ch.charCodeAt(0) - 48\n  if (ch >= 'A' && ch <= 'Z') return ch.charCodeAt(0) - 55 // A=10 … Z=35\n  return 0\n}\n\n/** ICAO 9303 check digit over a field (weights 7,3,1 repeating, mod 10). */\nfunction checkDigit(field: string): number {\n  const weights = [7, 3, 1]\n  let sum = 0\n  for (let i = 0; i < field.length; i += 1) sum += charValue(field[i]!) * weights[i % 3]!\n  return sum % 10\n}\n\n/** Whether `field`'s trailing check digit matches (`expected` is the digit char). */\nfunction checkOk(field: string, expected: string): boolean {\n  if (!/^[0-9]$/.test(expected)) return false\n  return checkDigit(field) === Number(expected)\n}\n\n/** Convert an MRZ `YYMMDD` to ISO `YYYY-MM-DD`, or `undefined` if implausible. */\nfunction mrzDate(yymmdd: string, kind: 'dob' | 'expiry'): IsoDate | undefined {\n  if (!/^[0-9]{6}$/.test(yymmdd)) return undefined\n  const yy = Number(yymmdd.slice(0, 2))\n  const mm = yymmdd.slice(2, 4)\n  const dd = yymmdd.slice(4, 6)\n  if (Number(mm) < 1 || Number(mm) > 12 || Number(dd) < 1 || Number(dd) > 31) return undefined\n  // DOB is in the past; expiry is generally in the present/future.\n  const pivot = 70\n  const century = kind === 'dob' ? (yy > pivot ? 1900 : 2000) : yy < pivot ? 2000 : 1900\n  return `${century + yy}-${mm}-${dd}` as IsoDate\n}\n\n/** Split an MRZ name field (`SURNAME<<GIVEN<NAMES`) into structured names. */\nfunction parseNameField(field: string): Pick<OcrFields, 'firstName' | 'lastName' | 'fullName'> {\n  const [surnameRaw = '', givenRaw = ''] = field.split('<<')\n  const surname = surnameRaw.replace(/</g, ' ').trim()\n  const given = givenRaw.replace(/</g, ' ').trim()\n  const firstName = given.split(/\\s+/)[0] || undefined\n  const fullName = [given, surname].filter(Boolean).join(' ') || undefined\n  return { firstName, lastName: surname || undefined, fullName }\n}\n\n/** Normalize raw OCR text into candidate MRZ lines (uppercase, MRZ charset). */\nfunction mrzLines(input: ParseInput): string[] {\n  const lines = input.lines ?? input.text.split(/\\r?\\n/)\n  return lines\n    .map((l) =>\n      l\n        .toUpperCase()\n        .replace(/\\s+/g, '')\n        .replace(/[^A-Z0-9<]/g, ''),\n    )\n    .filter((l) => l.length >= 28 && /^[A-Z0-9<]+$/.test(l) && l.includes(FILLER))\n}\n\n/** A field's value with whether its check digit validated. */\ninterface Checked {\n  value: string\n  ok: boolean\n}\n\nfunction field(raw: string, expected: string): Checked {\n  return { value: raw.replace(/</g, ''), ok: checkOk(raw, expected) }\n}\n\nfunction build(\n  name: ReturnType<typeof parseNameField>,\n  documentNumber: Checked,\n  nationality: string,\n  dob: { value: string; ok: boolean },\n  expiry: { value: string; ok: boolean },\n): ParseOutput {\n  const fields: OcrFields = {\n    ...name,\n    documentNumber: documentNumber.value || undefined,\n    nationality: nationality.replace(/</g, '') || undefined,\n    dateOfBirth: mrzDate(dob.value, 'dob'),\n    expiryDate: mrzDate(expiry.value, 'expiry'),\n  }\n  const checks = [documentNumber.ok, dob.ok, expiry.ok]\n  const passed = checks.filter(Boolean).length\n  const confidence = 0.5 + 0.5 * (passed / checks.length)\n  return {\n    fields,\n    confidence,\n    raw: { format: 'mrz', checks: { documentNumber: documentNumber.ok, dob: dob.ok, expiry: expiry.ok } },\n  }\n}\n\n/** TD3 (passport): two 44-char lines. */\nfunction parseTd3(l1: string, l2: string): ParseOutput {\n  const name = parseNameField(l1.slice(5))\n  const documentNumber = field(l2.slice(0, 9), l2[9]!)\n  const nationality = l2.slice(10, 13)\n  const dob = { value: l2.slice(13, 19), ok: checkOk(l2.slice(13, 19), l2[19]!) }\n  const expiry = { value: l2.slice(21, 27), ok: checkOk(l2.slice(21, 27), l2[27]!) }\n  return build(name, documentNumber, nationality, dob, expiry)\n}\n\n/** TD2: two 36-char lines. */\nfunction parseTd2(l1: string, l2: string): ParseOutput {\n  const name = parseNameField(l1.slice(5))\n  const documentNumber = field(l2.slice(0, 9), l2[9]!)\n  const nationality = l2.slice(10, 13)\n  const dob = { value: l2.slice(13, 19), ok: checkOk(l2.slice(13, 19), l2[19]!) }\n  const expiry = { value: l2.slice(21, 27), ok: checkOk(l2.slice(21, 27), l2[27]!) }\n  return build(name, documentNumber, nationality, dob, expiry)\n}\n\n/** TD1 (ID card): three 30-char lines. */\nfunction parseTd1(l1: string, l2: string, l3: string): ParseOutput {\n  const documentNumber = field(l1.slice(5, 14), l1[14]!)\n  const dob = { value: l2.slice(0, 6), ok: checkOk(l2.slice(0, 6), l2[6]!) }\n  const expiry = { value: l2.slice(8, 14), ok: checkOk(l2.slice(8, 14), l2[14]!) }\n  const nationality = l2.slice(15, 18)\n  const name = parseNameField(l3)\n  return build(name, documentNumber, nationality, dob, expiry)\n}\n\nclass MrzParser implements DocumentParser {\n  readonly name = 'mrz'\n\n  parse(input: ParseInput): ParseOutput | null {\n    const lines = mrzLines(input)\n    if (lines.length < 2) return null\n\n    // TD1: three ~30-char lines.\n    const td1 = lines.filter((l) => l.length >= 28 && l.length <= 32)\n    if (td1.length >= 3) return parseTd1(pad(td1[0]!, 30), pad(td1[1]!, 30), td1[2]!)\n\n    // TD3: two ~44-char lines.\n    const td3 = lines.filter((l) => l.length >= 42 && l.length <= 46)\n    if (td3.length >= 2) return parseTd3(pad(td3[0]!, 44), pad(td3[1]!, 44))\n\n    // TD2: two ~36-char lines.\n    const td2 = lines.filter((l) => l.length >= 34 && l.length <= 38)\n    if (td2.length >= 2) return parseTd2(pad(td2[0]!, 36), pad(td2[1]!, 36))\n\n    return null\n  }\n}\n\nfunction pad(line: string, len: number): string {\n  return line.length >= len ? line : line + FILLER.repeat(len - line.length)\n}\n\n/** The MRZ parser — a country-agnostic default for machine-readable documents. */\nexport function mrzParser(): DocumentParser {\n  return new MrzParser()\n}\n"],"mappings":";;;;;;;AAUA,MAAM,SAAS;;AAGf,SAAS,UAAU,IAAoB;CACrC,IAAI,OAAO,QAAQ,OAAO;CAC1B,IAAI,MAAM,OAAO,MAAM,KAAK,OAAO,GAAG,WAAW,CAAC,IAAI;CACtD,IAAI,MAAM,OAAO,MAAM,KAAK,OAAO,GAAG,WAAW,CAAC,IAAI;CACtD,OAAO;AACT;;AAGA,SAAS,WAAW,OAAuB;CACzC,MAAM,UAAU;EAAC;EAAG;EAAG;CAAC;CACxB,IAAI,MAAM;CACV,KAAK,IAAI,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK,GAAG,OAAO,UAAU,MAAM,EAAG,IAAI,QAAQ,IAAI;CACpF,OAAO,MAAM;AACf;;AAGA,SAAS,QAAQ,OAAe,UAA2B;CACzD,IAAI,CAAC,UAAU,KAAK,QAAQ,GAAG,OAAO;CACtC,OAAO,WAAW,KAAK,MAAM,OAAO,QAAQ;AAC9C;;AAGA,SAAS,QAAQ,QAAgB,MAA6C;CAC5E,IAAI,CAAC,aAAa,KAAK,MAAM,GAAG,OAAO,KAAA;CACvC,MAAM,KAAK,OAAO,OAAO,MAAM,GAAG,CAAC,CAAC;CACpC,MAAM,KAAK,OAAO,MAAM,GAAG,CAAC;CAC5B,MAAM,KAAK,OAAO,MAAM,GAAG,CAAC;CAC5B,IAAI,OAAO,EAAE,IAAI,KAAK,OAAO,EAAE,IAAI,MAAM,OAAO,EAAE,IAAI,KAAK,OAAO,EAAE,IAAI,IAAI,OAAO,KAAA;CAEnF,MAAM,QAAQ;CAEd,OAAO,IADS,SAAS,QAAS,KAAK,QAAQ,OAAO,MAAQ,KAAK,QAAQ,MAAO,QAC9D,GAAG,GAAG,GAAG,GAAG;AAClC;;AAGA,SAAS,eAAe,OAAuE;CAC7F,MAAM,CAAC,aAAa,IAAI,WAAW,MAAM,MAAM,MAAM,IAAI;CACzD,MAAM,UAAU,WAAW,QAAQ,MAAM,GAAG,CAAC,CAAC,KAAK;CACnD,MAAM,QAAQ,SAAS,QAAQ,MAAM,GAAG,CAAC,CAAC,KAAK;CAC/C,MAAM,YAAY,MAAM,MAAM,KAAK,CAAC,CAAC,MAAM,KAAA;CAC3C,MAAM,WAAW,CAAC,OAAO,OAAO,CAAC,CAAC,OAAO,OAAO,CAAC,CAAC,KAAK,GAAG,KAAK,KAAA;CAC/D,OAAO;EAAE;EAAW,UAAU,WAAW,KAAA;EAAW;CAAS;AAC/D;;AAGA,SAAS,SAAS,OAA6B;CAE7C,QADc,MAAM,SAAS,MAAM,KAAK,MAAM,OAAO,EAAA,CAElD,KAAK,MACJ,EACG,YAAY,CAAC,CACb,QAAQ,QAAQ,EAAE,CAAC,CACnB,QAAQ,eAAe,EAAE,CAC9B,CAAC,CACA,QAAQ,MAAM,EAAE,UAAU,MAAM,eAAe,KAAK,CAAC,KAAK,EAAE,SAAS,MAAM,CAAC;AACjF;AAQA,SAAS,MAAM,KAAa,UAA2B;CACrD,OAAO;EAAE,OAAO,IAAI,QAAQ,MAAM,EAAE;EAAG,IAAI,QAAQ,KAAK,QAAQ;CAAE;AACpE;AAEA,SAAS,MACP,MACA,gBACA,aACA,KACA,QACa;CACb,MAAM,SAAoB;EACxB,GAAG;EACH,gBAAgB,eAAe,SAAS,KAAA;EACxC,aAAa,YAAY,QAAQ,MAAM,EAAE,KAAK,KAAA;EAC9C,aAAa,QAAQ,IAAI,OAAO,KAAK;EACrC,YAAY,QAAQ,OAAO,OAAO,QAAQ;CAC5C;CACA,MAAM,SAAS;EAAC,eAAe;EAAI,IAAI;EAAI,OAAO;CAAE;CAGpD,OAAO;EACL;EACA,YAHiB,KAAM,MADV,OAAO,OAAO,OAAO,CAAC,CAAC,SACG,OAAO;EAI9C,KAAK;GAAE,QAAQ;GAAO,QAAQ;IAAE,gBAAgB,eAAe;IAAI,KAAK,IAAI;IAAI,QAAQ,OAAO;GAAG;EAAE;CACtG;AACF;;AAGA,SAAS,SAAS,IAAY,IAAyB;CAMrD,OAAO,MALM,eAAe,GAAG,MAAM,CAAC,CAKtB,GAJO,MAAM,GAAG,MAAM,GAAG,CAAC,GAAG,GAAG,EAIhB,GAHZ,GAAG,MAAM,IAAI,EAGY,GAAG;EAFlC,OAAO,GAAG,MAAM,IAAI,EAAE;EAAG,IAAI,QAAQ,GAAG,MAAM,IAAI,EAAE,GAAG,GAAG,GAAI;CAE1B,GAAG;EADpC,OAAO,GAAG,MAAM,IAAI,EAAE;EAAG,IAAI,QAAQ,GAAG,MAAM,IAAI,EAAE,GAAG,GAAG,GAAI;CACrB,CAAC;AAC7D;;AAGA,SAAS,SAAS,IAAY,IAAyB;CAMrD,OAAO,MALM,eAAe,GAAG,MAAM,CAAC,CAKtB,GAJO,MAAM,GAAG,MAAM,GAAG,CAAC,GAAG,GAAG,EAIhB,GAHZ,GAAG,MAAM,IAAI,EAGY,GAAG;EAFlC,OAAO,GAAG,MAAM,IAAI,EAAE;EAAG,IAAI,QAAQ,GAAG,MAAM,IAAI,EAAE,GAAG,GAAG,GAAI;CAE1B,GAAG;EADpC,OAAO,GAAG,MAAM,IAAI,EAAE;EAAG,IAAI,QAAQ,GAAG,MAAM,IAAI,EAAE,GAAG,GAAG,GAAI;CACrB,CAAC;AAC7D;;AAGA,SAAS,SAAS,IAAY,IAAY,IAAyB;CACjE,MAAM,iBAAiB,MAAM,GAAG,MAAM,GAAG,EAAE,GAAG,GAAG,GAAI;CACrD,MAAM,MAAM;EAAE,OAAO,GAAG,MAAM,GAAG,CAAC;EAAG,IAAI,QAAQ,GAAG,MAAM,GAAG,CAAC,GAAG,GAAG,EAAG;CAAE;CACzE,MAAM,SAAS;EAAE,OAAO,GAAG,MAAM,GAAG,EAAE;EAAG,IAAI,QAAQ,GAAG,MAAM,GAAG,EAAE,GAAG,GAAG,GAAI;CAAE;CAC/E,MAAM,cAAc,GAAG,MAAM,IAAI,EAAE;CAEnC,OAAO,MADM,eAAe,EACZ,GAAG,gBAAgB,aAAa,KAAK,MAAM;AAC7D;AAEA,IAAM,YAAN,MAA0C;CACxC,OAAgB;CAEhB,MAAM,OAAuC;EAC3C,MAAM,QAAQ,SAAS,KAAK;EAC5B,IAAI,MAAM,SAAS,GAAG,OAAO;EAG7B,MAAM,MAAM,MAAM,QAAQ,MAAM,EAAE,UAAU,MAAM,EAAE,UAAU,EAAE;EAChE,IAAI,IAAI,UAAU,GAAG,OAAO,SAAS,IAAI,IAAI,IAAK,EAAE,GAAG,IAAI,IAAI,IAAK,EAAE,GAAG,IAAI,EAAG;EAGhF,MAAM,MAAM,MAAM,QAAQ,MAAM,EAAE,UAAU,MAAM,EAAE,UAAU,EAAE;EAChE,IAAI,IAAI,UAAU,GAAG,OAAO,SAAS,IAAI,IAAI,IAAK,EAAE,GAAG,IAAI,IAAI,IAAK,EAAE,CAAC;EAGvE,MAAM,MAAM,MAAM,QAAQ,MAAM,EAAE,UAAU,MAAM,EAAE,UAAU,EAAE;EAChE,IAAI,IAAI,UAAU,GAAG,OAAO,SAAS,IAAI,IAAI,IAAK,EAAE,GAAG,IAAI,IAAI,IAAK,EAAE,CAAC;EAEvE,OAAO;CACT;AACF;AAEA,SAAS,IAAI,MAAc,KAAqB;CAC9C,OAAO,KAAK,UAAU,MAAM,OAAO,OAAO,OAAO,OAAO,MAAM,KAAK,MAAM;AAC3E;;AAGA,SAAgB,YAA4B;CAC1C,OAAO,IAAI,UAAU;AACvB"}

package/dist/parsers/registry.d.mts

@@ -0,0 +1,49 @@
+import { DocumentParser, ParseInput, ParseOutput } from "./types.mjs";
+import { DocumentType } from "@arkyc/types";
+
+//#region src/parsers/registry.d.ts
+/** Which stage of the pipeline produced a parse result. */
+type ParseStage = 'mrz' | 'custom' | 'generic';
+/**
+ * A registry of document parsers, resolved in three stages for each document:
+ *
+ *  1. **MRZ** — the machine-readable zone (reliable, check-digit-verified) when present.
+ *  2. **Custom** — country/document-type-specific parsers (registered via {@link register}),
+ *     matched against the user's selection (most specific first) when the MRZ fails.
+ *  3. **Generic** — a best-effort text scraper, the last resort.
+ *
+ * Custom parsers declare the countries and document types they apply to; register
+ * one per document layout the MRZ + generic stages can't read.
+ */
+declare class DocumentParserRegistry {
+  private readonly primary;
+  private readonly fallback;
+  private readonly parsers;
+  /**
+   * @param primary  Tried first for every document (the MRZ parser).
+   * @param fallback Tried last when nothing else extracts anything (generic).
+   */
+  constructor(primary: DocumentParser, fallback: DocumentParser);
+  /** Register a custom country/type-specific parser. Returns `this` for chaining. */
+  register(parser: DocumentParser): this;
+  /** The applicable custom parsers for a selection, most specific first. */
+  candidates(country?: string | null, documentType?: DocumentType | null): DocumentParser[];
+  /**
+   * Parse fields from OCR text: MRZ first, then the matched custom parsers, then
+   * the generic fallback. Always returns a result (empty fields, confidence 0,
+   * when nothing could parse).
+   */
+  parse(input: ParseInput): ParseOutput;
+}
+/**
+ * Create a registry with the MRZ parser as the primary stage and the generic text
+ * parser as the fallback. Register country/document-type-specific custom parsers
+ * on top for documents the MRZ + generic stages can't read.
+ */
+declare function createDocumentParserRegistry(options?: {
+  primary?: DocumentParser;
+  fallback?: DocumentParser;
+}): DocumentParserRegistry;
+//#endregion
+export { DocumentParserRegistry, ParseStage, createDocumentParserRegistry };
+//# sourceMappingURL=registry.d.mts.map

package/dist/parsers/registry.d.mts.map

@@ -0,0 +1 @@
+{"version":3,"file":"registry.d.mts","names":[],"sources":["../../src/parsers/registry.ts"],"mappings":";;;;;KA0BY,UAAA;AAAZ;;;;AAAsB;AAatB;;;;;;AAbA,cAaa,sBAAA;EAAA,iBAQQ,OAAA;EAAA,iBACA,QAAA;EAAA,iBARF,OAAA;EA+BoB;;;;cAxBlB,OAAA,EAAS,cAAA,EACT,QAAA,EAAU,cAAA;;EAI7B,QAAA,CAAS,MAAA,EAAQ,cAAA;EALE;EAWnB,UAAA,CAAW,OAAA,kBAAyB,YAAA,GAAe,YAAA,UAAsB,cAAA;EAVtD;;;;;EAuBnB,KAAA,CAAM,KAAA,EAAO,UAAA,GAAa,WAAA;AAAA;;;;;;iBA8BZ,4BAAA,CACd,OAAA;EAAW,OAAA,GAAU,cAAA;EAAgB,QAAA,GAAW,cAAA;AAAA,IAC/C,sBAAA"}

package/dist/parsers/registry.mjs

@@ -0,0 +1,100 @@
+import { mrzParser } from "./mrz.mjs";
+import { genericTextParser } from "./generic.mjs";
+//#region src/parsers/registry.ts
+function eqCountry(a, b) {
+	return a.trim().toUpperCase() === b.trim().toUpperCase();
+}
+/**
+* Specificity score for a parser against the user's selection, or -1 if it does
+* not apply. A parser scoped to a country scores higher than one scoped only to a
+* document type, so the most specific applicable parser wins.
+*/
+function matchScore(parser, country, documentType) {
+	const scopedCountries = parser.countries?.length ? parser.countries : null;
+	const scopedTypes = parser.documentTypes?.length ? parser.documentTypes : null;
+	const countryOk = !scopedCountries || !!country && scopedCountries.some((c) => eqCountry(c, country));
+	const typeOk = !scopedTypes || !!documentType && scopedTypes.includes(documentType);
+	if (!countryOk || !typeOk) return -1;
+	return (scopedCountries ? 2 : 0) + (scopedTypes ? 1 : 0);
+}
+/**
+* A registry of document parsers, resolved in three stages for each document:
+*
+*  1. **MRZ** — the machine-readable zone (reliable, check-digit-verified) when present.
+*  2. **Custom** — country/document-type-specific parsers (registered via {@link register}),
+*     matched against the user's selection (most specific first) when the MRZ fails.
+*  3. **Generic** — a best-effort text scraper, the last resort.
+*
+* Custom parsers declare the countries and document types they apply to; register
+* one per document layout the MRZ + generic stages can't read.
+*/
+var DocumentParserRegistry = class {
+	primary;
+	fallback;
+	parsers = [];
+	/**
+	* @param primary  Tried first for every document (the MRZ parser).
+	* @param fallback Tried last when nothing else extracts anything (generic).
+	*/
+	constructor(primary, fallback) {
+		this.primary = primary;
+		this.fallback = fallback;
+	}
+	/** Register a custom country/type-specific parser. Returns `this` for chaining. */
+	register(parser) {
+		this.parsers.push(parser);
+		return this;
+	}
+	/** The applicable custom parsers for a selection, most specific first. */
+	candidates(country, documentType) {
+		return this.parsers.map((parser, index) => ({
+			parser,
+			index,
+			score: matchScore(parser, country, documentType)
+		})).filter((entry) => entry.score >= 0).sort((a, b) => b.score - a.score || a.index - b.index).map((entry) => entry.parser);
+	}
+	/**
+	* Parse fields from OCR text: MRZ first, then the matched custom parsers, then
+	* the generic fallback. Always returns a result (empty fields, confidence 0,
+	* when nothing could parse).
+	*/
+	parse(input) {
+		const fromMrz = this.primary.parse(input);
+		if (fromMrz) return tag(fromMrz, this.primary.name, "mrz");
+		for (const parser of this.candidates(input.country, input.documentType)) {
+			const out = parser.parse(input);
+			if (out) return tag(out, parser.name, "custom");
+		}
+		const fromGeneric = this.fallback.parse(input);
+		if (fromGeneric) return tag(fromGeneric, this.fallback.name, "generic");
+		return {
+			fields: {},
+			confidence: 0,
+			raw: { parser: "none" }
+		};
+	}
+};
+/** Annotate a parse result with which parser + stage produced it. */
+function tag(out, parser, stage) {
+	const detail = out.raw && typeof out.raw === "object" ? out.raw : { value: out.raw };
+	return {
+		...out,
+		raw: {
+			parser,
+			stage,
+			...detail
+		}
+	};
+}
+/**
+* Create a registry with the MRZ parser as the primary stage and the generic text
+* parser as the fallback. Register country/document-type-specific custom parsers
+* on top for documents the MRZ + generic stages can't read.
+*/
+function createDocumentParserRegistry(options = {}) {
+	return new DocumentParserRegistry(options.primary ?? mrzParser(), options.fallback ?? genericTextParser());
+}
+//#endregion
+export { DocumentParserRegistry, createDocumentParserRegistry };
+
+//# sourceMappingURL=registry.mjs.map

package/dist/parsers/registry.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"registry.mjs","names":[],"sources":["../../src/parsers/registry.ts"],"sourcesContent":["import type { DocumentType } from '@arkyc/types'\nimport type { DocumentParser, ParseInput, ParseOutput } from './types'\nimport { mrzParser } from './mrz'\nimport { genericTextParser } from './generic'\n\nfunction eqCountry(a: string, b: string): boolean {\n  return a.trim().toUpperCase() === b.trim().toUpperCase()\n}\n\n/**\n * Specificity score for a parser against the user's selection, or -1 if it does\n * not apply. A parser scoped to a country scores higher than one scoped only to a\n * document type, so the most specific applicable parser wins.\n */\nfunction matchScore(parser: DocumentParser, country?: string | null, documentType?: DocumentType | null): number {\n  const scopedCountries = parser.countries?.length ? parser.countries : null\n  const scopedTypes = parser.documentTypes?.length ? parser.documentTypes : null\n\n  const countryOk = !scopedCountries || (!!country && scopedCountries.some((c) => eqCountry(c, country)))\n  const typeOk = !scopedTypes || (!!documentType && scopedTypes.includes(documentType))\n  if (!countryOk || !typeOk) return -1\n\n  return (scopedCountries ? 2 : 0) + (scopedTypes ? 1 : 0)\n}\n\n/** Which stage of the pipeline produced a parse result. */\nexport type ParseStage = 'mrz' | 'custom' | 'generic'\n\n/**\n * A registry of document parsers, resolved in three stages for each document:\n *\n *  1. **MRZ** — the machine-readable zone (reliable, check-digit-verified) when present.\n *  2. **Custom** — country/document-type-specific parsers (registered via {@link register}),\n *     matched against the user's selection (most specific first) when the MRZ fails.\n *  3. **Generic** — a best-effort text scraper, the last resort.\n *\n * Custom parsers declare the countries and document types they apply to; register\n * one per document layout the MRZ + generic stages can't read.\n */\nexport class DocumentParserRegistry {\n  private readonly parsers: DocumentParser[] = []\n\n  /**\n   * @param primary  Tried first for every document (the MRZ parser).\n   * @param fallback Tried last when nothing else extracts anything (generic).\n   */\n  constructor(\n    private readonly primary: DocumentParser,\n    private readonly fallback: DocumentParser,\n  ) {}\n\n  /** Register a custom country/type-specific parser. Returns `this` for chaining. */\n  register(parser: DocumentParser): this {\n    this.parsers.push(parser)\n    return this\n  }\n\n  /** The applicable custom parsers for a selection, most specific first. */\n  candidates(country?: string | null, documentType?: DocumentType | null): DocumentParser[] {\n    return this.parsers\n      .map((parser, index) => ({ parser, index, score: matchScore(parser, country, documentType) }))\n      .filter((entry) => entry.score >= 0)\n      .sort((a, b) => b.score - a.score || a.index - b.index)\n      .map((entry) => entry.parser)\n  }\n\n  /**\n   * Parse fields from OCR text: MRZ first, then the matched custom parsers, then\n   * the generic fallback. Always returns a result (empty fields, confidence 0,\n   * when nothing could parse).\n   */\n  parse(input: ParseInput): ParseOutput {\n    // 1. MRZ — the gold standard when a machine-readable zone is present.\n    const fromMrz = this.primary.parse(input)\n    if (fromMrz) return tag(fromMrz, this.primary.name, 'mrz')\n\n    // 2. Custom country/type-specific parsers, most specific first.\n    for (const parser of this.candidates(input.country, input.documentType)) {\n      const out = parser.parse(input)\n      if (out) return tag(out, parser.name, 'custom')\n    }\n\n    // 3. Generic best-effort extraction.\n    const fromGeneric = this.fallback.parse(input)\n    if (fromGeneric) return tag(fromGeneric, this.fallback.name, 'generic')\n\n    return { fields: {}, confidence: 0, raw: { parser: 'none' } }\n  }\n}\n\n/** Annotate a parse result with which parser + stage produced it. */\nfunction tag(out: ParseOutput, parser: string, stage: ParseStage): ParseOutput {\n  const detail = out.raw && typeof out.raw === 'object' ? out.raw : { value: out.raw }\n  return { ...out, raw: { parser, stage, ...detail } }\n}\n\n/**\n * Create a registry with the MRZ parser as the primary stage and the generic text\n * parser as the fallback. Register country/document-type-specific custom parsers\n * on top for documents the MRZ + generic stages can't read.\n */\nexport function createDocumentParserRegistry(\n  options: { primary?: DocumentParser; fallback?: DocumentParser } = {},\n): DocumentParserRegistry {\n  return new DocumentParserRegistry(options.primary ?? mrzParser(), options.fallback ?? genericTextParser())\n}\n"],"mappings":";;;AAKA,SAAS,UAAU,GAAW,GAAoB;CAChD,OAAO,EAAE,KAAK,CAAC,CAAC,YAAY,MAAM,EAAE,KAAK,CAAC,CAAC,YAAY;AACzD;;;;;;AAOA,SAAS,WAAW,QAAwB,SAAyB,cAA4C;CAC/G,MAAM,kBAAkB,OAAO,WAAW,SAAS,OAAO,YAAY;CACtE,MAAM,cAAc,OAAO,eAAe,SAAS,OAAO,gBAAgB;CAE1E,MAAM,YAAY,CAAC,mBAAoB,CAAC,CAAC,WAAW,gBAAgB,MAAM,MAAM,UAAU,GAAG,OAAO,CAAC;CACrG,MAAM,SAAS,CAAC,eAAgB,CAAC,CAAC,gBAAgB,YAAY,SAAS,YAAY;CACnF,IAAI,CAAC,aAAa,CAAC,QAAQ,OAAO;CAElC,QAAQ,kBAAkB,IAAI,MAAM,cAAc,IAAI;AACxD;;;;;;;;;;;;AAgBA,IAAa,yBAAb,MAAoC;CAQf;CACA;CARnB,UAA6C,CAAC;;;;;CAM9C,YACE,SACA,UACA;EAFiB,KAAA,UAAA;EACA,KAAA,WAAA;CAChB;;CAGH,SAAS,QAA8B;EACrC,KAAK,QAAQ,KAAK,MAAM;EACxB,OAAO;CACT;;CAGA,WAAW,SAAyB,cAAsD;EACxF,OAAO,KAAK,QACT,KAAK,QAAQ,WAAW;GAAE;GAAQ;GAAO,OAAO,WAAW,QAAQ,SAAS,YAAY;EAAE,EAAE,CAAC,CAC7F,QAAQ,UAAU,MAAM,SAAS,CAAC,CAAC,CACnC,MAAM,GAAG,MAAM,EAAE,QAAQ,EAAE,SAAS,EAAE,QAAQ,EAAE,KAAK,CAAC,CACtD,KAAK,UAAU,MAAM,MAAM;CAChC;;;;;;CAOA,MAAM,OAAgC;EAEpC,MAAM,UAAU,KAAK,QAAQ,MAAM,KAAK;EACxC,IAAI,SAAS,OAAO,IAAI,SAAS,KAAK,QAAQ,MAAM,KAAK;EAGzD,KAAK,MAAM,UAAU,KAAK,WAAW,MAAM,SAAS,MAAM,YAAY,GAAG;GACvE,MAAM,MAAM,OAAO,MAAM,KAAK;GAC9B,IAAI,KAAK,OAAO,IAAI,KAAK,OAAO,MAAM,QAAQ;EAChD;EAGA,MAAM,cAAc,KAAK,SAAS,MAAM,KAAK;EAC7C,IAAI,aAAa,OAAO,IAAI,aAAa,KAAK,SAAS,MAAM,SAAS;EAEtE,OAAO;GAAE,QAAQ,CAAC;GAAG,YAAY;GAAG,KAAK,EAAE,QAAQ,OAAO;EAAE;CAC9D;AACF;;AAGA,SAAS,IAAI,KAAkB,QAAgB,OAAgC;CAC7E,MAAM,SAAS,IAAI,OAAO,OAAO,IAAI,QAAQ,WAAW,IAAI,MAAM,EAAE,OAAO,IAAI,IAAI;CACnF,OAAO;EAAE,GAAG;EAAK,KAAK;GAAE;GAAQ;GAAO,GAAG;EAAO;CAAE;AACrD;;;;;;AAOA,SAAgB,6BACd,UAAmE,CAAC,GAC5C;CACxB,OAAO,IAAI,uBAAuB,QAAQ,WAAW,UAAU,GAAG,QAAQ,YAAY,kBAAkB,CAAC;AAC3G"}

package/dist/parsers/types.d.mts

@@ -0,0 +1,43 @@
+import { DocumentType, OcrFields } from "@arkyc/types";
+
+//#region src/parsers/types.d.ts
+/** The OCR text + context handed to a document parser. */
+interface ParseInput {
+  /** Full OCR text for the document (raw engine output). */
+  text: string;
+  /** Pre-split lines; derived from `text` when omitted. */
+  lines?: string[];
+  /** The document category the user selected (matched against parsers). */
+  documentType?: DocumentType | null;
+  /** The ISO country the user selected (matched against parsers). */
+  country?: string | null;
+}
+/** A parser's structured result. */
+interface ParseOutput {
+  fields: OcrFields;
+  /** Confidence in this parse, 0–1. */
+  confidence: number;
+  /** Anything worth retaining for audit (matched zone, raw lines, …). */
+  raw?: unknown;
+}
+/**
+ * A document parser turns raw OCR text into structured {@link OcrFields}. A parser
+ * declares the countries and document types it applies to; the registry matches
+ * those against the user's selection to pick the most specific parser, and falls
+ * back to the default when none match (or none can parse the text).
+ */
+interface DocumentParser {
+  readonly name: string;
+  /**
+   * ISO-3166 country codes (alpha-2 or alpha-3) this parser applies to. Omit (or
+   * leave empty) to apply to any country.
+   */
+  readonly countries?: readonly string[];
+  /** Document types this parser applies to. Omit/empty to apply to any type. */
+  readonly documentTypes?: readonly DocumentType[];
+  /** Parse fields from the input, or return `null` if it can't handle it. */
+  parse(input: ParseInput): ParseOutput | null;
+}
+//#endregion
+export { DocumentParser, ParseInput, ParseOutput };
+//# sourceMappingURL=types.d.mts.map

package/dist/parsers/types.d.mts.map

@@ -0,0 +1 @@
+{"version":3,"file":"types.d.mts","names":[],"sources":["../../src/parsers/types.ts"],"mappings":";;;;UAGiB,UAAA;EAAA;EAEf,IAAA;;EAEA,KAAA;EAFA;EAIA,YAAA,GAAe,YAAY;EAA3B;EAEA,OAAA;AAAA;;UAIe,WAAA;EACf,MAAA,EAAQ,SAAS;EADS;EAG1B,UAAA;EAFiB;EAIjB,GAAA;AAAA;;;;AAAG;AASL;;UAAiB,cAAA;EAAA,SACN,IAAA;EASI;;;;EAAA,SAJJ,SAAA;EAAA;EAAA,SAEA,aAAA,YAAyB,YAAA;EAAA;EAElC,KAAA,CAAM,KAAA,EAAO,UAAA,GAAa,WAAA;AAAA"}

package/dist/registry.d.mts

@@ -0,0 +1,20 @@
+import { OcrConfig, OcrDriver } from "./types.mjs";
+
+//#region src/registry.d.ts
+/**
+ * Selects an OCR driver from config. Call sites depend only on the
+ * {@link OcrDriver} interface, so swapping `config.driver` changes behaviour
+ * with no other changes.
+ */
+declare class OcrDriverFactory {
+  /**
+   * Resolve the OCR driver named by `config`.
+   *
+   * @param config
+   * @returns
+   */
+  static create(config: OcrConfig): OcrDriver;
+}
+//#endregion
+export { OcrDriverFactory };
+//# sourceMappingURL=registry.d.mts.map

package/dist/registry.d.mts.map

@@ -0,0 +1 @@
+{"version":3,"file":"registry.d.mts","names":[],"sources":["../src/registry.ts"],"mappings":";;;;;AAYA;;;cAAa,gBAAA;EAOJ;;;;;AAAoC;EAApC,OAAA,MAAA,CAAO,MAAA,EAAQ,SAAA,GAAY,SAAS;AAAA"}

package/dist/registry.mjs

@@ -0,0 +1,36 @@
+import { AnthropicOcrDriver } from "./drivers/ai.mjs";
+import { ExternalOcrDriver } from "./drivers/external.mjs";
+import { MockOcrDriver } from "./drivers/mock.mjs";
+import { TesseractOcrDriver } from "./drivers/tesseract.mjs";
+//#region src/registry.ts
+/**
+* Selects an OCR driver from config. Call sites depend only on the
+* {@link OcrDriver} interface, so swapping `config.driver` changes behaviour
+* with no other changes.
+*/
+var OcrDriverFactory = class {
+	/**
+	* Resolve the OCR driver named by `config`.
+	*
+	* @param config
+	* @returns
+	*/
+	static create(config) {
+		switch (config.driver) {
+			case "mock": return new MockOcrDriver();
+			case "tesseract": return new TesseractOcrDriver({ language: config.language });
+			case "external": return new ExternalOcrDriver(config);
+			case "ai": return new AnthropicOcrDriver({
+				apiKey: config.apiKey,
+				model: config.model,
+				baseUrl: config.endpoint,
+				maxEdge: config.maxEdge
+			});
+			default: throw new Error(`Unknown OCR driver: ${config.driver}`);
+		}
+	}
+};
+//#endregion
+export { OcrDriverFactory };
+
+//# sourceMappingURL=registry.mjs.map

package/dist/registry.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"registry.mjs","names":[],"sources":["../src/registry.ts"],"sourcesContent":["import type { OcrConfig, OcrDriver } from './types'\n\nimport { AnthropicOcrDriver } from './drivers/ai'\nimport { ExternalOcrDriver } from './drivers/external'\nimport { MockOcrDriver } from './drivers/mock'\nimport { TesseractOcrDriver } from './drivers/tesseract'\n\n/**\n * Selects an OCR driver from config. Call sites depend only on the\n * {@link OcrDriver} interface, so swapping `config.driver` changes behaviour\n * with no other changes.\n */\nexport class OcrDriverFactory {\n  /**\n   * Resolve the OCR driver named by `config`.\n   *\n   * @param config\n   * @returns\n   */\n  static create(config: OcrConfig): OcrDriver {\n    switch (config.driver) {\n      case 'mock':\n        return new MockOcrDriver()\n      case 'tesseract':\n        return new TesseractOcrDriver({ language: config.language })\n      case 'external':\n        return new ExternalOcrDriver(config)\n      case 'ai':\n        return new AnthropicOcrDriver({\n          apiKey: config.apiKey,\n          model: config.model,\n          baseUrl: config.endpoint,\n          maxEdge: config.maxEdge,\n        })\n      default:\n        throw new Error(`Unknown OCR driver: ${(config as OcrConfig).driver}`)\n    }\n  }\n}\n"],"mappings":";;;;;;;;;;AAYA,IAAa,mBAAb,MAA8B;;;;;;;CAO5B,OAAO,OAAO,QAA8B;EAC1C,QAAQ,OAAO,QAAf;GACE,KAAK,QACH,OAAO,IAAI,cAAc;GAC3B,KAAK,aACH,OAAO,IAAI,mBAAmB,EAAE,UAAU,OAAO,SAAS,CAAC;GAC7D,KAAK,YACH,OAAO,IAAI,kBAAkB,MAAM;GACrC,KAAK,MACH,OAAO,IAAI,mBAAmB;IAC5B,QAAQ,OAAO;IACf,OAAO,OAAO;IACd,SAAS,OAAO;IAChB,SAAS,OAAO;GAClB,CAAC;GACH,SACE,MAAM,IAAI,MAAM,uBAAwB,OAAqB,QAAQ;EACzE;CACF;AACF"}

package/dist/types.d.mts

@@ -0,0 +1,48 @@
+import { DocumentType, OcrResultData } from "@arkyc/types";
+
+//#region src/types.d.ts
+/** The bytes + context handed to an OCR driver for extraction. */
+interface OcrRequest {
+  /** Raw document-front image bytes. */
+  image: Uint8Array;
+  /**
+   * Raw document-back image bytes, when captured. Some documents (TD1 ID cards,
+   * residence permits) print the MRZ on the back, so both sides are read.
+   */
+  backImage?: Uint8Array;
+  /** Optional hint about the document category. */
+  documentType?: DocumentType | null;
+  /** Optional ISO country code hint. */
+  country?: string | null;
+  /**
+   * Optional deterministic signals (used by the `mock` driver and tests to
+   * steer the extracted confidence / expiry). Ignored by real drivers.
+   */
+  hints?: {
+    confidence?: number;
+    expired?: boolean;
+  };
+}
+/** A pluggable OCR provider. */
+interface OcrDriver {
+  readonly name: string;
+  extract(request: OcrRequest): Promise<OcrResultData>;
+}
+/** Identifier for a registered OCR driver. */
+type OcrDriverName = 'mock' | 'tesseract' | 'external' | 'ai';
+/** Configuration selecting + parameterising the active OCR driver. */
+interface OcrConfig {
+  driver: OcrDriverName;
+  /** Base URL for the `external` driver, or API base override for `ai`. */
+  endpoint?: string;
+  apiKey?: string;
+  /** Tesseract language(s), e.g. `eng` (the `tesseract` driver). */
+  language?: string;
+  /** Vision model id (the `ai` driver). */
+  model?: string;
+  /** Longest image edge (px) uploaded by the `ai` driver. */
+  maxEdge?: number;
+}
+//#endregion
+export { OcrConfig, OcrDriver, OcrDriverName, OcrRequest };
+//# sourceMappingURL=types.d.mts.map

package/dist/types.d.mts.map

@@ -0,0 +1 @@
+{"version":3,"file":"types.d.mts","names":[],"sources":["../src/types.ts"],"mappings":";;;;UAGiB,UAAA;EAAA;EAEf,KAAA,EAAO,UAAA;;;;;EAKP,SAAA,GAAY,UAAA;EAEe;EAA3B,YAAA,GAAe,YAAA;EAPR;EASP,OAAA;EAJY;;;;EASZ,KAAA;IAAU,UAAA;IAAqB,OAAA;EAAA;AAAA;AAIjC;AAAA,UAAiB,SAAA;EAAA,SACN,IAAA;EACT,OAAA,CAAQ,OAAA,EAAS,UAAA,GAAa,OAAA,CAAQ,aAAA;AAAA;;KAI5B,aAAA;;UAGK,SAAA;EACf,MAAA,EAAQ,aAAa;EARJ;EAUjB,QAAA;EACA,MAAA;EAXsC;EAatC,QAAA;EAbmD;EAenD,KAAA;EAXuB;EAavB,OAAA;AAAA"}

package/package.json

@@ -0,0 +1,32 @@
+{
+  "name": "@arkyc/ocr",
+  "version": "1.0.0",
+  "description": "Driver-based OCR extraction",
+  "license": "MIT",
+  "type": "module",
+  "main": "./dist/index.mjs",
+  "module": "./dist/index.mjs",
+  "types": "./dist/index.d.mts",
+  "exports": {
+    ".": {
+      "types": "./dist/index.d.mts",
+      "import": "./dist/index.mjs"
+    }
+  },
+  "files": [
+    "dist"
+  ],
+  "dependencies": {
+    "@arkyc/types": "^1.0.0"
+  },
+  "optionalDependencies": {
+    "sharp": "^0.33.5",
+    "tesseract.js": "^5.1.1"
+  },
+  "scripts": {
+    "typecheck": "tsc --noEmit",
+    "test": "vitest run",
+    "lint": "eslint src",
+    "clean": "rm -rf dist"
+  }
+}