@stll/anonymize-wasm 1.3.0 → 1.4.0

package/dist/constants.d.mts

@@ -0,0 +1,2 @@
+import { a as OPERATOR_TYPES, i as DetectionSource, n as DETECTION_SOURCES, o as OperatorType, r as DETECTOR_PRIORITY, t as DEFAULT_ENTITY_LABELS } from "./constants2.mjs";
+export { DEFAULT_ENTITY_LABELS, DETECTION_SOURCES, DETECTOR_PRIORITY, DetectionSource, OPERATOR_TYPES, OperatorType };

package/dist/constants.mjs

@@ -0,0 +1,85 @@
+//#region src/constants.ts
+/**
+* Runtime-free constants for the anonymization pipeline.
+*
+* This module is the SSR-safe / browser-safe entrypoint:
+* importing it must not pull in `@stll/text-search`,
+* `@stll/anonymize-wasm`, or any other runtime-bearing
+* module. Consumers that only need the static label list,
+* detection-source identifiers, or operator names import
+* from `@stll/anonymize/constants` (or
+* `@stll/anonymize-wasm/constants`) without paying the
+* wasm / regex-set startup cost.
+*
+* `types.ts` re-exports these for back-compat, so existing
+* `import { DEFAULT_ENTITY_LABELS } from "@stll/anonymize"`
+* call sites keep working.
+*/
+/**
+* Source of a detected entity span.
+* Ordered by detection layer in the pipeline.
+*/
+const DETECTION_SOURCES = {
+	TRIGGER: "trigger",
+	REGEX: "regex",
+	DENY_LIST: "deny-list",
+	LEGAL_FORM: "legal-form",
+	GAZETTEER: "gazetteer",
+	NER: "ner",
+	COREFERENCE: "coreference"
+};
+/**
+* Priority levels for detection sources.
+* Higher = more structurally reliable. Used during
+* overlap resolution so deterministic detectors beat
+* probabilistic ones regardless of raw score.
+*/
+const DETECTOR_PRIORITY = {
+	[DETECTION_SOURCES.GAZETTEER]: 5,
+	[DETECTION_SOURCES.TRIGGER]: 4,
+	[DETECTION_SOURCES.LEGAL_FORM]: 3,
+	[DETECTION_SOURCES.REGEX]: 3,
+	[DETECTION_SOURCES.DENY_LIST]: 2,
+	[DETECTION_SOURCES.COREFERENCE]: 2,
+	[DETECTION_SOURCES.NER]: 1
+};
+/**
+* Anonymization operator types. Each operator defines
+* how a confirmed entity is replaced in the output.
+*/
+const OPERATOR_TYPES = ["replace", "redact"];
+/**
+* Canonical entity labels used across the pipeline.
+* NER models may use different native labels; the bench
+* NER wrapper maps model output to these canonical names.
+*
+* These labels are ephemeral: entities are regenerated on
+* every pipeline run and never persisted to the database.
+* Renaming a label here requires no migration.
+*/
+const DEFAULT_ENTITY_LABELS = [
+	"person",
+	"organization",
+	"phone number",
+	"address",
+	"email address",
+	"date",
+	"date of birth",
+	"bank account number",
+	"iban",
+	"tax identification number",
+	"identity card number",
+	"birth number",
+	"national identification number",
+	"social security number",
+	"registration number",
+	"credit card number",
+	"passport number",
+	"monetary amount",
+	"land parcel",
+	"misc"
+];
+//#endregion
+export { DEFAULT_ENTITY_LABELS, DETECTION_SOURCES, DETECTOR_PRIORITY, OPERATOR_TYPES };
+
+//# sourceMappingURL=constants.mjs.map

package/dist/constants.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"constants.mjs","names":[],"sources":["../../src/constants.ts"],"sourcesContent":["/**\n * Runtime-free constants for the anonymization pipeline.\n *\n * This module is the SSR-safe / browser-safe entrypoint:\n * importing it must not pull in `@stll/text-search`,\n * `@stll/anonymize-wasm`, or any other runtime-bearing\n * module. Consumers that only need the static label list,\n * detection-source identifiers, or operator names import\n * from `@stll/anonymize/constants` (or\n * `@stll/anonymize-wasm/constants`) without paying the\n * wasm / regex-set startup cost.\n *\n * `types.ts` re-exports these for back-compat, so existing\n * `import { DEFAULT_ENTITY_LABELS } from \"@stll/anonymize\"`\n * call sites keep working.\n */\n\n/**\n * Source of a detected entity span.\n * Ordered by detection layer in the pipeline.\n */\nexport const DETECTION_SOURCES = {\n  TRIGGER: \"trigger\",\n  REGEX: \"regex\",\n  DENY_LIST: \"deny-list\",\n  LEGAL_FORM: \"legal-form\",\n  GAZETTEER: \"gazetteer\",\n  NER: \"ner\",\n  COREFERENCE: \"coreference\",\n} as const;\n\nexport type DetectionSource =\n  (typeof DETECTION_SOURCES)[keyof typeof DETECTION_SOURCES];\n\n/**\n * Priority levels for detection sources.\n * Higher = more structurally reliable. Used during\n * overlap resolution so deterministic detectors beat\n * probabilistic ones regardless of raw score.\n */\nexport const DETECTOR_PRIORITY = {\n  [DETECTION_SOURCES.GAZETTEER]: 5,\n  [DETECTION_SOURCES.TRIGGER]: 4,\n  [DETECTION_SOURCES.LEGAL_FORM]: 3,\n  [DETECTION_SOURCES.REGEX]: 3,\n  [DETECTION_SOURCES.DENY_LIST]: 2,\n  [DETECTION_SOURCES.COREFERENCE]: 2,\n  [DETECTION_SOURCES.NER]: 1,\n} as const satisfies Record<DetectionSource, number>;\n\n/**\n * Anonymization operator types. Each operator defines\n * how a confirmed entity is replaced in the output.\n */\nexport const OPERATOR_TYPES = [\"replace\", \"redact\"] as const;\n\nexport type OperatorType = (typeof OPERATOR_TYPES)[number];\n\n/**\n * Canonical entity labels used across the pipeline.\n * NER models may use different native labels; the bench\n * NER wrapper maps model output to these canonical names.\n *\n * These labels are ephemeral: entities are regenerated on\n * every pipeline run and never persisted to the database.\n * Renaming a label here requires no migration.\n */\nexport const DEFAULT_ENTITY_LABELS = [\n  \"person\",\n  \"organization\",\n  \"phone number\",\n  \"address\",\n  \"email address\",\n  \"date\",\n  \"date of birth\",\n  \"bank account number\",\n  \"iban\",\n  \"tax identification number\",\n  \"identity card number\",\n  \"birth number\",\n  \"national identification number\",\n  \"social security number\",\n  \"registration number\",\n  \"credit card number\",\n  \"passport number\",\n  \"monetary amount\",\n  \"land parcel\",\n  \"misc\",\n] as const;\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;AAqBA,MAAa,oBAAoB;CAC/B,SAAS;CACT,OAAO;CACP,WAAW;CACX,YAAY;CACZ,WAAW;CACX,KAAK;CACL,aAAa;CACd;;;;;;;AAWD,MAAa,oBAAoB;EAC9B,kBAAkB,YAAY;EAC9B,kBAAkB,UAAU;EAC5B,kBAAkB,aAAa;EAC/B,kBAAkB,QAAQ;EAC1B,kBAAkB,YAAY;EAC9B,kBAAkB,cAAc;EAChC,kBAAkB,MAAM;CAC1B;;;;;AAMD,MAAa,iBAAiB,CAAC,WAAW,SAAS;;;;;;;;;;AAanD,MAAa,wBAAwB;CACnC;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACD"}

package/dist/constants2.d.mts

@@ -0,0 +1,65 @@
+//#region src/constants.d.ts
+/**
+ * Runtime-free constants for the anonymization pipeline.
+ *
+ * This module is the SSR-safe / browser-safe entrypoint:
+ * importing it must not pull in `@stll/text-search`,
+ * `@stll/anonymize-wasm`, or any other runtime-bearing
+ * module. Consumers that only need the static label list,
+ * detection-source identifiers, or operator names import
+ * from `@stll/anonymize/constants` (or
+ * `@stll/anonymize-wasm/constants`) without paying the
+ * wasm / regex-set startup cost.
+ *
+ * `types.ts` re-exports these for back-compat, so existing
+ * `import { DEFAULT_ENTITY_LABELS } from "@stll/anonymize"`
+ * call sites keep working.
+ */
+/**
+ * Source of a detected entity span.
+ * Ordered by detection layer in the pipeline.
+ */
+declare const DETECTION_SOURCES: {
+  readonly TRIGGER: "trigger";
+  readonly REGEX: "regex";
+  readonly DENY_LIST: "deny-list";
+  readonly LEGAL_FORM: "legal-form";
+  readonly GAZETTEER: "gazetteer";
+  readonly NER: "ner";
+  readonly COREFERENCE: "coreference";
+};
+type DetectionSource = (typeof DETECTION_SOURCES)[keyof typeof DETECTION_SOURCES];
+/**
+ * Priority levels for detection sources.
+ * Higher = more structurally reliable. Used during
+ * overlap resolution so deterministic detectors beat
+ * probabilistic ones regardless of raw score.
+ */
+declare const DETECTOR_PRIORITY: {
+  readonly gazetteer: 5;
+  readonly trigger: 4;
+  readonly "legal-form": 3;
+  readonly regex: 3;
+  readonly "deny-list": 2;
+  readonly coreference: 2;
+  readonly ner: 1;
+};
+/**
+ * Anonymization operator types. Each operator defines
+ * how a confirmed entity is replaced in the output.
+ */
+declare const OPERATOR_TYPES: readonly ["replace", "redact"];
+type OperatorType = (typeof OPERATOR_TYPES)[number];
+/**
+ * Canonical entity labels used across the pipeline.
+ * NER models may use different native labels; the bench
+ * NER wrapper maps model output to these canonical names.
+ *
+ * These labels are ephemeral: entities are regenerated on
+ * every pipeline run and never persisted to the database.
+ * Renaming a label here requires no migration.
+ */
+declare const DEFAULT_ENTITY_LABELS: readonly ["person", "organization", "phone number", "address", "email address", "date", "date of birth", "bank account number", "iban", "tax identification number", "identity card number", "birth number", "national identification number", "social security number", "registration number", "credit card number", "passport number", "monetary amount", "land parcel", "misc"];
+//#endregion
+export { OPERATOR_TYPES as a, DetectionSource as i, DETECTION_SOURCES as n, OperatorType as o, DETECTOR_PRIORITY as r, DEFAULT_ENTITY_LABELS as t };
+//# sourceMappingURL=constants2.d.mts.map

package/dist/triggers.en.mjs

@@ -64,6 +64,38 @@ var triggers_en_default = [
 			"tax identification number",
 			"tax id"
 		]
+	},
+	{
+		"id": "en-uk-companies-house",
+		"label": "registration number",
+		"strategy": { "type": "company-id-value" },
+		"triggers": [
+			"Companies House",
+			"Company No.",
+			"Company No",
+			"CRN"
+		],
+		"extensions": ["add-colon", "add-colon-space"],
+		"validations": [{
+			"type": "matches-pattern",
+			"pattern": "^(?:SC|NI|OC|SO|R|FC|GE|GS|IC|IP|LP|NC|NF|NL|NO|NP|NR|NZ|RC|SA|SF|SI|SL|SP|SR|SZ)?\\d{6,8}$"
+		}]
+	},
+	{
+		"id": "en-uk-utr",
+		"label": "tax identification number",
+		"strategy": { "type": "company-id-value" },
+		"triggers": [
+			"UTR",
+			"Unique Tax Reference",
+			"Unique Taxpayer Reference",
+			"Self Assessment UTR"
+		],
+		"extensions": ["add-colon", "add-colon-space"],
+		"validations": [{
+			"type": "matches-pattern",
+			"pattern": "^\\d{10}$"
+		}]
 	}
 ];
 //#endregion

package/dist/wasm.d.mts

@@ -1,29 +1,9 @@
+import { a as OPERATOR_TYPES, i as DetectionSource, n as DETECTION_SOURCES, o as OperatorType, r as DETECTOR_PRIORITY, t as DEFAULT_ENTITY_LABELS } from "./constants2.mjs";
 import { Validator } from "@stll/stdnum";
 import { Match, PatternEntry, TextSearch } from "@stll/text-search";
 import { Tokenizer } from "@huggingface/tokenizers";
 
 //#region src/types.d.ts
-/**
- * Source of a detected entity span.
- * Ordered by detection layer in the pipeline.
- */
-declare const DETECTION_SOURCES: {
-  readonly TRIGGER: "trigger";
-  readonly REGEX: "regex";
-  readonly DENY_LIST: "deny-list";
-  readonly LEGAL_FORM: "legal-form";
-  readonly GAZETTEER: "gazetteer";
-  readonly NER: "ner";
-  readonly COREFERENCE: "coreference";
-};
-type DetectionSource = (typeof DETECTION_SOURCES)[keyof typeof DETECTION_SOURCES];
-/**
- * Priority levels for detection sources.
- * Higher = more structurally reliable. Used during
- * overlap resolution so deterministic detectors beat
- * probabilistic ones regardless of raw score.
- */
-declare const DETECTOR_PRIORITY: Record<DetectionSource, number>;
 /**
  * A detected PII entity span in the source text.
  * Every detection layer produces these.
@@ -176,12 +156,6 @@ type TriggerRule = {
   validations: CompiledValidation[];
   includeTrigger: boolean;
 };
-/**
- * Anonymisation operator types. Each operator defines
- * how a confirmed entity is replaced in the output.
- */
-declare const OPERATOR_TYPES: readonly ["replace", "redact"];
-type OperatorType = (typeof OPERATOR_TYPES)[number];
 /** Per-label operator selection. Key is the entity label. */
 type OperatorConfig = {
   /** Operator per label. Missing labels default to "replace". */operators: Record<string, OperatorType>; /** Custom replacement string for the redact operator. */
@@ -341,16 +315,6 @@ type PipelineConfig = {
    */
   dictionaries?: Dictionaries;
 };
-/**
- * Canonical entity labels used across the pipeline.
- * NER models may use different native labels; the bench
- * NER wrapper maps model output to these canonical names.
- *
- * These labels are ephemeral: entities are regenerated on
- * every pipeline run and never persisted to the database.
- * Renaming a label here requires no migration.
- */
-declare const DEFAULT_ENTITY_LABELS: readonly ["person", "organization", "phone number", "address", "email address", "date", "date of birth", "bank account number", "iban", "tax identification number", "identity card number", "birth number", "national identification number", "social security number", "registration number", "credit card number", "passport number", "monetary amount", "land parcel", "misc"];
 //#endregion
 //#region src/detectors/regex.d.ts
 type RegexMeta = {

package/dist/wasm.mjs

@@ -1,3 +1,4 @@
+import { DEFAULT_ENTITY_LABELS, DETECTION_SOURCES, DETECTOR_PRIORITY, OPERATOR_TYPES } from "./constants.mjs";
 import { TextSearch } from "@stll/text-search-wasm";
 import { at, be, bg, br, cy, cz, de, dk, ee, es, fi, fr, gb, gr, hr, hu, ie, it, lt, lu, lv, mt, nl, pl, pt, ro, se, si, sk } from "@stll/stdnum";
 import { toRegex } from "@stll/stdnum/patterns";
@@ -12,70 +13,6 @@ const getTextSearch = () => {
 };
 //#endregion
 //#region src/types.ts
-/**
-* Source of a detected entity span.
-* Ordered by detection layer in the pipeline.
-*/
-const DETECTION_SOURCES = {
-	TRIGGER: "trigger",
-	REGEX: "regex",
-	DENY_LIST: "deny-list",
-	LEGAL_FORM: "legal-form",
-	GAZETTEER: "gazetteer",
-	NER: "ner",
-	COREFERENCE: "coreference"
-};
-/**
-* Priority levels for detection sources.
-* Higher = more structurally reliable. Used during
-* overlap resolution so deterministic detectors beat
-* probabilistic ones regardless of raw score.
-*/
-const DETECTOR_PRIORITY = {
-	gazetteer: 5,
-	trigger: 4,
-	"legal-form": 3,
-	regex: 3,
-	"deny-list": 2,
-	coreference: 2,
-	ner: 1
-};
-/**
-* Anonymisation operator types. Each operator defines
-* how a confirmed entity is replaced in the output.
-*/
-const OPERATOR_TYPES = ["replace", "redact"];
-/**
-* Canonical entity labels used across the pipeline.
-* NER models may use different native labels; the bench
-* NER wrapper maps model output to these canonical names.
-*
-* These labels are ephemeral: entities are regenerated on
-* every pipeline run and never persisted to the database.
-* Renaming a label here requires no migration.
-*/
-const DEFAULT_ENTITY_LABELS = [
-	"person",
-	"organization",
-	"phone number",
-	"address",
-	"email address",
-	"date",
-	"date of birth",
-	"bank account number",
-	"iban",
-	"tax identification number",
-	"identity card number",
-	"birth number",
-	"national identification number",
-	"social security number",
-	"registration number",
-	"credit card number",
-	"passport number",
-	"monetary amount",
-	"land parcel",
-	"misc"
-];
 const isLegalFormsEnabled = (config) => config.enableLegalForms !== false;
 //#endregion
 //#region src/context.ts
@@ -2174,7 +2111,9 @@ const POST_NOMINALS = [
 	"CIPM",
 	"CIPT",
 	"CIPP/E",
-	"CIPP"
+	"CIPP",
+	"KC",
+	"QC"
 ];
 //#endregion
 //#region src/detectors/regex.ts
@@ -2277,6 +2216,11 @@ const HONORIFIC_PERSON = {
 	label: "person",
 	score: .95
 };
+const POSTNOMINAL_PERSON = {
+	pattern: `${NAME_WORD}(?:(?:${SP}|-){1,2}(?:${PARTICLE}${SP}+)?${NAME_WORD}){1,3},?${SP}+(?:KC|QC)\\b`,
+	label: "person",
+	score: .95
+};
 const IBAN = {
 	pattern: "\\b[A-Z]{2}\\d{2}\\s?[\\dA-Z]{4}\\s?[\\dA-Z]{4}\\s?[\\dA-Z]{4}\\s?[\\dA-Z]{4}\\s?[\\dA-Z]{0,14}\\b",
 	label: "iban",
@@ -2421,6 +2365,7 @@ const PATH_SUFFIX = `(?:[/?#][^\\s)\\]>]*[^\\s.,;:!?)\\]>])?`;
 const ALL_REGEX_DEFS = [
 	TITLED_PERSON,
 	HONORIFIC_PERSON,
+	POSTNOMINAL_PERSON,
 	IBAN,
 	EMAIL,
 	INTL_PHONE,
@@ -2458,6 +2403,17 @@ const ALL_REGEX_DEFS = [
 		label: "url",
 		score: .9
 	},
+	{
+		pattern: "\\b(?:GIR[^\\S\\n]?0AA|[A-PR-UWYZ](?:[A-HK-Y][0-9](?:[0-9]|[ABEHMNPRV-Y])?|[0-9](?:[0-9]|[A-HJKPS-UW])?)[^\\S\\n]?[0-9][ABD-HJLNP-UW-Z]{2})\\b",
+		label: "address",
+		score: .9
+	},
+	{
+		pattern: "\\b[A-CEGHJ-PR-TWXYZ][A-CEGHJ-NPR-TWXYZ][^\\S\\n]?\\d{2}[^\\S\\n]?\\d{2}[^\\S\\n]?\\d{2}[^\\S\\n]?[A-D]?\\b",
+		label: "social security number",
+		score: .95,
+		validator: gb.nino
+	},
 	{
 		pattern: "\\b(?:1[0-2]|0?[1-9]):[0-5]\\d[^\\S\\n]?(?:[aApP]\\.?[mM]\\.?)(?=[\\s,;!?)]|$)",
 		label: "date",