@redactpii/node

@redactpii/node

🔒 Simple PII redaction library for Node.js

A fast, zero-dependency library that redacts PII from text using regex patterns. Works completely offline. No API keys, no setup, just install and use.

🚀 Installation

npm install @redactpii/node
# or
pnpm add @redactpii/node
# or
yarn add @redactpii/node

🔥 Quick Start

import { Redactor } from '@redactpii/node';

const redactor = new Redactor();
const clean = redactor.redact('Hi David Johnson, call 555-555-5555');
// Result: "Hi PERSON_NAME, call PHONE_NUMBER"

🎯 Built-in PII Detection Patterns

The library includes regex patterns for:

• 👤 Names - Person identification (greeting-based detection)
• 📧 Emails - Email addresses
• 📞 Phones - US phone numbers (all formats)
• 💳 Credit Cards - Visa, Mastercard, Amex, Diners Club
• 🆔 SSN - US Social Security Numbers

🤖 Use with AI APIs

Protect user data before sending to OpenAI, Anthropic, or other LLM providers:

import { Redactor } from '@redactpii/node';
import OpenAI from 'openai';

const redactor = new Redactor();
const openai = new OpenAI();

// Redact before sending to OpenAI
const userMessage = 'Hi, my email is [e-mail] and my phone is 555-123-4567';
const cleanMessage = redactor.redact(userMessage);
// "Hi, my email is EMAIL_ADDRESS and my phone is PHONE_NUMBER"

const completion = await openai.chat.completions.create({
  messages: [{ role: 'user', content: cleanMessage }],
  model: 'gpt-4',
});

Works with any API that accepts JSON:

// Redact entire request payloads
const apiRequest = {
  user: {
    name: 'John Doe',
    email: '[e-mail]',
    notes: 'Call me at 555-123-4567',
  },
};

const cleanRequest = redactor.redactObject(apiRequest);
// Send cleanRequest to your AI service

🔍 Check for PII Without Redacting

const redactor = new Redactor({ rules: { EMAIL: true } });

if (redactor.hasPII('Contact [e-mail] for details')) {
  console.log('PII detected!');
  const clean = redactor.redact('Contact [e-mail] for details');
}

📦 Redact Objects

const redactor = new Redactor({ rules: { EMAIL: true } });

const user = {
  name: 'John Doe',
  email: '[e-mail]',
  profile: {
    contact: '[e-mail]',
  },
};

const clean = redactor.redactObject(user);
// {
//   name: 'John Doe',
//   email: 'EMAIL_ADDRESS',
//   profile: {
//     contact: 'EMAIL_ADDRESS',
//   },
// }

🎨 Customization

Configure Rules

Enable or disable specific PII detection patterns:

const redactor = new Redactor({
  rules: {
    CREDIT_CARD: true, // Enable credit card detection
    EMAIL: true, // Enable email detection
    NAME: false, // Disable name detection
    PHONE: true, // Enable phone detection
    SSN: false, // Disable SSN detection
  },
});

Custom Regex Patterns

Add your own regex patterns for domain-specific PII:

const redactor = new Redactor({
  rules: { EMAIL: true },
  customRules: [
    /\b\d{5}\b/g, // 5-digit codes
    /\bSECRET-\d+\b/g, // Secret codes
  ],
});

Global Replacement

Use a single replacement string for all PII types:

const redactor = new Redactor({
  rules: { EMAIL: true },
  globalReplaceWith: '[REDACTED]', // All PII types use this replacement
});

redactor.redact('[e-mail]'); // "[REDACTED]"

Anonymization with Unique IDs

Replace the same PII value with the same token throughout the text:

const redactor = new Redactor({
  rules: { EMAIL: true, NAME: true },
  anonymize: true, // Enable anonymization
});

// Same email gets same token
const text = 'Contact [e-mail]. Anne also uses [e-mail] for work.';
const result = redactor.redact(text);
// Result: "Contact EMAIL_1. PERSON_1 also uses EMAIL_1 for work."

// Works across objects too
const user = {
  primary: '[e-mail]',
  backup: '[e-mail]', // Same value → same token
  contact: '[e-mail]', // Different value → different token
};
const clean = redactor.redactObject(user);
// {
//   primary: 'EMAIL_1',
//   backup: 'EMAIL_1',    // Same as primary
//   contact: 'EMAIL_2'    // Different token
// }

Aggressive Mode

Use more permissive regex patterns to catch obfuscated or unusual PII formatting:

const redactor = new Redactor({
  rules: { EMAIL: true, CREDIT_CARD: true },
  aggressive: true, // Enable aggressive mode
});

// Catches obfuscated emails
redactor.redact('user [at] example [dot] com');
// Result: "EMAIL_ADDRESS"

// Catches partially masked credit cards
redactor.redact('Card ending in ****-****-****-1234');
// Result: "Card ending in CREDIT_CARD_NUMBER"

// Normal mode (aggressive: false) is more conservative
// and won't catch these variations

❓ FAQ

Is this regex-based?
Yes, this library uses regex patterns for detection. It's fast and works offline, but has limitations.

How does it handle misspellings or improperly formatted data?
It catches misspellings if the format is still valid (e.g., "[e-mail]" would be detected because it's still a valid email format). However, it won't catch obfuscated or non-standard formats like "john at example dot com" or "john[at]example[dot]com" unless you enable aggressive: true mode, which uses more permissive patterns.

What determines what counts as PII?
The built-in patterns cover common, obvious PII types (emails, SSNs, credit cards, phone numbers, names in greetings). These are based on standard formats, not a specific compliance framework. For your specific needs, use customRules to add domain-specific patterns.

Anonymization vs Redaction?
By default, this library does redaction (replacement with labels like EMAIL_ADDRESS). However, you can enable anonymization by setting anonymize: true, which replaces the same PII value with the same unique token (e.g., EMAIL_1, EMAIL_2) throughout the text, preserving relationships while protecting privacy.