Microsoft Presidio × Sentiqra

Run PII detection and redaction in your own environment before submitting text to Sentiqra. This pattern can reduce the amount of identifiable information sent downstream while keeping redaction policy under your control.

Why This Integration Matters

Every enterprise buyer with a GDPR compliance obligation asks the same question during vendor evaluation: "How do we make sure customer PII doesn't leave our infrastructure?" Without a documented PII-stripping layer, regulated industries - finance, healthcare, telecom - literally cannot deploy any text-processing API.

Microsoft Presidio is the industry standard for PII detection and anonymization. It runs locally in your infrastructure, meaning PII is detected and stripped before text is sent to Sentiqra. This solves the biggest deployment blocker for regulated industries and strengthens your zero-retention claim: even if Sentiqra did retain data - which it doesn't - the input received has already been sanitized locally by your Presidio deployment.

Publishing this reference architecture removes the #1 deployment blocker for regulated industries and signals to DPOs that you understand their consent architecture.

Reference Architecture

┌─────────────────────────────────────────────────────────────────┐
│                    Your Infrastructure                           │
│                                                                  │
│  ┌──────────┐     ┌──────────┐     ┌──────────┐                │
│  │  User     │────▶│ Presidio │────▶│ Sentiqra │                │
│  │  Input    │     │ (local)  │     │  API     │                │
│  │           │     │          │     │          │                │
│  │ "Hi I'm   │     │ Detect & │     │ Detect   │                │
│  │  John     │     │ Anonymize│     │ harmful  │                │
│  │  Smith,   │     │ PII      │     │ language │                │
│  │  john@    │     │          │     │          │                │
│  │  gmail.." │     │ "Hi I'm  │     │ Returns  │                │
│  │           │     │  <PERSON>│     │ groups + │                │
│  │           │     │  <EMAIL> │     │ concepts │                │
│  │           │     │  .."     │     │          │                │
│  └──────────┘     └──────────┘     └────┬─────┘                │
│                                         │                        │
│                                    ┌────▼─────┐                │
│                                    │ Your     │                │
│                                    │ Enforcement│               │
│                                    │ Pipeline │                │
│                                    └──────────┘                │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘
                              │
                    Sentiqra API (submitted classification content is not retained)
                    No PII ever leaves your infra

Data flow: User input enters your infrastructure → Presidio runs locally to detect and anonymize PII (names, emails, phone numbers, SSNs, credit cards, etc.) → sanitized text is sent to Sentiqra API → detection result returned → your enforcement pipeline acts on the result. PII never crosses the network boundary to Sentiqra.

Working Code

Actual, tested, copy-pasteable code. Not pseudocode. Not aspirational.

import os
import requests
from presidio_analyzer import AnalyzerEngine
from presidio_anonymizer import AnonymizerEngine

# 1. Initialise Presidio (runs locally in your infrastructure)
analyzer = AnalyzerEngine()
anonymizer = AnonymizerEngine()

# 2. User input - contains PII
user_input = (
    "Hi, I'm John Smith and my email is john.smith@gmail.com. "
    "You are a worthless idiot."
)

# 3. Detect PII with Presidio
analyzer_results = analyzer.analyze(text=user_input, language="en")

# 4. Anonymize PII - this happens LOCALLY, nothing leaves your infra
anonymized = anonymizer.anonymize(
    text=user_input,
    analyzer_results=analyzer_results,
)

# anonymized.text is now:
# "Hi, I'm <PERSON> and my email is <EMAIL_ADDRESS>. You are a worthless idiot."

# 5. Send sanitized text to Sentiqra - no PII in the request
response = requests.post(
    "https://sentiqra.tech/v1/analyze",
    headers={
        "Authorization": f"Bearer {os.environ['SENTIQRA_API_KEY']}",
        "Content-Type": "application/json",
    },
    json={"text": anonymized.text, "language": "en", "analyzers": ["trust-safety"]},
    timeout=10,
)
data = response.json()

# 6. Act on the detection result
top = data["signals"][0]
print(top["code"])    # 'insults_derogation'
print(top["score"])   # 0.99

// Presidio is Python-only - there is no official Node.js SDK.
// Run it as a local service instead:
//   docker run -d -p 5002:5000 mcr.microsoft.com/presidio-analyzer
//   docker run -d -p 5001:5000 mcr.microsoft.com/presidio-anonymizer

const PRESIDIO_ANALYZER = "http://localhost:5002";
const PRESIDIO_ANONYMIZER = "http://localhost:5001";

const userInput = "Hi, I'm John Smith (john.smith@gmail.com). You are a worthless idiot.";

// 1. Detect PII locally
const analyzerResults = await fetch(`${PRESIDIO_ANALYZER}/analyze`, {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({ text: userInput, language: "en" }),
}).then((r) => r.json());

// 2. Anonymize PII locally - nothing leaves your infra
const anonymized = await fetch(`${PRESIDIO_ANONYMIZER}/anonymize`, {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({ text: userInput, analyzer_results: analyzerResults }),
}).then((r) => r.json());

// 3. Send sanitized text to Sentiqra
const response = await fetch("https://sentiqra.tech/v1/analyze", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.SENTIQRA_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({ text: anonymized.text, language: "en", analyzers: ["trust-safety"] }),
});
const data = await response.json();

const top = data.signals[0];
console.log(top.code);    // "insults_derogation"
console.log(top.score);   // 0.99

# Python
pip install presidio-analyzer presidio-anonymizer requests
python -m spacy download en_core_web_lg

# JavaScript - Presidio runs as a local REST service (see comment above)
# Sentiqra is called over plain HTTPS - no SDK required

Security & Compliance Posture

• ✓
  GDPR Article 5(1)(c) - Data Minimisation: PII is stripped before data reaches Sentiqra. Only the minimum necessary text (without identifiers) is processed externally.
• ✓
  GDPR Article 32 - Security of Processing: Presidio runs locally in your infrastructure. No PII traverses the network to Sentiqra, reducing the attack surface.
• ✓
  Zero-Retention Reinforcement: Sentiqra already operates zero-retention. With Presidio, even the zero-retention input contains no PII - defence in depth.
• ✓
  DSA Article 6 - Risk Assessment: Provides evidence of proactive measures to protect user data during content moderation, which may support platform risk assessment documentation.
• ✓
  SOC 2 Type II - CC6.1: PII sanitisation before external processing can provide evidence relevant to logical access controls over customer data.
• ✓
  DLP Policy Support: Can help organizations implement data-loss-prevention policies that prohibit PII from being sent to third-party APIs.

Deployment Notes