Skip to main content
pii_detection:
  presidio_url: http://localhost:5002
  languages: ${PASTEGUARD_LANGUAGES:-en}  # Auto-configured per Docker image
  fallback_language: en
  score_threshold: 0.7
  entities:
    - PERSON
    - EMAIL_ADDRESS
    - PHONE_NUMBER
    - CREDIT_CARD
    - IBAN_CODE
    - IP_ADDRESS
    - LOCATION

​
Options

OptionDefaultDescription
presidio_urlhttp://localhost:5002Presidio analyzer URL
languages(per image)Languages to detect. Auto-configured in Docker images
fallback_languageenFallback if detected language not in list
score_threshold0.7Minimum confidence (0.0-1.0)
entitiesSee belowEntity types to detect

​
Languages

Languages are auto-configured per Docker image:
  • :en image → English only
  • :eu image → English, German, Spanish, French, Italian, Dutch, Polish, Portuguese, Romanian
Each language adds ~10s to startup time as spaCy models are loaded. For custom language builds: 
LANGUAGES=en,de,ja docker compose up -d --build
Available languages (24): ca, zh, hr, da, nl, en, fi, fr, de, el, it, ja, ko, lt, mk, nb, pl, pt, ro, ru, sl, es, sv, uk

​
Override Languages

For local development or custom setups, override via config: 
pii_detection:
  languages:
    - en
    - de

​
Fallback Language

If the detected language isn’t in your configured list, the fallback is used: 
pii_detection:
  fallback_language: en  # Used for unsupported languages

​
Performance

If only one language is configured, language detection is skipped for better performance.

​
Entities

EntityExamples
PERSONDr. Sarah Chen, John Smith
EMAIL_ADDRESSsarah.chen@hospital.org
PHONE_NUMBER+1-555-123-4567
CREDIT_CARD4111-1111-1111-1111
IBAN_CODEDE89 3704 0044 0532 0130 00
IP_ADDRESS192.168.1.1
LOCATIONNew York, 123 Main St
US_SSN123-45-6789
US_PASSPORT123456789
CRYPTOBitcoin addresses
URLhttps://example.com

​
Score Threshold

Higher = fewer false positives, might miss some PII. Lower = catches more PII, more false positives. 
pii_detection:
  score_threshold: 0.7  # Default, good balance
  # score_threshold: 0.5  # More aggressive
  # score_threshold: 0.9  # More conservative

​
Whitelist

Exclude specific text patterns from PII masking. Useful for preventing false positives on company names or product identifiers. 
masking:
  whitelist:
    - "Acme Corp"
    - "Product XYZ"
Patterns match bidirectionally - detected text containing a whitelist entry (or vice versa) is excluded.

​
Scan Roles

By default, all message roles are scanned. To scan only user-controlled content: 
pii_detection:
  scan_roles:
    - user
    - tool
    - function
RoleDescription
userUser messages (primary source of PII)
assistantAssistant responses
systemSystem prompts
toolTool/function call results
functionLegacy function results (OpenAI)
This reduces Presidio API calls for large system prompts and avoids false positives on app-controlled content.