AI & Copilot
✓ Zero-Leakage LLM Governance
Setup: ~3 mins
Copilot Studio GDPR / KVKK PII Redaction + Deterministic Broken AI JSON Repair
Strip emails, IBANs, phone numbers, credit cards, and national IDs BEFORE sending tickets to OpenAI/Copilot, then deterministically repair truncated or markdown-fenced LLM JSON responses.
Why Makers Use This Blueprint: Sending unmasked customer tickets to external LLMs violates EU GDPR / KVKK policies, and LLM markdown code fences frequently crash Power Automate Parse JSON steps.
Cloud Flow Architecture (Trigger → Zetaleap → Destination)
STEP 01 • Zetaleap Privacy Guard
POST /api/pp/mask-pii
Replaces sensitive PII tokens with deterministic placeholders ([EMAIL_1], [IBAN_1]).
STEP 02 • Copilot / Azure OpenAI
Run Classification & Extraction Prompt
LLM processes 100% anonymized text safely.
STEP 03 • Zetaleap AI JSON Repair
POST /api/hub/ai-json-repair
Strips ```json fences, fixes trailing commas/unclosed quotes, and guarantees valid JSON.
Power Automate HTTP Action Definition (
POST /api/pp/mask-pii)
Replace
zl_YOUR_API_KEY with your free Personal Access Token from /profile — or import the typed Custom Connector at /connectors.{
"type": "Http",
"inputs": {
"method": "POST",
"uri": "https://zetaleap.com/api/pp/mask-pii",
"headers": {
"Authorization": "Bearer zl_YOUR_API_KEY",
"Content-Type": "application/json"
},
"body": {
"text": "@{triggerOutputs()?['body/CustomerMessage']}",
"maskMode": "token"
}
}
}
Returned Dynamic Output Tokens
maskedTextredactionCountdetectedTypes[]