Privaro Python SDK

Privacy Infrastructure for Enterprise AI — Official Python SDK by iCommunity Labs

Protect PII in AI prompts with one line of code. Every interaction is tokenized, audited, and blockchain-certified.

---

🚀 Why Privaro

Control data before AI processing.

---

🔥 Example (OpenAI)

from openai import OpenAI
from privaro import Privaro

client = OpenAI()
privaro = Privaro(api_key="YOUR_API_KEY")

input_text = "My name is John Doe and my IBAN is ES76..."

protected = privaro.protect(prompt=input_text)

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role":"user","content":protected["protected_prompt"]}]
)

print(response)

---

❌ Without Privaro

Raw data → model

✅ With Privaro

Tokenized → safe → auditable

---

🤖 Agents

Supports multi-step agent protection.

---

🔗 Related

• Proxy
• JS SDK

---

⚡ Install

pip install privaro

No required dependencies — uses Python stdlib only.

# Optional: async support
pip install privaro[async]

---

Quick Start

import privaro

# Initialize once (e.g., at app startup)
privaro.init(
    api_key="prvr_your_api_key",
    pipeline_id="your-pipeline-uuid",
)

# Protect a prompt before sending to any LLM
result = privaro.protect("Patient: María García, DNI 34521789X, IBAN ES91 2100...")

print(result.protected)      # "Patient: [NM-0001], DNI [ID-0001], IBAN [BK-0001]..."
print(result.risk_score)     # 0.847
print(result.risk_level)     # "high"
print(result.gdpr_compliant) # True

# Send protected prompt to your LLM — no PII ever reaches the model
response = openai.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": result.protected}]
)

---

Usage Patterns

Protect + LLM (full pipeline)

import privaro
import openai

privaro.init(api_key="prvr_xxx", pipeline_id="uuid")

def ask_ai(user_input: str) -> str:
    # 1. Protect PII
    protected = privaro.protect(user_input)

    if not protected.is_safe:
        raise ValueError(f"PII leak detected: {protected.leaked} entities exposed")

    # 2. Call LLM with protected prompt
    response = openai.chat.completions.create(
        model="gpt-4o",
        messages=[{"role": "user", "content": protected.protected}]
    )

    return response.choices[0].message.content

Detect only (analysis mode)

result = privaro.detect("Call me at 612 345 678, email: user@company.es")

for detection in result.detections:
    print(f"{detection.type}: {detection.severity} ({detection.detector})")
# phone: high (regex)
# email: high (regex)

Agent mode (stricter policies)

result = privaro.protect(
    prompt=agent_input,
    agent_mode=True,   # Applies stricter policy preset
)

Multiple clients (multiple pipelines)

from privaro import PrivaroClient

legal_client = PrivaroClient(api_key="prvr_xxx", pipeline_id="legal-pipeline-uuid")
hr_client = PrivaroClient(api_key="prvr_xxx", pipeline_id="hr-pipeline-uuid")

legal_result = legal_client.protect(contract_text)
hr_result = hr_client.protect(employee_record)

Async support

from privaro.async_client import AsyncPrivaroClient

client = AsyncPrivaroClient(api_key="prvr_xxx", pipeline_id="uuid")

async def process(prompt: str):
    result = await client.protect(prompt)
    return result.protected

Error handling

from privaro.exceptions import AuthError, PolicyBlockError, ProxyUnavailableError

try:
    result = privaro.protect(prompt)
except PolicyBlockError as e:
    # Request blocked by policy (e.g., health data on non-approved provider)
    logger.warning(f"Request blocked: {e}")
    return "Request cannot be processed — sensitive data detected."
except ProxyUnavailableError:
    # Fallback: log and fail safely
    logger.error("Privaro proxy unavailable")
    raise
except AuthError:
    logger.error("Invalid Privaro API key")
    raise

---

ProtectResult Reference

Property	Type	Description
result.protected	str	Prompt with PII replaced by tokens
result.original	str	Original prompt (local only)
result.risk_score	float	0.0–1.0 composite risk score
result.risk_level	str	"high" / "medium" / "low"
result.gdpr_compliant	bool	True if no PII leaked
result.is_safe	bool	True if all PII masked
result.has_pii	bool	True if any entities detected
result.total_detected	int	Total PII entities found
result.total_masked	int	Entities successfully masked
result.leaked	int	Entities that passed through
result.detections	list[Detection]	Per-entity details
result.audit_log_id	str	Supabase audit log UUID
result.processing_ms	int	Proxy latency in ms
result.summary()	str	One-line log summary

---

Detection Reference

Property	Values
detection.type	dni iban email full_name phone health_record credit_card ip_address date_of_birth
detection.severity	critical high medium low
detection.action	tokenised anonymised blocked
detection.detector	regex presidio
detection.confidence	0.0–1.0
detection.is_high_risk	bool

---

Blockchain Verification

Every protect() call creates an immutable audit entry certified on Fantom Opera Mainnet via iBS. Verify any event at:

https://checker.icommunitylabs.com/check/fantom_opera_mainnet/{tx_hash}

Access the TX hash from your Privaro dashboard → Audit Logs → ⛓️ badge.

---

Requirements

• Python 3.9+
• Zero required dependencies (uses urllib from stdlib)
• Optional: aiohttp>=3.9 for async support

---

License

MIT — © 2026 by iCommunity Labs · privaro.ai icommunity.io