The Fine Print

from pydantic import BaseModel, Field
from langchain_openai import ChatOpenAI
class ZeroTrustPolicyDecision(BaseModel):
    access_granted: bool
    trust_score: float
    enforced_mitigations: list[str] = Field(description="MFA_REAUTHENTICATE, STEP_UP_AUTH, STEP_DOWN_PRIVILEGES, DENY")
    reasoning: str
def evaluate_zero_trust_access(user_role: str, device_is_managed: bool, disk_encrypted: bool, access_location: str, resource_sensitivity: str) -> ZeroTrustPolicyDecision:
    llm = ChatOpenAI(model="gpt-4o", temperature=0)
    structured_llm = llm.with_structured_output(ZeroTrustPolicyDecision)
    prompt = f"""
    Evaluate Zero Trust Access Request:
    User Role: {user_role}
    Managed Device: {device_is_managed}
    Disk Encrypted: {disk_encrypted}
    Location: {access_location}
    Target Resource: {resource_sensitivity}
    """
    return structured_llm.invoke(prompt)
if __name__ == "__main__":
    res = evaluate_zero_trust_access(
        user_role="Database Administrator",
        device_is_managed=False,
        disk_encrypted=False,
        access_location="Public Coffee Shop Wi-Fi",
        resource_sensitivity="Production Customer PII Database"
    )
    print(f"Access Granted: {res.access_granted}")
    print(f"Trust Score: {res.trust_score}")
    print(f"Enforced Actions: {res.enforced_mitigations}")
    print(f"Reasoning: {res.reasoning}")