hejdar 0.1.0

Hejdar AI governance SDK — policy evaluation for AI agents

Hejdar Python SDK

Policy evaluation for AI agents. Check actions against Hejdar governance rules before execution.

Install

pip install hejdar

Quickstart

from hejdar import Hejdar, HejdarDeniedError, HejdarNetworkError

client = Hejdar(api_key="hejdar_sk_...")

try:
    result = client.check(
        agent_id="finance-assistant",
        action_type="WRITE_DB",
        data_classification="PII",
        environment="production",
    )
    if result.allowed:
        # Proceed with action
        pass
except HejdarDeniedError as e:
    # Blocked in enforce mode
    print(f"Denied: {e.reason_code}")
except HejdarNetworkError:
    # API unreachable — fail closed (treat as deny)
    pass

Modes

• Observe (shadow): Returns WOULD_DENY / WOULD_REQUIRE_APPROVAL — logs a warning, never raises. Use to trial policies.
• Enforce: Returns DENY / REQUIRE_APPROVAL — raises HejdarDeniedError and blocks execution.

Async

result = await client.check_async(agent_id="...", action_type="READ_DB")

API Reference

check(agent_id, action_type, data_classification="INTERNAL", environment="production", ...) — see docstrings.