orkaia 0.1.0

Governance and audit SDK for AI agents

orka-sdk

Governance and audit SDK for AI agents. Add policy enforcement, immutable audit trails, and trust scoring to any agent with one decorator.

Install

pip install orka-sdk

Quickstart

import orka

# Initialize once at startup (get your API key at orka.ia.br)
orka.init(api_key="orka_your_key_here")

# Wrap any agent function
@orka.guard(agent_id="your-agent-uuid", task_type="summarize")
def summarize(text: str) -> str:
    return openai_client.chat(text)

# Works with async too
@orka.guard(agent_id="your-agent-uuid", task_type="financial_transfer", risk="HIGH")
async def transfer_funds(amount: float, to_account: str) -> dict:
    return await payment_api.transfer(amount, to_account)

That's it. Every call now:

• ✅ Checks your X-Shield policies before executing
• ✅ Logs to X-Ledger with immutable hash chain
• ✅ Updates the agent's trust score
• ✅ Blocks execution if policy denies it (raises OrkaPolicyBlocked)

Handle policy blocks

from orka import OrkaPolicyBlocked

try:
    result = transfer_funds(10000, "acc_123")
except OrkaPolicyBlocked as e:
    print(f"Blocked: {e.reason}")
    print(f"Policy: {e.policy_name}")

Works with LangChain, CrewAI, any framework

from langchain.agents import AgentExecutor
import orka

orka.init(api_key="orka_...")

@orka.guard(agent_id="langchain-agent-uuid", task_type="web_search")
def run_agent(query: str) -> str:
    return agent_executor.invoke({"input": query})

Never breaks your production code

If Orka is unreachable, your agent continues running. Governance failures are silent — your users never see Orka errors.

Risk levels

@orka.guard(agent_id="...", task_type="...", risk="HIGH")
# EU AI Act risk classification: UNACCEPTABLE | HIGH | LIMITED | MINIMAL | NONE

Environment variables

ORKA_API_KEY=orka_your_key
ORKA_BASE_URL=https://orka-backend.onrender.com/api/v1  # optional

import os, orka
orka.init(api_key=os.environ["ORKA_API_KEY"])