Quickstart

npm install @omegaengine/sdk

import { OmegaClient } from "@omegaengine/sdk";

const omega = new OmegaClient({
  baseUrl: "https://omegaengine.ai",
  apiKey: process.env.OMEGA_API_KEY!,
});

const result = await omega.decide({
  scenario: "Should I approve this $50k wire transfer?",
  context: { customer: "VIP", creditScore: 780 },
  domain: "general",
  riskTolerance: "LOW",
});

if (!result.ok) {
  throw new Error(result.error);
}

const { decision, meta } = result;

if (decision?.riskLevel === "CRITICAL") {
  throw new Error(decision.summary);
}

// Check if human review is needed
if (decision?.needsHumanReview) {
  await escalateToHuman(decision);
  return;
}

// Safe to proceed — decision.recommendedAction is one of:
// "ALLOW" | "PROCEED_WITH_GUARDRAILS" | "REJECT" | "ESCALATE"
console.log(`${decision?.recommendedAction} in ${meta?.durationMs}ms`);
await executeTransfer();

// Response from omega.decide() — typed as OmegaDecisionResponse
{
  ok: true,
  decision: {
    id: "d6dfb7a4-0af6-4a39...",
    recommendedAction: "PROCEED_WITH_GUARDRAILS",
    riskScore: 44,
    riskLevel: "MEDIUM",
    safetyScore: 80,
    needsHumanReview: false,
    regretScore: 48,
    omegaIQ: 77.4,
    ethicsConcern: false,
    summary: "Financial decision with moderate risk...",
    riskDrivers: [
      "Financial decision with material monetary exposure.",
      "Limited information: important details may be missing."
    ],
  },
  meta: {
    requestId: "d6dfb7a4-0af6-4a39...",
    durationMs: 6583,
    auditHash: "c45e7a6c5ec2af68...",
    hmacSignature: "38e294f316...",
    modelProvider: "google",
    modelName: "gemini-3.1-flash-lite",
  }
}

# The SDK calls /api/decision internally.
# For raw HTTP, you can also use /api/v2/judge (returns "judge" key).
curl -X POST https://omegaengine.ai/api/decision \
  -H "Content-Type: application/json" \
  -H "x-api-key: omega_your_api_key_here" \
  -d '{
    "scenario": "Should I approve this wire transfer?",
    "domain": "general",
    "riskTolerance": "LOW"
  }'

from omegaengine import OmegaClient, OmegaClientOptions

omega = OmegaClient(OmegaClientOptions(
    api_key="omega_your_api_key_here",
    base_url="https://omegaengine.ai",
))

result = omega.decide({
    "scenario": "Deploy v2.4 to production",
    "context": "All CI tests passed, staging verified",
    "domain": "TECHNICAL",
    "riskTolerance": "MEDIUM",
})

print(result["decision"]["recommendedAction"])