Nyra AI

The AI that returns typed decisions, not text.

Overview

Nyra AI is a System One decision engine. Feed it state plus questions and get back choice, score, or yes/no decisions with calibrated confidence
over a simple JSON API.

Features

• Three decision primitives:

  • noul - Calibrated yes/no (probability)
  • choice - Pick one of labelled options
  • score - Place input on ordered 2-10 scale

• Single round trip - All questions evaluated in one request

• Typed JSON output - Software can act directly on responses

• Confidence scores - Every answer includes calibrated confidence

• ~2s typical latency

API

POST /api/v1/guard
Content-Type: application/json

{
  "state": "context text here",
  "questions": [
    {"type": "noul", "key": "is_safe", "instruction": "Is this safe?"},
    {"type": "choice", "key": "department", "instruction": "Which dept?", "options": ["billing", "support"]},
    {"type": "score", "key": "urgency", "instruction": "Urgency 1-5", "levels": 5}
  ]
}

Response

{
  "model": "dev",
  "latency_ms": 1800,
  "is_safe": true,
  "decision": {...},
  "answers": {
    "department": {"value": "billing", "type": "choice", "confidence": 0.97},
    "urgency": {"value": 4, "type": "score", "confidence": 0.8},
    "is_safe": {"value": true, "type": "noul", "confidence": 0.95}
  }
}

Live Demo

Visit

https://carlostkd.ch/nyra

to try the live demo and see real-time API calls.

Code Examples

See the web interface at /nyra for curl, Python, and Node.js examples.