I built a decision-intelligence API: one POST, a verdict back

jaswant singh — Mon, 25 May 2026 10:40:37 +0000

Most "AI" endpoints give you text. I wanted one that gives you a decision — a verdict, a confidence score, the reasoning, and a receipt you can verify later. So I built Lumi.

The idea

You POST a question as plain JSON to one REST endpoint:

curl -X POST https://api.kauzio.com/public/v1/brain/query \
  -H "Authorization: Bearer kz_live_..." \
  -H "Content-Type: application/json" \
  -d '{ "question": "Should we expand to the US market in Q3?" }'

And you get back structured analysis, not prose:

{
  "answer": "Not yet. Two of your top three markets are still under-penetrated. Recommend a Q4 pilot with a 90-day proof window.",
  "confidence": 0.74,
  "verdict": "hold",
  "receipt_id": "rcpt_9f3a21c8",
  "latency_ms": 312
}

What's behind it

A "decide" endpoint that returns a verdict + confidence + reasoning
An "oppose" endpoint that argues the strongest case against your decision
A "verify" endpoint — every decision is signed, so you (or anyone) can confirm it later, no auth needed
Webhooks for decision.completed, certificate.issued, anomaly.detected

Why signed receipts?

The thing I cared most about: every verdict gets a tamper-proof certificate with a public verify URL. If your app makes an automated call, you can prove what was decided, when, and on what basis. Useful for audit, compliance, or just trust.

There's a free tier (no card) if you want to poke at it: https://lumi.kauzio.com

Would genuinely love feedback from this community — especially on the API shape. Does the request/response feel natural to drop into your own code? What endpoints would you want?

DEV Community: jaswant singh

I built a decision-intelligence API: one POST, a verdict back

The idea

What's behind it

Why signed receipts?