A payment engine that guarantees exactly-once processing under concurrent load — with an AI layer that reads its own audit trail and explains, in plain language, why every duplicate was caught.
—Same payment ID, same timestamp, one charge. Request 2 never reached the gateway — it was intercepted by the row lock and served from the idempotency cache. No retry logic on your end needed.
Endpoints
Everything below is live. Swap the API key for your own. Click any row to view full cURL or copy directly.
PII-scrubbed ledger of every attempt, with AI explanations distinguishing benign retries from anomalies.
curl -X GET https://payment-api-3fi4.onrender.com/api/ai/events/ \
-H "X-API-KEY: pay_demo.live_reviewer_test_key_2026"
AI layer
GET
/api/ai/risk-scores/
Weighted rule-based score, LLM adjustment clamped to ±0.1 with a written justification.
curl -X GET https://payment-api-3fi4.onrender.com/api/ai/risk-scores/ \
-H "X-API-KEY: pay_demo.live_reviewer_test_key_2026"
AI layer
POST
/api/ai/query/
Ask a question in English. SELECT-only, whitelisted tables, 500-row cap — every query logged.
curl -X POST https://payment-api-3fi4.onrender.com/api/ai/query/ \
-H "Content-Type: application/json" \
-H "X-API-KEY: pay_demo.live_reviewer_test_key_2026" \
-d '{"question": "Show me duplicate payments from the last hour", "auto_execute": true}'