Nessie v2 POST /simulations live mode 422 'rate_per_minute is capped at 10.' even when you omit rate_per_minute (spec default is 60)
Starting a live fraud-simulator run on Nessie v2 (POST https://v2-prod-api.nessieisreal.com/simulations) with {"mode":"live","rate_per_minute":60} fails with HTTP 422 and a problem+json body: {"type":"https://nessieisreal.com/problems/simulation-limit","title":"Unprocessable Entity","status":422,"detail":"rate_per_minute is capped at 10."}.
The surprising part: the OpenAPI spec (v2.1.3, SimulationCreateDTO) documents rate_per_minute as "Live: transactions per minute (default 60)" with example 60 and only a minimum of 1, no maximum. So a client generated from the spec, or a request that simply leaves the field out, gets the same 422. I verified that {"mode":"live","duration_seconds":5,"customers":1} with no rate_per_minute is rejected with exactly the same "capped at 10" detail, because the server-side default (60) is above the cap. A value of 11 is also rejected.
Fix: always send rate_per_minute explicitly with a value from 1 to 10 for live runs, e.g.
curl -X POST https://v2-prod-api.nessieisreal.com/simulations -H "Authorization: Bearer $NESSIE_KEY" -H "Content-Type: application/json" -d '{"mode":"live","rate_per_minute":10,"duration_seconds":300,"customers":3,"anomaly_rate":0.2}'
The response's rate_per_minute field then echoes 10. Historical runs are not affected (they report rate_per_minute 0 and use "transactions" instead), so if you just need a dataset to train a model on, mode "historical" avoids the cap entirely and is much faster. Plan live-mode demos around at most 10 transactions per minute: a default 300 second run produces roughly 50 transactions, not hundreds. The type URL ends in simulation-limit, which is a handy string to match on if you want to clamp and retry automatically.