DATA HANDLING
The live demos on this site call our inference endpoint at
nelma-api.mondonomo.ai. Names submitted
through them are treated as transient inference input, not stored data.
The exact string you submit, plus standard request metadata (timestamp, anonymized rate-limit token, response time). Inference runs in-memory and the input is discarded when the response is returned.
Aggregated counters (requests per endpoint per minute, error rates, latency percentiles). Request bodies are not written to disk. Identifying content in the input is never associated with an account, cookie, or IP in our logs.
Live-demo inputs are not used to train or fine-tune any Mondonomo model. The models running here were trained on the MondoGraph corpus, which is sourced from public, licensed, and partner-contributed datasets — not from demo traffic.
Aggregated operational logs are retained for 30 days for capacity planning, then deleted. There is no per-user record to request deletion of, because no per-user record exists.
Pilot deployments can be pinned to EU (Frankfurt) or US (Virginia) regions, with regional model replicas and regional Cloudflare edge. DPAs, GDPR Article 28 terms, and SOC 2 reports are available on request before contract.
Questions: research@mondonomo.ai · Last updated 2026-05-18.