AI Provider Settings
Admin draft-authoring support only. Customer assessment data is never sent to a provider.
Human review required
AI may propose draft question wording, anchors, translation candidates or validation hints. It cannot publish a form, change safety rules, access customer responses, or make clinical decisions.
OpenAI
EnabledCredential[REDACTED] · server-side only
Model catalog6 compatible models · sample cache
Default: question draftingSelected model · sample only
MiniMax
EnabledCredential[REDACTED] · server-side only
Model catalog6 compatible models · sample cache
Allowed model policyTranslation candidates only
Allowed model selection
Authoring purpose policy
A selected provider/model is captured on every AI draft. A later setting change cannot alter a completed draft or a published form.
| Purpose | Provider | Default model | Allowed models | Review status | Action |
|---|---|---|---|---|---|
| Question drafting | OpenAI | Selected sample model | 2 approved models | Reviewed | |
| Translation candidate | MiniMax | Selected sample model | 1 approved model | Reviewed | |
| Safety / clinical content | Not permitted | — | 0 models | Blocked | Human process only |