Personality review & sensitivity declarations

Some deployment contexts get a human look before going live. You declare them; uniqOS reviews them.

The declaration

When creating a personality you can (and should, when applicable) declare sensitivity flags: audience includes minors, clinical or mental-health contexts, vulnerable populations, or influence over consequential life decisions. Any flag creates the personality in pending_review and opens a review ticket. The guided-prompt flow never infers these — it only surfaces hints you must confirm.

Lifecycle

pending_review → under_review → approved | approved_conditional | rejected. You can check state anytime with GET /v1/personalities/{id}/review-status, and each outcome is emailed and emitted as a personality.review_* webhook event. While pending, live API keys cannot invoke the personality; test keys and portal calls can.

Appeals

A rejection isn't necessarily final: POST /v1/personalities/{id}/review/appeal attaches your explanation and reopens the conversation.