Probabilistic merge-candidate review queue
Confidence-scored merge candidates for the CDP Identity Rules review queue. Scans the tenant’s contacts, blocks them into phone-area / email-domain shards (so the pairwise comparison stays linear in the shard size), scores every in-shard pair with the probabilistic identity kernel, and returns the ranked candidates at/above min_confidence. Each candidate carries its band (auto_merge | review), the per-signal similarity breakdown, and a suggested primary/secondary so a reviewer can approve a merge in one click. READ-ONLY: it derives candidates on the fly and never merges or mutates a contact — approving a candidate calls the existing POST /contacts/merge. Owner / admin / developer with the contacts:read scope (the payload embeds masked contact PII). Off-spec query params degrade to their defaults rather than erroring.
Authorizations
Dashboard JWT token from Clerk
Query Parameters
Max candidates returned (post-ranking). Out-of-range values fall back to 50.
1 <= x <= 200Confidence floor, 0–100. Defaults to the review-band threshold so the queue surfaces every candidate a human should look at.
0 <= x <= 100When false, drops auto_merge-band candidates and returns only the human-review mid-band.