Warddocs
Patterns

Review queue

GET /api/v1/review/queue?days=7 lists statement decisions sent for review (STEP_UP or DECLINE) with no feedback yet, newest first and ranked by the score in force for your account. Each row also shows its rank under your learned model, so a team can see what the model would put first before it is switched on.

  1. Pull the queue; open each case's evidence pack.
  2. The analyst decides and sends feedback with right_decision and a label.
  3. Feedback trains your account's model (shadow first) and feeds threshold suggestions.

Turn on review sampling to also label a random share of APPROVE decisions: otherwise the model only ever learns from what Ward already flagged.