Settings
Detection thresholds and mode (observe-only in this phase)
⌕ Search exchange…
◔3
LIVE
Read-only
Viewing as a non-super_admin role — settings are read-only for you.
1. Rinser AI Mode
— — locked to observe in this phase. Recommend/approval/guarded_auto modes require an approval workflow, rollback snapshots, and MFA that don't exist yet (see RINSER_AI_PHASES.md).
2. Detection Thresholds
Edited as raw JSON in Phase 1 — a structured form arrives with the policy engine.
{}
3. Claude API Status
{}
4. Data Masking
{}
5. Incident Rules
Detection categories currently implemented.
carrier_outage, high_503, high_500, high_408, asr_drop, pdd_spike, no_final_response_spike, negative_margin, short_call_spike, capacity_exhaustion — evaluated every 5 minutes for global and carrier scope. See RINSER_AI_ARCHITECTURE.md.
6. Recommendation Rules
Recommendations come from Claude's analysis of each incident, not a deterministic rule engine — there is no separate recommendation-generation ruleset yet.
7. Approval Safety
All route-changing recommendations remain approval-only. No automatic route changes, carrier disables, rate changes, or service restarts exist in this build — approving a recommendation only records a human decision (see Recommendations page); there is no executor. Two-person approval and MFA enforcement are Phase 2 requirements not yet implemented.
8. Worker Status
Worker
not reporting
Last snapshot
—