Agent scores
Per-model exam leaderboard with cost-efficiency ranking from AI OS offline exams. Production and payment lanes stay blocked unless trusted scores exist.
Conservative mode
— incomplete totals detected; human review required before production routing.
Scores unavailable
AI-OS dashboard unreachable or empty — conservative mode for routing hints.
API providers health
Configured key counts only — no secrets exposed.
LLM
Agent runtime
Search
BI / embed