95%
Покрытие
Подключенные модули / всего
Agent efficiency
Read-only aggregate: how AI OS routes agents (exam permissions + ledger success). No per-task detail.
0%
Обновить
Свежесть тикера (окно 0–15 мин)
офлайн
Использование AI
Метрики времени выполнения модели (офлайн)
офлайн
Попадание в кеш
Коэффициент кеширования на границе (офлайн)
Self-improving loop
AI OS continuously adjusts agent routing based on exam scores and real-world outcomes.
1
Exam
Agent takes offline exam → gets score per category
→
2
Score → Routing
Score determines which tasks the agent can handle
→
3
LLM-маршрутизатор
Router picks cheaper or better backend based on cost/quality
→
4
Effectiveness Ledger
Real outcomes tracked: success rate, blocker rate, cost
→
5
Learning → RAG
Next session starts with context from previous results
Effectiveness ledger offline — aggregates shown only where available.
Generated 2026-06-27T05:23:16.048966+00:00
- Agents tracked
- 0
- Tasks recorded
- 0
- Overall success
- —
- Tests pass rate
- —
- Avg est. cost
- —
| Agent | Exam total | Success rate | Blocker rate | Recommended lane |
|---|