System and Monitoring
The System tab provides visibility into backend health, routing decisions, and usage metrics.
Status
The Status subtab shows a health grid — one card per discovered service. Each card displays:
| Field | Description |
|---|---|
| Service name | Backend identifier (e.g., ollama-macmini) |
| Health | Green (healthy), Yellow (degraded), Red (down) |
| Loaded model | Currently loaded model, if any |
| Circuit breaker | closed (normal), open (tripped), half-open (testing recovery) |
| Priority | Numeric priority used by the router (lower = preferred) |
Cards update in real time via polling.
Routing Activity Log
A reverse-chronological log of every routed request. Each entry shows:
| Column | Example |
|---|---|
| Timestamp | 14:32:07 |
| Service | ollama-macmini |
| Model | llama3:8b |
| Latency | 1.2s |
| Skipped | lmstudio-pc (circuit open) |
The skipped column lists services that were considered but bypassed, with the reason.
Usage Metrics
A summary panel shows today's usage:
- Total requests
- Total input / output tokens
- Estimated cost (see Cost Tracking)
Model Filter
Control which models the router considers using LobeChat-style filter syntax:
| Token | Meaning |
|---|---|
-all |
Exclude everything |
+gpt-4o |
Include gpt-4o |
-codellama |
Exclude codellama |
Filters are prefix-matched, so +llama3 matches llama3:8b, llama3:70b, etc. The filter is applied at the router level and persists across sessions.