Overview dashboard
Overview is the landing page after logging into the console, providing a global view: resource counts, instance running status, gated-request monitoring, and time-series trend charts.
Entry: Console → Overview (visible to admin and monitor).
Summary cards
Four cards at the top, reflecting the current configuration scale in real time:
| Card | Meaning |
|---|---|
| Total upstreams | Number of configured upstreams |
| Active protocols | Number of protocol types in use |
| Active models | Number of configured models |
| Active access keys | Number of issued access keys |
Instance panel
Shows the status of each running gateway instance, one row per instance:
| Info | Description |
|---|---|
| Instance ID | 8-character prefix (click to copy the full UUID) |
| Task status | R=running, P=crashed (highlighted red), S=stopped |
| Memory level | Normal / Warning / Pressure / Reclaim / Critical; shows used/limit, level ≥ Warning shows a ⚠ warning |
| Disk usage | Total disk usage (hover to view each storage breakdown) |
The instance panel's status refreshes in real time: in a multi-instance deployment with Redis configured, a gracefully-drained instance disappears from the panel within about 1 second; a single-instance deployment without Redis refreshes via periodic polling with slightly higher latency.
Disk details (hover to expand)
Hover over "Disk usage" to expand detailed metrics for each storage area:
Log queue (by role: memory / drain / live):
- Disk usage / limit
- Write pointer position
- Segment count × segment size
- Retention floor
- DB drain cursor position (drain role)
- SSE consumer count and minimum cursor
Other storage areas:
- Response body storage (used / limit)
- Request body storage (used / limit)
- Accumulator storage (used / limit)
- Ingress spill storage (used / limit)
Gated-request panel
Shows requests currently queued or being processed (under memory pressure the gateway queues new requests):
| Column | Description |
|---|---|
| Time | Enqueue time |
| Instance | The processing instance ID (8 characters) |
| Client | Access key name |
| Method | POST, etc. |
| Path | Request path |
| Status | Queued / Active / Completed / Rejected (timeout or queue full) |
| Elapsed | Queue waiting duration |
- By default, only requests in Active and Queued status are shown.
- Click "Show all history" to see snapshots of Completed and Rejected.
Time-series charts
Two time-series charts at the bottom, supporting time-range presets: Today / 12h / 24h / 2d / 7d / 30d.
Token usage chart (left): a line chart grouped by model, with hover showing each model's input/output token breakdown. The legend shows the model name and total; click a legend entry to hide that line.
Request count chart (right): three lines for total / success / error, with hover showing the specific value of each metric.
FAQ
Q: The instance count in Overview doesn't match my deployment? Only instances currently processing requests are shown. Idle instances (no in-flight tasks) do not appear in the list.
Q: What if the memory level shows Critical? The gateway is actively reclaiming memory, which may affect request latency. Check for memory leaks or traffic spikes, and scale up if necessary.
Q: Gated requests stay Queued and never move? The gateway is under high memory pressure, and queued requests are waiting for processing resources. Check the memory level, and rate-limit or scale up if necessary.
Q: Time-series charts have no data? Time-series data comes from persisted statistics. If the gateway just started or the statistics feature is not enabled, the charts may be empty. Confirm that the statistics-related configuration is correct.
Next: Console login and roles for login and roles; Statistics for usage statistics; Log viewer for request details.