Admin

Check gateway status below. Open Model operations to register or update a version.

Connection settings
Model operations Register → Promote → Warm pool

Register model versions and stages. Load and unload update the gateway warm pool; engine memory management remains with the backend.

  1. Enter a model ID and version, then Register.
  2. Choose a stage and Promote to update its metadata.
  3. Load / Unload changes the gateway warm-pool record, not GPU memory.

Registered versions

No registered versions yet.

Gateway health♥
—
/health · v? · up 0s
Batch size→
—
Average requests per micro-batch
Response cache hit rate◫
—
Gateway responses, not GPU KV tensors
Default backend≡
—
Backend adapter default
Advanced diagnostics Alerts, routing and release controls

GPU performance

Live GPU telemetry is not connected to this console. No synthetic chart is displayed as live data.

View the controlled scheduler experiment →

⚠ Active Alerts

Rule engine over live platform state.

Alert status unavailable
unknown
No active alert is not a GPU readiness or inference success check.

▷ Deployment Controls

Canary releases with auto-rollback on error-rate breach.

≡ Backend Adapter

Live backend routing and health.

◫ Warm Pool

—

→ Route Distribution

Requests handled per backend.

⟳ Releases & Registry

Deployment history and governed model versions.