Guides
Ports, Logs, and Metrics
Expose a port safely, tail the provision log live, and confirm GPU telemetry.
Port mappings
Default instance has no public ports beyond SSH. To expose (e.g. Jupyter, API):
- Instance → port-mappings → Create. Label must be unique.
- Dashboard returns mapping + access ticket (temporary credential).
- Delete mapping when done. Label conflict returns
409.
Tickets expire — re-issue, don't screenshot-share old ones.
Provision logs
- Snapshot: instance → logs. Static view.
- Stream: logs/stream (SSE). Live tail during create. If stuck in provision, stream tells why — paste relevant lines to Support, not whole dump.
Metrics
metrics/history,gpu-metrics,gpu-metrics/history— CPU/RAM/GPU telemetry per instance.- Batch endpoint aggregates across owned instances for dashboard cards.
- Verify GPU live with
nvidia-smi/nvtopafter launch. See Connect and Verify.
Console and desktop
Web console / SSH-WS / desktop-launch exist but may sit behind beta flag. Disabled = FeatureComingSoon screen, not a bug. SSH remains the supported path.