Usage
The Usage panel provides detailed analytics on how your API keys and controllers consume compute. Open it by clicking Usage in the navigation bar (visible when signed in).
Overview Stats
The top section shows four summary cards for the selected time window:
| Metric | Description |
|---|---|
| Jobs Submitted | Total submitted, with completed and still-running counts. |
| Core-hours Billed | Charged usage across managed and BYOC workloads. |
| Controller Jobs | Jobs that ran on BYOC controllers vs. managed compute. |
| Failure Pressure | Sum of failed and cancelled jobs in the window. |
Timeline
Below the overview cards, a 14-day trend chart plots daily submitted jobs (bars) and core-hours (line). Use the Download CSV button to export the raw data for your own analysis.
Time Windows
The time-window switcher in the hero bar lets you scope all metrics to:
- 24 hours — recent activity snapshot
- 7 days — weekly view
- 30 days — monthly roll-up (default)
- All time — lifetime usage
Switching windows instantly recalculates every card, chart, and metric on the page.
API Key Activity
Each API key that has submitted jobs gets a card showing its name, metrics, and attribution.
Lab workstation
sk-av••••••••x8m2
Unattributed jobs (submitted before key attribution was added) appear as a separate card so nothing is lost.
Controller Load
Every connected BYOC controller gets a card showing its status, capacity, and handled workload.
dgx-01
64 cores · 256 GB memory
If no controllers are connected yet, a placeholder card indicates when you can expect data to appear after your first BYOC controller comes online.
Session Chip
The top-right chip displays your account name, email, and current core-hour balance — the same chip shown across all dashboard pages.
Download the timeline as CSV to build your own reports or feed the data into external tooling.
Jobs submitted through the Python SDK carry your API key automatically. If a job lacks a key attribution (legacy submissions), it appears under the Unattributed card.