monitor-experiment
Monitor Beaker experiments until completion. Use when the user asks to monitor, check, or track a Beaker experiment.
When & Why to Use This Skill
The Monitor Beaker Experiment skill automates the tracking and management of computational jobs on the Beaker platform. It streamlines research workflows by providing real-time status updates, automated polling, and instant log retrieval upon experiment failure, significantly reducing the manual effort required for experiment oversight.
Use Cases
- Automated tracking of long-running machine learning training sessions, providing notifications exactly when the process finishes.
- Efficient debugging of failed experiments by automatically fetching and displaying error logs when a non-zero exit code is detected.
- Real-time progress monitoring of batch processing jobs on Beaker clusters without requiring manual CLI status checks.
- Hands-free experiment management where the agent handles the 'wait-and-retry' logic until specific success criteria are met.
| name | monitor-experiment |
|---|---|
| description | Monitor Beaker experiments until completion. Use when the user asks to monitor, check, or track a Beaker experiment. |
| allowed-tools | Bash(beaker:*) |
Monitor Beaker Experiment
Instructions
When monitoring a Beaker experiment:
- Get the experiment status using
beaker experiment get <experiment-id> - Check if the experiment has completed by looking at
status.exited - If still running, wait 30 seconds and check again
- When complete:
- If exitCode is 0: Report success
- If exitCode is non-zero: Fetch and display logs with
beaker experiment logs <experiment-id>
- Continue monitoring until the experiment finishes or the user asks you to stop
Examples
Check experiment status:
beaker experiment get 01KCW39T5JBZTYV69BXHWJJ83P
Get experiment logs on failure:
beaker experiment logs 01KCW39T5JBZTYV69BXHWJJ83P
Stream logs in real-time for running experiments:
beaker experiment logs --follow 01KCW39T5JBZTYV69BXHWJJ83P