AI-Assisted Workflows: Install the Konduktor Skills plugin to get AI help drafting task YAMLs (
/draft-task-yaml), writing Python SDK code (/draft-python-task), and constructing CLI commands (/konduktor-cli).npx skills add Trainy-ai/konduktor-skills
Create a Job
konduktor launch my_job.yaml
Fetch Job Statuses
konduktor status
konduktor status --limit <INT> # gets only <INT> most recent jobs
konduktor status --before <TIMESTAMP> # gets only jobs before <TIMESTAMP>
konduktor status --after <TIMESTAMP> # gets only jobs after <TIMESTAMP>
konduktor status --all-users # gets all jobs for all users
Stream Logs
konduktor logs <JOB_ID>
konduktor logs <JOB_ID> --no-follow # gets all logs without streaming
konduktor logs <JOB_ID> --num-lines <INT> # num of lines from the end of log file to show
konduktor logs <JOB_ID> --start-offset <TIME> # how much time from now to look back
konduktor logs <JOB_ID> --node-rank <INT> # node rank to tail logs from
Using with Claude Code: Ask Claude Code to run
konduktor logs <JOB_ID> as a background task. Claude Code can monitor the log output and identify errors, exceptions, or issues in your training runs while you continue working on other tasks.Suspend a Job
konduktor stop <JOB_ID>
konduktor stop --all # suspend all jobs
Unsuspend a Job
konduktor start <JOB_ID>
konduktor start --all # unsuspend all jobs
Delete a Job
konduktor down <JOB_ID>
konduktor down --all # delete all jobs
Manage Secrets
konduktor secret create <FLAGS> <SECRET_NAME>
konduktor secret delete <SECRET_NAME>
konduktor secret list
Manage Serving Deployments
konduktor serve launch my_deployment.yaml
konduktor serve down my_deployment.yaml
konduktor serve status
Enable Cloud Storage Filesync
konduktor check <CLOUD_STORAGE> # from {gs, s3}