Batch jobs that finish where the next one starts.

Flow runs scheduled jobs, keeps their artifacts addressable, and refuses to start a job whose inputs are not ready yet. It is a small internal tool, not a platform.

What it does

• Runs jobs on a cron expression, one instance at a time per job.
• Stores each run's output under a content hash, so a rerun with identical inputs is free.
• Blocks a downstream job until every input it declared has a completed run.
• Keeps the last 30 days of logs and exit codes; older runs keep only their metadata.

What it deliberately does not do

• No distributed execution. One host runs every job, and that is the point.
• No UI for creating jobs. A job is a file in the repository, reviewed like any other change.
• No retries that hide a broken job: two attempts, then the run fails and stays failed.

Where to go next

The usage page shows a job definition and the two commands worth knowing. Current component state is on the status page.