Every workflow begins with one Start node. Its only job is to decide when the workflow runs. The chat panel walks you through every choice in plain language and lets the AI fill in the technical bits (CRON expressions, file event flags, schedule descriptions) for you.
You pick one of three trigger types:
The workflow only runs when you click "Run" yourself. This is the default and is the right choice while you're building or iterating on a workflow.
The workflow runs automatically on a recurring schedule.
The workflow runs whenever a watched file or directory changes.
Scheduled workflows can fire on:
The chat builds a CRON expression and a human-readable description ("Every Monday at 9:30 am") side by side, so you can confirm the schedule reads the way you intended before saving.
Inputs are validated inline:
HH:MM form.You pick a path with the native file picker (single file or a whole directory). Then you pick which file-system events should fire the workflow:
Multi-select is allowed — for example, a watcher can fire on both "add" and "change". The available event list narrows based on whether the path you picked is a file or a directory.
At execution time, the Start node itself is a no-op — it doesn't read or write data. Its real role is to be registered with DagFlux's trigger system at startup, so that scheduled or file-watch triggers can fire the workflow even when you have it closed. When you change the trigger configuration, DagFlux automatically unregisters the old trigger before registering the new one, so you never end up with orphaned background jobs from a previous setup.
The trigger system never starts a second run of the same workflow while a previous run is still in progress. New trigger fires that arrive during an active run are skipped, and you'll see them noted in the execution history.