Skip to main content

​
Overview

Scheduled Jobs let Dexter run recurring tasks automatically — simulations, reports, data pipeline refreshes, standard analysis workflows. Each scheduled job runs a full Dexter conversation on the cadence you specify, so it has access to every capability Dexter has interactively. This is useful whenever you find yourself doing the same work on a regular cadence: a Monday morning simulation, a weekly exec-review report, a nightly pipeline re-run against fresh warehouse data. Set it up once and it runs itself.

​
Common Use Cases

  • Run a simulation every Monday morning and generate a PDF report with the results
  • Produce a weekly summary of the latest KPIs for distribution
  • Execute a standard analysis workflow on a recurring cadence
  • Re-run a data pipeline when fresh data is uploaded on a regular cycle
  • Monitor for anomalies in the latest run and flag them for review

​
How Scheduling Works

Scheduled jobs use cron expressions for flexible timing — daily, weekly, specific days and times, or custom intervals. Cron gives you precise control (e.g., “every weekday at 7am,” “the first of every month at midnight”) without being tied to a preset list of schedules. When a scheduled job fires, Dexter opens a conversation, executes the task you described, and saves any outputs (reports, exports, model changes) as if you’d run it yourself. Scheduled runs appear in your conversation history the same way interactive runs do, so you can review what Dexter did, what it produced, and whether any questions came up that need your follow-up.

​
Setting Up a Scheduled Job

​
Monitoring and Managing Jobs

​
Notifications

​
Best Practices

  • Keep prompts specific. The clearer the job prompt, the more reliable the run. “Generate a PDF report on last week’s throughput” is better than “send me an update.”
  • Include context Dexter might need. If the job depends on specific factories, models, or data sources, name them in the prompt so Dexter doesn’t have to guess.
  • Review early runs. For a newly-created job, review the first one or two runs in the conversation history to confirm Dexter is doing what you expect. Tune the prompt if anything’s off.
  • Use the Persistent Knowledge Base. The more Dexter knows about your operation (naming conventions, shift patterns, KPI targets), the better scheduled runs will match your expectations without re-explaining context every time.