Agent environment

​

Defer to a previous run

An Agent environment can defer to your production runs, so a background agent builds only the models it changed and reuses the rest from a previous Bolt run. This works like deferral in Bolt, but you set it once on the Agent environment and every agent that runs in it uses it.

​

Set it when you create the environment

In Workspace Settings > Environments, click New environment and create an Agent environment. The Defer to select lists your Bolt environments, plus No deferral. It starts on the workspace’s default Bolt environment, so a new Agent environment defers to that environment unless you change it.

​

Change it later

1. In Workspace Settings > Environments, open the menu on the Agent environment and select Environment settings. The Agent settings dialog opens.
2. Turn on Defer to a previous run.
3. Under Defer to, pick Environment or Schedule, then select the target:
   • Environment compares against the latest successful run of any schedule in that Bolt environment, so it keeps working when schedules are renamed or retired.
   • Schedule compares against the latest successful run of that one schedule.
4. Click Save. Save stays unavailable while Defer to a previous run is on and no target is selected.

An Agent environment that defers shows a defers to <name> badge on its row.

​

What the agent does with it

When an agent runs dbt™ in a deferring environment, Paradime sets DBT_DEFER=true and DBT_STATE to the target’s latest manifest, so plain dbt build or dbt run commands defer without extra flags.

• DBT_FAVOR_STATE isn’t set. The agent may add --favor-state itself for a one-off check against production data.
• Commands run through a wrapper script, such as make build or ./run.sh, don’t defer.

💡 This section contains details on how you can connect to each of the below Data Warehouse providers: