Job templates

A job template composes reusable tasks into a directed acyclic graph. Creating a job resolves template arguments, snapshots task definitions, binds stored connectors, creates dashboard instances, and seeds job configuration.

Structure

name: ticket-migration
description: Migrate tickets and attachments
 
connectors:
  source:
    type: truto
  destination:
    type: postgres
  mapping_db:
    type: sqlite
 
arguments:
  account_id:
    type: string
    required: true
 
job_config:
  migration:
    type: form
    title: Migration settings
    fields:
      - key: dry_run
        label: Preview only
        type: boolean
        default: true
 
tasks:
  prepare:
    task: prepare-destination
    connector_mapping:
      destination: destination
 
  tickets:
    task: migrate-tickets
    depends_on: [prepare]
    connector_mapping:
      source: source
      destination: destination
      mapping_db: mapping_db
    arguments:
      integrated_account_id: ${account_id}
 
on_failure: pause

Top-level fields

• name — stable template name.
• description — user-facing purpose.
• connectors — template connector names and types.
• arguments — create-time values substituted into the snapshot.
• job_config — runtime-editable schema copied into each job.
• tasks — task nodes keyed by job task key.
• on_failure — pause (default), abort, or continue.

Dashboard templates are associated with the job template as separate records and are instantiated with the job.

Task node fields

• task — referenced task definition name.
• depends_on — array of upstream task keys.
• connector_mapping — task slot to template connector name.
• arguments — task runtime arguments.
• description — optional node-specific description.
• type — optional command or validation behavior; omit for a regular task.

Task keys must be unique. Dependencies must exist and cannot form a cycle.

Failure behavior

Job execution schedules tasks when their dependencies satisfy the graph rules. A failed dependency can prevent downstream work. on_failure can pause for operator intervention, abort the graph, or allow independent work to continue.

Stopping a job does not roll back completed external writes.

Arguments and configuration

Use:

• template arguments for values fixed at job creation;
• job configuration for values editable between runs;
• task-node arguments for binding the first two into a particular task.

Targeted runs

A run can target selected task keys. Dependencies are included by default. skip_deps: true is an operator assertion that required upstream state already exists.

Snapshot behavior

Creating a job captures the current template and task versions. Starting a run captures the current job configuration. Later edits do not mutate historical jobs or runs.

An upgrade should compare resolved snapshots, not only top-level template version names.

Validation

Template validation checks:

• YAML shape;
• referenced tasks;
• connector type compatibility;
• connector slot mappings;
• missing and cyclic dependencies;
• argument definitions and substitutions;
• task override targets;
• dashboard definitions where applicable.

It cannot validate target credentials or external side effects.

See Override a task in a job template and Job templates and jobs.