Record Spawner

The Record Spawner primitive creates an entity record when a trigger fires, using a deterministic id template so retries are idempotent. It can also delete that record when a paired clear_when: predicate becomes true — the "spawn-when-broken, clear-when-recovered" pattern — and can optionally spawn a paired FSM job in the same step.

Configuration

Field	Type	Description
`spawn_entity`	string	The entity to create.
`id_template`	string	Mustache-style template that produces the spawned record's id (e.g. `inv-alert-{{entity.id}}`). Determinism is what makes upserts safe.
`payload_template`	object	Mustache-rendered JSON object that becomes the spawned record's payload.
`on_collision`	enum	`upsert` (overwrite if exists), `skip` (no-op if exists), or `error`.
`clear_when`	JSONLogic object	Optional. When the predicate evaluates true, the spawned record is deleted instead of created.
`also_spawn_job`	object	Optional. `{ schema_name, initial_state }` — also creates a job bound to the spawned record.

Example: Inventory low-stock alert (spawn when below reorder point, clear when restocked)

[[bindings]]
primitive = "fastyoke.record_spawner"
trigger = {
  kind = "on_field_change",
  entity = "InventoryItem",
  field = "qty_on_hand",
  when = { "<": [{ var = "after.qty_on_hand" }, { var = "after.reorder_point" }] }
}
target_entity = "InventoryItem"
config = {
  spawn_entity = "InventoryAlert",
  id_template = "inv-alert-{{entity.id}}",
  payload_template = {
    item_id = "{{entity.id}}",
    on_hand = "{{after.qty_on_hand}}",
    reorder_point = "{{entity.reorder_point}}",
    status = "open"
  },
  on_collision = "upsert",
  clear_when = {
    ">=": [{ var = "after.qty_on_hand" }, { var = "after.reorder_point" }]
  }
}

When qty_on_hand drops below reorder_point, the engine upserts an InventoryAlert row with a deterministic id. When qty_on_hand rises back above reorder_point later, the same row is deleted. Retrying the original event is safe because the id is derived from the source record.

When to use it

Watchdog patterns: spawn an alert row when a condition goes bad, clear it when the condition recovers.
Cascade workflows: spawn a follow-up record (and its job) as a side effect of a transition.
Bulk materialization: pair with Scheduler to run a periodic sweep that creates rows for everything matching a predicate.

What it can't do

Trigger on entity reads. Reads must not create records; only write-side triggers are supported.
Resolve multi-hop FK paths. V1 paths are single-hop.
Bypass the cascade budget. Each spawned record fires its own on_record_create bindings — the engine caps total invocations per originating event to keep accidental loops from running away.