Reads the current stack slug from application config.

A "stack" is one logical cyclium cluster sharing a database with other clusters. The slug identifies which cluster owns a given episode, trigger request, or workflow instance and is used by Recovery and the trigger-request poller to avoid stealing work across clusters.

Stacks and supervised actors

Cyclium itself exposes no per-actor DSL option for stacks — the slug is read once per cluster and stamped onto every row that cluster's actors produce. Which actors actually start on a given cluster is the consumer's decision, implemented in their own supervisor.

The simplest pattern is to keep a stack-specific actor list per deployment:

# On the stack_a cluster's host app:
config :cyclium, :stack_slug, "stack_a"
config :my_app, :cyclium_actors, [StackAOnlyActor, SharedActor]

# On the stack_b cluster's host app:
config :cyclium, :stack_slug, "stack_b"
config :my_app, :cyclium_actors, [StackBOnlyActor, SharedActor]

A richer pattern declares the allowed stacks on each actor's child-spec and has the supervisor filter the list at init time:

config :my_app, :cyclium_actors, [
  {SharedActor, []},
  {StackAOnlyActor, stacks: [:stack_a]},
  {CrossStackActor, stacks: [:stack_a, :stack_b]}
]

# In the supervisor's init/1:
stack = Application.get_env(:cyclium, :stack_slug)
actors = Enum.filter(children, &actor_matches_stack?(&1, stack))

Either way, every actor that does end up starting inherits that cluster's :stack_slug. Any episode, workflow instance, or deferred trigger request those actors create is stamped with the slug at insertion time, so Recovery on that cluster only sees its own work and does not reschedule episodes whose in-memory state lives on another cluster's nodes.

Configuration

Set the slug statically:

config :cyclium, :stack_slug, "stack_a"

or from an env var so the same release can be deployed into multiple stacks:

config :cyclium, :stack_slug, System.get_env("CYCLIUM_STACK_SLUG")

A nil / unset slug means "single-stack deploy" — rows are stamped NULL and Recovery is unscoped (current behavior for consumers that have not adopted multi-stack).