Oi means Orchid integration — lightweight glue layer between Orchid workflows and OrchidSymbiont runtimes.
-
Graph → Compile → Execute — three-phase pipeline:
- Build a
Graphof step nodes and edges Oi.compile(graph)— topology → staticCompiled(bundles + plan, reusable)Oi.execute(compiled, opts)— resolve inputs, apply orchid_adapters, run via pluggable executor
- Build a
-
Compiled — immutable compilation product. Compile once, dispatch many times with different inputs / interventions / executors.
-
Inputs vs Interventions — two kinds of external data:
inputs:—%{"io_key" => payload}seeded into drafting memory before dispatchinterventions:—%{{:port, node, port} => {type, payload}}merged into Orchid baggage, consumed byOrchid.Hook.ApplyInterventionsper-step
-
Executor — pluggable task fan-out per stage. Built-in:
Sync(serial),TaskSup(Task.Supervisor).Pool(NimblePool) planned. -
Session — optional multi-tenant isolation via
Oi.Runtime.Session, wrapping per-tenantOrchidSymbiont.Runtime+Task.Supervisor. -
Extension model — Oi intentionally does not implement its own workflow plugin system. Use Orchid hooks through
:orchid_optsfor step-level concerns such as auth, logging, metrics, retry, and caching.
defmodule MyApp.Steps.Upcase do
use Oi.Step, name: :upcase
manifest(
inputs: [:text],
outputs: [result: :string]
)
routine text, _opts do
text |> String.upcase() |> ok()
end
endalias Oi.Topology.Graph
alias Oi.Topology.Graph.Node
graph =
Graph.new()
|> Graph.add_node(%Node{
id: :up,
container: MyApp.Steps.Upcase,
inputs: [:text],
outputs: [:result]
})
{:ok, compiled} = Oi.compile(graph)
{:ok, result} = Oi.execute(compiled,
inputs: %{"up|text" => "hello"}
)
Oi.Result.reify(result, "up|result")
# => {:ok, "HELLO"}# Use vanilla orchid with custom opts(without orchid_adapters)
{:ok, result} = Oi.execute(compiled,
inputs: %{"step1|in" => "hello"},
interventions: %{
{:port, :step2, :in} => {:override, "forced_value"}
},
orchid_opts: [
global_hooks_stack: [Orchid.Hook.ApplyInterventions]
]
)Oi.Runtime.Session.start("tenant-1")
graph = build_graph()
{:ok, compiled} = Oi.compile(graph)
{:ok, result} = Oi.execute(compiled,
inputs: %{"step|in" => "data"},
executor: Oi.Executor.TaskSup,
executor_opts: [sup: Oi.Runtime.Session.tasks_tuple("tenant-1")],
orchid_baggage: %{scope_id: "tenant-1"},
orchid_opts: [global_hooks_stack: [OrchidSymbiont.Hooks.Injector]]
)
Oi.Runtime.Session.stop("tenant-1")defmodule MyApp.Steps.Predict do
use Oi.Step, name: :predict, symbiont?: true
manifest(
inputs: [:features],
outputs: [prediction: :string],
models: [:model]
)
routine features, models, _opts do
{:ok, result} = OrchidSymbiont.call(models.model, {:predict, features})
ok(result)
end
end- Graph compilation + staged dispatch
- Pluggable Executor (Sync / TaskSup)
- Multi-tenant Session
- Oi.Step macro DSL
- Interventions (via Orchid baggage)
- Executor.Pool (NimblePool for GPU-intensive tasks)
MIT