Daemon

⚠️ Example only — placeholder content that demonstrates the docs platform. Not final; to be replaced by Agent A. Do not treat as canonical.

A Daemon is long-running or scheduled work — a poller, a supervisor, a periodic sync. You define it as a tag, then run it as a layer that ticks on a schedule and records its execution history.

Define and run

Daemon.Tag names the daemon and types its result. Daemon.layer runs it, driving the effect on a Polling cadence.

import * as Daemon from "hyperlink-ts/Daemon"
import { Polling } from "hyperlink-ts"
import { Duration, Effect, Schema } from "effect"

class Prices extends Daemon.Tag<Prices>()("app/Prices", {
  success: Schema.Number,
}) {}

const live = Daemon.layer(Prices, {
  effect: Effect.succeed(100_000),          // one tick's work
  polling: Polling.spaced(Duration.seconds(5)),
})

Execution history

Run under Daemon.layer and each terminal tick is auto-appended to the Soft journal as a typed event — Started, Completed, Failed, or Interrupted — which the dashboards read back as a timeline. Default is in-memory; override with an app Store.Service that registers Daemon.store(Prices) (see Stores).

import * as Store from "hyperlink-ts/Store"
import * as Daemon from "hyperlink-ts/Daemon"

class AppStore extends Store.Service<AppStore>("@app/Store")(
  Daemon.store(Prices),
) {}

const store = yield* AppStore.at(Prices)
const events = yield* store.events()       // [{ _tag: "Completed", success, … }, …]

Daemon.layer auto-writes history; a bare engine make without the layer does not. Reach for the layer form when you want the run timeline recorded. There is no ProcessStore.

Polling cadences

Polling shapes when a daemon ticks — fixed spacing, or accelerating toward an event and relaxing afterward.

Polling.spaced(Duration.seconds(30))       // steady 30s
Polling.accelerating({                     // fast near the action, slow when quiet
  fast: Duration.seconds(2),
  slow: Duration.minutes(5),
})

Edit this page on GitHub