@action-flow/scenario
v0.1.0
Published
Isolated, typed debug scenarios for ActionFlow
Maintainers
Readme
@action-flow/scenario
Typed debug scenarios for @action-flow/core. Requires a Runtime created with environment: "development", environment: "test" or an explicit allowDebug: true.
npm install @action-flow/core @action-flow/scenarioimport { createRuntime, defineAction, defineModule } from "@action-flow/core";
import { createScenarioRunner } from "@action-flow/scenario";
const runtime = createRuntime({
providers: {}, environment: "development",
modules: [defineModule("math", { double: defineAction({ handler: (n: number) => n * 2 }) })],
});
const runner = createScenarioRunner(runtime);
const scenario = runner.defineScenario({ id: "double", action: "math.double", input: 3 });
const result = await runner.run(scenario);
if (result.status === "success") console.log(result.output); // number: 6Definitions support providers, mocks, setup, cleanup and mutually exclusive input / inputFactory. Input and return mocks are checked against the selected registered Action. providers and mocks can be async factories receiving a per-run context. A provider factory initially sees the Runtime's base providers; subsequent hooks see the scoped bindings. Call ctx.defer(release) immediately after allocating a fixture to release it even if setup later fails.
runner.runtime exposes the associated Runtime. runner.run(definition, { onTrace }) streams this run's live events only; synchronous and asynchronous observer failures are isolated. The separate @action-flow/devtools package uses this interface for its browser panel. Scenario does not import DevTools or access DOM globals.
Overrides and mocks propagate to nested Dispatch through an execution scope. They never replace Runtime bindings or handlers. Static objects retain their business object references; use factories for fresh mutable providers, inputs and mock values on repeated or concurrent runs. Overrides replace top-level bindings and are not deep merged or cloned.
The runner drains all started descendants before cleanup, then calls definition cleanup (if setup was entered) and all deferred releases in reverse order. Cancellation is cooperative and does not force a pending handler or hook to settle. Escaped calls made after scope closure reject. A caught or detached child failure stays in the trace; the root outcome determines execution success.
Results include status, scenarioId, runId, optional root IDs, a bounded local trace, droppedRecords and cleanupErrors. Failures preserve the thrown value. Cleanup failures return an AggregateError containing the primary failure (if any) followed by cleanup errors, and set status to error. Invalid/foreign definitions and disabled debug execution reject before lifecycle callbacks. The local trace defaults to 1,000 records and no payload capture, independently of Runtime retention and capture settings; configure createScenarioRunner(runtime, { trace: ... }) when needed.
ESM only. The sole runtime dependency is @action-flow/core. This package does not manage application state. Production debug opt-in is an execution policy, not authentication or a security boundary for an exposed remote endpoint. See the changelog for 0.x API changes.
