Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import { expect, test } from '@playwright/test'
import {
E2E_EMAIL,
RECIPE_NS,
cleanupRecipeRuntimeState,
clearSession,
launchAndLogin,
loginAs,
openWorkflowsPage,
rendererListWorkflowRuns,
seedAllowlist,
selectWorkflow,
shortRunId,
waitForNewRun,
} from './workflowUi'
const RECIPE_NAME = 'e2e-ondemand-simple'
test('desktop-triggered runs are attributed to the logged-in user', async () => {
await clearSession()
cleanupRecipeRuntimeState(RECIPE_NAME)
const auth = await loginAs(E2E_EMAIL)
seedAllowlist(auth.userId, RECIPE_NAME)
const { app, page } = await launchAndLogin()
try {
await openWorkflowsPage(page)
const detailCard = await selectWorkflow(page, RECIPE_NAME, RECIPE_NS)
const beforeRuns = await rendererListWorkflowRuns(page, RECIPE_NS, RECIPE_NAME, 20)
const beforeIds = beforeRuns.items.map(item => item.id)
await detailCard.getByRole('button', { name: 'Trigger' }).click()
const newRun = await waitForNewRun(page, RECIPE_NS, RECIPE_NAME, beforeIds)
expect(newRun.actor?.type).toBe('user-session')
expect(newRun.actor?.userId).toBe(auth.userId)
await detailCard.getByRole('button', { name: 'Refresh' }).click()
await expect(
detailCard
.locator('.context-resource-row')
.filter({ hasText: shortRunId(newRun.id) })
.first()
).toBeVisible({ timeout: 30_000 })
} finally {
await app.close()
cleanupRecipeRuntimeState(RECIPE_NAME)
}
})
test('desktop-triggered runs eventually carry an executionRef', async () => {
await clearSession()
cleanupRecipeRuntimeState(RECIPE_NAME)
const auth = await loginAs(E2E_EMAIL)
seedAllowlist(auth.userId, RECIPE_NAME)
const { app, page } = await launchAndLogin()
try {
await openWorkflowsPage(page)
const detailCard = await selectWorkflow(page, RECIPE_NAME, RECIPE_NS)
const beforeRuns = await rendererListWorkflowRuns(page, RECIPE_NS, RECIPE_NAME, 20)
const beforeIds = beforeRuns.items.map(item => item.id)
await detailCard.getByRole('button', { name: 'Trigger' }).click()
const newRun = await waitForNewRun(page, RECIPE_NS, RECIPE_NAME, beforeIds)
let claimedRun = newRun
await expect
.poll(
async () => {
const runs = await rendererListWorkflowRuns(page, RECIPE_NS, RECIPE_NAME, 20)
const match = runs.items.find(item => item.id === newRun.id)
if (match?.executionRef?.namespace && match?.executionRef?.name) {
claimedRun = match
return `${match.executionRef.namespace}/${match.executionRef.name}`
}
return null
},
{
timeout: 60_000,
intervals: [500, 1_000, 2_000],
message: `run ${newRun.id} did not claim an executionRef`,
}
)
.not.toBeNull()
expect(claimedRun.executionRef).not.toBeNull()
expect(claimedRun.executionRef?.namespace).toBe(RECIPE_NS)
// Per docs/architecture/workflow-recipe-naming.md, WRC's child-recipe
// naming uses an 8-char short run-id suffix (first hex octet of the
// UUIDv4) — NOT the full 36-char UUID — so the name is 63-byte-safe for
// every downstream K8s label surface. Match against the short suffix
// only; `runUUID[:8]` is unique across `workflow_runs` per parent
// thanks to idx_wr_idempotency, so it is a sufficient identifier.
const shortRunId = newRun.id.toLowerCase().slice(0, 8)
expect(claimedRun.executionRef?.name).toContain(shortRunId)
} finally {
await app.close()
cleanupRecipeRuntimeState(RECIPE_NAME)
}
})