Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
/**
* E2E -- HostApprovalSection editor flow
*
* Validates the per-tool approval override UI on the host detail page:
* Scenario 1: Toggle http_request → Skip → Save → override persists.
* Scenario 2: Toggle http_request back to Default → Save → override clears.
* Scenario 3: Cancel discards the in-flight shell_exec change.
*
* Prerequisites:
* 1. Port-forwards running (control-ui :3000, control-api :8090)
* 2. Admin credentials: ADMIN_USER, ADMIN_PASS
* 3. A host named E2E_HOST_NAME (default "chatllm") exists in the cluster
*/
import { type Page, expect, test } from '@playwright/test'
const BASE_API = process.env.CONTROL_API_URL || 'http://localhost:8090'
const BASE_UI = process.env.CONTROL_UI_URL || 'http://localhost:3000'
const ADMIN_USER = process.env.ADMIN_USER || 'admin'
const ADMIN_PASS = process.env.ADMIN_PASS || 'changeme123!'
const HOST_NAME = process.env.E2E_HOST_NAME || 'chatllm'
async function getToken(page: Page): Promise<string> {
return page.evaluate(() => localStorage.getItem('controlUiAdminToken') ?? '')
}
async function api(
token: string,
method: string,
path: string,
body?: Record<string, unknown>
): Promise<{ status: number; data: Record<string, unknown> }> {
const opts: RequestInit = {
method,
headers: {
Authorization: `Bearer ${token}`,
'Content-Type': 'application/json',
},
}
if (body) opts.body = JSON.stringify(body)
const resp = await fetch(`${BASE_API}${path}`, opts)
const text = await resp.text()
let data: Record<string, unknown> = {}
try {
data = JSON.parse(text)
} catch {
data = { raw: text }
}
return { status: resp.status, data }
}
async function login(page: Page) {
await page.goto(BASE_UI)
await page.waitForSelector('text=Sign in', { timeout: 15_000 })
const inputs = page.locator('input')
await inputs.nth(0).fill(ADMIN_USER)
await inputs.nth(1).fill(ADMIN_PASS)
await page.locator('button:has-text("Sign in")').last().click()
await page.waitForSelector('text=Marketplace', { timeout: 20_000 })
}
/**
* Enter edit mode when the embedding surface did not open it by default.
*/
async function clickSectionEdit(page: Page) {
if ((await page.locator('#approval-http_request').count()) > 0) return
await page
.locator('.cu-host-approval-section__actions')
.locator('button:has-text("Edit")')
.click()
}
/** Navigate to the host detail page and wait for the Per-tool approval section to load. */
async function gotoHostDetails(page: Page) {
// The Per-tool approval editor now lives under the consolidated Advanced
// tab. The old `/approvals` slug is still served via a 308 redirect, but
// navigating to the canonical URL directly keeps this E2E independent of
// that compatibility shim. Advanced opens on the Hooks sub-tab, so select
// Per-tool approval explicitly — the guardrails panel reuses the
// `.cu-host-approval-section` class, and waiting on it before switching
// sub-tabs would match the wrong section.
await page.goto(`${BASE_UI}/hosts/${encodeURIComponent(HOST_NAME)}/advanced`)
await page.locator('.cu-advanced-tabs button:has-text("Per-tool approval")').click()
await page.waitForSelector('.cu-host-approval-section', { timeout: 20_000 })
}
/**
* Reset approval tools for the test host via the API so each scenario starts
* from a known state (no overrides). Tolerates 404 when the host has no
* approval subtree yet.
*/
async function clearApprovalTools(token: string) {
// Fetch current host spec
const { status, data } = await api(token, 'GET', `/api/v1/admin/hosts/${HOST_NAME}`)
if (status !== 200) return // host not found — skip cleanup
const spec = (data as { spec?: Record<string, unknown> }).spec ?? {}
const approval = (spec.approval as Record<string, unknown> | undefined) ?? {}
// Write back without the tools sub-key
const { tools: _removed, ...approvalWithoutTools } = approval
const nextApproval =
Object.keys(approvalWithoutTools).length > 0 ? approvalWithoutTools : undefined
const nextSpec = nextApproval !== undefined ? { ...spec, approval: nextApproval } : { ...spec }
// Remove approval key entirely when empty
if (nextApproval === undefined) {
delete nextSpec.approval
}
await api(token, 'PUT', `/api/v1/admin/hosts/${HOST_NAME}`, { spec: nextSpec })
}
test.describe('HostApprovalSection E2E -- per-tool override editor flow', () => {
test.describe.configure({ mode: 'serial' })
let page: Page
let token: string
test.beforeAll(async ({ browser }) => {
page = await browser.newPage()
})
test.afterAll(async () => {
await page?.close()
})
// ── HAT0. Login ──────────────────────────────────────────────────────────
test('HAT0. Admin login', async () => {
await login(page)
token = await getToken(page)
expect(token).toBeTruthy()
})
// ── HAT1. Toggle http_request → Skip → Save ─────────────────────────────
test('HAT1. Toggle http_request to Skip, Save — override persists with warning icon', async () => {
test.skip(!token, 'Login failed — skipping')
// Start from clean slate
await clearApprovalTools(token)
await gotoHostDetails(page)
await clickSectionEdit(page)
// Wait for the edit-mode selects to appear
await page.waitForSelector('#approval-http_request', { timeout: 10_000 })
// Verify warning icon is NOT yet visible (we haven't changed to Skip)
// The warning <span> appears when state === 'skip' && codeDefault === 'required'
// Its title contains CLERUM_HTTP_ALLOWLIST
const warningBeforeChange = page.locator('[title*="CLERUM_HTTP_ALLOWLIST"]')
expect(await warningBeforeChange.count()).toBe(0)
// Change http_request from "Default" to "Skip"
await page.selectOption('#approval-http_request', 'skip')
// Warning icon should now appear (title="Skipping approval relies on CLERUM_HTTP_ALLOWLIST…")
await page.waitForSelector('[title*="CLERUM_HTTP_ALLOWLIST"]', { timeout: 5_000 })
const warningIcon = page.locator('[title*="CLERUM_HTTP_ALLOWLIST"]')
expect(await warningIcon.count()).toBeGreaterThan(0)
// Click Save (enabled because isDirty)
const saveBtn = page
.locator('button.cu-btn--primary:has-text("Save"), button:has-text("Save")')
.filter({ hasNot: page.locator(':disabled') })
.last()
await saveBtn.click()
await expect(page.locator('#approval-http_request')).toHaveValue('skip')
await expect(
page.locator('.cu-host-approval-section__actions button:has-text("Save")')
).toBeDisabled()
// Bonus: verify via API that the CRD was patched
const { status, data } = await api(token, 'GET', `/api/v1/admin/hosts/${HOST_NAME}`)
if (status === 200) {
const tools = (data as { spec?: { approval?: { tools?: Record<string, boolean> } } }).spec
?.approval?.tools
// http_request = false means Skip in the CRD
if (tools !== undefined) {
expect(tools['http_request']).toBe(false)
}
}
})
// ── HAT2. Toggle back to Default → Save → empty state ───────────────────
test('HAT2. Toggle http_request back to Default, Save — override clears', async () => {
test.skip(!token, 'Login failed — skipping')
await gotoHostDetails(page)
// Click Edit
await clickSectionEdit(page)
// Wait for the select to appear
await page.waitForSelector('#approval-http_request', { timeout: 10_000 })
// The http_request row should currently be "skip" (set by HAT1)
const currentValue = await page.locator('#approval-http_request').inputValue()
// If the cluster already has the value, proceed; otherwise the test is still valid
// — we change it to Default either way.
expect(['skip', 'default', 'required']).toContain(currentValue)
// Change back to Default
await page.selectOption('#approval-http_request', 'default')
// Save button: only enabled when dirty
const saveBtn = page
.locator('button.cu-btn--primary:has-text("Save"), button:has-text("Save")')
.filter({ hasNot: page.locator(':disabled') })
.last()
await saveBtn.click()
// Wait for re-render
await page.waitForTimeout(2_000)
await expect(page.locator('#approval-http_request')).toHaveValue('default')
// API verification
const { status, data } = await api(token, 'GET', `/api/v1/admin/hosts/${HOST_NAME}`)
if (status === 200) {
const tools = (data as { spec?: { approval?: { tools?: Record<string, boolean> } } }).spec
?.approval?.tools
if (tools !== undefined) {
// http_request key should be absent (Default means no override)
expect(Object.prototype.hasOwnProperty.call(tools, 'http_request')).toBe(false)
}
}
})
// ── HAT3. Cancel discards in-flight changes ───────────────────────────────
test('HAT3. Cancel discards shell_exec change without saving', async () => {
test.skip(!token, 'Login failed — skipping')
// Ensure clean state before this scenario
await clearApprovalTools(token)
await gotoHostDetails(page)
await clickSectionEdit(page)
await page.waitForSelector('#approval-shell_exec', { timeout: 10_000 })
const valueBefore = await page.locator('#approval-shell_exec').inputValue()
// Change shell_exec to Skip (a dirty change)
await page.selectOption('#approval-shell_exec', 'skip')
// Click Cancel
await page.locator('button:has-text("Cancel")').last().click()
await expect(page.locator('#approval-shell_exec')).toHaveValue(valueBefore)
await expect(
page.locator('.cu-host-approval-section__actions button:has-text("Save")')
).toBeDisabled()
})
})