Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import fs from 'node:fs/promises'
import os from 'node:os'
import path from 'node:path'
import { AppService } from '../appService.js'
import { __setChatStoreBaseDirForTests } from '../chatStoreBinding.js'
// R1-M4 (desktop-app): cover the producer-side fallback branch of Decision #6 —
// `agentDisplayByName[name]` is filled total over `agentNames`, falling back to
// the identifier whenever the wire entry carries no usable `displayName`. Two
// under-tested inputs exercise it:
// (a) an agent with NO `displayName` field at all (pre-display API build), and
// (b) an agent with `displayName: ''` (empty string → the `candidate &&`
// guard treats it as absent).
// The catalog is derived from the real producer (AppService.refreshAccessCatalog,
// T1) rather than a hand-built map, so the assertion certifies the producer's own
// output, not the author's idea of it.
const ME = {
id: '00000000-0000-4000-8000-000000000001',
email: '[email protected]',
name: 'Test User',
picture: null,
teamId: '00000000-0000-4000-8000-0000000000aa',
teamName: 'Test Team',
role: 'member',
}
describe('AppService.refreshAccessCatalog — agentDisplayByName fallback (R1-M4)', () => {
let chatStoreBaseDir: string
beforeEach(async () => {
chatStoreBaseDir = await fs.mkdtemp(path.join(os.tmpdir(), 'clerum-app-service-display-'))
__setChatStoreBaseDirForTests(chatStoreBaseDir)
})
afterEach(async () => {
__setChatStoreBaseDirForTests(null)
await fs.rm(chatStoreBaseDir, { recursive: true, force: true })
})
it('falls back to the identifier for an agent with no displayName and for displayName: ""', async () => {
const service = new AppService() as any
service.sessionToken = 'session-token'
service.me = ME
service.rpcTokenManager = {
getOrIssue: vi.fn().mockResolvedValue({ token: 'rpc-token' }),
clear: vi.fn(),
}
service.authClient = {
getMe: vi.fn().mockResolvedValue(ME),
getMyContexts: vi.fn().mockResolvedValue({ contextIds: [] }),
getMyAgents: vi.fn().mockResolvedValue({
agentNames: ['no-display', 'empty-display', 'with-display'],
agents: [
// (a) pre-display build: the wire entry has no displayName field.
{ name: 'no-display', mcpServers: [] },
// (b) displayName present but empty → treated as absent by the guard.
{ name: 'empty-display', displayName: '', mcpServers: [] },
// Control: a real display maps through so the fallback is distinguishable.
{ name: 'with-display', displayName: 'With Display Host', mcpServers: [] },
],
}),
getTeamContexts: vi.fn().mockResolvedValue({ contextIds: [] }),
getTeamAgents: vi.fn().mockResolvedValue({ agentNames: [], agents: [] }),
}
const catalog = await service.refreshAccessCatalog()
// The map is total over agentNames — no undefined lookups.
expect(Object.keys(catalog.agentDisplayByName).sort()).toEqual([
'empty-display',
'no-display',
'with-display',
])
// (a) no displayName → id.
expect(catalog.agentDisplayByName['no-display']).toBe('no-display')
// (b) displayName: '' → id (empty string must NOT survive as the display).
expect(catalog.agentDisplayByName['empty-display']).toBe('empty-display')
// Control: a genuine display is preserved.
expect(catalog.agentDisplayByName['with-display']).toBe('With Display Host')
})
})