The all-in-one local AI studio for your desktop: chat, image and video generation and a coding agent in one free, open source app. Windows and Linux. No Docker, no terminal, no cloud required.
TypeScript agentai-chatai-studioall-in-onecoding-agentdesktop-appfluximage-generationlinuxllmlocal-ailocal-llmopen-sourceprivacyself-hostedstable-diffusionuncensoreduncensored-aivideo-generationwindows
/**
* A14 (2.6.8), der Inhaber: "Use auf einer LU-Engine-Kachel, waehrend Ollama aktiv
* ist: schaltet den Chat-Provider auf die LU Engine um. Wenn die LU Engine als
* Provider gar nicht angelegt ist, muss Use sie anlegen."
*
* Both halves are one write, because `ProviderId` is a fixed set of four slots
* and every OpenAI-protocol backend shares the `openai` one: there is no fifth
* entry to add, taking the slot IS setting the engine up. This exercises the
* real store, so what is proven is the state the app is left in, not the shape
* of the call.
*
* Run: npx vitest run src/api/__tests__/lu-engine-switch.test.ts
*/
import { describe, it, expect, beforeEach, vi } from 'vitest'
vi.mock('../providers/registry', () => ({ clearProviderCache: vi.fn() }))
vi.mock('../backend', () => ({
isTauri: () => false,
backendCall: vi.fn(async () => null),
secretGet: vi.fn().mockRejectedValue(new Error('no vault')),
secretSet: vi.fn(),
secretDelete: vi.fn(),
}))
const { ensureLuEngineIsChatProvider, LU_ENGINE_SWITCH_NOTE } = await import('../lu-engine-switch')
const { useProviderStore } = await import('../../stores/providerStore')
const JAN = { enabled: true, name: 'Jan', baseUrl: 'http://localhost:1337/v1', isLocal: true, managed: false, displaced: undefined }
function slot() {
return useProviderStore.getState().providers.openai
}
beforeEach(() => {
useProviderStore.getState().resetProvidersToDefaults()
})
describe('a pick on an LU Engine model hands it the chat slot', () => {
it('turns an off slot back into the engine and says the user has to be told', () => {
// The Mac in the report: Ollama in front, the openai slot parked by
// onboarding, so Settings shows no engine card at all.
useProviderStore.getState().setProviderConfig('openai', { enabled: false, managed: false })
expect(ensureLuEngineIsChatProvider()).toBe(true)
expect(slot().enabled).toBe(true)
expect(slot().managed).toBe(true)
expect(slot().name).toBe('LU Engine')
expect(slot().baseUrl).toBe('http://127.0.0.1:8127/v1')
})
it('takes the slot from another backend and leaves that one a way back', () => {
useProviderStore.getState().setProviderConfig('openai', JAN)
expect(ensureLuEngineIsChatProvider()).toBe(true)
expect(slot().managed).toBe(true)
expect(slot().name).toBe('LU Engine')
// The displaced backend keeps a standby card with an Enable button, the
// same route the provider card uses. Without this the user's own entry
// would vanish without a word, which is the bug R10 exists for.
expect(slot().displaced?.name).toBe('Jan')
expect(slot().displaced?.baseUrl).toBe('http://localhost:1337/v1')
})
// NEGATIVE CONTROL: an engine that already holds the slot is left completely
// alone. Writing the config again would restart the standby bookkeeping and
// could drop a card that is still owed, and the user would be told about a
// switch that never happened.
it('changes nothing and says nothing when the engine already holds the slot', () => {
useProviderStore.getState().setProviderConfig('openai', {
enabled: true, managed: true, name: 'LU Engine',
displaced: { name: 'Jan', baseUrl: 'http://localhost:1337/v1', isLocal: true },
})
const before = { ...slot() }
expect(ensureLuEngineIsChatProvider()).toBe(false)
expect(slot()).toEqual(before)
expect(slot().displaced?.name).toBe('Jan')
})
// NEGATIVE CONTROL: Ollama lives in its own slot and is not the thing being
// replaced. Switching it off here would take the user's other models away
// from him for no reason, and the way back is his provider card.
it('leaves Ollama switched on', () => {
useProviderStore.getState().setProviderConfig('ollama', { enabled: true })
useProviderStore.getState().setProviderConfig('openai', { enabled: false, managed: false })
ensureLuEngineIsChatProvider()
expect(useProviderStore.getState().providers.ollama.enabled).toBe(true)
})
it('the sentence the user reads names the engine and the reason', () => {
expect(LU_ENGINE_SWITCH_NOTE).toBe('Switched your chat provider to the LU Engine for this model.')
})
})