7.4 billion tokens per month. 34 free LLM providers. 635 free model endpoints. All behind one /v1 endpoint, plus any custom OpenAI-compatible endpoint. Smart routing, automatic failover, encrypted keys. Personal experimentation only.
TypeScript
// @vitest-environment jsdom
// Models → Decision lists the catalog's decision models, toggles one with
// PATCH, and "Try it" runs one choice question through /api/decisions/try.
import { afterEach, beforeAll, beforeEach, expect, it, vi } from 'vitest'
import { act } from 'react'
import { createRoot, type Root } from 'react-dom/client'
import { MemoryRouter } from 'react-router-dom'
import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
import { I18nProvider } from '@/i18n'
import { apiFetch } from '@/lib/api'
import type { DecisionModel } from '@/lib/decision'
import { DecisionModelsView } from './decision-models'
import { ModelsTabs } from './models-tabs'
vi.mock('@/lib/api', () => ({ apiFetch: vi.fn() }))
const model = (over: Partial<DecisionModel>): DecisionModel => ({
id: 1, platform: 'cloudflare', modelId: '@cf/cloudflare/clef-flash', displayName: 'Clef Flash', priority: 0,
enabled: true, quotaLabel: 'Workers AI free daily neurons', contextWindow: 24576,
keyCount: 1, healthyKeyCount: 1, decisionsThisMonth: 4, ...over,
})
let root: Root
let container: HTMLDivElement
let client: QueryClient
beforeAll(() => { (globalThis as unknown as { IS_REACT_ACT_ENVIRONMENT: boolean }).IS_REACT_ACT_ENVIRONMENT = true })
beforeEach(() => {
client = new QueryClient({ defaultOptions: { queries: { retry: false, gcTime: 0 }, mutations: { retry: false } } })
container = document.createElement('div'); document.body.appendChild(container); root = createRoot(container)
})
afterEach(() => { act(() => root.unmount()); client.clear(); container.remove() })
async function flush() { for (let i = 0; i < 4; i++) await act(async () => { await new Promise(r => setTimeout(r, 0)) }) }
async function mount(models: DecisionModel[]) {
vi.mocked(apiFetch).mockReset().mockImplementation(async (path: string) => {
if (path === '/api/decisions/models') return models
if (path === '/api/decisions/try') {
return {
model: '@cf/cloudflare/clef-flash', provider: 'cloudflare', usage: { input_tokens: 10, output_tokens: 0 },
answers: { answer: { type: 'choice', choice: 'code', confidence: 0.91, probabilities: { chat: 0.04, code: 0.93, reasoning: 0.03 } } },
}
}
return {}
})
act(() => root.render(
<QueryClientProvider client={client}>
<I18nProvider initialLocale="en">
<MemoryRouter><DecisionModelsView /></MemoryRouter>
</I18nProvider>
</QueryClientProvider>
))
await flush()
}
it('lists the decision models with usage and a no-key badge', async () => {
await mount([
model({}),
model({ id: 2, platform: 'openrouter', modelId: 'inception/mercury-decide:free', displayName: 'Mercury Decide (free)', priority: 1, keyCount: 0, healthyKeyCount: 0, decisionsThisMonth: 0 }),
])
const text = container.textContent ?? ''
expect(text).toContain('Clef Flash')
expect(text).toContain('inception/mercury-decide:free')
expect(text).toContain('4 this month')
expect(text).toContain('POST /v1/decisions')
expect(text).toContain('Chat routing only uses them when the Decision routing strategy is on')
expect(container.querySelectorAll('[role="switch"]')).toHaveLength(2)
})
it('turns a model off with PATCH', async () => {
await mount([model({})])
const toggle = container.querySelector('[role="switch"]') as HTMLElement
await act(async () => { toggle.click() })
await flush()
expect(vi.mocked(apiFetch)).toHaveBeenCalledWith('/api/decisions/models/1', { method: 'PATCH', body: JSON.stringify({ enabled: false }) })
})
it('runs a test decision and shows the answer with probabilities', async () => {
await mount([model({})])
const run = [...container.querySelectorAll('button')].find(b => b.textContent === 'Decide') as HTMLButtonElement
await act(async () => { run.click() })
await flush()
const call = vi.mocked(apiFetch).mock.calls.find(c => c[0] === '/api/decisions/try')!
expect(JSON.parse((call[1] as RequestInit).body as string)).toEqual({
state: 'my react useEffect runs twice, why?',
question: 'Which kind of model should answer this?',
options: ['chat', 'code', 'reasoning', 'writing'],
})
const text = container.textContent ?? ''
expect(text).toContain('Answer: code')
expect(text).toContain('91% confident')
expect(text).toContain('93%')
})
it('shows the empty state before the catalog syncs', async () => {
await mount([])
expect(container.textContent).toContain('No decision models yet')
})
it('orders the Models tabs with Embeddings after Audio and Decision next to it', () => {
act(() => root.render(
<I18nProvider initialLocale="en"><MemoryRouter><ModelsTabs /></MemoryRouter></I18nProvider>
))
const labels = [...container.querySelectorAll('a')].map(a => a.textContent)
expect(labels).toEqual(['Chat models', 'Image', 'Video', 'Audio', 'Embeddings', 'Decision', 'Search', 'Fusion'])
})