Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import { describe, expect, it, vi } from 'vitest'
import { createSandboxFindResultGate } from '../sandboxFindSession.js'
describe('sandbox native find result gate', () => {
it('keeps a provisional final zero pending when a real final result follows', () => {
vi.useFakeTimers()
const deliver = vi.fn()
const gate = createSandboxFindResultGate(deliver)
gate.accept({ requestId: 1, activeMatchOrdinal: 0, matches: 0, finalUpdate: true })
expect(deliver).not.toHaveBeenCalled()
gate.accept({ requestId: 1, activeMatchOrdinal: 1, matches: 2, finalUpdate: true })
expect(deliver).toHaveBeenCalledOnce()
expect(deliver).toHaveBeenLastCalledWith(
expect.objectContaining({ activeMatchOrdinal: 1, matches: 2 })
)
vi.runAllTimers()
expect(deliver).toHaveBeenCalledOnce()
vi.useRealTimers()
})
it('delivers a genuine final zero after the producer settling window', () => {
vi.useFakeTimers()
const deliver = vi.fn()
const gate = createSandboxFindResultGate(deliver)
gate.accept({ requestId: 2, activeMatchOrdinal: 0, matches: 0, finalUpdate: true })
vi.advanceTimersByTime(49)
expect(deliver).not.toHaveBeenCalled()
vi.advanceTimersByTime(1)
expect(deliver).toHaveBeenCalledWith(
expect.objectContaining({ requestId: 2, matches: 0, finalUpdate: true })
)
vi.useRealTimers()
})
it('drops a pending zero when the session is disposed', () => {
vi.useFakeTimers()
const deliver = vi.fn()
const gate = createSandboxFindResultGate(deliver)
gate.accept({ requestId: 3, activeMatchOrdinal: 0, matches: 0, finalUpdate: true })
gate.dispose()
vi.runAllTimers()
expect(deliver).not.toHaveBeenCalled()
vi.useRealTimers()
})
})