d3ro-voice/apps/desktop/tests/unit/realtime-session-redteam-r1-30.test.ts

529 lines
18 KiB
TypeScript

// tests/unit/realtime-session-redteam-r1-30.test.ts
// Realtime 대화 세션: 시작 중 취소 누수, 끊김 후 재연결, open 전 첫 메시지 유실 회귀 테스트.
// node 환경에서 fake 전송 계층/WebRTC 의존성으로 검증한다.
import { describe, it, expect, vi } from 'vitest'
import {
INITIAL_REALTIME_MODEL,
reduceRealtimeServerEvent,
type RealtimeModelContext,
type RealtimeServerEvent,
} from '../../src/renderer/services/realtimeConversationModel'
import {
RealtimeAbortError,
WebRtcRealtimeTransport,
type RealtimeClientEvent,
type RealtimeTransport,
type WebRtcTransportDeps,
} from '../../src/renderer/services/realtimeSession'
import {
REALTIME_SESSION_UPDATE,
RealtimeSessionController,
} from '../../src/renderer/services/realtimeSessionController'
// ── helpers ──
interface Deferred<T> {
promise: Promise<T>
resolve: (value: T) => void
reject: (err: Error) => void
}
function deferred<T>(): Deferred<T> {
let resolve!: (value: T) => void
let reject!: (err: Error) => void
const promise = new Promise<T>((res, rej) => {
resolve = res
reject = rej
})
return { promise, resolve, reject }
}
const flush = async (): Promise<void> => {
for (let i = 0; i < 10; i++) await Promise.resolve()
}
function makeCtx(): RealtimeModelContext {
let seq = 0
return {
nextId: (prefix) => `${prefix}-${++seq}`,
now: () => 1000,
}
}
// ── 순수 reducer ──
describe('reduceRealtimeServerEvent', () => {
it('maps a full assistant turn to thinking → speaking → message + listening', () => {
const ctx = makeCtx()
let m = reduceRealtimeServerEvent(INITIAL_REALTIME_MODEL, { type: 'response.created' }, ctx)
expect(m.conversationState).toBe('thinking')
expect(m.assistantMsgId).toBe('assistant-1')
m = reduceRealtimeServerEvent(m, { type: 'response.output_audio_transcript.delta', delta: '안녕' }, ctx)
m = reduceRealtimeServerEvent(m, { type: 'response.audio_transcript.delta', delta: '하세요' }, ctx)
expect(m.conversationState).toBe('speaking')
expect(m.streamingText).toBe('안녕하세요')
expect(m.streamingMsgId).toBe('assistant-1')
m = reduceRealtimeServerEvent(m, { type: 'response.done' }, ctx)
expect(m.conversationState).toBe('listening')
expect(m.streamingText).toBe('')
expect(m.streamingMsgId).toBeNull()
expect(m.messages).toEqual([
{ id: 'assistant-1', role: 'assistant', content: '안녕하세요', timestamp: 1000 },
])
})
it('adds trimmed user transcripts and ignores empty ones', () => {
const ctx = makeCtx()
const ev = (transcript: string): RealtimeServerEvent => ({
type: 'conversation.item.input_audio_transcription.completed',
transcript,
})
let m = reduceRealtimeServerEvent(INITIAL_REALTIME_MODEL, ev(' hi '), ctx)
expect(m.messages).toEqual([{ id: 'user-1', role: 'user', content: 'hi', timestamp: 1000 }])
const same = reduceRealtimeServerEvent(m, ev(' '), ctx)
expect(same).toBe(m)
m = reduceRealtimeServerEvent(m, { type: 'error', error: { message: 'boom' } }, ctx)
expect(m.error).toBe('boom')
expect(reduceRealtimeServerEvent(m, { type: 'error' }, ctx).error).toBe('Realtime error')
})
it('returns the same reference for unknown events and empty deltas', () => {
const ctx = makeCtx()
expect(reduceRealtimeServerEvent(INITIAL_REALTIME_MODEL, { type: 'session.created' }, ctx)).toBe(
INITIAL_REALTIME_MODEL,
)
expect(
reduceRealtimeServerEvent(INITIAL_REALTIME_MODEL, { type: 'response.text.delta', delta: '' }, ctx),
).toBe(INITIAL_REALTIME_MODEL)
})
})
// ── WebRTC transport (fake 의존성) ──
class FakeTrack {
stop = vi.fn()
}
class FakeStream {
readonly track = new FakeTrack()
getTracks = (): FakeTrack[] => [this.track]
getAudioTracks = (): FakeTrack[] => [this.track]
}
class FakeDataChannel {
readyState = 'connecting'
onopen: (() => void) | null = null
onclose: (() => void) | null = null
onmessage: ((e: { data: unknown }) => void) | null = null
send = vi.fn()
close = vi.fn(() => {
this.readyState = 'closed'
})
open(): void {
this.readyState = 'open'
this.onopen?.()
}
receive(data: unknown): void {
this.onmessage?.({ data })
}
}
class FakePeerConnection {
connectionState = 'new'
ontrack: ((e: { streams: unknown[] }) => void) | null = null
onconnectionstatechange: (() => void) | null = null
readonly dc = new FakeDataChannel()
readonly remote = deferred<void>()
addTrack = vi.fn()
createDataChannel = vi.fn(() => this.dc)
createOffer = vi.fn(async () => ({ type: 'offer', sdp: 'offer-sdp' }))
setLocalDescription = vi.fn(async () => undefined)
setRemoteDescription = vi.fn(() => this.remote.promise)
close = vi.fn()
setState(state: string): void {
this.connectionState = state
this.onconnectionstatechange?.()
}
}
interface TransportHarness {
transport: WebRtcRealtimeTransport
token: Deferred<{ value: string; model: string }>
media: Deferred<FakeStream>
stream: FakeStream
pcs: FakePeerConnection[]
sink: { setStream: ReturnType<typeof vi.fn>; dispose: ReturnType<typeof vi.fn> }
deps: {
getUserMedia: ReturnType<typeof vi.fn>
createPeerConnection: ReturnType<typeof vi.fn>
postSdp: ReturnType<typeof vi.fn>
}
}
function makeTransport(sdpOk = true): TransportHarness {
const token = deferred<{ value: string; model: string }>()
const media = deferred<FakeStream>()
const stream = new FakeStream()
const pcs: FakePeerConnection[] = []
const sink = { setStream: vi.fn(), dispose: vi.fn() }
const getUserMedia = vi.fn(() => media.promise)
const createPeerConnection = vi.fn(() => {
const pc = new FakePeerConnection()
pcs.push(pc)
return pc
})
const postSdp = vi.fn(async () => ({
ok: sdpOk,
status: sdpOk ? 201 : 401,
text: async () => (sdpOk ? 'answer-sdp' : 'unauthorized'),
}))
const deps = {
getToken: () => token.promise,
getUserMedia,
createPeerConnection,
postSdp,
createAudioSink: () => sink,
callsUrl: 'https://example.test/calls',
} as unknown as WebRtcTransportDeps
return {
transport: new WebRtcRealtimeTransport(deps),
token,
media,
stream,
pcs,
sink,
deps: { getUserMedia, createPeerConnection, postSdp },
}
}
async function driveToRemoteDescription(h: TransportHarness): Promise<FakePeerConnection> {
h.token.resolve({ value: 'ek_1', model: 'gpt-realtime' })
await flush()
h.media.resolve(h.stream)
await flush()
const pc = h.pcs[0]
expect(pc).toBeDefined()
return pc
}
describe('WebRtcRealtimeTransport', () => {
it('resolves connect only after the data channel opens', async () => {
const h = makeTransport()
const ac = new AbortController()
let settled = false
const p = h.transport.connect(ac.signal).then(() => {
settled = true
})
const pc = await driveToRemoteDescription(h)
expect(h.deps.postSdp).toHaveBeenCalledWith(
'https://example.test/calls?model=gpt-realtime',
expect.objectContaining({ body: 'offer-sdp' }),
)
pc.remote.resolve()
await flush()
expect(settled).toBe(false)
expect(h.transport.send({ type: 'x' })).toBe(false)
pc.dc.open()
await p
expect(settled).toBe(true)
expect(h.transport.send({ type: 'x' })).toBe(true)
expect(pc.dc.send).toHaveBeenCalledWith(JSON.stringify({ type: 'x' }))
})
it('abort during the token request never opens the microphone', async () => {
const h = makeTransport()
const ac = new AbortController()
const p = h.transport.connect(ac.signal)
ac.abort()
h.token.resolve({ value: 'ek', model: 'm' })
await expect(p).rejects.toBeInstanceOf(RealtimeAbortError)
expect(h.deps.getUserMedia).not.toHaveBeenCalled()
expect(h.deps.createPeerConnection).not.toHaveBeenCalled()
})
it('abort during getUserMedia stops the late microphone stream and creates no peer connection', async () => {
const h = makeTransport()
const ac = new AbortController()
const p = h.transport.connect(ac.signal)
h.token.resolve({ value: 'ek', model: 'm' })
await flush()
expect(h.deps.getUserMedia).toHaveBeenCalledTimes(1)
ac.abort()
h.media.resolve(h.stream)
await expect(p).rejects.toBeInstanceOf(RealtimeAbortError)
expect(h.stream.track.stop).toHaveBeenCalled()
expect(h.deps.createPeerConnection).not.toHaveBeenCalled()
})
it('abort while waiting for the channel releases pc, channel, mic and audio sink', async () => {
const h = makeTransport()
const ac = new AbortController()
const p = h.transport.connect(ac.signal)
const pc = await driveToRemoteDescription(h)
pc.remote.resolve()
await flush()
ac.abort()
await expect(p).rejects.toBeInstanceOf(RealtimeAbortError)
expect(pc.close).toHaveBeenCalled()
expect(pc.dc.close).toHaveBeenCalled()
expect(h.stream.track.stop).toHaveBeenCalled()
expect(h.sink.dispose).toHaveBeenCalled()
})
it('a failed SDP exchange rejects with the status and releases resources', async () => {
const h = makeTransport(false)
const p = h.transport.connect(new AbortController().signal)
const pc = await driveToRemoteDescription(h)
await expect(p).rejects.toThrow('SDP exchange failed (401): unauthorized')
expect(pc.close).toHaveBeenCalled()
expect(h.stream.track.stop).toHaveBeenCalled()
})
it('a peer connection failure before open rejects connect', async () => {
const h = makeTransport()
const p = h.transport.connect(new AbortController().signal)
const pc = await driveToRemoteDescription(h)
pc.remote.resolve()
await flush()
pc.setState('failed')
await expect(p).rejects.toThrow('Realtime connection lost')
expect(h.stream.track.stop).toHaveBeenCalled()
})
it('after open, a failed connection closes itself and notifies onClosed exactly once', async () => {
const h = makeTransport()
const onClosed = vi.fn()
h.transport.onClosed(onClosed)
const p = h.transport.connect(new AbortController().signal)
const pc = await driveToRemoteDescription(h)
pc.remote.resolve()
await flush()
pc.dc.open()
await p
pc.setState('disconnected') // 일시적 끊김은 종료로 보지 않는다
expect(onClosed).not.toHaveBeenCalled()
pc.setState('failed')
pc.setState('closed')
expect(onClosed).toHaveBeenCalledTimes(1)
expect(pc.close).toHaveBeenCalled()
expect(h.stream.track.stop).toHaveBeenCalled()
expect(h.transport.send({ type: 'x' })).toBe(false)
})
it('explicit close does not report onClosed, and forwards only valid JSON events', async () => {
const h = makeTransport()
const onClosed = vi.fn()
const onEvent = vi.fn()
h.transport.onClosed(onClosed)
h.transport.onEvent(onEvent)
const p = h.transport.connect(new AbortController().signal)
const pc = await driveToRemoteDescription(h)
pc.remote.resolve()
await flush()
pc.dc.open()
await p
pc.dc.receive('not json')
pc.dc.receive(JSON.stringify({ nope: true }))
pc.dc.receive(JSON.stringify({ type: 'response.created' }))
expect(onEvent).toHaveBeenCalledTimes(1)
expect(onEvent).toHaveBeenCalledWith({ type: 'response.created' })
h.transport.close()
expect(onClosed).not.toHaveBeenCalled()
expect(pc.close).toHaveBeenCalled()
})
it('is single-use', async () => {
const h = makeTransport()
h.transport.close()
await expect(h.transport.connect(new AbortController().signal)).rejects.toThrow('single-use')
})
})
// ── 컨트롤러 (fake transport) ──
class FakeTransport implements RealtimeTransport {
readonly connectCall = deferred<void>()
open = false
ignoreAbort = false
sent: RealtimeClientEvent[] = []
close = vi.fn(() => {
this.open = false
})
private eventCb: ((e: RealtimeServerEvent) => void) | null = null
private closedCb: (() => void) | null = null
connect(signal: AbortSignal): Promise<void> {
signal.addEventListener('abort', () => {
if (!this.ignoreAbort) this.connectCall.reject(new RealtimeAbortError())
})
return this.connectCall.promise.then(() => {
this.open = true
})
}
send(event: RealtimeClientEvent): boolean {
if (!this.open) return false
this.sent.push(event)
return true
}
onEvent(cb: (e: RealtimeServerEvent) => void): () => void {
this.eventCb = cb
return () => undefined
}
onClosed(cb: () => void): () => void {
this.closedCb = cb
return () => undefined
}
emit(e: RealtimeServerEvent): void {
this.eventCb?.(e)
}
drop(): void {
this.open = false
this.closedCb?.()
}
}
function makeController(): { controller: RealtimeSessionController; transports: FakeTransport[] } {
const transports: FakeTransport[] = []
const controller = new RealtimeSessionController(() => {
const t = new FakeTransport()
transports.push(t)
return t
}, makeCtx())
return { controller, transports }
}
describe('RealtimeSessionController', () => {
it('stop during connecting cancels the attempt, closes the transport and never goes live', async () => {
const { controller, transports } = makeController()
const p = controller.start()
expect(controller.getSnapshot().connectionState).toBe('connecting')
controller.stop()
expect(transports[0].close).toHaveBeenCalled()
await expect(p).resolves.toBe('cancelled')
expect(controller.getSnapshot().connectionState).toBe('idle')
})
it('a connect that completes after stop (abort ignored) is closed and never goes live', async () => {
const { controller, transports } = makeController()
const p = controller.start()
const t = transports[0]
t.ignoreAbort = true
controller.stop()
t.close.mockClear()
t.connectCall.resolve()
await expect(p).resolves.toBe('cancelled')
expect(t.close).toHaveBeenCalled()
expect(t.sent).toEqual([])
expect(controller.getSnapshot().connectionState).toBe('idle')
})
it('concurrent start calls share one connection attempt', async () => {
const { controller, transports } = makeController()
const a = controller.start()
const b = controller.start()
expect(transports).toHaveLength(1)
transports[0].connectCall.resolve()
await expect(a).resolves.toBe('live')
await expect(b).resolves.toBe('live')
await expect(controller.start()).resolves.toBe('live')
expect(transports).toHaveLength(1)
})
it('sends session.update when live and sendText right after start is delivered', async () => {
const { controller, transports } = makeController()
const p = controller.start()
transports[0].connectCall.resolve()
await expect(p).resolves.toBe('live')
controller.sendText('hello')
const t = transports[0]
expect(t.sent[0]).toEqual(REALTIME_SESSION_UPDATE)
expect(t.sent.slice(1).map((e) => e.type)).toEqual(['conversation.item.create', 'response.create'])
expect(controller.getSnapshot().messages.map((m) => m.content)).toEqual(['hello'])
expect(controller.getSnapshot().conversationState).toBe('listening')
})
it('queues text sent while connecting and flushes it after session.update', async () => {
const { controller, transports } = makeController()
const p = controller.start()
controller.sendText('first')
expect(controller.getSnapshot().messages.map((m) => m.content)).toEqual(['first'])
expect(transports[0].sent).toEqual([])
transports[0].connectCall.resolve()
await p
expect(transports[0].sent.map((e) => e.type)).toEqual([
'session.update',
'conversation.item.create',
'response.create',
])
})
it('after the connection drops, the session reports error and start reconnects with a new transport', async () => {
const { controller, transports } = makeController()
const p = controller.start()
transports[0].connectCall.resolve()
await p
transports[0].emit({ type: 'response.created' })
transports[0].emit({ type: 'response.text.delta', delta: 'partial' })
transports[0].drop()
const snap = controller.getSnapshot()
expect(snap.connectionState).toBe('error')
expect(snap.conversationState).toBe('idle')
expect(snap.messages.map((m) => m.content)).toEqual(['partial'])
const again = controller.start()
expect(transports).toHaveLength(2)
transports[1].connectCall.resolve()
await expect(again).resolves.toBe('live')
// 이전 전송 계층의 늦은 이벤트는 무시된다
transports[0].emit({ type: 'response.created' })
expect(controller.getSnapshot().conversationState).toBe('listening')
})
it('a failed connect reports failed with the error message and allows retry', async () => {
const { controller, transports } = makeController()
const p = controller.start()
transports[0].connectCall.reject(new Error('token request failed'))
await expect(p).resolves.toBe('failed')
expect(controller.getSnapshot()).toMatchObject({
connectionState: 'error',
conversationState: 'idle',
error: 'token request failed',
})
void controller.start()
expect(transports).toHaveLength(2)
expect(controller.getSnapshot().error).toBeNull()
})
it('notifies subscribers and applies server events only from the current transport', async () => {
const { controller, transports } = makeController()
const listener = vi.fn()
const unsubscribe = controller.subscribe(listener)
const p = controller.start()
transports[0].connectCall.resolve()
await p
listener.mockClear()
transports[0].emit({ type: 'response.created' })
expect(listener).toHaveBeenCalledTimes(1)
expect(controller.getSnapshot().conversationState).toBe('thinking')
controller.cancelResponse()
expect(transports[0].sent.at(-1)).toEqual({ type: 'response.cancel' })
expect(controller.getSnapshot().conversationState).toBe('listening')
unsubscribe()
})
it('sendText is ignored when no session exists', () => {
const { controller, transports } = makeController()
controller.sendText('x')
expect(transports).toHaveLength(0)
expect(controller.getSnapshot().messages).toEqual([])
})
})