라이브룸 V2V 번역 자막 기능을 추가한다

라이브룸에서 진행자 언어와 기기 언어가 다를 때 자막 토글을 제공한다.
룸 정보 응답에 V2V 워커 토큰과 진행자 언어 코드를 포함한다.
Agora V2V 에이전트 참여와 종료 API 연동을 추가한다
This commit is contained in:
Yu Sung
2026-02-09 21:11:17 +09:00
parent 7f703024d8
commit b796f6d9c5
11 changed files with 816 additions and 2 deletions

View File

@@ -13,6 +13,7 @@ struct GetRoomInfoResponse: Decodable {
let channelName: String
let rtcToken: String
let rtmToken: String
let v2vWorkerToken: String
let creatorId: Int
let creatorNickname: String
let creatorProfileUrl: String
@@ -25,6 +26,7 @@ struct GetRoomInfoResponse: Decodable {
let managerList: [LiveRoomMember]
let donationRankingTop3UserIds: [Int]
let menuPan: String
let creatorLanguageCode: String?
let isActiveRoulette: Bool
let isPrivateRoom: Bool
let password: String?

View File

@@ -34,6 +34,7 @@ final class LiveRoomViewModel: NSObject, ObservableObject {
private var agora: Agora = Agora.shared
private let repository = LiveRepository()
private let v2vRepository: V2VRepository
private let userRepository = UserRepository()
private let reportRepository = ReportRepository()
private let rouletteRepository = RouletteRepository()
@@ -55,6 +56,7 @@ final class LiveRoomViewModel: NSObject, ObservableObject {
@Published var isLoadingLikeHeart = false
@Published var isLoading = false
@Published var isV2VLoading = false
@Published var errorMessage = ""
@Published var reportMessage = ""
@Published var isShowReportPopup = false
@@ -283,9 +285,33 @@ final class LiveRoomViewModel: NSObject, ObservableObject {
private var blockedMemberIdList = Set<Int>()
private var hasInvokedJoinChannel = false
private var v2vMessageAssembler = V2vMessageAssembler()
private var v2vAgentId: String?
private var v2vSourceLanguage: String?
private var v2vTargetLanguage: String?
private var isV2VJoinInProgress = false
private var isV2VLeaveInProgress = false
@Published var isV2VAvailable = false
@Published var isV2VCaptionOn = false
@Published var v2vCaptionText = ""
// BIG_HEART :
private var suppressNextRemoteWaterFill = false
init(v2vRepository: V2VRepository = V2VRepositoryImpl()) {
self.v2vRepository = v2vRepository
super.init()
}
private var isV2VJoined: Bool {
v2vAgentId != nil
}
func stopV2VTranslationIfJoined(clearCaptionText: Bool = true) {
guard isV2VJoined else { return }
stopV2VTranslation(clearCaptionText: clearCaptionText)
}
func getBlockedMemberIdList() {
userRepository.getBlockedMemberIdList()
@@ -332,8 +358,149 @@ final class LiveRoomViewModel: NSObject, ObservableObject {
}
private func deInitAgoraEngine() {
stopV2VTranslationIfJoined()
agora.deInit()
}
func updateV2VAvailability(roomInfo: GetRoomInfoResponse?, deviceLanguage: String? = nil) {
guard let roomInfo = roomInfo else {
disableV2VIfNeeded()
return
}
let source = V2vLanguageMapper.mapToAgoraLanguage(roomInfo.creatorLanguageCode)
let target = V2vLanguageMapper.mapToAgoraLanguage(deviceLanguage ?? V2vAppLanguageResolver.currentLanguageCode())
let available = source != nil && target != nil && source != target
v2vSourceLanguage = source
v2vTargetLanguage = target
isV2VAvailable = available
if !available {
disableV2VIfNeeded()
}
}
func toggleV2VCaption() {
guard isV2VAvailable else { return }
if isV2VCaptionOn {
stopV2VTranslation()
} else {
isV2VCaptionOn = true
startV2VTranslation()
}
}
func startV2VTranslation() {
guard !isV2VJoinInProgress,
!isV2VLeaveInProgress,
isV2VAvailable,
v2vAgentId == nil,
let roomInfo = liveRoomInfo,
let sourceLanguage = v2vSourceLanguage,
let targetLanguage = v2vTargetLanguage else {
isV2VCaptionOn = false
return
}
isV2VJoinInProgress = true
isV2VLoading = true
let request = V2VJoinRequest(
roomInfo: roomInfo,
sourceLanguage: sourceLanguage,
targetLanguage: targetLanguage
)
v2vRepository.join(request: request)
.sink { [weak self] completion in
guard let self else { return }
self.isV2VJoinInProgress = false
self.isV2VLoading = self.isV2VLeaveInProgress
if case .failure(let error) = completion {
ERROR_LOG("[V2V] join failed: \(error)")
self.isV2VCaptionOn = false
self.errorMessage = error.userMessage
self.isShowErrorPopup = true
}
} receiveValue: { [weak self] agentId in
guard let self else { return }
self.v2vAgentId = agentId
DEBUG_LOG("[V2V] join success. agentId=\(agentId)")
if self.isV2VAvailable {
self.isV2VCaptionOn = true
} else {
self.stopV2VTranslation()
}
}
.store(in: &subscription)
}
func stopV2VTranslation(clearCaptionText: Bool = true) {
if clearCaptionText {
v2vCaptionText = ""
}
isV2VCaptionOn = false
v2vMessageAssembler.reset()
guard !isV2VLeaveInProgress,
!isV2VJoinInProgress else {
return
}
guard let agentId = v2vAgentId else {
return
}
DEBUG_LOG("[V2V] leave start. agentId=\(agentId)")
isV2VLeaveInProgress = true
isV2VLoading = true
v2vRepository.leave(agentId: agentId)
.sink { [weak self] completion in
guard let self else { return }
self.isV2VLeaveInProgress = false
self.isV2VLoading = self.isV2VJoinInProgress
switch completion {
case .finished:
self.v2vAgentId = nil
self.isV2VCaptionOn = false
DEBUG_LOG("[V2V] leave success")
case .failure(let error):
ERROR_LOG("[V2V] leave failed: \(error)")
self.errorMessage = error.userMessage
self.isShowErrorPopup = true
}
} receiveValue: { _ in }
.store(in: &subscription)
}
func handleV2VIncomingData(_ data: Data) {
guard isV2VAvailable, isV2VCaptionOn, isV2VJoined else { return }
guard let parsedText = v2vMessageAssembler.consume(data: data) else {
return
}
let normalized = parsedText.trimmingCharacters(in: .whitespacesAndNewlines)
guard !normalized.isEmpty else {
return
}
DEBUG_LOG("[V2V] subtitle received. text=\(normalized)")
v2vCaptionText = normalized
}
private func disableV2VIfNeeded() {
isV2VAvailable = false
stopV2VTranslation(clearCaptionText: true)
}
func agoraConnectSuccess(isManager: Bool) {
self.isLoading = false
@@ -360,6 +527,7 @@ final class LiveRoomViewModel: NSObject, ObservableObject {
}
func quitRoom() {
stopV2VTranslationIfJoined()
isLoading = true
if let index = muteSpeakers.firstIndex(of: UInt(UserDefaults.int(forKey: .userId))) {
@@ -424,6 +592,7 @@ final class LiveRoomViewModel: NSObject, ObservableObject {
if let data = decoded.data, decoded.success {
self.liveRoomInfo = data
self.updateV2VAvailability(roomInfo: data)
if self.coverImageUrl != data.coverImageUrl {
self.coverImageUrl = data.coverImageUrl
@@ -2497,6 +2666,7 @@ extension LiveRoomViewModel: AgoraRtcEngineDelegate {
DispatchQueue.main.async {[unowned self] in
if uid == UInt(self.liveRoomInfo!.creatorId) {
//
self.stopV2VTranslationIfJoined()
self.liveRoomInfo = nil
self.errorMessage = "라이브가 종료되었습니다."
self.isShowErrorPopup = true
@@ -2506,6 +2676,16 @@ extension LiveRoomViewModel: AgoraRtcEngineDelegate {
}
}
}
func rtcEngine(_ engine: AgoraRtcEngineKit, receiveStreamMessageFromUid uid: UInt, streamId: Int, data: Data) {
DispatchQueue.main.async { [weak self] in
guard let self,
self.isV2VJoined,
self.isV2VCaptionOn else { return }
DEBUG_LOG("[V2V] rtc stream message received uid=\(uid), streamId=\(streamId), size=\(data.count)")
self.handleV2VIncomingData(data)
}
}
}
extension LiveRoomViewModel: AgoraRtmClientDelegate {
@@ -2521,6 +2701,11 @@ extension LiveRoomViewModel: AgoraRtmClientDelegate {
let rawMessageString = String(data: message, encoding: .utf8)
DispatchQueue.main.async { [unowned self] in
if self.isV2VJoined && self.isV2VCaptionOn {
DEBUG_LOG("[V2V] rtm raw message received size=\(message.count)")
self.handleV2VIncomingData(message)
}
if rawMessageString == LiveRoomRequestType.CHANGE_LISTENER.rawValue {
self.setListener()
return

View File

@@ -17,7 +17,9 @@ struct LiveRoomInfoGuestView: View {
let isOnNotice: Bool
let isOnMenuPan: Bool
let isOnSignature: Bool
let isOnV2VCaption: Bool
let isShowMenuPanButton: Bool
let isShowV2VCaptionButton: Bool
let creatorId: Int
let creatorNickname: String
@@ -37,6 +39,7 @@ struct LiveRoomInfoGuestView: View {
let onClickTotalHeart: () -> Void
let onClickTotalDonation: () -> Void
let onClickChangeListener: () -> Void
let onClickToggleV2VCaption: () -> Void
let onClickToggleSignature: () -> Void
var body: some View {
@@ -62,6 +65,20 @@ struct LiveRoomInfoGuestView: View {
strokeCornerRadius: 5.3
) { onClickChangeListener() }
}
if isShowV2VCaptionButton {
LiveRoomOverlayStrokeTextToggleButton(
isOn: isOnV2VCaption,
onText: I18n.LiveRoom.captionOn,
onTextColor: Color.button,
onStrokeColor: Color.button,
offText: I18n.LiveRoom.captionOff,
offTextColor: Color.graybb,
offStrokeColor: Color.graybb,
strokeWidth: 1,
strokeCornerRadius: 5.3
) { onClickToggleV2VCaption() }
}
LiveRoomOverlayStrokeTextToggleButton(
isOn: isOnSignature,
@@ -222,7 +239,9 @@ struct LiveRoomInfoGuestView_Previews: PreviewProvider {
isOnNotice: false,
isOnMenuPan: false,
isOnSignature: false,
isOnV2VCaption: false,
isShowMenuPanButton: false,
isShowV2VCaptionButton: true,
creatorId: 1,
creatorNickname: "도화",
creatorProfileUrl: "https://cf.sodalive.net/profile/26/26-profile-ddf78b4d-0300-4c50-9c84-5d8a95fd5fe2-4892-1705256364320",
@@ -258,6 +277,7 @@ struct LiveRoomInfoGuestView_Previews: PreviewProvider {
onClickTotalHeart: {},
onClickTotalDonation: {},
onClickChangeListener: {},
onClickToggleV2VCaption: {},
onClickToggleSignature: {}
)
}

View File

@@ -95,7 +95,9 @@ struct LiveRoomViewV2: View {
isOnNotice: viewModel.isShowNotice,
isOnMenuPan: viewModel.isShowMenuPan,
isOnSignature: viewModel.isSignatureOn,
isOnV2VCaption: viewModel.isV2VCaptionOn,
isShowMenuPanButton: !liveRoomInfo.menuPan.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty,
isShowV2VCaptionButton: viewModel.isV2VAvailable,
creatorId: liveRoomInfo.creatorId,
creatorNickname: liveRoomInfo.creatorNickname,
creatorProfileUrl: liveRoomInfo.creatorProfileUrl,
@@ -132,6 +134,9 @@ struct LiveRoomViewV2: View {
onClickChangeListener: {
viewModel.setListener()
},
onClickToggleV2VCaption: {
viewModel.toggleV2VCaption()
},
onClickToggleSignature: {
viewModel.isSignatureOn.toggle()
}
@@ -182,7 +187,7 @@ struct LiveRoomViewV2: View {
.onPreferenceChange(ScrollOffsetKey.self) {
viewModel.setOffset($0)
}
.padding(.bottom, 70)
.padding(.bottom, v2vCaptionBottomInset)
}
.padding(.top, 16)
@@ -308,6 +313,19 @@ struct LiveRoomViewV2: View {
}
.padding(.trailing, 13.3)
if isV2VCaptionVisible {
Text(viewModel.v2vCaptionText)
.appFont(size: 12, weight: .medium)
.foregroundColor(.white)
.lineLimit(2)
.padding(.horizontal, 12)
.padding(.vertical, 10)
.frame(maxWidth: .infinity, alignment: .center)
.background(Color.black.opacity(0.75))
.cornerRadius(10)
.padding(.horizontal, 13.3)
}
LiveRoomInputChatView {
viewModel.sendMessage(chatMessage: $0) {
viewModel.isShowingNewChat = false
@@ -316,6 +334,7 @@ struct LiveRoomViewV2: View {
return true
}
.padding(.top, isV2VCaptionVisible ? -13.3 : 0)
.padding(.bottom, 10)
}
@@ -323,7 +342,7 @@ struct LiveRoomViewV2: View {
LiveRoomNewChatView{
viewModel.isShowingNewChat = false
proxy.scrollTo(viewModel.messages.count - 1, anchor: .center)
}.padding(.bottom, 70)
}.padding(.bottom, v2vCaptionBottomInset)
}
if viewModel.isSignatureOn && viewModel.signatureImageUrl.trimmingCharacters(in: .whitespacesAndNewlines).count > 0 {
@@ -474,6 +493,7 @@ struct LiveRoomViewV2: View {
.onDisappear {
UIApplication.shared.isIdleTimerDisabled = false
NotificationCenter.default.removeObserver(self)
viewModel.stopV2VTranslationIfJoined()
viewModel.stopPeriodicPlaybackValidation()
}
@@ -744,6 +764,10 @@ struct LiveRoomViewV2: View {
if viewModel.isLoading && viewModel.liveRoomInfo == nil {
LoadingView()
}
if viewModel.isV2VLoading {
LoadingView()
}
}
.overlay(alignment: .center) {
ZStack {
@@ -916,6 +940,17 @@ struct LiveRoomViewV2: View {
}
}
private extension LiveRoomViewV2 {
var isV2VCaptionVisible: Bool {
viewModel.isV2VCaptionOn &&
!viewModel.v2vCaptionText.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
}
var v2vCaptionBottomInset: CGFloat {
isV2VCaptionVisible ? 120 : 70
}
}
struct LiveRoomViewV2_Previews: PreviewProvider {
static var previews: some View {
LiveRoomViewV2()

View File

@@ -0,0 +1,53 @@
//
// V2vApi.swift
// SodaLive
//
// Created by klaus on 2/9/26.
//
import Foundation
import Moya
enum V2vApi {
case join(request: V2VJoinRequest)
case leave(agentId: String)
}
extension V2vApi: TargetType {
var baseURL: URL {
URL(string: "https://api.agora.io/api/speech-to-speech-translation/v2/")!
}
var path: String {
switch self {
case .join:
return "projects/\(AGORA_APP_ID)/join"
case .leave(let agentId):
return "projects/\(AGORA_APP_ID)/agents/\(agentId)/leave"
}
}
var method: Moya.Method {
.post
}
var task: Task {
switch self {
case .join(let request):
return .requestJSONEncodable(request)
case .leave:
return .requestPlain
}
}
var headers: [String: String]? {
let credential = "\(AGORA_CUSTOMER_ID):\(AGORA_CUSTOMER_SECRET)"
let basicToken = Data(credential.utf8).base64EncodedString()
return [
"Authorization": "Basic \(basicToken)",
"X-Request-Id": UUID().uuidString,
"Content-Type": "application/json"
]
}
}

View File

@@ -0,0 +1,109 @@
//
// V2vModels.swift
// SodaLive
//
// Created by klaus on 2/9/26.
//
import Foundation
struct V2VJoinRequest: Encodable {
let name: String
let preset: String
let properties: Properties
init(roomInfo: GetRoomInfoResponse, sourceLanguage: String, targetLanguage: String) {
let rtcUid = UserDefaults.int(forKey: .userId)
self.name = "sodalive-v2v-\(roomInfo.roomId)-\(UUID().uuidString)"
self.preset = "v2vt_base"
self.properties = .init(
channel: roomInfo.channelName,
token: roomInfo.v2vWorkerToken,
agentRtcUid: "\(rtcUid)333",
remoteRtcUids: ["\(roomInfo.creatorId)"],
idleTimeout: 300,
advancedFeatures: .init(enableRtm: false),
parameters: .init(dataChannel: "datastream"),
asr: .init(language: sourceLanguage),
translation: .init(language: targetLanguage),
tts: .init(enable: false)
)
}
struct Properties: Encodable {
let channel: String
let token: String
let agentRtcUid: String
let remoteRtcUids: [String]
let idleTimeout: Int
let advancedFeatures: AdvancedFeatures
let parameters: Parameters
let asr: Asr
let translation: Translation
let tts: Tts
enum CodingKeys: String, CodingKey {
case channel
case token
case agentRtcUid = "agent_rtc_uid"
case remoteRtcUids = "remote_rtc_uids"
case idleTimeout = "idle_timeout"
case advancedFeatures = "advanced_features"
case parameters
case asr
case translation
case tts
}
}
struct AdvancedFeatures: Encodable {
let enableRtm: Bool
enum CodingKeys: String, CodingKey {
case enableRtm = "enable_rtm"
}
}
struct Parameters: Encodable {
let dataChannel: String
enum CodingKeys: String, CodingKey {
case dataChannel = "data_channel"
}
}
struct Asr: Encodable {
let language: String
}
struct Translation: Encodable {
let language: String
}
struct Tts: Encodable {
let enable: Bool
}
}
struct V2VJoinResponse: Decodable {
let agentId: String
let createTs: Int
let status: String
enum CodingKeys: String, CodingKey {
case agentId = "agent_id"
case createTs = "create_ts"
case status
}
}
struct V2VLeaveResponse: Decodable {
let agentId: String
let status: String
enum CodingKeys: String, CodingKey {
case agentId = "agent_id"
case status
}
}

View File

@@ -0,0 +1,128 @@
//
// V2vRepository.swift
// SodaLive
//
// Created by klaus on 2/9/26.
//
import Foundation
import Combine
import Moya
import CombineMoya
protocol V2VRepository {
func join(request: V2VJoinRequest) -> AnyPublisher<String, V2VRepositoryError>
func leave(agentId: String) -> AnyPublisher<Void, V2VRepositoryError>
}
enum V2VRepositoryError: Error {
case network(message: String)
case decoding
case business(message: String)
var userMessage: String {
switch self {
case .network(let message):
return message
case .decoding:
return I18n.Common.commonError
case .business(let message):
return message
}
}
}
final class V2VRepositoryImpl: V2VRepository {
private let api: MoyaProvider<V2vApi>
init(api: MoyaProvider<V2vApi> = MoyaProvider<V2vApi>()) {
self.api = api
}
func join(request: V2VJoinRequest) -> AnyPublisher<String, V2VRepositoryError> {
api.requestPublisher(.join(request: request))
.tryMap { response in
try Self.validateStatusCode(response)
do {
let decoded = try JSONDecoder().decode(V2VJoinResponse.self, from: response.data)
guard !decoded.agentId.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty else {
throw V2VRepositoryError.decoding
}
return decoded.agentId
} catch let error as V2VRepositoryError {
throw error
} catch {
throw V2VRepositoryError.decoding
}
}
.mapError { Self.mapError($0) }
.eraseToAnyPublisher()
}
func leave(agentId: String) -> AnyPublisher<Void, V2VRepositoryError> {
api.requestPublisher(.leave(agentId: agentId))
.tryMap { response in
try Self.validateStatusCode(response)
if !response.data.isEmpty {
if let text = String(data: response.data, encoding: .utf8) {
DEBUG_LOG("[V2V] leave response: \(text)")
}
_ = try? JSONDecoder().decode(V2VLeaveResponse.self, from: response.data)
}
return ()
}
.mapError { Self.mapError($0) }
.eraseToAnyPublisher()
}
private static func validateStatusCode(_ response: Response) throws {
guard (200..<300).contains(response.statusCode) else {
throw V2VRepositoryError.business(message: parseBusinessMessage(from: response.data) ?? I18n.Common.commonError)
}
}
private static func mapError(_ error: Error) -> V2VRepositoryError {
if let mapped = error as? V2VRepositoryError {
return mapped
}
guard let moyaError = error as? MoyaError else {
return .network(message: I18n.Common.commonError)
}
switch moyaError {
case .statusCode(let response):
if let message = parseBusinessMessage(from: response.data) {
return .business(message: message)
}
return .business(message: I18n.Common.commonError)
case .objectMapping, .jsonMapping, .encodableMapping, .stringMapping, .imageMapping:
return .decoding
default:
return .network(message: I18n.Common.commonError)
}
}
private static func parseBusinessMessage(from data: Data) -> String? {
if let decoded = try? JSONDecoder().decode(V2VErrorResponse.self, from: data),
let message = decoded.message?.trimmingCharacters(in: .whitespacesAndNewlines),
!message.isEmpty {
return message
}
if let decoded = try? JSONDecoder().decode(ApiResponseWithoutData.self, from: data),
let message = decoded.message?.trimmingCharacters(in: .whitespacesAndNewlines),
!message.isEmpty {
return message
}
return nil
}
}
private struct V2VErrorResponse: Decodable {
let message: String?
}

View File

@@ -0,0 +1,274 @@
//
// V2vState.swift
// SodaLive
//
// Created by klaus on 2/9/26.
//
import Foundation
struct V2vState {
var isAvailable: Bool = false
var isCaptionOn: Bool = false
var captionText: String = ""
var agentId: String? = nil
var sourceLanguage: String? = nil
var targetLanguage: String? = nil
}
enum V2vLanguageMapper {
static func mapToAgoraLanguage(_ code: String?) -> String? {
guard let normalized = code?.trimmingCharacters(in: .whitespacesAndNewlines).lowercased() else {
return nil
}
if normalized.hasPrefix("ko") { return "ko-KR" }
if normalized.hasPrefix("ja") { return "ja-JP" }
if normalized.hasPrefix("en") { return "en-US" }
return nil
}
}
enum V2vAppLanguageResolver {
static func currentLanguageCode() -> String {
let headerCode = LanguageHeaderProvider.current.trimmingCharacters(in: .whitespacesAndNewlines).lowercased()
if headerCode == "ko" || headerCode == "ja" || headerCode == "en" {
return headerCode
}
let saved = UserDefaults.standard.string(forKey: "app.language")
if let saved,
let option = LanguageOption(rawValue: saved),
option != .system {
return option.rawValue
}
guard let preferred = Locale.preferredLanguages.first?.lowercased() else {
return "ko"
}
if preferred.hasPrefix("ko") { return "ko" }
if preferred.hasPrefix("ja") { return "ja" }
return "en"
}
}
final class V2vMessageAssembler {
private struct Chunk {
let messageId: String
let partIdx: Int
let partSum: Int
let content: String
}
private struct Buffer {
var partSum: Int
var createdAt: Date
var parts: [Int: String]
}
private var buffers: [String: Buffer] = [:]
private let timeout: TimeInterval = 10
func consume(data: Data) -> String? {
cleanupExpiredBuffers()
guard let chunk = parseChunk(data: data) else {
DEBUG_LOG("[V2V] chunk parsing failed. raw=\(preview(data))")
return nil
}
DEBUG_LOG("[V2V] chunk received messageId=\(chunk.messageId), partIdx=\(chunk.partIdx), partSum=\(chunk.partSum), contentLength=\(chunk.content.count)")
var buffer = buffers[chunk.messageId] ?? Buffer(
partSum: chunk.partSum,
createdAt: Date(),
parts: [:]
)
buffer.partSum = max(buffer.partSum, chunk.partSum)
// partIdx
if buffer.parts[chunk.partIdx] == nil {
buffer.parts[chunk.partIdx] = chunk.content
}
buffers[chunk.messageId] = buffer
guard buffer.parts.count >= buffer.partSum else {
DEBUG_LOG("[V2V] chunk buffering \(chunk.messageId): \(buffer.parts.count)/\(buffer.partSum)")
return nil
}
let sortedParts = buffer.parts
.sorted { $0.key < $1.key }
.prefix(buffer.partSum)
.map { $0.value }
guard sortedParts.count == buffer.partSum else {
return nil
}
let combined = sortedParts.joined()
DEBUG_LOG("[V2V] chunk assembled messageId=\(chunk.messageId), assembledLength=\(combined.count)")
buffers[chunk.messageId] = nil
return decodeCaptionText(from: combined)
}
func reset() {
buffers.removeAll()
}
private func cleanupExpiredBuffers() {
let now = Date()
buffers = buffers.filter { _, value in
now.timeIntervalSince(value.createdAt) <= timeout
}
}
private func parseChunk(data: Data) -> Chunk? {
if let text = String(data: data, encoding: .utf8) {
if let parsed = parsePipeChunk(text) {
return parsed
}
if let parsed = parseJSONChunk(text) {
return parsed
}
}
return nil
}
private func parsePipeChunk(_ raw: String) -> Chunk? {
let items = raw.split(separator: "|", maxSplits: 3, omittingEmptySubsequences: false)
guard items.count == 4,
let partIdx = Int(items[1]),
let partSum = Int(items[2]),
partIdx >= 0,
partSum > 0 else {
return nil
}
return Chunk(
messageId: String(items[0]),
partIdx: partIdx,
partSum: partSum,
content: String(items[3])
)
}
private func parseJSONChunk(_ raw: String) -> Chunk? {
guard let data = raw.data(using: .utf8),
let object = try? JSONSerialization.jsonObject(with: data) as? [String: Any] else {
return nil
}
guard let messageId = object["message_id"] as? String,
let partIdx = anyToInt(object["part_idx"]),
let partSum = anyToInt(object["part_sum"]),
let content = object["content"] as? String,
partIdx >= 0,
partSum > 0 else {
return nil
}
return Chunk(
messageId: messageId,
partIdx: partIdx,
partSum: partSum,
content: content
)
}
private func anyToInt(_ value: Any?) -> Int? {
if let intValue = value as? Int {
return intValue
}
if let stringValue = value as? String {
return Int(stringValue)
}
if let number = value as? NSNumber {
return number.intValue
}
return nil
}
private func decodeCaptionText(from base64String: String) -> String? {
if let plainData = base64String.data(using: .utf8),
let plainJson = try? JSONSerialization.jsonObject(with: plainData),
let text = extractTranslationText(from: plainJson) {
DEBUG_LOG("[V2V] plain json subtitle parsed")
return text
}
guard let decodedData = decodeBase64Payload(base64String),
let jsonObject = try? JSONSerialization.jsonObject(with: decodedData) else {
DEBUG_LOG("[V2V] base64 or final json parsing failed. payloadPrefix=\(String(base64String.prefix(80)))")
return nil
}
let text = extractTranslationText(from: jsonObject)
if text == nil {
DEBUG_LOG("[V2V] final json parsed but translation text not found")
}
return text
}
private func extractTranslationText(from object: Any) -> String? {
if let dict = object as? [String: Any] {
let eventType = (dict["object"] as? String) ?? (dict["type"] as? String) ?? (dict["event"] as? String)
if let eventType,
(eventType == "user.translation" || eventType == "agent.translation"),
let text = (dict["text"] as? String)?.trimmingCharacters(in: .whitespacesAndNewlines),
!text.isEmpty {
return text
}
for value in dict.values {
if let text = extractTranslationText(from: value) {
return text
}
}
}
if let array = object as? [Any] {
for value in array {
if let text = extractTranslationText(from: value) {
return text
}
}
}
return nil
}
private func decodeBase64Payload(_ raw: String) -> Data? {
let trimmed = raw.trimmingCharacters(in: .whitespacesAndNewlines)
if let data = Data(base64Encoded: trimmed) {
return data
}
let normalized = trimmed
.replacingOccurrences(of: "-", with: "+")
.replacingOccurrences(of: "_", with: "/")
let remainder = normalized.count % 4
let padded = remainder == 0 ? normalized : normalized + String(repeating: "=", count: 4 - remainder)
return Data(base64Encoded: padded)
}
private func preview(_ data: Data) -> String {
if let text = String(data: data, encoding: .utf8) {
return String(text.prefix(80))
}
return "\(data.count)bytes"
}
}