added ios module

ios/OralTrainer/Models/AssessmentModels.swift

@@ -0,0 +1,98 @@
+import Foundation
+
+struct ImitationAssessmentRequest {
+    let mediaId: String
+    let sentence: SentenceBoundary
+    let recordingURL: URL
+    let locale: String?
+}
+
+struct TextSubstitution: Decodable {
+    let expected: String
+    let actual: String
+}
+
+struct ImitationAssessmentResult: Decodable {
+    let attemptId: String?
+    let scoringVersion: String?
+    let overallScore: Float
+    let passed: Bool?
+    let passScore: Float?
+    let contentScore: Float?
+    let completenessScore: Float?
+    let fluencyScore: Float?
+    let pronunciationScore: Float?
+    let prosodyScore: Float?
+    let durationScore: Float?
+    let pauseScore: Float?
+    let speechRateScore: Float?
+    let referenceText: String?
+    let recognizedText: String?
+    let referenceDurationMs: Int64?
+    let referenceSpeechDurationMs: Int64?
+    let studentRecordingDurationMs: Int64?
+    let studentSpeechDurationMs: Int64?
+    let durationRatio: Float?
+    let missingTokens: [String]
+    let extraTokens: [String]
+    let substitutions: [TextSubstitution]
+    let feedback: String?
+    let details: [String: String]
+
+    private enum CodingKeys: String, CodingKey {
+        case attemptId = "attempt_id"
+        case scoringVersion = "scoring_version"
+        case overallScore = "overall_score"
+        case passed
+        case passScore = "pass_score"
+        case contentScore = "content_score"
+        case completenessScore = "completeness_score"
+        case fluencyScore = "fluency_score"
+        case pronunciationScore = "pronunciation_score"
+        case prosodyScore = "prosody_score"
+        case durationScore = "duration_score"
+        case pauseScore = "pause_score"
+        case speechRateScore = "speech_rate_score"
+        case referenceText = "reference_text"
+        case recognizedText = "recognized_text"
+        case referenceDurationMs = "reference_duration_ms"
+        case referenceSpeechDurationMs = "reference_speech_duration_ms"
+        case studentRecordingDurationMs = "student_recording_duration_ms"
+        case studentSpeechDurationMs = "student_speech_duration_ms"
+        case durationRatio = "duration_ratio"
+        case missingTokens = "missing_tokens"
+        case extraTokens = "extra_tokens"
+        case substitutions
+        case feedback
+        case details
+    }
+
+    init(from decoder: Decoder) throws {
+        let container = try decoder.container(keyedBy: CodingKeys.self)
+        attemptId = try container.decodeIfPresent(String.self, forKey: .attemptId)
+        scoringVersion = try container.decodeIfPresent(String.self, forKey: .scoringVersion)
+        overallScore = try container.decodeIfPresent(Float.self, forKey: .overallScore) ?? 0
+        passed = try container.decodeIfPresent(Bool.self, forKey: .passed)
+        passScore = try container.decodeIfPresent(Float.self, forKey: .passScore)
+        contentScore = try container.decodeIfPresent(Float.self, forKey: .contentScore)
+        completenessScore = try container.decodeIfPresent(Float.self, forKey: .completenessScore)
+        fluencyScore = try container.decodeIfPresent(Float.self, forKey: .fluencyScore)
+        pronunciationScore = try container.decodeIfPresent(Float.self, forKey: .pronunciationScore)
+        prosodyScore = try container.decodeIfPresent(Float.self, forKey: .prosodyScore)
+        durationScore = try container.decodeIfPresent(Float.self, forKey: .durationScore)
+        pauseScore = try container.decodeIfPresent(Float.self, forKey: .pauseScore)
+        speechRateScore = try container.decodeIfPresent(Float.self, forKey: .speechRateScore)
+        referenceText = try container.decodeIfPresent(String.self, forKey: .referenceText)
+        recognizedText = try container.decodeIfPresent(String.self, forKey: .recognizedText)
+        referenceDurationMs = try container.decodeIfPresent(Int64.self, forKey: .referenceDurationMs)
+        referenceSpeechDurationMs = try container.decodeIfPresent(Int64.self, forKey: .referenceSpeechDurationMs)
+        studentRecordingDurationMs = try container.decodeIfPresent(Int64.self, forKey: .studentRecordingDurationMs)
+        studentSpeechDurationMs = try container.decodeIfPresent(Int64.self, forKey: .studentSpeechDurationMs)
+        durationRatio = try container.decodeIfPresent(Float.self, forKey: .durationRatio)
+        missingTokens = try container.decodeIfPresent([String].self, forKey: .missingTokens) ?? []
+        extraTokens = try container.decodeIfPresent([String].self, forKey: .extraTokens) ?? []
+        substitutions = try container.decodeIfPresent([TextSubstitution].self, forKey: .substitutions) ?? []
+        feedback = try container.decodeIfPresent(String.self, forKey: .feedback)
+        details = try container.decodeIfPresent([String: String].self, forKey: .details) ?? [:]
+    }
+}

ios/OralTrainer/Models/PlaybackModels.swift

@@ -0,0 +1,58 @@
+import Foundation
+
+enum PlaybackState: String {
+    case idle
+    case buffering
+    case ready
+    case ended
+}
+
+struct PlaybackSnapshot: Equatable {
+    var mediaId: String?
+    var positionMs: Int64 = 0
+    var durationMs: Int64 = -1
+    var bufferedPositionMs: Int64 = 0
+    var isPlaying = false
+    var state: PlaybackState = .idle
+    var speed: Float = 1
+    var sentenceIndex: Int?
+}
+
+struct PlayerConfig {
+    var sentenceMode = true
+    var defaultSeekStepMs: Int64 = 10_000
+    var autoPlay = false
+    var continuousPlayback = true
+    var minPlaybackSpeed: Float = 0.5
+    var maxPlaybackSpeed: Float = 2.0
+}
+
+enum LoopMode {
+    case off
+    case one
+    case all
+}
+
+enum SwipeAction {
+    case none
+    case rewind
+    case forward
+    case previousSentence
+    case nextSentence
+    case previousSentenceOrRewind
+    case nextSentenceOrForward
+}
+
+enum GestureKind {
+    case singleTap
+    case swipeLeft
+    case swipeRight
+    case longPressSpeed
+}
+
+enum BoundaryState: Equatable {
+    case idle
+    case loading(isLocal: Bool)
+    case loaded(count: Int)
+    case failed(message: String)
+}

ios/OralTrainer/Models/TrainingModels.swift

@@ -0,0 +1,95 @@
+import Foundation
+
+struct SentenceBoundary: Identifiable, Equatable, Decodable {
+    let index: Int
+    let startMs: Int64
+    let endMs: Int64
+    let text: String?
+    let language: String?
+    let referenceSpeechDurationMs: Int64?
+
+    var id: Int { index }
+
+    private enum CodingKeys: String, CodingKey {
+        case index, text, language
+        case startMs = "start_ms"
+        case endMs = "end_ms"
+        case referenceSpeechDurationMs = "reference_speech_duration_ms"
+    }
+}
+
+struct SentenceBoundaryDocument: Decodable {
+    let videoHash: String
+    let durationMs: Int64
+    let algorithmVersion: String
+    let sentences: [SentenceBoundary]
+
+    private enum CodingKeys: String, CodingKey {
+        case videoHash = "video_hash"
+        case durationMs = "duration_ms"
+        case algorithmVersion = "algorithm_version"
+        case sentences
+    }
+}
+
+struct TrainingVideoSummary: Identifiable, Decodable {
+    let videoHash: String
+    let title: String
+    let streamUrl: String
+    let durationMs: Int64?
+    let sizeBytes: Int64
+    let language: String?
+    let sentenceCount: Int
+    let status: String
+
+    var id: String { videoHash }
+
+    private enum CodingKeys: String, CodingKey {
+        case title, language, status
+        case videoHash = "video_hash"
+        case streamUrl = "stream_url"
+        case durationMs = "duration_ms"
+        case sizeBytes = "size_bytes"
+        case sentenceCount = "sentence_count"
+    }
+
+    /// 与安卓端一致：相对路径基于服务地址解析；服务地址为 http 时强制使用 http 流地址。
+    func resolvedStreamURL(baseURL: String) -> URL? {
+        guard let base = URL(string: baseURL) else { return nil }
+        var absolute = URL(string: streamUrl, relativeTo: base)?.absoluteString ?? streamUrl
+        if base.scheme == "http" {
+            absolute = absolute.replacingOccurrences(
+                of: "https://",
+                with: "http://",
+                options: [.anchored]
+            )
+        }
+        return URL(string: absolute)
+    }
+}
+
+struct TrainingMediaItem: Identifiable, Equatable {
+    let id: String
+    let title: String
+    let url: URL
+    let sentences: [SentenceBoundary]
+    let isLocalFile: Bool
+
+    init(
+        id: String,
+        title: String,
+        url: URL,
+        sentences: [SentenceBoundary] = [],
+        isLocalFile: Bool = false
+    ) {
+        self.id = id
+        self.title = title
+        self.url = url
+        self.sentences = sentences
+        self.isLocalFile = isLocalFile
+    }
+
+    static func == (lhs: TrainingMediaItem, rhs: TrainingMediaItem) -> Bool {
+        lhs.id == rhs.id
+    }
+}