import Foundation
import SstCore
import WhisperKit

public final class WhisperKitTranscriber: Transcriber {
    public enum TranscriberError: Error, LocalizedError {
        case notLoaded

        public var errorDescription: String? { "Spraakmodel is nog niet geladen." }
    }

    private let modelName: String
    private var pipeline: WhisperKit?

    public init(model: String = "large-v3-v20240930_turbo") {
        self.modelName = model
    }

    /// Downloads the model on first run (≈1.5 GB) and loads it into memory.
    public func load() async throws {
        pipeline = try await WhisperKit(WhisperKitConfig(model: modelName))
    }

    public func transcribe(_ samples: [Float]) async throws -> String {
        guard let pipeline else { throw TranscriberError.notLoaded }
        // language nil + detectLanguage: Whisper picks NL or EN per utterance.
        let options = DecodingOptions(
            task: .transcribe,
            language: nil,
            detectLanguage: true,
            skipSpecialTokens: true,
            withoutTimestamps: true
        )
        let results = try await pipeline.transcribe(audioArray: samples, decodeOptions: options)
        return results
            .map(\.text)
            .joined(separator: " ")
            .trimmingCharacters(in: .whitespacesAndNewlines)
    }
}