Skip to content

Integration.md

Latest commit

 

History

History
127 lines (95 loc) · 4.03 KB

Integration.md

File metadata and controls

127 lines (95 loc) · 4.03 KB
Raw

🚀 Integration & Concurrency: Architectural Guide

This guide is for developers and system architects integrating the AudioIntelligence SDK into enterprise ecosystems or professional Mastering Dashboards.

1. Professional Installation (SPM)

AudioIntelligence is a modular SDK optimized for the Swift Package Manager.

Configuration

Update your Package.swift to include the Infinity Engine:

let package = Package(
    name: "MyProAudioApp",
    dependencies: [
        .package(url: "https://github.com/trgysvc/audiointelligence.git", from: "8.2.3")
    ],
    targets: [
        .target(
            name: "MyTarget",
            dependencies: [
                .product(name: "AudioIntelligence", package: "AudioIntelligence"),
                .product(name: "AudioIntelligenceUI", package: "AudioIntelligence") // Add for UI views
            ]
        )
    ]
)

2. Structured Concurrency (Swift 6)

The SDK utilizes the Actor Model to ensure that heavy DSP calculations never block the UI thread or interfere with low-latency audio tasks.

The Infinity Actor

All analysis is performed within the AudioIntelligence actor. This provides compile-time protection against data races.

Async Analysis Lifecycle

import AudioIntelligence

let sdk = AudioIntelligence()

Task {
    do {
        let report = try await sdk.analyze(url: songURL) { progress, stage, _ in
            print("[\(stage)] \(Int(progress * 100))%")
        }

        // Measurement (validated) vs estimation (statistical):
        print("Loudness: \(report.measurements.loudness.integrated.value) LUFS")  // EBU R128
        print("Structure: \(report.estimations.structure.count) sections detected")
        print("Tempo: \(report.estimations.tempo.value) BPM @ \(report.estimations.tempo.confidence)")
    } catch {
        print("Analysis Error: \(error)")
    }
}

3. 🎨 Native SwiftUI Visualizations

The AudioIntelligenceUI module provides high-performance, Metal-accelerated views for your dashboard.

Dashboard & spectral views

The module ships MainDashboardView (a full engineering dashboard) and the SpectralLandscapeView component. Both consume the typed AudioReport:

import SwiftUI
import AudioIntelligenceUI

struct AnalysisDashboard: View {
    @State private var report: AudioReport?

    var body: some View {
        VStack {
            if let report {
                // Full engineering dashboard
                MainDashboardView(report: report)

                // Or drive an individual component from the heavy feature series
                SpectralLandscapeView(magnitudes: report.features?.magnitudeSpectrogram ?? [])
                    .frame(height: 300)
            }
        }
    }
}

4. Advanced Memory & Cache Management

Hybrid 4GB Persistent Cache

AudioIntelligence implements an industrial-grade persistent store:

  • Identifier: Keyed to a SHA256 content hash. Moving or renaming files does NOT trigger re-analysis.
  • Auto-Eviction: Maintains a strict 4GB disk limit with an LRU policy.

Manual Invalidation

await sdk.invalidateCache() // Clears the hybrid store

5. Hardware Optimization (Apple Silicon)

AudioIntelligence optimizes for the modern M-series SoC:

  • Accelerate / vDSP (incl. AMX): High-throughput FFT and matrix/vector math.
  • Metal GPU: Parallelized DSP kernels and UI rendering, with a CPU fallback when Metal is unavailable.

The analysis pipeline (analyze()) is pure Swift on Accelerate/AVFoundation/Metal — no Core ML, no ANE inference, no network. (There is a separate NeuralSeparationEngine interface that can host a Core ML model for stem separation, but no model ships and it is not part of analyze(). The instrument layer IS built and data-derived — InstrumentEngine, fit and calibrated on real OpenMIC-2018 audio, no Core ML/ANE — but held-out accuracy is class-dependent; see README Validation Status.)

For technical specs on specific analysis engines, see Engines.md.