#Apple MLX Support (apple_mlx)

Local package collecting experimental Dart wrappers around lib_mlx, speech_mlx, and mlx_audio.

Development status: Native LLM inference is disabled in LocalMind. The current lib_mlx 0.1.1 dependency ships a mock model core that returns stub responses; it does not execute downloaded MLX weights. Its bridge targets iOS only. The speech dependencies target macOS Apple Silicon and are not integrated into the LocalMind mobile app.

#Capabilities

#1. LLM Text Generation (AppleMlxLlmManager & lib_mlx)

• Prototype only: Exercises the native lifecycle and local OpenAI-compatible HTTP/SSE contract against the upstream mock core.
• Not user-facing: AppleMlxCapabilities.nativeLlmInferenceAvailable stays false until a real, device-validated backend replaces the mock.

#2. Speech-To-Text (AppleMlxSpeechManager & speech_mlx / mlx_audio)

• macOS only: These dependencies do not provide iOS speech inference.
• Parakeet-TDT & Whisper: High-accuracy local speech recognition.
• Isolate Offloading: Dedicated worker isolates keep audio processing and resampling off the UI isolate.

#3. Text-To-Speech (PocketTTS)

• Neural Voice Synthesis: Real-time voice generation with incremental PCM streaming.
• Voices: Built-in multi-voice support (alba, marius, javert, jean).

#4. Conversational Turn Detection (SmartTurn)

• Smart Turn v3.2: Sub-millisecond endpoint classifier to determine whether a speaker finished speaking during a voice conversation.

#Quickstart

#LLM Text Generation

The following example is for bridge development and contract testing only. It does not perform real inference with lib_mlx 0.1.1.

import 'package:apple_mlx/apple_mlx.dart';

Future<void> main() async {
  final llm = AppleMlxLlmManager();

  // 1. Load MLX model and launch local server
  final serverInfo = await llm.loadAndStartServer(
    modelPath: '/path/to/mlx-community/gemma-4-e2b-it-4bit',
    modelId: AppleMlxModelIds.gemma4E2b,
    thinkingEnabled: true,
  );
  print('MLX Server running on ${serverInfo.baseUrl}');

  // 2. Stream generation
  final stream = llm.generateStream(
    messages: [
      {'role': 'user', 'content': 'Explain quantum computing in simple terms.'},
    ],
    temperature: 0.7,
  );

  await for (final delta in stream) {
    if (delta.reasoningContent != null) {
      print('[Thinking] ${delta.reasoningContent}');
    }
    if (delta.content != null) {
      print(delta.content);
    }
  }

  // 3. Clean up
  await llm.dispose();
}

#STT and TTS Audio

import 'package:apple_mlx/apple_mlx.dart';

Future<void> main() async {
  final speech = AppleMlxSpeechManager();
  await speech.initialize();

  // Transcribe audio
  final text = await speech.transcribeWavFile('input.wav');
  print('Transcript: $text');

  // Synthesize voice
  await speech.synthesizeToWav(
    text: 'Hello from Apple MLX on LocalMind!',
    outputPath: 'output.wav',
    voice: 'alba',
  );

  await speech.dispose();
}