Set Gemma 2 2B MLX as default local model and add --mlx-model flag

bot.py

@@ -157,6 +157,11 @@ def parse_args() -> argparse.Namespace:
         default="apple",
         help="LLM engine to use: apple/macos for native on-device macOS LLM, claude for Claude Code.",
     )
+    parser.add_argument(
+        "--mlx-model",
+        default="mlx-community/gemma-2-2b-it-4bit",
+        help="MLX model repo or path for local macOS execution (e.g. mlx-community/gemma-2-2b-it-4bit).",
+    )
     parser.add_argument(
         "--claude-model",
         default=DEFAULT_CLAUDE_MODEL,
@@ -490,7 +495,7 @@ def build_llm(args: argparse.Namespace, vocabulary=None, brain=None, observer=No
         logger.info(f"LLM: macOS native model ({reason})")
         personality = read_personality(args.cwd) or "You are a helpful macOS voice assistant."
         system_prompt = personality + "\n\n" + VOICE_STYLE
-        return MacOSLLM(system_prompt=system_prompt, observer=observer)
+        return MacOSLLM(model=args.mlx_model, system_prompt=system_prompt, observer=observer)
 
     logger.info("LLM: Claude Code")
     return ClaudeCodeLLM(