Add macOS LLM model engine support and environment setup

bot.py

@@ -151,6 +151,12 @@ def parse_args() -> argparse.Namespace:
     parser.add_argument(
         "--list-voices", action="store_true", help="Print macOS system voices and exit."
     )
+    parser.add_argument(
+        "--llm-engine",
+        choices=["apple", "macos", "claude"],
+        default="apple",
+        help="LLM engine to use: apple/macos for native on-device macOS LLM, claude for Claude Code.",
+    )
     parser.add_argument(
         "--claude-model",
         default=DEFAULT_CLAUDE_MODEL,
@@ -475,6 +481,23 @@ def build_claude_options(args: argparse.Namespace, vocabulary=None, brain=None)
     )
 
 
+def build_llm(args: argparse.Namespace, vocabulary=None, brain=None, observer=None):
+    if args.llm_engine in ("apple", "macos"):
+        from apple_llm import MacOSLLM, probe_apple_llm
+
+        available, reason = probe_apple_llm()
+        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)
+
+    logger.info("LLM: Claude Code")
+    return ClaudeCodeLLM(
+        options=build_claude_options(args, vocabulary, brain),
+        observer=observer,
+    )
+
+
 async def main() -> int:
     args = parse_args()
 
@@ -523,10 +546,7 @@ async def main() -> int:
         if vocabulary:
             vocabulary.observe(text)
 
-    llm = ClaudeCodeLLM(
-        options=build_claude_options(args, vocabulary, brain),
-        observer=on_reply,
-    )
+    llm = build_llm(args, vocabulary, brain, observer=on_reply)
 
     tts = build_tts(args)