Set Qwen2.5-7B-Instruct-4bit as default MLX model and add spoken text filtering

apple_llm.py

@@ -86,13 +86,34 @@ def _latest_user_text(context) -> str:
     return ""
 
 
+import re
+
+def _clean_spoken_text(text: str) -> str:
+    """Clean text for speech output and truncate fake turn generations."""
+    if not text:
+        return ""
+    # Truncate if model hallucinates fake turn markers
+    for marker in ("User:", "Human:", "Assistant:", "\nUser", "\nHuman", "\nAssistant"):
+        if marker in text:
+            text = text.split(marker)[0]
+    # Remove markdown code blocks
+    text = re.sub(r"```[\s\S]*?```", "", text)
+    # Remove inline code ticks
+    text = re.sub(r"`[^`]*`", "", text)
+    # Remove markdown syntax characters
+    text = re.sub(r"[\#\*\_\~]", "", text)
+    # Flatten newlines into clear speech
+    lines = [line.strip() for line in text.splitlines() if line.strip()]
+    return " ".join(lines).strip()
+
+
 class MacOSLLM(FrameProcessor):
     """Runs user turns through macOS native LLM (Apple Intelligence or MLX)."""
 
     def __init__(
         self,
         *,
-        model: str = "mlx-community/gemma-2-2b-it-4bit",
+        model: str = "mlx-community/Qwen2.5-7B-Instruct-4bit",
         system_prompt: str | None = None,
         observer=None,
         **kwargs,

bot.py

@@ -159,8 +159,8 @@ def parse_args() -> argparse.Namespace:
     )
     parser.add_argument(
         "--mlx-model",
-        default="mlx-community/gemma-2-2b-it-4bit",
+        default="mlx-community/Qwen2.5-7B-Instruct-4bit",
-        help="MLX model repo or path for local macOS execution (e.g. mlx-community/gemma-2-2b-it-4bit).",
+        help="MLX model repo or path for local macOS execution (e.g. mlx-community/Qwen2.5-7B-Instruct-4bit).",
     )
     parser.add_argument(
         "--claude-model",
@@ -495,6 +495,8 @@ 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
+        if brain and (memory := brain.prompt_block()):
+            system_prompt += "\n\n" + memory
         return MacOSLLM(model=args.mlx_model, system_prompt=system_prompt, observer=observer)
 
     logger.info("LLM: Claude Code")