chore(voice): default NOTIFY_VOICE_STYLE to max

SETUP.md

@@ -123,7 +123,7 @@ Verify with `/mcp` inside Claude Code.
 
 Expose an HTTP service that accepts JSON `{"text","style","seed"}` and returns
 WAV audio (and optionally SSE `/tts/stream`). Style default used by this MCP:
-`charlie-t` (`NOTIFY_VOICE_STYLE`).
+`max` (`NOTIFY_VOICE_STYLE`).
 
 Ensure the service is reachable only on trusted networks or behind auth.
 

docs/VOICE-BACKENDS.md

@@ -41,7 +41,7 @@ You may substitute any server that matches the **API contracts** below.
 | GET | `/audio/{name}` | WAV bytes for that job |
 | POST | `/tts/stream` | Optional SSE stream for low-latency speak |
 
-Default style used by the MCP: `charlie-t` (`NOTIFY_VOICE_STYLE`). Provide a
+Default style used by the MCP: `max` (`NOTIFY_VOICE_STYLE`). Provide a
 matching voice-clone reference on the server, or change the style env.
 
 Optional Bearer: server `TTS_API_TOKEN`; client `ASK_QUESTION_TTS_TOKEN` or

src/ask_question_mcp/doctor.py

@@ -912,7 +912,7 @@ def setup_guide(topic: str) -> dict[str, Any]:
             "Provision a host with a suitable GPU (ROCm/CUDA as required by Qwen3-TTS) "
             "or accept CPU-only if your stack supports it.",
             "Install Qwen3-TTS + deps in a venv; place voice reference clips for style "
-            "`charlie-t` (or set NOTIFY_VOICE_STYLE to a style you provide).",
+            "`max` (or set NOTIFY_VOICE_STYLE to a style you provide).",
             "Run a FastAPI (or similar) server that implements:",
             "  - GET /health → 200 when ready",
             "  - POST /tts JSON {text, style, seed} → {name, style, …}; then GET /audio/{name} WAV",

src/ask_question_mcp/voice_acks.py

@@ -1,6 +1,6 @@
 """Voice helpers for ask-question-mcp (acks + question speak).
 
-Never the user's voice: style is always ``charlie-t``, with Piper
+Never the user's voice: style defaults to ``max`` (``NOTIFY_VOICE_STYLE``), with Piper
 fallback. Question lines are cached under ``~/.cache/ask-question-mcp/`` so
 repeats skip the GPU; acks are a fixed phrase set (also cached).
 
@@ -218,7 +218,7 @@ def _tts_svc() -> str:
     ).rstrip("/")
 
 
-_TTS_STYLE = os.environ.get("NOTIFY_VOICE_STYLE", "charlie-t")
+_TTS_STYLE = os.environ.get("NOTIFY_VOICE_STYLE", "max")
 _TTS_SEED = int(os.environ.get("NOTIFY_VOICE_SEED", "2"))
 _TTS_TIMEOUT = float(os.environ.get("NOTIFY_VOICE_TTS_TIMEOUT", "15"))
 # Whole-stream budget (sentence streaming); first chunk still bounded by _TTS_TIMEOUT.