voice-mode 2.28.1__tar.gz → 2.28.3__tar.gz

{voice_mode-2.28.1 → voice_mode-2.28.3}/CHANGELOG.md

@@ -7,6 +7,37 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [2.28.3] - 2025-08-24
+
+### Fixed
+- **Parameter type handling for MCP tools**
+  - Fixed vad_aggressiveness parameter to accept string values from LLMs
+  - Fixed port parameters in kokoro_install and livekit_install
+  - Fixed lines parameter in service management tool
+  - All numeric parameters now properly convert strings to integers
+  - Addresses systemic issue where Claude Code MCP client passes strings
+
+- **Installer script uvx command corrections**
+  - Fixed MCP configuration to use correct command `uvx voice-mode` (without --refresh)
+  - Installer now always refreshes to latest version at start
+  - Removed unnecessary --refresh flags from runtime commands
+  - Updated user-facing command examples to show correct usage
+
+## [2.28.2] - 2025-08-24
+
+### Added
+- **Configurable audio feedback pip delays**
+  - Added VOICEMODE_PIP_LEADING_SILENCE and VOICEMODE_PIP_TRAILING_SILENCE environment variables
+  - Allows customization of silence before and after audio feedback chimes
+  - Configurable via converse tool parameters pip_leading_silence and pip_trailing_silence
+  - Helps prevent audio cutoff on Bluetooth devices and other audio systems with delay
+
+### Fixed
+- **Audio feedback for Bluetooth devices**
+  - Added silence buffer before chimes to prevent Bluetooth audio cutoff
+  - Improved compatibility with devices that have audio activation delay
+  - Better audio feedback experience across different output devices
+
 ## [2.28.1] - 2025-08-24
 
 ### Added

{voice_mode-2.28.1 → voice_mode-2.28.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: voice-mode
-Version: 2.28.1
+Version: 2.28.3
 Summary: VoiceMode - Voice interaction capabilities for AI assistants (formerly voice-mcp)
 Project-URL: Homepage, https://github.com/mbailey/voicemode
 Project-URL: Repository, https://github.com/mbailey/voicemode

{voice_mode-2.28.1 → voice_mode-2.28.3}/voice_mode/__version__.py

@@ -1,3 +1,3 @@
 # This file is automatically updated by 'make release'
 # Do not edit manually
-__version__ = "2.28.1"
+__version__ = "2.28.3"

{voice_mode-2.28.1 → voice_mode-2.28.3}/voice_mode/config.py

@@ -291,6 +291,12 @@ INITIAL_SILENCE_GRACE_PERIOD = float(os.getenv("VOICEMODE_INITIAL_SILENCE_GRACE_
 # Default listen duration for converse tool
 DEFAULT_LISTEN_DURATION = float(os.getenv("VOICEMODE_DEFAULT_LISTEN_DURATION", "120.0"))  # Default 120s listening time
 
+# Audio feedback pip/chime configuration
+# Leading silence before pips to allow Bluetooth devices to wake up
+PIP_LEADING_SILENCE = float(os.getenv("VOICEMODE_PIP_LEADING_SILENCE", "0.1"))  # Default 0.1s - minimal delay for Bluetooth
+# Trailing silence after pips to prevent cutoff
+PIP_TRAILING_SILENCE = float(os.getenv("VOICEMODE_PIP_TRAILING_SILENCE", "0.2"))  # Default 0.2s - reduced for responsiveness
+
 # Audio format configuration
 AUDIO_FORMAT = os.getenv("VOICEMODE_AUDIO_FORMAT", "pcm").lower()
 TTS_AUDIO_FORMAT = os.getenv("VOICEMODE_TTS_AUDIO_FORMAT", "pcm").lower()  # Default to PCM for optimal streaming

{voice_mode-2.28.1 → voice_mode-2.28.3}/voice_mode/core.py

@@ -395,8 +395,9 @@ async def text_to_speech(
                         if event_logger:
                             event_logger.log_event(event_logger.TTS_PLAYBACK_START)
                         
-                        # Add 100ms of silence at the beginning to prevent clipping
-                        silence_duration = 0.1  # seconds
+                        # Add configurable silence at the beginning to prevent clipping
+                        from .config import PIP_LEADING_SILENCE
+                        silence_duration = PIP_LEADING_SILENCE  # seconds
                         silence_samples = int(audio.frame_rate * silence_duration)
                         # Match the shape of the samples array exactly
                         if samples.ndim == 1:
@@ -507,13 +508,21 @@ async def text_to_speech(
         return False, metrics
 
 
-def generate_chime(frequencies: list, duration: float = 0.1, sample_rate: int = SAMPLE_RATE) -> np.ndarray:
+def generate_chime(
+    frequencies: list, 
+    duration: float = 0.1, 
+    sample_rate: int = SAMPLE_RATE,
+    leading_silence: Optional[float] = None,
+    trailing_silence: Optional[float] = None
+) -> np.ndarray:
     """Generate a chime sound with given frequencies.
     
     Args:
         frequencies: List of frequencies to play in sequence
         duration: Duration of each tone in seconds
         sample_rate: Sample rate for audio generation
+        leading_silence: Optional override for leading silence duration (seconds)
+        trailing_silence: Optional override for trailing silence duration (seconds)
         
     Returns:
         Numpy array of audio samples
@@ -521,12 +530,30 @@ def generate_chime(frequencies: list, duration: float = 0.1, sample_rate: int =
     samples_per_tone = int(sample_rate * duration)
     fade_samples = int(sample_rate * 0.01)  # 10ms fade
     
+    # Determine amplitude based on output device
+    amplitude = 0.0375  # Default (very quiet)
+    try:
+        import sounddevice as sd
+        default_output = sd.default.device[1]
+        if default_output is not None:
+            devices = sd.query_devices()
+            device_name = devices[default_output]['name'].lower()
+            # Check for Bluetooth devices (AirPods, Bluetooth headphones, etc)
+            if 'airpod' in device_name or 'bluetooth' in device_name or 'bt' in device_name:
+                amplitude = 0.15  # Higher amplitude for Bluetooth devices
+                logger.debug(f"Bluetooth device detected ({devices[default_output]['name']}), using amplitude {amplitude}")
+            else:
+                amplitude = 0.075  # Moderate amplitude for built-in speakers
+                logger.debug(f"Built-in speaker detected ({devices[default_output]['name']}), using amplitude {amplitude}")
+    except Exception as e:
+        logger.debug(f"Could not detect output device type: {e}, using default amplitude {amplitude}")
+    
     all_samples = []
     
     for freq in frequencies:
         # Generate sine wave
         t = np.linspace(0, duration, samples_per_tone, False)
-        tone = 0.0375 * np.sin(2 * np.pi * freq * t)  # 0.0375 amplitude for extremely quiet volume
+        tone = amplitude * np.sin(2 * np.pi * freq * t)
         
         # Apply fade in/out to prevent clicks
         fade_in = np.linspace(0, 1, fade_samples)
@@ -540,21 +567,55 @@ def generate_chime(frequencies: list, duration: float = 0.1, sample_rate: int =
     # Concatenate all tones
     chime = np.concatenate(all_samples)
     
+    # Import config values if not overridden
+    from .config import PIP_LEADING_SILENCE, PIP_TRAILING_SILENCE
+    
+    # Use parameter overrides or fall back to config
+    actual_leading_silence = leading_silence if leading_silence is not None else PIP_LEADING_SILENCE
+    actual_trailing_silence = trailing_silence if trailing_silence is not None else PIP_TRAILING_SILENCE
+    
+    # Add leading silence for Bluetooth wake-up time
+    # This prevents the beginning of the chime from being cut off
+    silence_samples = int(sample_rate * actual_leading_silence)
+    silence = np.zeros(silence_samples)
+    
+    # Add trailing silence to prevent end cutoff
+    trailing_silence_samples = int(sample_rate * actual_trailing_silence)
+    trailing_silence = np.zeros(trailing_silence_samples)
+    
+    # Combine: leading silence + chime + trailing silence
+    chime_with_buffer = np.concatenate([silence, chime, trailing_silence])
+    
     # Convert to 16-bit integer
-    chime_int16 = (chime * 32767).astype(np.int16)
+    chime_int16 = (chime_with_buffer * 32767).astype(np.int16)
     
     return chime_int16
 
 
-async def play_chime_start(sample_rate: int = SAMPLE_RATE) -> bool:
+async def play_chime_start(
+    sample_rate: int = SAMPLE_RATE,
+    leading_silence: Optional[float] = None,
+    trailing_silence: Optional[float] = None
+) -> bool:
     """Play the recording start chime (ascending tones).
     
+    Args:
+        sample_rate: Sample rate for audio
+        leading_silence: Optional override for leading silence duration (seconds)
+        trailing_silence: Optional override for trailing silence duration (seconds)
+    
     Returns:
         True if chime played successfully, False otherwise
     """
     try:
         import sounddevice as sd
-        chime = generate_chime([800, 1000], duration=0.1, sample_rate=sample_rate)
+        chime = generate_chime(
+            [800, 1000], 
+            duration=0.1, 
+            sample_rate=sample_rate,
+            leading_silence=leading_silence,
+            trailing_silence=trailing_silence
+        )
         sd.play(chime, sample_rate)
         sd.wait()
         return True
@@ -563,15 +624,30 @@ async def play_chime_start(sample_rate: int = SAMPLE_RATE) -> bool:
         return False
 
 
-async def play_chime_end(sample_rate: int = SAMPLE_RATE) -> bool:
+async def play_chime_end(
+    sample_rate: int = SAMPLE_RATE,
+    leading_silence: Optional[float] = None,
+    trailing_silence: Optional[float] = None
+) -> bool:
     """Play the recording end chime (descending tones).
     
+    Args:
+        sample_rate: Sample rate for audio
+        leading_silence: Optional override for leading silence duration (seconds)
+        trailing_silence: Optional override for trailing silence duration (seconds)
+    
     Returns:
         True if chime played successfully, False otherwise
     """
     try:
         import sounddevice as sd
-        chime = generate_chime([1000, 800], duration=0.1, sample_rate=sample_rate)
+        chime = generate_chime(
+            [1000, 800], 
+            duration=0.1, 
+            sample_rate=sample_rate,
+            leading_silence=leading_silence,
+            trailing_silence=trailing_silence
+        )
         sd.play(chime, sample_rate)
         sd.wait()
         return True

{voice_mode-2.28.1 → voice_mode-2.28.3}/voice_mode/tools/converse.py

@@ -784,7 +784,9 @@ async def play_audio_feedback(
     style: str = "whisper", 
     feedback_type: Optional[str] = None,
     voice: str = "nova",
-    model: str = "gpt-4o-mini-tts"
+    model: str = "gpt-4o-mini-tts",
+    pip_leading_silence: Optional[float] = None,
+    pip_trailing_silence: Optional[float] = None
 ) -> None:
     """Play an audio feedback chime
     
@@ -796,6 +798,8 @@ async def play_audio_feedback(
         feedback_type: Kept for compatibility, not used
         voice: Kept for compatibility, not used
         model: Kept for compatibility, not used
+        pip_leading_silence: Optional override for leading silence duration
+        pip_trailing_silence: Optional override for trailing silence duration
     """
     # Use parameter override if provided, otherwise use global setting
     if enabled is False:
@@ -810,11 +814,17 @@ async def play_audio_feedback(
         return
     
     try:
-        # Play appropriate chime
+        # Play appropriate chime with optional delay overrides
         if text == "listening":
-            await play_chime_start()
+            await play_chime_start(
+                leading_silence=pip_leading_silence,
+                trailing_silence=pip_trailing_silence
+            )
         elif text == "finished":
-            await play_chime_end()
+            await play_chime_end(
+                leading_silence=pip_leading_silence,
+                trailing_silence=pip_trailing_silence
+            )
     except Exception as e:
         logger.debug(f"Audio feedback failed: {e}")
         # Don't interrupt the main flow if feedback fails
@@ -1313,8 +1323,10 @@ async def converse(
     audio_format: Optional[str] = None,
     disable_silence_detection: Union[bool, str] = False,
     speed: Optional[float] = None,
-    vad_aggressiveness: Optional[int] = None,
-    skip_tts: Optional[Union[bool, str]] = None
+    vad_aggressiveness: Optional[Union[int, str]] = None,
+    skip_tts: Optional[Union[bool, str]] = None,
+    pip_leading_silence: Optional[float] = None,
+    pip_trailing_silence: Optional[float] = None
 ) -> str:
     """Have a voice conversation - speak a message and optionally listen for response.
     
@@ -1391,6 +1403,10 @@ async def converse(
                   When False: Always use TTS regardless of environment setting
                   When None: Follow VOICEMODE_SKIP_TTS environment variable
                   Useful for rapid development iterations or when voice isn't needed
+        pip_leading_silence: Override leading silence before chimes (default: None uses VOICEMODE_PIP_LEADING_SILENCE env var)
+                             Time in seconds to add before the chime starts (e.g., 1.0 for Bluetooth devices)
+        pip_trailing_silence: Override trailing silence after chimes (default: None uses VOICEMODE_PIP_TRAILING_SILENCE env var)
+                              Time in seconds to add after the chime ends (e.g., 0.5 to prevent cutoff)
         If wait_for_response is False: Confirmation that message was spoken
         If wait_for_response is True: The voice response received (or error/timeout message)
     
@@ -1479,6 +1495,15 @@ async def converse(
     if skip_tts is not None and isinstance(skip_tts, str):
         skip_tts = skip_tts.lower() in ('true', '1', 'yes', 'on')
     
+    # Convert vad_aggressiveness to integer if provided as string
+    if vad_aggressiveness is not None and isinstance(vad_aggressiveness, str):
+        try:
+            vad_aggressiveness = int(vad_aggressiveness)
+            # Validation will happen later in the function
+        except ValueError:
+            logger.warning(f"Invalid VAD aggressiveness value '{vad_aggressiveness}', using default")
+            vad_aggressiveness = None
+    
     # Determine whether to skip TTS
     if skip_tts is not None:
         # Parameter explicitly set, use it
@@ -1797,7 +1822,14 @@ async def converse(
                     await asyncio.sleep(0.5)
                     
                     # Play "listening" feedback sound
-                    await play_audio_feedback("listening", openai_clients, audio_feedback, audio_feedback_style or "whisper")
+                    await play_audio_feedback(
+                        "listening", 
+                        openai_clients, 
+                        audio_feedback, 
+                        audio_feedback_style or "whisper",
+                        pip_leading_silence=pip_leading_silence,
+                        pip_trailing_silence=pip_trailing_silence
+                    )
                     
                     # Record response
                     logger.info(f"🎤 Listening for {listen_duration} seconds...")
@@ -1821,7 +1853,14 @@ async def converse(
                         })
                     
                     # Play "finished" feedback sound
-                    await play_audio_feedback("finished", openai_clients, audio_feedback, audio_feedback_style or "whisper")
+                    await play_audio_feedback(
+                        "finished", 
+                        openai_clients, 
+                        audio_feedback, 
+                        audio_feedback_style or "whisper",
+                        pip_leading_silence=pip_leading_silence,
+                        pip_trailing_silence=pip_trailing_silence
+                    )
                     
                     # Mark the end of recording - this is when user expects response to start
                     user_done_time = time.perf_counter()

{voice_mode-2.28.1 → voice_mode-2.28.3}/voice_mode/tools/service.py

@@ -8,7 +8,7 @@ import platform
 import subprocess
 import time
 from pathlib import Path
-from typing import Literal, Optional, Dict, Any
+from typing import Literal, Optional, Dict, Any, Union
 
 import psutil
 
@@ -980,7 +980,7 @@ async def view_logs(service_name: str, lines: Optional[int] = None) -> str:
 async def service(
     service_name: Literal["whisper", "kokoro", "livekit", "frontend"],
     action: Literal["status", "start", "stop", "restart", "enable", "disable", "logs", "update-service-files"] = "status",
-    lines: Optional[int] = None
+    lines: Optional[Union[int, str]] = None
 ) -> str:
     """Unified service management tool for voice mode services.
     
@@ -1007,6 +1007,14 @@ async def service(
         service("kokoro", "start")    # Start Kokoro service
         service("whisper", "logs", 100)  # View last 100 lines of Whisper logs
     """
+    # Convert lines to integer if provided as string
+    if lines is not None and isinstance(lines, str):
+        try:
+            lines = int(lines)
+        except ValueError:
+            logger.warning(f"Invalid lines value '{lines}', using default 50")
+            lines = 50
+    
     # Route to appropriate handler
     if action == "status":
         return await status_service(service_name)

{voice_mode-2.28.1 → voice_mode-2.28.3}/voice_mode/tools/services/kokoro/install.py

@@ -26,7 +26,7 @@ logger = logging.getLogger("voice-mode")
 async def kokoro_install(
     install_dir: Optional[str] = None,
     models_dir: Optional[str] = None,
-    port: int = 8880,
+    port: Union[int, str] = 8880,
     auto_start: Union[bool, str] = True,
     install_models: Union[bool, str] = True,
     force_reinstall: Union[bool, str] = False,
@@ -54,6 +54,14 @@ async def kokoro_install(
         Installation status with service configuration details
     """
     try:
+        # Convert port to integer if provided as string
+        if isinstance(port, str):
+            try:
+                port = int(port)
+            except ValueError:
+                logger.warning(f"Invalid port value '{port}', using default 8880")
+                port = 8880
+        
         # Check for and migrate old installations
         migration_msg = auto_migrate_if_needed("kokoro")
         

{voice_mode-2.28.1 → voice_mode-2.28.3}/voice_mode/tools/services/livekit/install.py

@@ -124,7 +124,7 @@ room:
 @mcp.tool()
 async def livekit_install(
     install_dir: Optional[str] = None,
-    port: int = 7880,
+    port: Union[int, str] = 7880,
     force_reinstall: Union[bool, str] = False,
     auto_enable: Optional[Union[bool, str]] = None,
     version: str = "latest"
@@ -159,6 +159,14 @@ async def livekit_install(
         # Check system
         system = platform.system()
         
+        # Convert port to integer if provided as string
+        if isinstance(port, str):
+            try:
+                port = int(port)
+            except ValueError:
+                logger.warning(f"Invalid port value '{port}', using default 7880")
+                port = 7880
+        
         # Handle string boolean conversions
         if isinstance(force_reinstall, str):
             force_reinstall = force_reinstall.lower() in ("true", "1", "yes", "on")