agentvibes 1.1.2 → 2.0.0

package/.claude/hooks/play-tts-piper.sh

@@ -0,0 +1,175 @@
+#!/bin/bash
+#
+# @fileoverview Piper TTS Provider Implementation
+# @context Free, offline neural TTS for WSL/Linux
+# @architecture Implements provider contract for Piper binary
+# @dependencies piper (pipx), piper-voice-manager.sh, mpv/aplay
+# @entrypoints Called by play-tts.sh router
+# @patterns Provider contract: text/voice → audio file path
+# @related play-tts.sh, piper-voice-manager.sh, GitHub Issue #25
+#
+
+# Fix locale warnings
+export LC_ALL=C
+
+TEXT="$1"
+VOICE_OVERRIDE="$2"  # Optional: voice model name
+
+# Source voice manager and language manager
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "$SCRIPT_DIR/piper-voice-manager.sh"
+source "$SCRIPT_DIR/language-manager.sh"
+
+# Default voice for Piper
+DEFAULT_VOICE="en_US-lessac-medium"
+
+# @function determine_voice_model
+# @intent Resolve voice name to Piper model name with language support
+# @why Support voice override, language-specific voices, and default fallback
+# @param Uses global: $VOICE_OVERRIDE
+# @returns Sets $VOICE_MODEL global variable
+# @sideeffects None
+VOICE_MODEL=""
+
+# Get current language setting
+CURRENT_LANGUAGE=$(get_current_language)
+
+if [[ -n "$VOICE_OVERRIDE" ]]; then
+  # Use override if provided
+  VOICE_MODEL="$VOICE_OVERRIDE"
+  echo "🎤 Using voice: $VOICE_OVERRIDE (session-specific)"
+else
+  # Try to get language-specific voice
+  LANG_VOICE=$(get_voice_for_language "$CURRENT_LANGUAGE" "piper" 2>/dev/null)
+
+  if [[ -n "$LANG_VOICE" ]]; then
+    VOICE_MODEL="$LANG_VOICE"
+    echo "🌍 Using $CURRENT_LANGUAGE voice: $LANG_VOICE (Piper)"
+  else
+    # Use default voice
+    VOICE_MODEL="$DEFAULT_VOICE"
+  fi
+fi
+
+# @function validate_inputs
+# @intent Check required parameters
+# @why Fail fast with clear errors if inputs missing
+# @exitcode 1=missing text, 2=missing piper binary
+if [[ -z "$TEXT" ]]; then
+  echo "Usage: $0 \"text to speak\" [voice_model_name]"
+  exit 1
+fi
+
+# Check if Piper is installed
+if ! command -v piper &> /dev/null; then
+  echo "❌ Error: Piper TTS not installed"
+  echo "Install with: pipx install piper-tts"
+  echo "Or run: .claude/hooks/piper-installer.sh"
+  exit 2
+fi
+
+# @function ensure_voice_downloaded
+# @intent Download voice model if not cached
+# @why Provide seamless experience with automatic downloads
+# @param Uses global: $VOICE_MODEL
+# @sideeffects Downloads voice model files
+# @edgecases Prompts user for consent before downloading
+if ! verify_voice "$VOICE_MODEL"; then
+  echo "📥 Voice model not found: $VOICE_MODEL"
+  echo "   File size: ~25MB"
+  echo "   Preview: https://huggingface.co/rhasspy/piper-voices"
+  echo ""
+  read -p "   Download this voice model? [y/N]: " -n 1 -r
+  echo
+
+  if [[ $REPLY =~ ^[Yy]$ ]]; then
+    if ! download_voice "$VOICE_MODEL"; then
+      echo "❌ Failed to download voice model"
+      echo "Fix: Download manually or choose different voice"
+      exit 3
+    fi
+  else
+    echo "❌ Voice download cancelled"
+    exit 3
+  fi
+fi
+
+# Get voice model path
+VOICE_PATH=$(get_voice_path "$VOICE_MODEL")
+if [[ $? -ne 0 ]]; then
+  echo "❌ Voice model path not found: $VOICE_MODEL"
+  exit 3
+fi
+
+# @function determine_audio_directory
+# @intent Find appropriate directory for audio file storage
+# @why Supports project-local and global storage
+# @returns Sets $AUDIO_DIR global variable
+if [[ -n "$CLAUDE_PROJECT_DIR" ]]; then
+  AUDIO_DIR="$CLAUDE_PROJECT_DIR/.claude/audio"
+else
+  # Fallback: try to find .claude directory in current path
+  CURRENT_DIR="$PWD"
+  while [[ "$CURRENT_DIR" != "/" ]]; do
+    if [[ -d "$CURRENT_DIR/.claude" ]]; then
+      AUDIO_DIR="$CURRENT_DIR/.claude/audio"
+      break
+    fi
+    CURRENT_DIR=$(dirname "$CURRENT_DIR")
+  done
+  # Final fallback to global if no project .claude found
+  if [[ -z "$AUDIO_DIR" ]]; then
+    AUDIO_DIR="$HOME/.claude/audio"
+  fi
+fi
+
+mkdir -p "$AUDIO_DIR"
+TEMP_FILE="$AUDIO_DIR/tts-$(date +%s).wav"
+
+# @function synthesize_with_piper
+# @intent Generate speech using Piper TTS
+# @why Provides free, offline TTS alternative
+# @param Uses globals: $TEXT, $VOICE_PATH
+# @returns Creates WAV file at $TEMP_FILE
+# @exitcode 0=success, 4=synthesis error
+# @sideeffects Creates audio file
+# @edgecases Handles piper errors, invalid models
+echo "$TEXT" | piper --model "$VOICE_PATH" --output_file "$TEMP_FILE" 2>/dev/null
+
+if [[ ! -f "$TEMP_FILE" ]] || [[ ! -s "$TEMP_FILE" ]]; then
+  echo "❌ Failed to synthesize speech with Piper"
+  echo "Voice model: $VOICE_MODEL"
+  echo "Check that voice model is valid"
+  exit 4
+fi
+
+# @function add_silence_padding
+# @intent Add silence to prevent WSL audio static
+# @why WSL audio subsystem cuts off first ~200ms
+# @param Uses global: $TEMP_FILE
+# @returns Updates $TEMP_FILE to padded version
+# @sideeffects Modifies audio file
+# AI NOTE: Use ffmpeg if available, otherwise skip padding (degraded experience)
+if command -v ffmpeg &> /dev/null; then
+  PADDED_FILE="$AUDIO_DIR/tts-padded-$(date +%s).wav"
+  # Add 200ms of silence at the beginning
+  ffmpeg -f lavfi -i anullsrc=r=44100:cl=stereo:d=0.2 -i "$TEMP_FILE" \
+    -filter_complex "[0:a][1:a]concat=n=2:v=0:a=1[out]" \
+    -map "[out]" -y "$PADDED_FILE" 2>/dev/null
+
+  if [[ -f "$PADDED_FILE" ]]; then
+    rm -f "$TEMP_FILE"
+    TEMP_FILE="$PADDED_FILE"
+  fi
+fi
+
+# @function play_audio
+# @intent Play generated audio using available player
+# @why Support multiple audio players
+# @param Uses global: $TEMP_FILE
+# @sideeffects Plays audio in background
+# Play audio (WSL/Linux) in background
+(mpv "$TEMP_FILE" 2>/dev/null || aplay "$TEMP_FILE" 2>/dev/null || paplay "$TEMP_FILE" 2>/dev/null) &
+
+echo "🎵 Saved to: $TEMP_FILE"
+echo "🎤 Voice used: $VOICE_MODEL (Piper TTS)"

package/.claude/hooks/play-tts.sh

@@ -109,8 +109,25 @@ curl -s -X POST "https://api.elevenlabs.io/v1/text-to-speech/${VOICE_ID}" \
   -d "{\"text\":\"${TEXT}\",\"model_id\":\"eleven_monolingual_v1\",\"voice_settings\":{\"stability\":0.5,\"similarity_boost\":0.75}}" \
   -o "${TEMP_FILE}"
 
-# Play audio (WSL/Linux) in background to avoid blocking
+# Add silence padding to prevent WSL audio static
 if [ -f "${TEMP_FILE}" ]; then
+  # Check if ffmpeg is available for adding padding
+  if command -v ffmpeg &> /dev/null; then
+    PADDED_FILE="$AUDIO_DIR/tts-padded-$(date +%s).mp3"
+    # Add 200ms of silence at the beginning to prevent static
+    ffmpeg -f lavfi -i anullsrc=r=44100:cl=stereo:d=0.2 -i "${TEMP_FILE}" \
+      -filter_complex "[0:a][1:a]concat=n=2:v=0:a=1[out]" \
+      -map "[out]" -y "${PADDED_FILE}" 2>/dev/null
+
+    if [ -f "${PADDED_FILE}" ]; then
+      # Use padded file and clean up original
+      rm -f "${TEMP_FILE}"
+      TEMP_FILE="${PADDED_FILE}"
+    fi
+    # If padding failed, just use original file
+  fi
+
+  # Play audio (WSL/Linux) in background to avoid blocking
   (paplay "${TEMP_FILE}" 2>/dev/null || aplay "${TEMP_FILE}" 2>/dev/null || mpg123 "${TEMP_FILE}" 2>/dev/null) &
   # Keep temp files for later review - cleaned up weekly by cron
   echo "🎵 Saved to: ${TEMP_FILE}"

package/.claude/hooks/play-tts.sh.backup-20251005-163851

@@ -0,0 +1,138 @@
+#!/bin/bash
+# Quick TTS playback script with session-specific voice support
+# Usage: play-tts.sh "Text to speak" [voice_name_or_id]
+#
+# Examples:
+#   play-tts.sh "Hello world"                           # Uses default voice from voice manager
+#   play-tts.sh "Hello world" "Sarah"                   # Uses Sarah voice by name
+#   play-tts.sh "Hello world" "KTPVrSVAEUSJRClDzBw7"   # Uses voice by direct ID
+#
+# This allows different sessions to use different voices for easy identification!
+
+# Fix locale warnings
+export LC_ALL=C
+
+TEXT="$1"
+VOICE_OVERRIDE="$2"  # Optional: voice name or direct voice ID
+API_KEY="${ELEVENLABS_API_KEY}"
+
+# Check for project-local pretext configuration
+CONFIG_DIR="${CLAUDE_PROJECT_DIR:-.}/.claude/config"
+CONFIG_FILE="$CONFIG_DIR/agentvibes.json"
+
+if [[ -f "$CONFIG_FILE" ]] && command -v jq &> /dev/null; then
+  PRETEXT=$(jq -r '.pretext // empty' "$CONFIG_FILE" 2>/dev/null)
+  if [[ -n "$PRETEXT" ]]; then
+    TEXT="$PRETEXT: $TEXT"
+  fi
+fi
+
+# Limit text length to prevent API issues (max 500 chars for safety)
+if [ ${#TEXT} -gt 500 ]; then
+  TEXT="${TEXT:0:497}..."
+  echo "⚠️ Text truncated to 500 characters for API safety"
+fi
+
+# Source the single voice configuration file
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "$SCRIPT_DIR/voices-config.sh"
+
+# Determine which voice to use
+VOICE_ID=""
+
+if [[ -n "$VOICE_OVERRIDE" ]]; then
+  # Check if override is a voice name (lookup in mapping)
+  if [[ -n "${VOICES[$VOICE_OVERRIDE]}" ]]; then
+    VOICE_ID="${VOICES[$VOICE_OVERRIDE]}"
+    echo "🎤 Using voice: $VOICE_OVERRIDE (session-specific)"
+  # Check if override looks like a voice ID (alphanumeric string ~20 chars)
+  elif [[ "$VOICE_OVERRIDE" =~ ^[a-zA-Z0-9]{15,30}$ ]]; then
+    VOICE_ID="$VOICE_OVERRIDE"
+    echo "🎤 Using custom voice ID (session-specific)"
+  else
+    echo "⚠️ Unknown voice '$VOICE_OVERRIDE', using default"
+  fi
+fi
+
+# If no override or invalid override, use default from voice manager
+if [[ -z "$VOICE_ID" ]]; then
+  VOICE_MANAGER_SCRIPT="$(dirname "$0")/voice-manager.sh"
+  if [[ -f "$VOICE_MANAGER_SCRIPT" ]]; then
+    VOICE_NAME=$("$VOICE_MANAGER_SCRIPT" get)
+    VOICE_ID="${VOICES[$VOICE_NAME]}"
+  fi
+
+  # Final fallback to Cowboy Bob default
+  if [[ -z "$VOICE_ID" ]]; then
+    echo "⚠️ No voice configured, using Cowboy Bob default"
+    VOICE_ID="${VOICES[Cowboy Bob]}"
+  fi
+fi
+
+if [ -z "$TEXT" ]; then
+  echo "Usage: $0 \"text to speak\""
+  exit 1
+fi
+
+if [ -z "$API_KEY" ]; then
+  echo "Error: ELEVENLABS_API_KEY not set"
+  exit 1
+fi
+
+# Create audio file in project-local storage
+# Use project directory if available, otherwise fall back to global
+if [[ -n "$CLAUDE_PROJECT_DIR" ]]; then
+  AUDIO_DIR="$CLAUDE_PROJECT_DIR/.claude/audio"
+else
+  # Fallback: try to find .claude directory in current path
+  CURRENT_DIR="$PWD"
+  while [[ "$CURRENT_DIR" != "/" ]]; do
+    if [[ -d "$CURRENT_DIR/.claude" ]]; then
+      AUDIO_DIR="$CURRENT_DIR/.claude/audio"
+      break
+    fi
+    CURRENT_DIR=$(dirname "$CURRENT_DIR")
+  done
+  # Final fallback to global if no project .claude found
+  if [[ -z "$AUDIO_DIR" ]]; then
+    AUDIO_DIR="$HOME/.claude/audio"
+  fi
+fi
+
+mkdir -p "$AUDIO_DIR"
+TEMP_FILE="$AUDIO_DIR/tts-$(date +%s).mp3"
+
+# Generate audio
+curl -s -X POST "https://api.elevenlabs.io/v1/text-to-speech/${VOICE_ID}" \
+  -H "xi-api-key: ${API_KEY}" \
+  -H "Content-Type: application/json" \
+  -d "{\"text\":\"${TEXT}\",\"model_id\":\"eleven_monolingual_v1\",\"voice_settings\":{\"stability\":0.5,\"similarity_boost\":0.75}}" \
+  -o "${TEMP_FILE}"
+
+# Add silence padding to prevent WSL audio static
+if [ -f "${TEMP_FILE}" ]; then
+  # Check if ffmpeg is available for adding padding
+  if command -v ffmpeg &> /dev/null; then
+    PADDED_FILE="$AUDIO_DIR/tts-padded-$(date +%s).mp3"
+    # Add 200ms of silence at the beginning to prevent static
+    ffmpeg -f lavfi -i anullsrc=r=44100:cl=stereo:d=0.2 -i "${TEMP_FILE}" \
+      -filter_complex "[0:a][1:a]concat=n=2:v=0:a=1[out]" \
+      -map "[out]" -y "${PADDED_FILE}" 2>/dev/null
+
+    if [ -f "${PADDED_FILE}" ]; then
+      # Use padded file and clean up original
+      rm -f "${TEMP_FILE}"
+      TEMP_FILE="${PADDED_FILE}"
+    fi
+    # If padding failed, just use original file
+  fi
+
+  # Play audio (WSL/Linux) in background to avoid blocking
+  (paplay "${TEMP_FILE}" 2>/dev/null || aplay "${TEMP_FILE}" 2>/dev/null || mpg123 "${TEMP_FILE}" 2>/dev/null) &
+  # Keep temp files for later review - cleaned up weekly by cron
+  echo "🎵 Saved to: ${TEMP_FILE}"
+  echo "🎤 Voice used: ${VOICE_NAME} (${VOICE_ID})"
+else
+  echo "Failed to generate audio"
+  exit 1
+fi