agentvibes 2.0.3 → 2.0.5

package/.claude/commands/agent-vibes/provider.md

@@ -0,0 +1,54 @@
+---
+description: Manage TTS providers (list, switch, info, test)
+argument-hint: [command] [args...]
+---
+
+# Provider Management Commands
+
+Manage TTS providers (ElevenLabs, Piper) - switch between providers, view details, and test.
+
+## Usage
+
+```bash
+/agent-vibes:provider list              # Show all available providers
+/agent-vibes:provider switch <name>     # Switch to a different provider
+/agent-vibes:provider info <name>       # Show detailed provider information
+/agent-vibes:provider test              # Test current provider
+/agent-vibes:provider get               # Show current active provider
+/agent-vibes:provider help              # Show this help
+```
+
+## Examples
+
+```bash
+# List available providers
+/agent-vibes:provider list
+
+# Switch to Piper (free, offline)
+/agent-vibes:provider switch piper
+
+# Switch to ElevenLabs (premium quality)
+/agent-vibes:provider switch elevenlabs
+
+# Get info about a provider
+/agent-vibes:provider info piper
+
+# Test current provider
+/agent-vibes:provider test
+
+# Show current provider
+/agent-vibes:provider get
+```
+
+## Provider Comparison
+
+| Feature | ElevenLabs | Piper |
+|---------|------------|-------|
+| Quality | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ |
+| Cost | Free tier + $5-22/mo | Free forever |
+| Offline | No | Yes |
+| Platform | All | WSL/Linux only |
+
+Learn more: agentvibes.org/providers
+
+!bash .claude/hooks/provider-commands.sh $ARGUMENTS

package/.claude/hooks/piper-download-voices.sh

@@ -0,0 +1,133 @@
+#!/bin/bash
+#
+# @fileoverview Piper Voice Model Downloader
+# @context Downloads Piper TTS voice models from HuggingFace
+# @purpose Batch download popular voices after installation
+# @dependencies piper-voice-manager.sh, piper binary
+# @usage ./piper-download-voices.sh [--yes|-y]
+#   --yes|-y: Skip confirmation prompt (auto-download)
+#
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "$SCRIPT_DIR/piper-voice-manager.sh"
+
+# Parse command line arguments
+AUTO_YES=false
+if [[ "$1" == "--yes" ]] || [[ "$1" == "-y" ]]; then
+  AUTO_YES=true
+fi
+
+# Common voice models to download
+COMMON_VOICES=(
+  "en_US-lessac-medium"      # Default, clear male
+  "en_US-amy-medium"         # Warm female
+  "en_US-joe-medium"         # Professional male
+  "en_US-ryan-high"          # Expressive male
+  "en_US-libritts-high"      # Premium quality
+)
+
+echo "🎙️  Piper Voice Model Downloader"
+echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+echo ""
+echo "This will download the most commonly used Piper voice models."
+echo "Each voice is approximately 25MB."
+echo ""
+
+# Check if piper is installed
+if ! command -v piper &> /dev/null; then
+  echo "❌ Error: Piper TTS not installed"
+  echo "Install with: pipx install piper-tts"
+  exit 1
+fi
+
+# Get storage directory
+VOICE_DIR=$(get_voice_storage_dir)
+
+echo "📂 Storage location: $VOICE_DIR"
+echo ""
+
+# Count already downloaded
+ALREADY_DOWNLOADED=0
+ALREADY_DOWNLOADED_LIST=()
+NEED_DOWNLOAD=()
+
+for voice in "${COMMON_VOICES[@]}"; do
+  if verify_voice "$voice" 2>/dev/null; then
+    ((ALREADY_DOWNLOADED++))
+    ALREADY_DOWNLOADED_LIST+=("$voice")
+  else
+    NEED_DOWNLOAD+=("$voice")
+  fi
+done
+
+echo "📊 Status:"
+echo "   Already downloaded: $ALREADY_DOWNLOADED voice(s)"
+echo "   Need to download: ${#NEED_DOWNLOAD[@]} voice(s)"
+echo ""
+
+# Show already downloaded voices
+if [[ $ALREADY_DOWNLOADED -gt 0 ]]; then
+  echo "✅ Already downloaded (skipped):"
+  for voice in "${ALREADY_DOWNLOADED_LIST[@]}"; do
+    echo "   ✓ $voice"
+  done
+  echo ""
+fi
+
+if [[ ${#NEED_DOWNLOAD[@]} -eq 0 ]]; then
+  echo "🎉 All common voices ready to use!"
+  exit 0
+fi
+
+echo "Voices to download:"
+for voice in "${NEED_DOWNLOAD[@]}"; do
+  echo "  • $voice (~25MB)"
+done
+echo ""
+
+# Ask for confirmation (skip if --yes flag provided)
+if [[ "$AUTO_YES" == "false" ]]; then
+  read -p "Download ${#NEED_DOWNLOAD[@]} voice model(s)? [Y/n]: " -n 1 -r
+  echo
+
+  if [[ ! $REPLY =~ ^[Yy]$ ]] && [[ -n $REPLY ]]; then
+    echo "❌ Download cancelled"
+    exit 0
+  fi
+else
+  echo "Auto-downloading ${#NEED_DOWNLOAD[@]} voice model(s)..."
+  echo ""
+fi
+
+# Download each voice
+DOWNLOADED=0
+FAILED=0
+
+for voice in "${NEED_DOWNLOAD[@]}"; do
+  echo ""
+  echo "📥 Downloading: $voice..."
+
+  if download_voice "$voice"; then
+    ((DOWNLOADED++))
+    echo "✅ Downloaded: $voice"
+  else
+    ((FAILED++))
+    echo "❌ Failed: $voice"
+  fi
+done
+
+echo ""
+echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+echo "📊 Download Summary:"
+echo "   ✅ Successfully downloaded: $DOWNLOADED"
+echo "   ❌ Failed: $FAILED"
+echo "   📦 Total voices available: $((ALREADY_DOWNLOADED + DOWNLOADED))"
+echo ""
+
+if [[ $DOWNLOADED -gt 0 ]]; then
+  echo "✨ Ready to use Piper TTS with downloaded voices!"
+  echo ""
+  echo "Try it:"
+  echo "  /agent-vibes:provider switch piper"
+  echo "  /agent-vibes:preview"
+fi

package/.claude/hooks/piper-voice-manager.sh

@@ -0,0 +1,194 @@
+#!/bin/bash
+#
+# @fileoverview Piper Voice Model Management
+# @context Manages downloading, caching, and validating Piper ONNX voice models
+# @architecture Voice model lifecycle management for Piper provider
+# @dependencies curl, piper binary
+# @entrypoints Sourced by play-tts-piper.sh and provider management commands
+# @patterns HuggingFace model repository integration, file-based caching
+# @related play-tts-piper.sh, provider-manager.sh, GitHub Issue #25
+#
+
+# Base URL for Piper voice models on HuggingFace
+PIPER_VOICES_BASE_URL="https://huggingface.co/rhasspy/piper-voices/resolve/main"
+
+# @function get_voice_storage_dir
+# @intent Determine directory for storing Piper voice models
+# @why Voice models are large (~25MB each) and should be shared globally across all projects
+# @returns Echoes path to voice storage directory (~/.claude/piper-voices)
+# @sideeffects Creates directory if it doesn't exist
+# @architecture Always uses global storage to avoid redundant downloads per project
+get_voice_storage_dir() {
+  # Always use global storage for voice models
+  # This prevents downloading 125MB+ of models into every project
+  local voice_dir="$HOME/.claude/piper-voices"
+
+  mkdir -p "$voice_dir"
+  echo "$voice_dir"
+}
+
+# @function verify_voice
+# @intent Check if voice model files exist locally
+# @why Avoid redundant downloads, detect missing models
+# @param $1 {string} voice_name - Voice model name (e.g., en_US-lessac-medium)
+# @returns None
+# @exitcode 0=voice exists, 1=voice missing
+# @sideeffects None
+verify_voice() {
+  local voice_name="$1"
+  local voice_dir
+  voice_dir=$(get_voice_storage_dir)
+
+  local onnx_file="$voice_dir/${voice_name}.onnx"
+  local json_file="$voice_dir/${voice_name}.onnx.json"
+
+  [[ -f "$onnx_file" ]] && [[ -f "$json_file" ]]
+}
+
+# @function get_voice_path
+# @intent Get absolute path to voice model ONNX file
+# @why Piper binary requires full path to model file
+# @param $1 {string} voice_name - Voice model name
+# @returns Echoes path to .onnx file
+# @exitcode 0=success, 1=voice not found
+# @sideeffects None
+get_voice_path() {
+  local voice_name="$1"
+  local voice_dir
+  voice_dir=$(get_voice_storage_dir)
+
+  local onnx_file="$voice_dir/${voice_name}.onnx"
+
+  if [[ ! -f "$onnx_file" ]]; then
+    echo "❌ Voice model not found: $voice_name" >&2
+    return 1
+  fi
+
+  echo "$onnx_file"
+}
+
+# @function parse_voice_components
+# @intent Extract language, locale, speaker, quality from voice name
+# @why HuggingFace uses directory structure: lang/locale/speaker/quality
+# @param $1 {string} voice_name - Voice name (e.g., en_US-lessac-medium)
+# @returns Sets global variables: LANG, LOCALE, SPEAKER, QUALITY
+# @sideeffects Sets global variables
+# AI NOTE: Voice name format is: lang_LOCALE-speaker-quality
+parse_voice_components() {
+  local voice_name="$1"
+
+  # Extract components from voice name
+  # Format: en_US-lessac-medium
+  #         lang_LOCALE-speaker-quality
+
+  local lang_locale="${voice_name%%-*}"  # en_US
+  local speaker_quality="${voice_name#*-}"  # lessac-medium
+
+  LANG="${lang_locale%%_*}"  # en
+  LOCALE="${lang_locale#*_}"  # US
+  SPEAKER="${speaker_quality%%-*}"  # lessac
+  QUALITY="${speaker_quality#*-}"  # medium
+}
+
+# @function download_voice
+# @intent Download Piper voice model from HuggingFace
+# @why Provide free offline TTS voices
+# @param $1 {string} voice_name - Voice model name
+# @param $2 {string} lang_code - Language code (optional, inferred from voice_name)
+# @returns None
+# @exitcode 0=success, 1=download failed
+# @sideeffects Downloads .onnx and .onnx.json files
+# @edgecases Handles network failures, validates file integrity
+download_voice() {
+  local voice_name="$1"
+  local lang_code="${2:-}"
+
+  local voice_dir
+  voice_dir=$(get_voice_storage_dir)
+
+  # Check if already downloaded
+  if verify_voice "$voice_name"; then
+    echo "✅ Voice already downloaded: $voice_name"
+    return 0
+  fi
+
+  # Parse voice components
+  parse_voice_components "$voice_name"
+
+  # Construct download URLs
+  # Path format: {language}/{language}_{locale}/{speaker}/{quality}/{speaker}-{quality}.onnx
+  local model_path="${LANG}/${LANG}_${LOCALE}/${SPEAKER}/${QUALITY}/${voice_name}"
+  local onnx_url="${PIPER_VOICES_BASE_URL}/${model_path}.onnx"
+  local json_url="${PIPER_VOICES_BASE_URL}/${model_path}.onnx.json"
+
+  echo "📥 Downloading Piper voice: $voice_name"
+  echo "   Source: HuggingFace (rhasspy/piper-voices)"
+  echo "   Size: ~25MB"
+  echo ""
+
+  # Download ONNX model
+  echo "   Downloading model file..."
+  if ! curl -L --progress-bar -o "$voice_dir/${voice_name}.onnx" "$onnx_url"; then
+    echo "❌ Failed to download voice model"
+    rm -f "$voice_dir/${voice_name}.onnx"
+    return 1
+  fi
+
+  # Download JSON config
+  echo "   Downloading config file..."
+  if ! curl -L -s -o "$voice_dir/${voice_name}.onnx.json" "$json_url"; then
+    echo "❌ Failed to download voice config"
+    rm -f "$voice_dir/${voice_name}.onnx" "$voice_dir/${voice_name}.onnx.json"
+    return 1
+  fi
+
+  # Verify file integrity (basic check - file size > 0)
+  if [[ ! -s "$voice_dir/${voice_name}.onnx" ]]; then
+    echo "❌ Downloaded file is empty or corrupt"
+    rm -f "$voice_dir/${voice_name}.onnx" "$voice_dir/${voice_name}.onnx.json"
+    return 1
+  fi
+
+  echo "✅ Voice downloaded successfully: $voice_name"
+  echo "   Location: $voice_dir/${voice_name}.onnx"
+}
+
+# @function list_downloaded_voices
+# @intent Show all locally cached voice models
+# @why Help users see what voices they have available
+# @returns Echoes voice names (one per line)
+# @exitcode 0=success
+# @sideeffects None
+list_downloaded_voices() {
+  local voice_dir
+  voice_dir=$(get_voice_storage_dir)
+
+  echo "📦 Downloaded Piper Voices:"
+  echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+
+  local count=0
+  shopt -s nullglob
+  for onnx_file in "$voice_dir"/*.onnx; do
+    if [[ -f "$onnx_file" ]]; then
+      local voice_name
+      voice_name=$(basename "$onnx_file" .onnx)
+      local file_size
+      file_size=$(du -h "$onnx_file" | cut -f1)
+      echo "  • $voice_name ($file_size)"
+      ((count++))
+    fi
+  done
+  shopt -u nullglob
+
+  if [[ $count -eq 0 ]]; then
+    echo "  (No voices downloaded yet)"
+  fi
+
+  echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+  echo "Total: $count voices"
+}
+
+# AI NOTE: This file manages the lifecycle of Piper voice models
+# Voice models are ONNX files (~20-30MB each) downloaded from HuggingFace
+# Files are cached locally to avoid repeated downloads
+# Project-local storage preferred over global for isolation

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

@@ -0,0 +1,201 @@
+#!/bin/bash
+#
+# @fileoverview ElevenLabs TTS Provider Implementation
+# @context Provider-specific implementation for ElevenLabs API integration
+# @architecture Part of multi-provider TTS system - implements provider interface
+# @dependencies Requires ELEVENLABS_API_KEY, curl, ffmpeg, paplay/aplay/mpg123, jq
+# @entrypoints Called by play-tts.sh router with ($1=text, $2=voice_name)
+# @patterns Follows provider contract: accept text/voice, output audio file path
+# @related play-tts.sh, provider-manager.sh, GitHub Issue #25
+#
+
+# 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"
+source "$SCRIPT_DIR/language-manager.sh"
+
+# @function determine_voice_and_language
+# @intent Resolve voice name/ID and language for multilingual support
+# @why Supports both voice names and direct IDs, plus language-specific voices
+# @param $VOICE_OVERRIDE {string} Voice name or ID (optional)
+# @returns Sets $VOICE_ID and $LANGUAGE_CODE global variables
+# @sideeffects None
+# @edgecases Handles unknown voices, falls back to default
+VOICE_ID=""
+LANGUAGE_CODE="en"  # Default to English
+
+# Get current language setting
+CURRENT_LANGUAGE=$(get_current_language)
+
+# Get language code for API
+LANGUAGE_CODE=$(get_language_code_for_name "$CURRENT_LANGUAGE")
+[[ -z "$LANGUAGE_CODE" ]] && LANGUAGE_CODE="en"
+
+if [[ -n "$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', trying language-specific voice"
+  fi
+fi
+
+# If no override or invalid override, use language-specific voice
+if [[ -z "$VOICE_ID" ]]; then
+  # Try to get voice for current language
+  LANG_VOICE=$(get_voice_for_language "$CURRENT_LANGUAGE" "elevenlabs" 2>/dev/null)
+
+  if [[ -n "$LANG_VOICE" ]] && [[ -n "${VOICES[$LANG_VOICE]}" ]]; then
+    VOICE_ID="${VOICES[$LANG_VOICE]}"
+    echo "🌍 Using $CURRENT_LANGUAGE voice: $LANG_VOICE"
+  else
+    # Fall back to voice manager
+    VOICE_MANAGER_SCRIPT="$(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 default
+    if [[ -z "$VOICE_ID" ]]; then
+      echo "⚠️ No voice configured, using default"
+      VOICE_ID="${VOICES[Aria]}"
+    fi
+  fi
+fi
+
+# @function validate_inputs
+# @intent Check required parameters and API key
+# @why Fail fast with clear errors if inputs missing
+# @exitcode 1=missing text, 2=missing API key
+if [ -z "$TEXT" ]; then
+  echo "Usage: $0 \"text to speak\" [voice_name_or_id]"
+  exit 1
+fi
+
+if [ -z "$API_KEY" ]; then
+  echo "Error: ELEVENLABS_API_KEY not set"
+  echo "Set your API key: export ELEVENLABS_API_KEY=your_key_here"
+  exit 2
+fi
+
+# @function determine_audio_directory
+# @intent Find appropriate directory for audio file storage
+# @why Supports project-local and global storage
+# @returns Sets $AUDIO_DIR global variable
+# @sideeffects None
+# @edgecases Handles missing directories, creates if needed
+# AI NOTE: Check project dir first, then search up tree, finally fall back to global
+if [[ -n "$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"
+
+# @function synthesize_with_elevenlabs
+# @intent Call ElevenLabs API to generate speech
+# @why Encapsulates API call with error handling
+# @param Uses globals: $TEXT, $VOICE_ID, $API_KEY
+# @returns Creates audio file at $TEMP_FILE
+# @exitcode 0=success, 3=API error
+# @sideeffects Creates MP3 file in audio directory
+# @edgecases Handles network failures, API errors, rate limiting
+# Choose model based on language
+if [[ "$LANGUAGE_CODE" == "en" ]]; then
+  MODEL_ID="eleven_monolingual_v1"
+else
+  MODEL_ID="eleven_multilingual_v2"
+fi
+
+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\":\"${MODEL_ID}\",\"language_code\":\"${LANGUAGE_CODE}\",\"voice_settings\":{\"stability\":0.5,\"similarity_boost\":0.75}}" \
+  -o "${TEMP_FILE}"
+
+# @function add_silence_padding
+# @intent Add silence to beginning of audio to prevent WSL static
+# @why WSL audio subsystem cuts off first ~200ms, causing static/clipping
+# @param Uses global: $TEMP_FILE
+# @returns Updates $TEMP_FILE to padded version
+# @sideeffects Modifies audio file, removes original
+# @edgecases Gracefully falls back to unpadded if ffmpeg unavailable
+# Add silence padding to prevent WSL audio static
+if [ -f "${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
+
+  # @function play_audio
+  # @intent Play generated audio file using available player
+  # @why Support multiple audio players (paplay, aplay, mpg123)
+  # @param Uses global: $TEMP_FILE
+  # @sideeffects Plays audio in background
+  # @edgecases Falls through players until one works
+  # Play audio (WSL/Linux) in background to avoid blocking
+  (paplay "${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 - API may be unavailable"
+  echo "Check your API key and network connection"
+  exit 3
+fi