agentvibes 5.12.0 → 5.13.0

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

@@ -31,13 +31,13 @@
 #
 # ---
 #
-# @fileoverview TTS Provider Router with Translation and Language Learning Support
+# @fileoverview TTS Provider Router with Translation Support
 # @context Routes TTS requests to active provider (Piper or macOS) with optional translation
-# @architecture Provider abstraction layer - single entry point for all TTS, handles translation and learning mode
-# @dependencies provider-manager.sh, play-tts-piper.sh, translator.py, translate-manager.sh, learn-manager.sh
+# @architecture Provider abstraction layer - single entry point for all TTS, handles translation
+# @dependencies provider-manager.sh, play-tts-piper.sh, translator.py, translate-manager.sh
 # @entrypoints Called by hooks, slash commands, personality-manager.sh, and all TTS features
 # @patterns Provider pattern - delegates to provider-specific implementations, auto-detects provider from voice name
-# @related provider-manager.sh, play-tts-piper.sh, learn-manager.sh, translate-manager.sh
+# @related provider-manager.sh, play-tts-piper.sh, translate-manager.sh
 #
 # **IMPORTANT: This script should be called inline (NOT in background) in Bash tool**
 # Wait for TTS playback to complete before continuing.
@@ -113,6 +113,12 @@ elif [[ -f "$GLOBAL_MUTE_FILE" ]]; then
   exit 0
 fi
 
+# Resolve a working Python interpreter once (translator and transport config
+# both need it). Sourced AFTER the mute exits so muted calls
+# pay nothing. Windows git-bash frequently has no python3 on PATH — see
+# python-resolver.sh; $PYTHON_BIN is empty when none is usable.
+source "$SCRIPT_DIR/python-resolver.sh"
+
 # Parse named flags (e.g. --llm) before positional arguments.
 # This allows callers to pass: play-tts.sh --llm claude-code "text to speak"
 # Named args are extracted; remaining positional args are shifted into $1/$2/$3.
@@ -241,6 +247,53 @@ if [[ -z "$_PRETEXT" ]]; then
     fi
   done
 fi
+# Dynamic session self-ID: a per-LLM PRETEXT containing the {{session}} token
+# expands to "<LLM> on <Project> in <Terminal>" so multi-session users can tell
+# which tab just spoke. Announced once per session (best-effort — see the key
+# note below); on later utterances the token is dropped so it doesn't prefix
+# every line. Opt-in — only fires when the pretext actually contains the token.
+if [[ "$_PRETEXT" == *"{{session}}"* ]]; then
+  # Session key. There is no true Claude-session id available to a plain Bash
+  # command, so this is best-effort: a terminal-session id if the emulator
+  # exports one (stable for the tab's life), else the parent PID. Worst case the
+  # self-ID repeats or is skipped — cosmetic, and only when a pretext opts in.
+  _SID_RAW="${WT_SESSION:-${TERM_SESSION_ID:-${TMUX:-$PPID}}}"
+  _SID_KEY="$(printf '%s' "${LLM_PROVIDER:-}|$(basename "${CLAUDE_PROJECT_DIR:-$PROJECT_ROOT}")|$_SID_RAW" | tr -c 'A-Za-z0-9' '_')"
+  # Announce-marker dir: an OWNED subdir, per the project's secure-temp rule.
+  # If the dir exists but isn't ours (pre-created symlink attack on a shared
+  # host), skip the marker entirely and just strip the token — never chmod or
+  # write through someone else's path.
+  _ANN_BASE="${XDG_RUNTIME_DIR:-/tmp/agentvibes-$(id -u 2>/dev/null || echo 0)}"
+  _ANN_DIR="$_ANN_BASE/agentvibes-session"
+  _ANN_OK=0
+  if mkdir -p "$_ANN_DIR" 2>/dev/null && chmod 700 "$_ANN_DIR" 2>/dev/null; then
+    _ANN_OWNER="$(stat -c '%u' "$_ANN_DIR" 2>/dev/null || stat -f '%u' "$_ANN_DIR" 2>/dev/null || echo '')"
+    [[ -z "$_ANN_OWNER" || "$_ANN_OWNER" == "$(id -u 2>/dev/null || echo 0)" ]] && _ANN_OK=1
+  fi
+  _ANN_FILE="$_ANN_DIR/announced-$_SID_KEY"
+  # First utterance → replacement is the self-ID; later utterances → empty (drop).
+  if [[ "$_ANN_OK" == "1" && -f "$_ANN_FILE" ]]; then
+    _SESSION_ID=""
+  else
+    _SESSION_ID="$(bash "$SCRIPT_DIR/agentvibes-session-id.sh" "${LLM_PROVIDER:-claude-code}" "${CLAUDE_PROJECT_DIR:-$PROJECT_ROOT}" 2>/dev/null || true)"
+    if [[ "$_ANN_OK" == "1" ]]; then
+      touch "$_ANN_FILE" 2>/dev/null || true
+      # Bound growth: drop announce-markers older than a day (find is optional).
+      find "$_ANN_DIR" -maxdepth 1 -name 'announced-*' -mtime +1 -delete 2>/dev/null || true
+    fi
+  fi
+  if [[ -n "$_SESSION_ID" ]]; then
+    _PRETEXT="${_PRETEXT//"{{session}}"/$_SESSION_ID}"
+  else
+    # Empty replacement: also swallow one adjacent ", " so "Hey {{session}}, ready"
+    # becomes "Hey ready", not "Hey , ready"; then drop any bare token.
+    _PRETEXT="${_PRETEXT//"{{session}}, "/}"
+    _PRETEXT="${_PRETEXT//", {{session}}"/}"
+    _PRETEXT="${_PRETEXT//"{{session}}"/}"
+  fi
+  # Final tidy: collapse doubled/edge commas and runs of whitespace.
+  _PRETEXT="$(printf '%s' "$_PRETEXT" | sed -E 's/,[[:space:]]*,/,/g; s/^[[:space:]]*,[[:space:]]*//; s/[[:space:]]*,[[:space:]]*$//; s/[[:space:]]{2,}/ /g; s/^[[:space:]]+//; s/[[:space:]]+$//')"
+fi
 if [[ -n "$_PRETEXT" ]]; then
   TEXT="${_PRETEXT}, ${TEXT}"
 fi
@@ -332,7 +385,7 @@ case "$ACTIVE_PROVIDER" in
     if [[ "$VOICE_OVERRIDE" =~ ^(af|am|bf|bm|jf|jm|kf|km|zf|zm|ff|fm|hf|hm|if|im|pf|pm|ef|em|nf|nm)_[a-zA-Z0-9]+$ ]]; then
       # Kokoro-shaped voice override (e.g. af_heart) must win over the LLM row's
       # engine column, else it is synthesized as a Piper voice, 404s downloading
-      # a model that doesn't exist, and stays silent. (Grafted from master WIP.)
+      # a model that doesn't exist, and stays silent.
       ACTIVE_PROVIDER="kokoro"
     elif [[ -n "$_LLM_ENGINE" ]]; then
       ACTIVE_PROVIDER="$_LLM_ENGINE"
@@ -356,8 +409,8 @@ fi
 _TRANSPORT_CFG="$HOME/.agentvibes/transport-config.json"
 if [[ "$ACTIVE_PROVIDER" != "ssh-remote" && "$ACTIVE_PROVIDER" != "agentvibes-receiver" && "$ACTIVE_PROVIDER" != "termux-ssh" ]] \
    && [[ -n "$LLM_PROVIDER" && "$LLM_PROVIDER" != "default" ]] \
-   && [[ -f "$_TRANSPORT_CFG" ]] && command -v python3 &>/dev/null; then
-  _LLM_SSH_MODE=$(AGENTVIBES_CFG="$_TRANSPORT_CFG" AGENTVIBES_KEY="$LLM_PROVIDER" python3 - <<'PYEOF'
+   && [[ -f "$_TRANSPORT_CFG" ]] && [[ -n "$PYTHON_BIN" ]]; then
+  _LLM_SSH_MODE=$(AGENTVIBES_CFG="$_TRANSPORT_CFG" AGENTVIBES_KEY="$LLM_PROVIDER" "$PYTHON_BIN" - <<'PYEOF'
 import json, os, sys
 try:
     d = json.load(open(os.environ['AGENTVIBES_CFG'], encoding='utf-8'))
@@ -368,7 +421,7 @@ PYEOF
 )
   if [[ "$_LLM_SSH_MODE" == "remote" ]]; then
     # Redirect this LLM's audio through ssh-remote using its own SSH config
-    _llm_remote_data=$(AGENTVIBES_CFG="$_TRANSPORT_CFG" AGENTVIBES_KEY="$LLM_PROVIDER" python3 - <<'PYEOF'
+    _llm_remote_data=$(AGENTVIBES_CFG="$_TRANSPORT_CFG" AGENTVIBES_KEY="$LLM_PROVIDER" "$PYTHON_BIN" - <<'PYEOF'
 import json, os, sys
 try:
     d = json.load(open(os.environ['AGENTVIBES_CFG'], encoding='utf-8'))
@@ -536,50 +589,8 @@ speak_text() {
   esac
 }
 
-# Note: learn-manager.sh and translate-manager.sh are sourced inside their
-# respective handler functions to avoid triggering their main handlers
-
-# @function handle_learning_mode
-# @intent Speak in both main language and target language for learning
-# @why Issue #51 - Auto-translate and speak twice for immersive language learning
-# @returns 0 if learning mode handled, 1 if not in learning mode
-handle_learning_mode() {
-  # Source learn-manager for learning mode functions
-  source "$SCRIPT_DIR/learn-manager.sh" 2>/dev/null || return 1
-
-  # Check if learning mode is enabled
-  if ! is_learn_mode_enabled 2>/dev/null; then
-    return 1
-  fi
-
-  local target_lang
-  target_lang=$(get_target_language 2>/dev/null || echo "")
-  local target_voice
-  target_voice=$(get_target_voice 2>/dev/null || echo "")
-
-  # Need both target language and voice for learning mode
-  if [[ -z "$target_lang" ]] || [[ -z "$target_voice" ]]; then
-    return 1
-  fi
-
-  # 1. Speak in main language (current voice)
-  speak_text "$TEXT" "$VOICE_OVERRIDE" "$ACTIVE_PROVIDER"
-
-  # 2. Auto-translate to target language
-  local translated
-  # SECURITY: Add timeout to prevent hanging (#134)
-  translated=$(timeout 5 python3 "$SCRIPT_DIR/translator.py" "$TEXT" "$target_lang" 2>/dev/null) || translated="$TEXT"
-
-  # Small pause between languages
-  sleep 0.5
-
-  # 3. Speak translated text with target voice
-  local target_provider
-  target_provider=$(detect_voice_provider "$target_voice")
-  speak_text "$translated" "$target_voice" "$target_provider"
-
-  return 0
-}
+# Note: translate-manager.sh is sourced inside its handler function to avoid
+# triggering its main handler
 
 # @function handle_translation_mode
 # @intent Translate and speak in target language (non-learning mode)
@@ -604,7 +615,7 @@ handle_translation_mode() {
   # Translate text
   local translated
   # SECURITY: Add timeout to prevent hanging (#134)
-  translated=$(timeout 5 python3 "$SCRIPT_DIR/translator.py" "$TEXT" "$translate_to" 2>/dev/null) || translated="$TEXT"
+  translated=$(timeout 5 "${PYTHON_BIN:-python3}" "$SCRIPT_DIR/translator.py" "$TEXT" "$translate_to" 2>/dev/null) || translated="$TEXT"
 
   # Get voice for target language if no override specified
   local voice_to_use="$VOICE_OVERRIDE"
@@ -625,14 +636,8 @@ handle_translation_mode() {
 }
 
 # Mode priority:
-# 1. Learning mode (speaks twice: main + translated)
-# 2. Translation mode (speaks translated only)
-# 3. Normal mode (speaks as-is)
-
-# Try learning mode first (Issue #51)
-if handle_learning_mode; then
-  exit 0
-fi
+# 1. Translation mode (speaks translated only)
+# 2. Normal mode (speaks as-is)
 
 # Try translation mode (Issue #50)
 if handle_translation_mode; then

package/.claude/hooks/provider-catalog.json

@@ -0,0 +1,352 @@
+{
+  "_generated": "DO NOT EDIT — generated from src/services/provider-catalog.js",
+  "_contentHash": "sha256:5e717024f1250a8f54d07291256e532606dba7b4a15ab9589305d1af6bba7f35",
+  "version": "5e717024f1250a8f",
+  "providers": [
+    {
+      "id": "soprano",
+      "engineId": "soprano",
+      "aliases": [],
+      "displayName": "Soprano TTS",
+      "voiceModel": "single",
+      "defaultVoice": "soprano-default",
+      "requires": "pip",
+      "runtime": {
+        "unix": "play-tts-soprano.sh",
+        "windows": "play-tts-soprano.ps1",
+        "darwinOnly": false
+      },
+      "voices": [
+        "soprano-default"
+      ]
+    },
+    {
+      "id": "piper",
+      "engineId": "piper",
+      "aliases": [],
+      "displayName": "Piper TTS",
+      "voiceModel": "discovered",
+      "defaultVoice": "en_US-lessac-medium",
+      "requires": "pip",
+      "runtime": {
+        "unix": "play-tts-piper.sh",
+        "windows": null,
+        "darwinOnly": false
+      },
+      "voices": null
+    },
+    {
+      "id": "kokoro",
+      "engineId": "kokoro",
+      "aliases": [],
+      "displayName": "Kokoro TTS",
+      "voiceModel": "static",
+      "defaultVoice": "af_heart",
+      "requires": "pip",
+      "runtime": {
+        "unix": "play-tts-kokoro.sh",
+        "windows": "play-tts-kokoro.ps1",
+        "darwinOnly": false
+      },
+      "voices": [
+        "af_heart",
+        "af_alloy",
+        "af_aoede",
+        "af_bella",
+        "af_jessica",
+        "af_kore",
+        "af_nicole",
+        "af_nova",
+        "af_river",
+        "af_sarah",
+        "af_sky",
+        "am_adam",
+        "am_echo",
+        "am_eric",
+        "am_fenrir",
+        "am_liam",
+        "am_michael",
+        "am_onyx",
+        "am_puck",
+        "bf_alice",
+        "bf_emma",
+        "bf_isabella",
+        "bf_lily",
+        "bm_daniel",
+        "bm_fable",
+        "bm_george",
+        "bm_lewis",
+        "jf_alpha",
+        "jf_gongitsune",
+        "jf_nezumi",
+        "jf_tebukuro",
+        "jm_kumo",
+        "zf_xiaobei",
+        "zf_xiaoni",
+        "zf_xiaoxiao",
+        "zf_xiaoyi",
+        "zm_yunxi",
+        "zm_yunxia",
+        "zm_yunyang",
+        "ef_dora",
+        "em_alex",
+        "em_santa",
+        "ff_siwis",
+        "hf_alpha",
+        "hm_omega",
+        "if_sara",
+        "im_nicola",
+        "pf_dora",
+        "pm_alex",
+        "pm_santa",
+        "kf_alpha",
+        "km_hyunsu"
+      ]
+    },
+    {
+      "id": "elevenlabs",
+      "engineId": "elevenlabs",
+      "aliases": [],
+      "displayName": "ElevenLabs",
+      "voiceModel": "name-to-id",
+      "defaultVoice": "Sarah",
+      "requires": "api-key",
+      "runtime": {
+        "unix": "play-tts-elevenlabs.sh",
+        "windows": null,
+        "darwinOnly": false
+      },
+      "voices": [
+        {
+          "id": "EXAVITQu4vr4xnSDxMaL",
+          "name": "Sarah",
+          "gender": "Female",
+          "lang": "en-US",
+          "desc": "Mature, reassuring"
+        },
+        {
+          "id": "CwhRBWXzGAHq8TQ4Fs17",
+          "name": "Roger",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Laid-back, casual"
+        },
+        {
+          "id": "FGY2WhTYpPnrIDTdsKH5",
+          "name": "Laura",
+          "gender": "Female",
+          "lang": "en-US",
+          "desc": "Enthusiast, quirky"
+        },
+        {
+          "id": "IKne3meq5aSn9XLyUdCD",
+          "name": "Charlie",
+          "gender": "Male",
+          "lang": "en-AU",
+          "desc": "Deep, confident"
+        },
+        {
+          "id": "JBFqnCBsd6RMkjVDRZzb",
+          "name": "George",
+          "gender": "Male",
+          "lang": "en-GB",
+          "desc": "Warm storyteller"
+        },
+        {
+          "id": "N2lVS1w4EtoT3dr4eOWO",
+          "name": "Callum",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Husky trickster"
+        },
+        {
+          "id": "SAz9YHcvj6GT2YYXdXww",
+          "name": "River",
+          "gender": "",
+          "lang": "en-US",
+          "desc": "Relaxed, neutral"
+        },
+        {
+          "id": "SOYHLrjzK2X1ezoPC6cr",
+          "name": "Harry",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Fierce warrior"
+        },
+        {
+          "id": "TX3LPaxmHKxFdv7VOQHJ",
+          "name": "Liam",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Energetic creator"
+        },
+        {
+          "id": "Xb7hH8MSUJpSbSDYk0k2",
+          "name": "Alice",
+          "gender": "Female",
+          "lang": "en-GB",
+          "desc": "Clear educator"
+        },
+        {
+          "id": "XrExE9yKIg1WjnnlVkGX",
+          "name": "Matilda",
+          "gender": "Female",
+          "lang": "en-US",
+          "desc": "Knowledgable, pro"
+        },
+        {
+          "id": "bIHbv24MWmeRgasZH58o",
+          "name": "Will",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Relaxed optimist"
+        },
+        {
+          "id": "cgSgspJ2msm6clMCkdW9",
+          "name": "Jessica",
+          "gender": "Female",
+          "lang": "en-US",
+          "desc": "Playful, bright"
+        },
+        {
+          "id": "cjVigY5qzO86Huf0OWal",
+          "name": "Eric",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Smooth, trustworthy"
+        },
+        {
+          "id": "hpp4J3VqNfWAUOO0d1Us",
+          "name": "Bella",
+          "gender": "Female",
+          "lang": "en-US",
+          "desc": "Professional, warm"
+        },
+        {
+          "id": "iP95p4xoKVk53GoZ742B",
+          "name": "Chris",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Charming, down-to-earth"
+        },
+        {
+          "id": "nPczCjzI2devNBz1zQrb",
+          "name": "Brian",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Deep, comforting"
+        },
+        {
+          "id": "onwK4e9ZLuTAKqWW03F9",
+          "name": "Daniel",
+          "gender": "Male",
+          "lang": "en-GB",
+          "desc": "Steady broadcaster"
+        },
+        {
+          "id": "pFZP5JQG7iQjIQuC4Bku",
+          "name": "Lily",
+          "gender": "Female",
+          "lang": "en-GB",
+          "desc": "Velvety actress"
+        },
+        {
+          "id": "pNInz6obpgDQGcFmaJgB",
+          "name": "Adam",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Dominant, firm"
+        },
+        {
+          "id": "pqHfZKP75CvOlQylNhV4",
+          "name": "Bill",
+          "gender": "Male",
+          "lang": "en-US",
+          "desc": "Wise, mature"
+        }
+      ]
+    },
+    {
+      "id": "macos",
+      "engineId": "macos",
+      "aliases": [
+        "macos-say",
+        "say"
+      ],
+      "displayName": "macOS Say",
+      "voiceModel": "discovered",
+      "defaultVoice": "Samantha",
+      "requires": "builtin",
+      "runtime": {
+        "unix": "play-tts-macos.sh",
+        "windows": null,
+        "darwinOnly": true
+      },
+      "voices": null
+    },
+    {
+      "id": "windows-sapi",
+      "engineId": "sapi",
+      "aliases": [
+        "sapi"
+      ],
+      "displayName": "Windows SAPI",
+      "voiceModel": "discovered",
+      "defaultVoice": "",
+      "requires": "builtin",
+      "runtime": {
+        "unix": null,
+        "windows": "play-tts-sapi.ps1",
+        "darwinOnly": false
+      },
+      "voices": null
+    },
+    {
+      "id": "windows-piper",
+      "engineId": "piper",
+      "aliases": [
+        "piper"
+      ],
+      "displayName": "Piper TTS",
+      "voiceModel": "discovered",
+      "defaultVoice": "en_US-lessac-medium",
+      "requires": "exe",
+      "runtime": {
+        "unix": null,
+        "windows": "play-tts-windows-piper.ps1",
+        "darwinOnly": false
+      },
+      "voices": null
+    }
+  ],
+  "platforms": {
+    "unix": [
+      "soprano",
+      "piper",
+      "kokoro",
+      "elevenlabs"
+    ],
+    "darwin": [
+      "soprano",
+      "piper",
+      "kokoro",
+      "elevenlabs",
+      "macos"
+    ],
+    "windows": [
+      "soprano",
+      "kokoro",
+      "windows-sapi",
+      "windows-piper"
+    ]
+  },
+  "displayNames": {
+    "soprano": "Soprano TTS",
+    "piper": "Piper TTS",
+    "kokoro": "Kokoro TTS",
+    "elevenlabs": "ElevenLabs",
+    "macos": "macOS Say",
+    "windows-sapi": "Windows SAPI",
+    "windows-piper": "Piper TTS"
+  }
+}