OpenCode Smart Voice Notify

Disclaimer: This project is not built by the OpenCode team and is not affiliated with OpenCode in any way. It is an independent community plugin.

A smart voice notification plugin for OpenCode with multiple TTS engines, native desktop notifications, and an intelligent reminder system.

OpenCode's Built-in Notifications

OpenCode provides basic built-in notifications via its internal:notifications interface. These cover simple toast alerts for common events (task completion, errors, permission requests).

This plugin is for users who want more: multi-engine TTS voice narration, intelligent delayed reminders with exponential backoff, AI-generated contextual messages, Discord/webhook integration, custom sound themes, focus-aware suppression, and per-project sound customization. If you only need a simple desktop toast, OpenCode's built-in notifications may be sufficient. If you want a fully-featured voice notification experience, this plugin delivers it.

Features

Smart TTS Engine Selection

The plugin automatically tries multiple TTS engines in order, falling back if one fails:

1. OpenAI-Compatible (Cloud/Self-hosted) - Any OpenAI-compatible /v1/audio/speech endpoint (Kokoro, LocalAI, Coqui, AllTalk, OpenAI API, etc.)
2. ElevenLabs (Online) - High-quality, anime-like voices with natural expression
3. Edge TTS (Free) - Microsoft's neural voices via Python CLI (recommended) or native npm fallback
4. Windows SAPI (Offline) - Built-in Windows speech synthesis
5. macOS Say (Offline) - Built-in macOS speech synthesis
6. Local Sound Files (Fallback) - Plays bundled MP3 files if all TTS fails

Smart Notification System

• Sound-first mode: Play a sound immediately, then speak a TTS reminder if user doesn't respond
• TTS-first mode: Speak immediately using TTS
• Both mode: Play sound AND speak TTS at the same time
• Sound-only mode: Just play sounds, no TTS

Intelligent Reminders

• Granular Control: Enable or disable notifications and reminders for specific event types (Idle, Permission, Question, Error) via configuration.
• Delayed TTS reminders if user doesn't respond within configurable time
• Follow-up reminders with exponential backoff
• Automatic cancellation when user responds
• Per-notification type delays (permission requests are more urgent)
• Smart Quota Handling: Automatically falls back to free Edge TTS if ElevenLabs quota is exceeded
• Permission Batching: Multiple simultaneous permission requests are batched into a single notification (e.g., "5 permission requests require your attention")
• Question Tool Support (SDK v2 event): Notifies when the agent asks questions and needs user input

AI-Generated Messages

• Dynamic notifications: Use a local AI to generate unique, contextual messages instead of preset static ones
• OpenAI-compatible: Works with Ollama, LM Studio, LocalAI, vLLM, llama.cpp, Jan.ai, or any OpenAI-compatible endpoint
• User-hosted: You provide your own AI endpoint - no cloud API keys required
• Custom prompts: Configure prompts per notification type for full control over AI personality
• Smart fallback: Automatically falls back to static messages if AI is unavailable

System Integration

• Native Desktop Notifications: Windows (Toast), macOS (Notification Center), and Linux (notify-send) support
• Native Edge TTS: No external dependencies (Python/pip) required
• Focus Detection (Cross-platform): Suppresses notifications when terminal is focused (Windows, macOS, Linux)
• Webhook Integration: Receive notifications on Discord or any custom webhook endpoint when tasks finish or need attention
• Themed Sound Packs: Use custom sound collections (e.g., Warcraft, StarCraft) by simply pointing to a directory
• Per-Project Sounds: Assign unique sounds to different projects for easy identification
• Wake monitor from sleep before notifying
• Auto-boost volume if too low
• TUI toast notifications

Installation

Option 1: From npm/Bun (Recommended)

Add to your OpenCode config file (~/.config/opencode/opencode.json):

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["opencode-smart-voice-notify@latest"]
}

Note: OpenCode will automatically install the plugin using your system's package manager (npm or bun).

Option 2: From GitHub

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["github:MasuRii/opencode-smart-voice-notify"]
}

Option 3: Local Development

1. Clone the repository:

   git clone https://github.com/MasuRii/opencode-smart-voice-notify.git
   

2. Reference the local path in your config:

   {
     "plugin": ["file:///path/to/opencode-smart-voice-notify"]
   }
   

Configuration

Automatic Setup

When you first run OpenCode with this plugin installed, it will automatically create:

1. ~/.config/opencode/smart-voice-notify.jsonc - A comprehensive configuration file with all available options fully documented.
2. ~/.config/opencode/assets/*.mp3 - Bundled notification sound files.
3. ~/.config/opencode/logs/ - Debug log folder (created when debug logging is enabled).

The auto-generated configuration includes all advanced settings, message arrays, and engine options, so you don't have to refer back to the documentation for available settings.

Manual Configuration

If you prefer to create the config manually, add a smart-voice-notify.jsonc file in your OpenCode config directory (~/.config/opencode/):

{
    // Master switch to enable/disable the plugin without uninstalling
    "enabled": true,

    // Notification mode: 'sound-first', 'tts-first', 'both', 'sound-only'
    "notificationMode": "sound-first",
    
    // TTS engine: 'openai', 'elevenlabs', 'edge', 'sapi'
    "ttsEngine": "openai",
    "enableTTS": true,
    
    // ElevenLabs settings (get API key from https://elevenlabs.io/app/settings/api-keys)
    "elevenLabsApiKey": "YOUR_API_KEY_HERE",
    "elevenLabsVoiceId": "cgSgspJ2msm6clMCkdW9",  // Jessica - Playful, Bright
    
    // OpenAI-compatible TTS (Kokoro, LocalAI, OpenAI, Coqui, AllTalk, etc.)
    "openaiTtsEndpoint": "http://localhost:8880",
    "openaiTtsVoice": "af_heart",
    "openaiTtsModel": "kokoro",
    
    // Edge TTS settings (free, no API key required)
    "edgeVoice": "en-US-AnaNeural",
    "edgePitch": "+50Hz",
    "edgeRate": "+10%",
    
    // Desktop Notifications
    "enableDesktopNotification": true,
    "desktopNotificationTimeout": 5,
    "showProjectInNotification": true,

    // TTS reminder settings
    "enableTTSReminder": true,
    "ttsReminderDelaySeconds": 30,
    "enableFollowUpReminders": true,
    
    // Focus Detection (suppress notifications when terminal is focused)
    // Default: false (notifications always play)
    "suppressWhenFocused": false,
    "alwaysNotify": false,

    // AI-generated messages (optional - requires local AI server)
    "enableAIMessages": false,
    "aiEndpoint": "http://localhost:11434/v1",
    
    // Webhook settings (optional - works with Discord)
    "enableWebhook": false,
    "webhookUrl": "",
    "webhookUsername": "OpenCode Notify",
    
    // Sound theme settings (optional)
    "soundThemeDir": "", // Path to custom sound theme directory
    
    // Per-project sounds
    "perProjectSounds": false,
    "projectSoundSeed": 0,

    // General settings
    "wakeMonitor": true,
    "forceVolume": false,
    "volumeThreshold": 50,
    "enableToast": true,
    "enableSound": true,
    "debugLog": false
}

For the complete configuration with all TTS engine settings, message arrays, AI prompts, and advanced options, see example.config.jsonc in the plugin directory.

OpenAI-Compatible TTS Setup (Kokoro, LocalAI, OpenAI API, etc.)

For cloud-based or self-hosted TTS using any OpenAI-compatible /v1/audio/speech endpoint:

{
  "ttsEngine": "openai",
  "openaiTtsEndpoint": "http://192.168.86.43:8880",  // Your TTS server
  "openaiTtsVoice": "af_heart",                      // Server-dependent
  "openaiTtsModel": "kokoro",                        // Server-dependent
  "openaiTtsApiKey": "",                             // Optional, if server requires auth
  "openaiTtsSpeed": 1.0                              // 0.25 to 4.0
}

Supported OpenAI-Compatible TTS Servers:

Server	Example Endpoint	Voices
Kokoro	http://localhost:8880	af_heart, af_bella, am_adam, etc.
LocalAI	http://localhost:8080	Model-dependent
AllTalk	http://localhost:7851	Model-dependent
OpenAI	https://api.openai.com	alloy, echo, fable, onyx, nova, shimmer
Coqui	http://localhost:5002	Model-dependent

AI Message Generation (Optional)

If you want dynamic, AI-generated notification messages instead of preset ones, you can connect to a local AI server:

1. Install a local AI server (e.g., Ollama):

   # Install Ollama and pull a model
   ollama pull llama3
   

2. Enable AI messages in your config:

   {
     "enableAIMessages": true,
     "aiEndpoint": "http://localhost:11434/v1",
     "aiModel": "llama3",
     "aiApiKey": "",
     "aiFallbackToStatic": true,
     "enableContextAwareAI": false  // Set to true for personalized messages with project/task context
   }
   

3. The AI will generate unique messages for each notification, which are then spoken by your TTS engine.

4. Context-Aware Messages (optional): Enable enableContextAwareAI for personalized notifications that include project name, task title, and change summary (e.g., "Your work on MyProject is complete!").

Supported AI Servers:

Server	Default Endpoint	API Key
Ollama	http://localhost:11434/v1	Not needed
LM Studio	http://localhost:1234/v1	Not needed
LocalAI	http://localhost:8080/v1	Not needed
vLLM	http://localhost:8000/v1	Use "EMPTY"
Jan.ai	http://localhost:1337/v1	Required

Discord / Webhook Integration (Optional)

Receive remote notifications on Discord or any custom endpoint. This is perfect for long-running tasks when you're away from your computer.

1. Create a Discord Webhook:

   • In Discord, go to Server Settings > Integrations > Webhooks.
   • Click New Webhook, choose a channel, and click Copy Webhook URL.

2. Enable Webhooks in your config:

   {
     "enableWebhook": true,
     "webhookUrl": "https://discord.com/api/webhooks/...",
     "webhookUsername": "OpenCode Notify",
     "webhookEvents": ["idle", "permission", "error", "question"],
     "webhookMentionOnPermission": true
   }
   

3. Features:

   • Color-coded Embeds: Different colors for task completion (green), permissions (orange), errors (red), and questions (blue).
   • Smart Mentions: Automatically @everyone on Discord for urgent permission requests.
   • Rate Limiting: Intelligent retry logic with backoff if Discord's rate limits are hit.
   • Fire-and-forget: Webhook requests never block local sound or TTS playback.

Supported Webhook Events:

Event	Trigger
idle	Agent finished working
permission	Agent needs permission for a tool
error	Agent encountered an error
question	Agent is asking you a question

Custom Sound Themes (Optional)

You can replace individual sound files with entire "Sound Themes" (like the classic Warcraft II or StarCraft sound packs).

1. Set up your theme directory: Create a folder (e.g., ~/.config/opencode/themes/warcraft2/) with the following structure:

   warcraft2/
   ├── idle/          # Sounds for when the agent finishes
   │   ├── job_done.mp3
   │   └── alright.wav
   ├── permission/    # Sounds for permission requests
   │   ├── help.mp3
   │   └── need_orders.wav
   ├── error/         # Sounds for agent errors
   │   └── alert.mp3
   └── question/      # Sounds for agent questions
       └── yes_milord.mp3
   

2. Configure the theme in your config:

   {
     "soundThemeDir": "themes/warcraft2",
     "randomizeSoundFromTheme": true
   }
   

3. Features:

   • Automatic Fallback: If a theme subdirectory or sound is missing, the plugin automatically falls back to your default sound files.
   • Randomization: If multiple sounds are in a subdirectory, the plugin will pick one at random each time (if randomizeSoundFromTheme is true).
   • Relative Paths: Paths are relative to your OpenCode config directory (~/.config/opencode/).

Requirements

Platform Support Matrix

Feature	Windows	macOS	Linux
Sound Playback	✅	✅	✅
TTS (Cloud/Edge)	✅	✅	✅
TTS (Windows SAPI)	✅	❌	❌
TTS (macOS Say)	❌	✅	❌
Desktop Notifications	✅	✅	✅ (req libnotify)
Focus Detection	✅	✅	✅
Webhook Integration	✅	✅	✅
Wake Monitor	✅	✅	✅ (X11/Gnome)
Volume Control	✅	✅	✅ (Pulse/ALSA)

For OpenAI-Compatible TTS

• Any server implementing the /v1/audio/speech endpoint
• Examples: Kokoro, LocalAI, AllTalk, OpenAI API, etc.
• Works with both local self-hosted servers and cloud-based providers.

For ElevenLabs TTS

• ElevenLabs API key (free tier: 10,000 characters/month)
• Internet connection

For Edge TTS

• Internet connection required
• Recommended: Install Python edge-tts for best reliability: pip install edge-tts
• Fallback: Works without Python (uses bundled npm package), but may be less reliable
• If Edge TTS fails, automatically falls back to SAPI (Windows) or Say (macOS)

For Windows SAPI

• Windows OS (uses built-in System.Speech)

For macOS Say

• macOS (uses built-in say command)
• Serves as fallback when other TTS engines fail

For Desktop Notifications

• Windows: Built-in (uses Toast notifications)
• macOS: Built-in (uses Notification Center)
• Linux: Requires notify-send (libnotify)

  # Ubuntu/Debian
  sudo apt install libnotify-bin
  
  # Fedora
  sudo dnf install libnotify
  
  # Arch Linux
  sudo pacman -S libnotify
  

For Sound Playback

• Windows: Built-in (uses Windows Media Player)
• macOS: Built-in (afplay)
• Linux: paplay or aplay

For Focus Detection

Focus detection suppresses sound and desktop notifications when the terminal is focused. Also detects minimized or hidden terminal windows.

Platform	Support	Notes
macOS	✅ Full	Uses AppleScript to detect frontmost application
Windows	✅ Full	Uses native window focus and visibility detection
Linux	✅ Full	Uses xdotool / xprop (X11) or gdbus (Wayland)

Note: If focus detection fails on any platform, notifications are still sent (fail-open behavior). TTS reminders are never suppressed, even when focused, since users may step away after seeing the toast.

For Webhook Notifications

• Discord: Full support for Discord's webhook embed format.
• Generic: Works with any endpoint that accepts a POST request with a JSON body (though formatting is optimized for Discord).
• Rate Limits: The plugin handles HTTP 429 (Too Many Requests) automatically with retries and a 250ms queue delay.

Events Handled

Event	Action
session.idle	Agent finished working - notify user
session.error	Agent encountered an error - alert user
permission.asked	Permission request (SDK v2 event) - alert user
permission.updated	Permission request (SDK v1 event) - alert user
permission.replied	User responded - cancel pending reminders
question.asked	Agent asks question (SDK v2 event) - notify user
question.replied	User answered question - cancel pending reminders
question.rejected	User dismissed question - cancel pending reminders
message.updated	New user message - cancel pending reminders
session.created	New session - reset state

Note: OpenCode exposes both v1 and v2 SDK event surfaces. This plugin handles both to maintain backward compatibility. permission.asked and question.asked are v2 SDK events; permission.updated is a v1 SDK event. Both are supported regardless of which version @opencode-ai/plugin you use (upstream is currently 1.17.3).

Development

To develop on this plugin locally:

1. Clone the repository:

   git clone https://github.com/MasuRii/opencode-smart-voice-notify.git
   cd opencode-smart-voice-notify
   

2. Install dependencies:

   # Using Bun (recommended)
   bun install
   
   # Or using npm
   npm install
   

3. Link to your OpenCode config:

   {
     "plugin": ["file:///absolute/path/to/opencode-smart-voice-notify"]
   }
   

Testing

The plugin uses Bun's built-in test runner for unit and E2E tests.

# Run all tests
bun test

# Run tests with coverage
bun test --coverage

# Run tests in watch mode
bun test --watch

For more detailed testing guidelines and mock usage examples, see CONTRIBUTING.md.

Updating

OpenCode does not automatically update plugins. To update to the latest version:

# Clear the cached plugin
rm -rf ~/.cache/opencode/node_modules/opencode-smart-voice-notify

# Run OpenCode to trigger a fresh install
opencode

License

MIT

Support

• Open an issue on GitHub
• Check the OpenCode docs