OpenCode Voice Plugin

A plugin for OpenCode that provides text-to-speech (TTS) functionality, allowing the AI assistant to speak responses aloud.

Features

• speak tool - AI can call this tool to speak text aloud
• Speech lock support - Automatically checks if user is recording before speaking
• HTTP API support - Connects to any TTS server with a simple REST API
• Fallback support - Falls back to shell script if API is unavailable
• Session events - Optional automatic announcements on session completion
• Configurable - Customize via environment variables

Installation

1. Clone the repository

git clone https://github.com/Olbrasoft/opencode-voice-plugin.git
cd opencode-voice-plugin

2. Install dependencies and build

npm install
npm run build

3. Link to OpenCode plugins directory

# Create plugin directory if it doesn't exist
mkdir -p ~/.config/opencode/plugin

# Create symlink
ln -s /path/to/opencode-voice-plugin/dist/index.js ~/.config/opencode/plugin/voice.js

4. Restart OpenCode

The plugin will be loaded automatically on next startup.

Configuration

Configure the plugin using environment variables:

Variable	Default	Description
OPENCODE_TTS_API_URL	http://localhost:5555/api/speech/speak	TTS API endpoint
OPENCODE_TTS_CAN_SPEAK_URL	http://localhost:5555/api/speech/can-speak	Speech lock check endpoint
OPENCODE_TTS_FALLBACK_SCRIPT	~/voice-assistant/voice-output/tts-api.sh	Fallback shell script
OPENCODE_TTS_ANNOUNCE_IDLE	false	Announce when session becomes idle
OPENCODE_TTS_IDLE_MESSAGE	Úkol dokončen.	Message to speak on idle

Example: Using with EdgeTTS WebSocket Server

This plugin is designed to work with EdgeTTS WebSocket Server (part of VoiceAssistant project):

# Set your TTS API endpoint
export OPENCODE_TTS_API_URL="http://localhost:5555/api/speech/speak"
export OPENCODE_TTS_CAN_SPEAK_URL="http://localhost:5555/api/speech/can-speak"

Example: Using with a shell script

If you prefer using a shell script for TTS:

# Point to your TTS script
export OPENCODE_TTS_FALLBACK_SCRIPT="/path/to/your/tts-script.sh"

Usage

Once installed, the AI can use the speak tool to speak text aloud:

AI: I'll confirm this with voice output.
⚙ speak [text=Úkol byl úspěšně dokončen.]
⚙ „Úkol byl úspěšně dokončen."

The output uses Czech quotation marks („") for clean, readable display.

TTS API Contract

The plugin expects the TTS API to accept POST requests with JSON body:

{
  "text": "Text to speak"
}

The API should return:

• 200 OK on success
• Any other status code on failure (plugin will use fallback)

Development

Build

npm run build

Watch mode

npm run watch

Clean

npm run clean

License

MIT License - see LICENSE for details.

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

Related Projects

• OpenCode - The AI coding assistant
• EdgeTTS - Microsoft Edge TTS Python library